1 2018-11-14 Jakub Jelinek <jakub@redhat.com>
3 PR rtl-optimization/87817
4 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
5 last argument has low 8 bits clear, fold to 0.
7 2018-11-14 Iain Sandoe <iain@sandoe.co.uk>
9 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
10 MAX_OFILE_ALIGNMENT as needed.
12 2018-11-14 Richard Biener <rguenther@suse.de>
14 PR tree-optimization/88019
15 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
16 COND_EXPR arguments to non-trapping overflow.
18 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
20 * config/arc/arc.h (reg_class): Reorder registers classes, remove
21 unused register classes.
22 (REG_CLASS_NAMES): Likewise.
23 (REG_CLASS_CONTENTS): Likewise.
24 (FIXED_REGISTERS): Make lp_count fixed.
25 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
26 (PROGRAM_COUNTER_REGNO): Remove.
27 * config/arc/arc.c (arc_conditional_register_usage): Remove unused
28 register classes, use constants for register numbers, remove
29 reg_class_contents references.
30 (arc_process_double_reg_moves): Add asserts.
31 (arc_secondary_reload): Remove LPCOUNT_REG reference, use
32 lra_in_progress predicate.
33 (arc_init_reg_tables): Remove unused register classes.
34 (arc_register_move_cost): Likewise.
35 (arc_preferred_reload_class): Likewise.
36 (hwloop_optimize): Update rtx patterns involving lp_count
38 (arc_return_address_register): Rename ILINK1, INLINK2 regnums
40 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
41 CHEAP_CORE_REGS. Former one will be used for LRA.
42 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former
43 one will be used for LRA.
44 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former
45 one will be used for LRA.
46 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
47 Former one will be used for LRA.
48 ("f"): Delete constraint.
51 ("l"): Change it from register constraint to constraint.
52 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
53 (movhi_insn): Likewise.
54 (movsi_insn): Update pattern.
57 (stack_tie): Remove 'b' constraint letter.
59 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
60 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
61 (R57_REG, R59_REG, PCL_REG): Likewise.
62 (ILINK1_REGNUM): Renamed to ILINK1_REG.
63 (ILINK2_REGNUM): Renamed to ILINK2_REG.
66 (Rcw): Remove unused reg classes.
67 * config/arc/predicates.md (dest_reg_operand): Just default on
68 register_operand predicate.
69 (mpy_dest_reg_operand): Likewise.
70 (move_dest_operand): Use macros instead of constants.
71 * config/arc/arc.opt (mlra): Switch to lra as default.
73 2018-11-14 Richard Biener <rguenther@suse.de>
75 PR tree-optimization/87974
76 * tree-vect-loop.c (vectorizable_reduction): When computing
77 the vectorized reduction PHI vector type ignore constant
80 2018-11-14 Jakub Jelinek <jakub@redhat.com>
82 PR tree-optimization/87977
83 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
84 stmt, build a new one and replace the old one with it. Formatting fix.
85 Call release_ssa_name (x) if !has_other_use and !delete_div.
86 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
87 verify lhs of stmt is still def.
89 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
91 PR rtl-optimization/87507
92 * lower-subreg.c (operand_for_swap_move_operator): New function.
93 (simple_move): Strip simple operators.
94 (find_pseudo_copy): Likewise.
95 (resolve_operand_for_swap_move_operator): New function.
96 (resolve_simple_move): Strip simple operators and swap operands.
98 2018-11-13 Jakub Jelinek <jakub@redhat.com>
100 PR tree-optimization/87898
101 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
102 (ipa_simd_modify_function_body): Remove debug stmts where the first
103 argument was changed into a non-decl.
105 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
107 PR rtl-optimization/87899
108 * lra-lives.c (start_living): Update white space in comment.
109 (enum point_type): New.
110 (sparseset_contains_pseudos_p): New function.
111 (update_pseudo_point): Likewise.
112 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
113 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts
114 handling. Move early exit after adding conflicts.
115 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit
116 if regno is already live. Remove all handling of program points.
117 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit
118 after adding conflicts. Remove all handling of program points and
119 ignore_reg_for_conflicts.
120 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value
121 and do not guard call to mark_pseudo_live.
122 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value
123 and do not guard call to mark_pseudo_dead.
124 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
125 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
126 Use new function update_pseudo_point. Handle register copies by
127 removing the source register from the live set. Handle INOUT operands.
128 Update to the next program point using the unused_set, dead_set and
130 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
132 2018-11-13 David Malcolm <dmalcolm@redhat.com>
134 * builtins.c: Replace "source_location" with "location_t".
135 * diagnostic-show-locus.c: Likewise.
136 * diagnostic.c: Likewise.
137 * dumpfile.c: Likewise.
138 * gcc-rich-location.h: Likewise.
139 * genmatch.c: Likewise.
140 * gimple.h: Likewise.
141 * gimplify.c: Likewise.
143 * input.h: Likewise. Eliminate the typedef.
144 * omp-expand.c: Likewise.
145 * selftest.h: Likewise.
146 * substring-locations.h (get_source_location_for_substring):
148 (get_location_within_string): ...this.
149 * tree-cfg.c: Replace "source_location" with "location_t".
150 * tree-cfgcleanup.c: Likewise.
151 * tree-diagnostic.c: Likewise.
152 * tree-into-ssa.c: Likewise.
153 * tree-outof-ssa.c: Likewise.
154 * tree-parloops.c: Likewise.
155 * tree-phinodes.c: Likewise.
156 * tree-phinodes.h: Likewise.
157 * tree-ssa-loop-ivopts.c: Likewise.
158 * tree-ssa-loop-manip.c: Likewise.
159 * tree-ssa-phiopt.c: Likewise.
160 * tree-ssa-phiprop.c: Likewise.
161 * tree-ssa-threadupdate.c: Likewise.
162 * tree-ssa.c: Likewise.
163 * tree-ssa.h: Likewise.
164 * tree-vect-loop-manip.c: Likewise.
166 2018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
168 * doc/extend.texi: Fix typo in the weakref description.
170 2018-11-13 Richard Biener <rguenther@suse.de>
172 PR tree-optimization/86991
173 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
174 group building until we have successfully detected the SLP
176 (vect_is_simple_reduction): Remove fixup code here.
178 2018-11-13 David Malcolm <dmalcolm@redhat.com>
180 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
181 (dump_gimple_stmt): Use it.
182 (dump_gimple_stmt_loc): Likewise.
183 (dump_gimple_expr): Likewise.
184 (dump_gimple_expr_loc): Likewise.
185 (dump_generic_expr): Likewise.
186 (dump_generic_expr_loc): Likewise.
187 (dump_printf): Likewise.
188 (dump_printf_loc): Likewise.
189 (dump_dec): Likewise.
190 (dump_dec): Likewise.
191 (dump_hex): Likewise.
192 (dump_symtab_node): Likewise.
193 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
194 Guard dump call with dump_enabled_p.
195 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
196 * graphite-optimize-isl.c (optimize_isl): Likewise.
197 * graphite.c (graphite_transform_loops): Likewise.
198 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
199 * tree-parloops.c (parallelize_loops): Likewise.
200 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
201 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
202 (vect_prune_runtime_alias_test_list): Likewise.
203 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
204 (vect_estimate_min_profitable_iters): Likewise.
205 * tree-vect-slp.c (vect_record_max_nunits): Likewise.
206 (vect_build_slp_tree_2): Likewise.
207 (vect_supported_load_permutation_p): Likewise.
208 (vect_slp_analyze_operations): Likewise.
209 (vect_slp_analyze_bb_1): Likewise.
210 (vect_slp_bb): Likewise.
211 * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
212 * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
213 (pass_slp_vectorize::execute): Likewise.
214 (increase_alignment): Likewise.
216 2018-11-13 David Malcolm <dmalcolm@redhat.com>
219 * ipa-inline.c (report_inline_failed_reason): Guard calls to
220 cl_target_option_print_diff and cl_optimization_print_diff with
223 2018-11-13 David Malcolm <dmalcolm@redhat.com>
225 * doc/invoke.texi (-fsave-optimization-record): Note that the
226 output is compressed.
227 * optinfo-emit-json.cc: Include <zlib.h>.
228 (optrecord_json_writer::write): Compress the output.
230 2018-11-13 Aldy Hernandez <aldyh@redhat.com>
232 * tree-vrp.c (value_range_base::dump): Dump type.
233 Do not use INF nomenclature for 1-bit types.
234 (dump_value_range): Group all variants to common dumping code.
235 (debug): New overloaded functions for value_ranges.
236 (value_range_base::dump): Remove no argument version.
237 (value_range::dump): Same.
239 2018-11-13 Richard Biener <rguenther@suse.de>
241 PR tree-optimization/87931
242 * tree-vect-loop.c (vect_is_simple_reduction): Restrict
243 nested cycles we support to latch computations vectorizable_reduction
246 2018-11-13 Martin Liska <mliska@suse.cz>
248 PR tree-optimization/87885
249 * cfghooks.c (account_profile_record): Rename
251 (profile_record_check_consistency): ... this.
252 Calculate missing num_mismatched_freq_in.
253 (profile_record_account_profile): New function
254 that calculates time and size of a function.
255 * cfghooks.h (struct profile_record): Remove
257 (struct cfg_hooks): Remove after_pass flag.
258 (account_profile_record): Rename to ...
259 (profile_record_check_consistency): ... this.
260 (profile_record_account_profile): New.
261 * cfgrtl.c (rtl_account_profile_record): Remove
263 * passes.c (check_profile_consistency): Do only
265 (account_profile): Calculate size and time of
267 (pass_manager::dump_profile_report): Reformat
269 (execute_one_ipa_transform_pass): Call
270 consistency check before clean upand call account_profile
271 after a clean up is done.
272 (execute_one_pass): Call check_profile_consistency and
273 account_profile instead of using after_pass flag..
274 * tree-cfg.c (gimple_account_profile_record): Likewise.
276 2018-11-13 Martin Liska <mliska@suse.cz>
279 * config/i386/i386.c (ix86_option_override_internal): Error
280 about usage -mabi=ms and -fsanitize={,kernel-}address.
282 2018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
284 * config/aarch64/aarch64.c
285 (aarch64_vectorize_preferred_vector_alignment): Change return type to
287 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
288 alignment being a poly int.
289 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
290 return type to poly_uint64.
291 * target.def (default_preferred_vector_alignment): Likewise.
292 * targhooks.c (default_preferred_vector_alignment): Likewise.
293 * targhooks.h (default_preferred_vector_alignment): Likewise.
294 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
295 (vect_compute_data_ref_alignment): Adapt to vector alignment being a
297 (vect_update_misalignment_for_peel): Likewise.
298 (vect_enhance_data_refs_alignment): Likewise.
299 (vect_find_same_alignment_drs): Likewise.
300 (vect_duplicate_ssa_name_ptr_info): Likewise.
301 (vect_setup_realignment): Likewise.
302 (vect_can_force_dr_alignment_p): Change alignment parameter type to
304 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
305 mask with a compile time variable vector alignment.
306 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
308 (vect_do_peeling): Exit early if vector alignment is not constant.
309 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
311 (vectorizable_store): Likewise.
312 (vectorizable_load): Likweise.
313 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
315 (vect_known_alignment_in_bytes): Adapt to vector alignment being a
317 (vect_can_force_dr_alignment_p): Change alignment parameter type to
320 2018-11-13 Richard Biener <rguenther@suse.de>
322 PR tree-optimization/87962
323 * tree-vect-loop.c (vect_is_simple_reduction): More reliably
324 detect outer reduction for disqualifying in-loop uses.
326 2018-11-13 Richard Biener <rguenther@suse.de>
328 PR tree-optimization/87967
329 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
330 for constants for the scalar loop.
332 2018-11-13 Alan Modra <amodra@gmail.com>
334 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
335 integer constants when -mcmodel=medium.
337 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
339 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
340 * config/arc/elf.h (PROFILE_HOOK): Define.
341 * config/arc/linux.h (PROFILE_HOOK): Likewise.
343 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
345 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
346 jump table data in the text section.
348 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
350 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
351 to fit the eh_return pattern.
352 * config/arc/arc.md (eh_return): Define.
353 (VUNSPEC_ARC_EH_RETURN): Likewise.
354 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
356 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
359 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
361 * common/config/arc/arc-common.c (arc_option_optimization_table):
362 Millicode optimization is default on for size optimizations.
363 * config/arc/arc-protos.h (arc_check_multi): New function.
364 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
365 (ENTER_LEAVE_START_REG): Define.
366 (ENTER_LEAVE_END_REG): Likewise.
367 (arc_override_options): Disable millicode when long calls option
369 (arc_frame_info): Change it from int to bool.
370 (arc_compute_frame_size): Clean up.
371 (arc_save_restore): Remove.
372 (frame_save_reg): New function.
373 (frame_restore_reg): Likewise.
374 (arc_enter_leave_p): Likewise.
375 (arc_save_callee_saves): Likewise.
376 (arc_restore_callee_saves): Likewise.
377 (arc_save_callee_enter): Likewise.
378 (arc_restore_callee_leave): Likewise.
379 (arc_save_callee_milli): Likewise.
380 (arc_restore_callee_milli): Likewise.
381 (arc_expand_prologue): Reimplement to emit enter/leave
383 (arc_expand_epilogue): Likewise.
384 (arc_check_multi): New function.
385 * config/arc/arc.md (push_multi_fp): New pattern.
386 (push_multi_fp_blink): Likewise.
387 (pop_multi_fp): Likewise.
388 (pop_multi_fp_blink): Likewise.
389 (pop_multi_fp_ret): Likewise.
390 (pop_multi_fp_blink_ret): Likewise.
391 * config/arc/arc.opt (mmillicode): Update option.
392 (mcode-density-frame): New option.
393 * config/arc/predicates.md (push_multi_operand): New predicate.
394 (pop_multi_operand): Likewise.
395 * doc/invoke.texi (ARC): Update ARC options information.
398 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
400 * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
401 * config/arc/arc.c (operands_ok_ldd_std): New function.
402 (mem_ok_for_ldd_std): Likewise.
403 (gen_operands_ldd_std): Likewise.
404 * config/arc/arc.md: Add peephole2 rules for std/ldd.
406 2018-11-13 Eric Botcazou <ebotcazou@adacore.com>
408 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
409 test on stack_usage_file.
410 (lang_dependent_init): Do not open the .su file if generating LTO.
412 2018-11-13 Jakub Jelinek <jakub@redhat.com>
414 PR rtl-optimization/87918
415 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
416 simplify_gen_relational rather than simplify_gen_binary.
418 2018-11-13 Richard Biener <rguenther@suse.de>
420 * tree-ssanames.h (set_range_info): Use value_range_base.
421 (get_range_info): Likewise.
422 * tree-ssanames.c (set_range_info): Likewise.
423 (get_range_info): Likewise.
424 * tree-vrp.c (value_range_base::union_helper): Split
425 out common parts of value_range[_base]::union_.
426 (value_range_base::union_): Update.
427 (value_range::union_): Likewise.
428 (determine_value_range_1): Use value_range_base.
429 (determine_value_range): Likewise.
430 * tree-vrp.h (value_range_base::union_helper): Move ...
431 (value_range::union_helper): ... from here.
433 2018-11-13 Alan Modra <amodra@gmail.com>
435 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
437 (rs6000_secondary_reload_gpr): Don't call find_replacement.
439 2018-11-13 Sandra Loosemore <sandra@codesourcery.com>
442 * doc/invoke.texi (Optimize Options): Clarify that the
443 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
444 apply to data cache size.
446 2018-11-13 Alan Modra <amodra@gmail.com>
448 * config/rs6000/predicates.md (logical_const_operand),
449 (logical_operand): Correct comment.
450 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
452 2018-11-13 Alan Modra <amodra@gmail.com>
454 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
456 2018-11-13 Alan Modra <amodra@gmail.com>
458 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
460 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
461 covered by alternative.
462 (movcc_internal1): Ignore h for register preference.
463 (mov<mode>_hardfloat64): Likewise.
464 (mov<mode>_softfloat): Ignore c, l, h for register preference.
466 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
468 PR preprocessor/47823
469 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
472 * doc/extend.texi (Pragmas): Note additional pragmas documented
475 2018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org>
479 * tree-scalar-evolution.c (expression_expensive_p): Make BUILTIN POPCOUNT
480 as expensive when backend does not define it.
482 2018-11-12 Fredrik Noring <noring@nocrew.org>
484 * config.gcc: Update with-llsc defaults for MIPS r5900.
486 2018-11-12 Martin Liska <mliska@suse.cz>
488 PR gcov-profile/87442
489 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
491 * doc/invoke.texi: Document them.
492 * tree-profile.c (parse_profile_filter): New.
493 (parse_profile_file_filtering): Likewise.
494 (release_profile_file_filtering): Likewise.
495 (include_source_file_for_profile): Likewise.
496 (tree_profiling): Filter source files based on the
499 2018-11-12 Martin Liska <mliska@suse.cz>
502 * doc/extend.texi: Add missing values for __builtin_cpu_is and
503 __builtin_cpu_supports for x86 target.
505 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
508 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
510 (Arithmetic): Likewise for "lo_sum".
512 2018-11-12 Renlin Li <renlin.li@arm.com>
515 * dse.c (get_stored_val): Add check for compile-time constantness
518 2018-11-12 Sudakshina Das <sudi.das@arm.com>
520 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
521 (ARMv8_5a): New fgroup.
522 (armv8.5-a): New arch.
523 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
525 * config/arm/arm-tables.opt: Regenerate.
526 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
527 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
528 * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
529 Add matching rules for -march=armv8.5-a and extensions.
530 * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
531 Add sb and predres to all armv8-a except armv8.5-a.
533 2018-11-12 Richard Biener <rguenther@suse.de>
535 * tree-vrp.h (value_range[_base]::set): Make public. Provide
536 overload for single value.
537 (value_range[_base]::set_nonnull): New.
538 (value_range[_base]::set_null): Likewise.
539 (value_range): Document bitmap copying behavior, mark
540 copy constructor and assignment operator deleted.
541 (value_range::move): New.
542 (value_range::set_and_canonicalize): Default bitmap to zero.
543 (set_value_range_to_nonnull): Remove.
544 (set_value_range_to_null): Likewise.
545 (set_value_range): Likewise.
546 (set_value_range_to_value): Likewise.
547 (extract_range_from_unary_expr): Work on value_range_base.
548 (extract_range_from_binary_expr_1): Likewise. Rename to...
549 (extract_range_from_binary_expr): ... this.
550 * tree-vrp.c (value_range::update): Clear equiv bitmap
552 (value_range::move): New, move equiv bitmap.
553 (value_range_base::set_undefined): Avoid assignment.
554 (value_range::set_undefined): Likewise.
555 (value_range_base::set_varying): Likewise.
556 (value_range::set_varying): Likewise.
557 (set_value_range): Remove.
558 (value_range_base::set): New overload for value.
559 (value_range::set): Likewise.
560 (set_value_range_to_nonnull): Remove.
561 (value_range_base::set_nonnull): New.
562 (value_range::set_nonnull): Likewise.
563 (set_value_range_to_null): Remove.
564 (value_range_base::set_null): New.
565 (value_range::set_null): Likewise.
566 (range_is_null): Work on value_range_base.
567 (range_is_nonnull): Likewise.
568 (ranges_from_anti_range): Likewise.
569 (extract_range_into_wide_ints): Likewise.
570 (extract_range_from_multiplicative_op): Likewise.
571 (extract_range_from_binary_expr): Likewise. Update for API changes.
572 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF
574 (value_range::intersect_helper): Avoid copy and assignment.
575 (value_range::union_helper): Likewise.
576 (determine_value_range_1): Adjust.
577 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
578 Avoid assignment by using move.
579 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
580 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
582 * tree-ssanames.c (get_range_info): Likewise.
583 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
584 * vr-values.c (vr_values::get_value_range): Adjust.
585 (vr_values::update_value_range): Likewise.
586 (symbolic_range_based_on_p): Work on value_range_base.
587 (vr_values::extract_range_from_binary_expr): Use value_range_base.
588 (vr_values::extract_range_from_unary_expr): Likewise.
589 (vr_values::extract_range_from_cond_expr): Avoid assignment.
590 (vr_values::extract_range_from_comparison): Adjust.
591 (vr_values::check_for_binary_op_overflow): Use value_range_base.
592 (vr_values::extract_range_basic): Adjust.
593 (vr_values::adjust_range_with_scev): Likewise.
594 (vr_values::vrp_visit_assignment_or_call): Likewise.
595 (vr_values::get_vr_for_comparison): Change API to avoid
596 assignment and copy construction.
597 (vr_values::compare_name_with_value): Adjust accordingly.
598 (vr_values::compare_names): Likewise.
599 (vr_values::extract_range_from_phi_node): Avoid assignment and
600 bogus in-place modify of equiv bitmap.
601 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
602 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
603 for extract_range_from_unary_expr API change.
604 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
606 2018-11-12 Eric Botcazou <ebotcazou@adacore.com>
608 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
610 2018-11-12 Richard Biener <rguenther@suse.de>
612 * tree-vrp.h (value_range_base::symbolic_p,
613 value_range_base::constant_p, value_range_base::zero_p,
614 value_range_base::singleton_p): Move from value_range.
615 (value_range::dump): Add.
616 * gimple-ssa-evrp-analyze.c
617 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
618 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
619 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
621 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
622 (value_range::constant_p): Likewise.
623 (value_range::singleton_p): Likewise.
624 (value_range_base::dump): Add.
625 (set_value_range_to_undefined): Remove.
626 (set_value_range_to_varying): Likewise.
627 (range_int_cst_p): Take value_range_base argument.
628 (range_int_cst_singleton_p): Likewise.
629 (value_range_constant_singleton): Likewise.
630 (vrp_set_zero_nonzero_bits): Likewise.
631 (extract_range_from_multiplicative_op): Use set_varying.
632 (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
633 (extract_range_from_unary_expr): Likewise.
634 (dump_value_range_base): Change to overload of dump_value_range.
635 (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
636 (vrp_prop::visit_stmt): Likewise.
637 (value_range::intersect_helper): Likewise.
638 (value_range::union_helper): Likewise.
639 (determine_value_range_1): Likewise.
641 2018-11-12 Richard Biener <rguenther@suse.de>
643 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
644 (set_value_range_to_null): Likewise.
645 * vr-values.c (vr_values::extract_range_from_comparison):
646 Clear equiv for constant singleton ranges.
648 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
650 * config/i386/sse.md: Combine VFIXUPIMM* patterns
651 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
652 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
653 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
654 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
655 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
656 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
658 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
661 * doc/extend.texi (Common Type Attributes): For the align type
662 attribute, copy language about decreasing alignment from the
663 corresponding variable attribute.
665 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
667 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
670 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
673 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
674 -frtti and -fno-rtti code.
676 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
679 * doc/extend.texi (Other Builtins): Document probability associated
680 with __builtin_expect.
682 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
685 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
686 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
687 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
688 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
690 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
692 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
694 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
696 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
698 2018-11-11 Richard Biener <rguenther@suse.de>
700 * tree-vrp.h (class value_range_base): New base class for
701 value_range containing all but the m_equiv member.
702 (dump_value_range_base): Add.
703 (range_includes_zero_p): Work on value_range_base.
704 * tree-vrp.c (value_range_base::set): Split out base handling
706 (value_range::set): this.
707 (value_range::set_equiv): New.
708 (value_range_base::value_range_base): New constructors.
709 (value_range_base::check): Split out base handling from...
710 (value_range::check): this.
711 (value_range::equal_p): Refactor in terms of
712 ignore_equivs_equal_p which is now member of the base.
713 (value_range_base::set_undefined): New.
714 (value_range_base::set_varying): Likewise.
715 (value_range_base::dump):Split out base handling from...
716 (value_range::dump): this.
717 (value_range_base::set_and_canonicalize): Split out base handling
719 (value_range::set_and_canonicalize): this.
720 (value_range_base::union_): New.
721 * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
723 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
724 instead of value_range everywhere.
725 (ipcp_vr_lattice::print): Use dump_value_range_base.
726 (ipcp_vr_lattice::meet_with): Adjust.
727 (ipcp_vr_lattice::meet_with_1): Likewise.
728 (ipa_vr_operation_and_type_effects): Likewise.
729 (propagate_vr_across_jump_function): Likewise.
730 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
731 (ipa_get_value_range): Likewise.
732 (ipa_set_jfunc_vr): Likewise.
733 (ipa_compute_jump_functions_for_edge): Likewise.
735 2018-11-10 Sandra Loosemore <sandra@codesourcery.com>
738 * doc/invoke.texi (Optimize Options): Add @opindex entries
739 for the positive forms of -fno-xxx and -mno-xxx options
740 that were lacking them.
742 2018-11-10 Segher Boessenkool <segher@kernel.crashing.org>
744 * combine.c (make_more_copies): Only make an intermediate copy if the
745 dest of a move is a pseudo.
747 2018-11-09 Maya Rashish <coypu@sdf.org>
750 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
751 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
753 2018-11-09 Sandra Loosemore <sandra@codesourcery.com>
757 * doc/invoke.texi (Optimize Options): Clarify default behavior
758 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
760 2018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de>
762 PR tree-optimization/87940
763 * expr.c (string_constant): Don't strip NOPS in subexpressions.
764 Fold PLUS_EXPR correctly.
766 2018-11-09 Ilya Leoshkevich <iii@linux.ibm.com>
769 * config/s390/s390.c (s390_safe_relative_long_p): New function.
770 (annotate_constant_pool_refs): Skip insns which support
772 (annotate_constant_pool_refs_1): New helper function.
773 (find_constant_pool_ref): Skip insns which support relative
775 (find_constant_pool_ref_1): New helper function.
776 (replace_constant_pool_ref): Skip insns which support
778 (replace_constant_pool_ref_1): New helper function.
779 (s390_mainpool_start): Adapt to the new signature.
780 (s390_mainpool_finish): Likewise.
781 (s390_chunkify_start): Likewise.
782 (s390_chunkify_finish): Likewise.
783 (pass_s390_early_mach::execute): Likewise.
784 (s390_prologue_plus_offset): Likewise.
785 (s390_emit_prologue): Likewise.
786 (s390_emit_epilogue): Likewise.
788 2018-11-09 Jakub Jelinek <jakub@redhat.com>
790 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
791 but unsupported lastprivate with conditional modifier.
793 2018-11-09 Jeff Law <law@redhat.com>
795 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
796 unused argument better. Add gcc_unreachable to silence warning.
798 2018-11-09 Martin Sebor <msebor@redhat.com>
801 * attribs.c (has_attribute): New helper function.
802 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
803 * attribs.h (decls_mismatched_attributes): Declare.
804 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
805 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
806 * common.opt (-Wattribute-alias): Take an argument.
807 (-Wno-attribute-alias): New option.
808 * doc/extend.texi (Common Function Attributes): Document copy.
809 (Common Variable Attributes): Same.
810 * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
811 (-Wattribute-alias): Document new option argument.
813 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
815 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
816 matching criteria. Remove unused array initializer.
818 2018-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
819 Jinsong Ji <jji@us.ibm.com>
821 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
822 constraints by introducing a new temporary.
823 (_mm_cvtss_si64): Likewise.
825 2018-11-09 Martin Liska <mliska@suse.cz>
827 * common.opt: Add -fipa-stack-alignment flag.
828 * doc/invoke.texi: Document it.
829 * final.c (rest_of_clean_state): Guard stack
832 2018-11-09 Martin Liska <mliska@suse.cz>
834 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
836 (ipa_discover_variable_flags): ... this.
837 * common.opt: Come up with new flag -fipa-reference-addressable.
838 * doc/invoke.texi: Document it.
839 * ipa-reference.c (propagate): Call the renamed fn.
840 * ipa-visibility.c (whole_program_function_and_variable_visibility):
842 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
844 (ipa_discover_variable_flags): ... this. Discover
845 non-addressable variables only with the newly added flag.
846 * opts.c: Enable the newly added flag with -O1 and higher
849 2018-11-09 David Malcolm <dmalcolm@redhat.com>
851 * json.cc (selftest::test_writing_literals): Fix comment.
853 2018-11-09 Martin Liska <mliska@suse.cz>
855 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
856 string to a stack buffer.
857 (aarch64_parse_cpu): Likewise.
858 (aarch64_parse_tune): Likewise.
860 2018-11-09 Richard Biener <rguenther@suse.de>
862 PR tree-optimization/87953
863 * tree-vect-loop.c (vectorizable_reduction): For analysis
864 always pass ops[0] to vectorizable_condition.
866 2018-11-09 Stafford Horne <shorne@gmail.com>
867 Richard Henderson <rth@twiddle.net>
868 Joel Sherrill <joel@rtems.org>
870 * common/config/or1k/or1k-common.c: New file.
871 * config/or1k/*: New.
872 * config.gcc (or1k*-*-*): New.
873 * configure.ac (or1k*-*-*): New test for openrisc tls.
874 * configure: Regenerated.
875 * doc/install.texi: Document OpenRISC triplets.
876 * doc/invoke.texi: Document OpenRISC arguments.
877 * doc/md.texi: Document OpenRISC.
879 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
881 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
882 (arm7tdmi-s): Delete CPU.
883 (arm710t): Add aliases for arm720t and arm740t.
884 (arm720t, arm740t): Delete CPUs.
885 (arm920t): Add aliases for arm920, arm922t and arm940t.
886 (arm920, arm922t, arm940t): Delete CPUs.
887 (arm10tdmi): Add alias for arm1020t.
888 (arm1020t): Delete CPU.
889 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
890 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
891 (arm10e): Add aliases for arm1020e and arm1022e.
892 (arm1020e, arm1022e): Delete CPU.
893 * config/arm/arm.md (generic_sched): Remove entries that are now
895 (generic_vfp): Likewise.
896 * config/arm/arm1020e.md: Simplify tuning selection based on alias
898 * config/arm/arm-tune.md: Regenerated.
899 * config/arm/arm-tables.opt: Regenerated.
901 2018-11-09 Richard Biener <rguenther@suse.de>
903 PR tree-optimization/87621
904 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
905 op with only phi inputs.
906 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
907 (ch_base::copy_headers): Run CSE on copied loop headers.
908 (pass_ch_vect::process_loop_p): Simplify.
910 2018-11-09 Alexandre Oliva <oliva@adacore.com>
912 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
913 for non-w64 x86_64 biarch.
915 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
917 PR rtl-optimization/86438
918 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
919 of in_b for the compare if in_b is SET_DEST.
922 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
925 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
927 * tree-vrp.c (value_range::check): Do not access internals
929 (value_range::singleton_p): Same.
930 (value_range::type): Same.
931 (vrp_finalize): Use value_range API.
933 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
935 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
937 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
939 * vr-values.c (vr_values::get_value_range): Use value_range API
940 instead of piecing together ranges.
941 (vr_values::update_value_range): Same.
943 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
945 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
946 of performing ad-hoc calculations.
947 * tree-ssanames.c (set_range_info): New overloaded function
948 accepting value_range &.
949 (get_range_info): Same.
950 * tree-ssanames.h (set_range_info_raw): Remove.
951 (set_range_info): New prototype.
952 (get_range_info): Same.
953 * tree-vrp.h (value_range::null_p): Rename to zero_p.
954 * tree-vrp.c (value_range::null_p): Same.
956 2018-11-09 Jan Hubicka <jh@suse.cz>
958 * tree.c (fld_type_variant_equal_p): Test user align flag.
959 (flt_type_variant): Copy user align flag.
960 (fld_incomplete_type_of): Clear it.
962 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
964 * config/arm/neon.md (div<mode>3): New pattern.
966 2018-11-08 Andi Kleen <ak@linux.intel.com>
968 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
969 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
970 (ix86_handle_option): Handle OPT_mptwrite.
971 * config/i386/cpuid.h (bit_PTWRITE): Add.
972 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
973 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
974 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
975 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
976 (ix86_option_override_internal): Handle PTA_PTWRITE.
977 (ix86_valid_target_attribute_inner_p): Define ptwrite.
978 (def_builtin2): Force UINT64 to be 64bit only.
979 * config/i386/i386.h (TARGET_PTWRITE): Add.
980 (TARGET_PTWRITE_P): Add.
982 * config/i386/i386.md: Define ptwrite.
983 * config/i386/i386.opt: Add -mptwrite.
984 * config/i386/immintrin.h (_ptwrite64): Add.
986 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
987 * doc/invoke.texi: Document -mptwrite.
989 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
991 PR rtl-optimization/87600
992 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
993 * lra-constraints.c (process_alt_operands): Skip illegal hard
994 register usage. Prefer reloading non hard register operands.
996 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
999 * doc/invoke.texi (Optimize Options): Clarify default behavior
1000 for -fno-sched-interblock and -fno-sched-spec.
1002 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
1004 * collect2.c (linker_select): Add USE_LLD_LD.
1005 (ld_suffixes): Add ld.lld.
1006 (main): Handle -fuse-ld=lld.
1007 * common.opt (-fuse-ld=lld): New option.
1008 * doc/invoke.texi (-fuse-ld=lld): Document.
1009 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1011 2018-11-08 Paul Koning <ni1d@arrl.net>
1013 * config/pdp11/constraints.md: Add "Z" series constraints for use
1014 with pre-dec and post-inc addressing.
1015 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1016 (pdp11_expand_operands): Add int argument (word count).
1017 (pdp11_sp_frame_offset): Delete.
1018 (pdp11_cmp_length): New function.
1019 (pushpop_regeq): New function.
1020 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1022 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1024 (pdp11_initial_elimination_offset): Ditto.
1025 (pdp11_expand_operands): Add word count argument. Bugfixes.
1026 (output_move_multiple): Change how pointer adjustment is done.
1027 (pdp11_gen_int_label): Correct format.
1028 (output_ascii): Ditto.
1029 (pdp11_asm_output_var): Add code for DEC assembler case.
1030 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1032 (legitimate_const_double_p): Ditto.
1033 (pdp11_register_move_cost): Adjust for new register classes.
1034 (pdp11_regno_reg_class): Ditto.
1035 (expand_block_move): Delete.
1036 (pushpop_regeq): New function.
1037 (pdp11_legitimate_address_p): Bugfix in check for constant
1039 (pdp11_sp_frame_offset): Delete.
1040 (pdp11_reg_save_size): New helper function for new frame layout.
1041 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1042 (pdp11_expand_shift): Bugfix in check for constant shift count.
1043 (pdp11_shift_length): Ditto.
1044 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1045 (pdp11_cmp_length): New function.
1046 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1047 some compile options.
1048 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1049 (CALL_USED_REGISTERS): Ditto.
1050 (ELIMINABLE_REGS): Ditto.
1051 (REGISTER_NAMES): Ditto.
1052 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1054 (REG_CLASS_NAMES): Ditto.
1055 (REG_CLASS_CONTENTS): Ditto. Also remove
1056 HARD_FRAME_POINTER_REGNUM.
1057 (CPU_REG_CLASS): New macro.
1058 (CLASS_MAX_NREGS): Adjust for new register classes.
1059 (FUNCTION_PROFILER): Make no-op.
1060 (may_call_alloca): Remove unused declaration.
1061 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1062 (ASM_OUTPUT_SKIP): Fix format.
1063 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1064 (HARD_FRAME_POINTER_REGNUM): Remove.
1066 (*rts): Rename. Remove epilogue related checks.
1067 (cmpsi, cmpdi): New insn.
1068 (cbranch<mode>4): Change to apply to SI and DI modes as well.
1069 (mov<mode>): Change constraints to enforce that push/pop
1070 destination cannot use the same register as source.
1071 (*mov<mode><cc_cc>): Ditto.
1072 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1073 at assembly output rather than as RTL expander.
1074 (zero_extendqihi2): Bugfix in check for same registers.
1075 (adddi3_nocc): Bugfix in check for constant operand.
1076 (addsi3_nocc): Ditto.
1077 (subdi3_nocc): Ditto.
1078 (subsi3_nocc): Ditto.
1079 (negdi2_nocc): Copy input to pdp11_expand_operands.
1080 (negsi2_nocc): Ditto.
1081 (bswap2_nocc): Ditto.
1082 * config/pdp11/pdp11.opt (mlra): Fix documentation.
1083 * config/pdp11/t-pdp11: Use -Os.
1085 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
1087 * config/arm/parsecpu.awk (/alias/): New parsing rule.
1088 (/begin cpu/): Check that the cpu name hasn't been previously defined.
1089 (gen_comm_data): Print out CPU alias tables.
1090 (check_cpu): Match aliases when checking the CPU name.
1091 * config/arm/arm-protos.h (cpu_alias): New structure.
1092 (cpu_option): Add entry for aliases.
1093 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1094 strongarm1100 and strongarm1110.
1095 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1096 (config/arm/arm-generic.md): Remove redundant references to
1097 strongarm110, strongarm1100 and strongarm1110.
1098 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1099 Scan aliases for additional hints.
1100 (arm_parse_cpu_option_name): Also match a cpu name against the list
1102 * config/arm/arm-tables.opt: Regenerated.
1103 * config/arm/arm-tune.md: Regenerated.
1105 2018-11-08 Jakub Jelinek <jakub@redhat.com>
1107 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1108 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1109 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1110 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1111 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1112 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1113 * gengtype.c (open_base_files): Add omp-general.h.
1114 * gimple.c (gimple_build_omp_critical):
1115 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
1116 gimple_omp_taskgroup_set_clauses.
1117 (gimple_build_omp_atomic_load): Add mo argument, call
1118 gimple_omp_atomic_set_memory_order.
1119 (gimple_build_omp_atomic_store): Likewise.
1120 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1121 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1123 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1124 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1125 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1126 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
1127 different value for GF_OMP_ATOMIC_NEED_VALUE.
1128 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1130 (struct gimple_statement_omp_single_layout): And remove here.
1131 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1132 than gimple_statement_omp_single_layout.
1133 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1135 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1136 (gimple_omp_subcode): Formatting fix.
1137 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1138 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1139 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1140 gimple_omp_teams_host, gimple_omp_teams_set_host,
1141 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1142 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1143 gimple_omp_taskgroup_set_clauses): New inline functions.
1144 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1145 (gimple_build_omp_atomic_store): Likewise.
1146 (gimple_omp_atomic_seq_cst_p): Remove.
1147 (gimple_omp_atomic_memory_order): New function.
1148 (gimple_omp_atomic_set_seq_cst): Remove.
1149 (gimple_omp_atomic_set_memory_order): New function.
1150 (gimple_build_omp_taskgroup): Add clauses argument.
1151 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1152 (dump_gimple_omp_task): Print taskwait with depend clauses.
1153 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1154 dump_omp_atomic_memory_order.
1155 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1156 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1157 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1158 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1159 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1160 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1161 ORT_UNTIED_TASKLOOP enumerators.
1162 (enum gimplify_defaultmap_kind): New.
1163 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1164 target_map_pointers_as_0len_arrays members, add defaultmap.
1165 (new_omp_context): Initialize defaultmap member.
1166 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1167 (maybe_fold_stmt): Don't fold even in host teams regions.
1168 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1169 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
1170 ctx->omp_firstprivatize_variable.
1171 (omp_add_variable): Don't add private/firstprivate for VLAs in
1173 (omp_default_clause): Print "taskloop" rather than "task" if
1175 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1176 Handle new defaultmap clause kinds.
1177 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
1178 iterator to be lastprivate or private. Fix up diagnostics if linear
1179 is used on collapse>1 simd iterator.
1180 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1181 (gimplify_omp_depend): New function.
1182 (gimplify_scan_omp_clauses): Add shared clause on parallel for
1183 combined parallel master taskloop{, simd} if taskloop has
1184 firstprivate, lastprivate or reduction clause. Handle
1185 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
1186 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
1187 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
1188 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
1189 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
1191 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
1192 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
1193 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
1194 OMP_CLAUSE_{TASK,IN}_REDUCTION.
1195 (gimplify_omp_task): Handle taskwait with depend clauses.
1196 (gimplify_omp_for): Add shared clause on parallel for combined
1197 parallel master taskloop{, simd} if taskloop has firstprivate,
1198 lastprivate or reduction clause. Use ORT_TASKLOOP or
1199 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
1200 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
1201 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
1202 __for_end and __for_range temporaries on OMP_PARALLEL for
1203 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
1204 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1205 sandwiched in between two taskloops.
1206 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1207 instead of ctx->omp_firstprivatize_variable.
1208 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1209 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
1210 host teams, use gimplify_and_return_first etc. for body like
1211 for target or target data constructs, and at the end call
1212 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1213 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1214 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1215 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1216 gimple_omp_atomic_set_seq_cst calls.
1217 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1218 case, handle taskgroup clauses.
1219 * lto-streamer-out.c (hash_tree): Handle
1220 OMP_CLAUSE_{TASK,IN}_REDUCTION.
1221 * Makefile.in (GTFILES): Add omp-general.h.
1222 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1223 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1224 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1225 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1226 BUILT_IN_GOMP_LOOP_DOACROSS_START,
1227 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1228 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1229 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1230 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1231 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1232 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1233 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1234 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1235 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1236 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1237 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1238 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1239 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1240 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1241 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1242 * omp-expand.c (workshare_safe_to_combine_p): Return false for
1243 non-worksharing loops.
1244 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1245 (determine_parallel_type): Don't combine parallel with worksharing
1246 which has _reductemp_ clause.
1247 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1248 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1249 if there is nonmonotonic modifier or if there is no modifier and no
1250 ordered clause. For dynamic and guided schedule without monotonic
1251 and nonmonotonic modifier, default to nonmonotonic.
1252 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
1253 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1255 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1256 there are any reduction clauses.
1257 (expand_taskwait_call): New function.
1258 (expand_teams_call): New function.
1259 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1260 expand_teams_call for it. Formatting fix. Handle taskwait with
1262 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
1263 of worksharing loops with task reductions.
1264 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1265 expansion of worksharing loops with task reductions.
1266 (expand_omp_sections): Handle expansion of sections with task
1268 (expand_omp_synch): For host teams call expand_omp_taskreg.
1269 (omp_memory_order_to_memmodel): New function.
1270 (expand_omp_atomic_load, expand_omp_atomic_store,
1271 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1272 instead of gimple_omp_atomic_seq_cst_p.
1273 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1274 depend clauses as a standalone directive.
1275 * omp-general.c (enum omp_requires): New variable.
1276 (omp_extract_for_data): Initialize have_reductemp member. Allow
1277 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1278 GT_EXPR loops depending on incr sign. Formatting fixes.
1279 * omp-general.h (struct omp_for_data): Add have_reductemp member.
1280 (enum omp_requires): New enum.
1281 (omp_requires_mask): Declare.
1282 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1284 * omp-low.c (struct omp_context): Add task_reductions and
1285 task_reduction_map fields.
1286 (is_host_teams_ctx): New function.
1287 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1288 (use_pointer_for_field): Use is_global_var instead of
1289 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1291 (build_outer_var_ref): Ignore taskgroup outer contexts.
1292 (delete_omp_context): Release task_reductions and task_reduction_map.
1293 (scan_sharing_clauses): Don't add any fields for reduction clause on
1294 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
1295 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1296 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
1297 Handle OMP_CLAUSE_NONTEMPORAL.
1298 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1300 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1301 clauses with task modifier.
1302 (scan_omp_task): Handle taskwait with depend clauses.
1303 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1304 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1305 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1306 (scan_omp_for): Fix comment formatting.
1307 (scan_omp_teams): Handle host teams constructs.
1308 (check_omp_nesting_restrictions): Allow teams with no outer
1309 OpenMP context. Adjust diagnostics for teams strictly nested into
1310 some explicit OpenMP construct other than target. Allow OpenMP atomics
1311 inside of simd regions.
1312 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1313 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1314 taskreg_nesting_level while scanning host teams construct.
1315 (task_reduction_read): New function.
1316 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1317 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1318 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
1319 second argument create_tmp_var if it is NULL. Don't ignore shared
1320 clauses in is_host_teams_ctx contexts. Handle
1321 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1323 (lower_reduction_clauses): Ignore reduction clauses with task
1324 modifier. Remove second argument create_tmp_var if it is NULL.
1325 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1326 (lower_send_clauses): Ignore reduction clauses with task modifier.
1327 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
1328 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
1329 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1330 rely that it is the last stmt in body so far. Ignore outer taskgroup
1332 (omp_task_reductions_find_first, omp_task_reduction_iterate,
1333 lower_omp_task_reductions): New functions.
1334 (lower_omp_sections): Handle reduction clauses with taskgroup
1335 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
1336 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1337 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
1339 (lower_omp_taskgroup): Handle taskgroup reductions.
1340 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1341 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1342 (lower_depend_clauses): If there are any
1343 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1344 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
1345 seen, assume lowering is done already and return early. Set kind
1346 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1347 (lower_omp_taskreg): Handle reduction clauses with task modifier on
1348 parallel construct. Handle reduction clause on taskloop construct.
1349 Handle taskwait with depend clauses.
1350 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1351 for host teams constructs.
1352 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1353 nontemporal and _reductemp_ clause entries.
1354 (omp_clause_code_name): Likewise.
1355 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1356 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1357 * tree-core.h (enum omp_clause_code): Add
1358 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1359 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1360 (struct tree_base): Add omp_atomic_memory_order field into union.
1361 Remove OMP_ATOMIC_SEQ_CST comment.
1362 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1363 and OMP_CLAUSE_DEPEND_DEPOBJ.
1364 (struct tree_omp_clause): Add subcode.defaultmap_kind.
1365 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1366 OpenMP constructs with body and clauses operands.
1367 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1368 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1369 (OMP_TASKGROUP_CLAUSES): Define.
1370 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1371 OMP_CLAUSE__LOOPTEMP_.
1372 (OMP_ATOMIC_SEQ_CST): Remove.
1373 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1374 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1375 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1376 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1377 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1378 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1379 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1380 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1381 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1382 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1384 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1385 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1386 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1387 (convert_local_omp_clauses): Likewise. Remove useless test.
1388 * tree-parloops.c (create_call_for_reduction_1): Pass
1389 OMP_MEMORY_ORDER_RELAXED as new argument to
1390 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1391 * tree-pretty-print.c (dump_omp_iterators): New function.
1392 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1393 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
1394 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1395 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
1396 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
1397 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
1398 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1399 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1400 (dump_omp_atomic_memory_order): New function.
1401 (dump_generic_node): Use it. Print taskgroup clauses. Print
1402 taskwait with depend clauses.
1403 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1404 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1405 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1406 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1407 write_ts_omp_clause_tree_pointers): Likewise.
1409 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1412 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1413 "-missed", "-note", and "-all" sub-options.
1414 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1416 (can_early_inline_edge_p): Likewise.
1417 (want_early_inline_function_p): Likewise.
1418 (want_inline_self_recursive_call_p): Likewise.
1419 (recursive_inlining): Likewise.
1420 (inline_small_functions): Likewise.
1421 (flatten_function): Likewise.
1422 (ipa_inline): Likewise.
1423 (inline_always_inline_functions): Likewise.
1424 (early_inline_small_functions): Likewise.
1425 (early_inliner): Likewise.
1426 * tree-inline.c (expand_call_inline): Likewise.
1428 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1430 * pretty-print.c (pp_format): Handle %f.
1431 (selftest::test_pp_format): Add test of %f.
1432 * pretty-print.h (pp_double): New macro.
1434 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1436 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1437 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1438 (dump_pretty_printer::decode_format): Implement "%C" for
1440 (selftest::test_capture_of_dump_calls): Rename "where" to
1441 "stmt_loc". Convert test_decl to a function decl and set its
1442 location. Add a symbol_table_test RAII instance and a
1443 cgraph_node, using it to test "%C" and dump_symtab_node.
1445 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
1448 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1450 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1452 * cgraph.c: Include "selftest.h".
1453 (saved_symtab): New variable.
1454 (selftest::symbol_table_test::symbol_table_test): New ctor.
1455 (selftest::symbol_table_test::~symbol_table_test): New dtor.
1456 (selftest::test_symbol_table_test): New test.
1457 (selftest::cgraph_c_tests): New.
1458 * cgraph.h (saved_symtab): New decl.
1459 (selftest::symbol_table_test): New class.
1460 * selftest-run-tests.c (selftest::run_tests): Call
1461 selftest::cgraph_c_tests.
1462 * selftest.h (selftest::cgraph_c_tests): New decl.
1464 2018-11-08 Richard Biener <rguenther@suse.de>
1466 * tree-data-ref.h (lambda_int): New typedef.
1467 (lambda_vector_gcd): Adjust.
1468 (lambda_vector_new): Likewise.
1469 (lambda_matrix_new): Likewise.
1470 * tree-data-ref.c (print_lambda_vector): Adjust.
1472 2018-11-08 Richard Biener <rguenther@suse.de>
1474 PR tree-optimization/87929
1475 * tree-complex.c (expand_complex_comparison): Clean EH.
1477 2018-11-08 Martin Liska <mliska@suse.cz>
1479 * doc/extend.texi: Reword.
1480 * predict.c (expr_expected_value_1): Likewise.
1482 2018-11-08 Richard Biener <rguenther@suse.de>
1484 PR tree-optimization/87913
1485 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1486 of extreme values to ordered comparisons.
1488 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1491 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1493 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1495 * doc/invoke.texi: Remove leading dash from @opindex entries
1496 throughout the file.
1498 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1501 * doc/invoke.texi (Option Summary): Add -e and --entry.
1502 (Link Options): Likewise.
1504 2018-11-07 Nathan Sidwell <nathan@acm.org>
1507 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1508 --disable-checking bootstrap.
1510 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
1512 * configure: Regenerated.
1514 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1517 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1518 to the mode of the widest field iff the widest field has mode class
1519 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1522 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
1524 * common.opt: Add -Wattribute-warning.
1525 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1526 * expr.c (expand_expr_real_1): Add new attribute to warning_at
1527 call to allow user configure behavior of "warning" attribute.
1529 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
1531 * target.def: Put @: after every vs., e.g., and i.e. where it is
1532 followed by whitespace.
1533 * doc/extend.texi: Ditto.
1534 * doc/fragments.texi: Ditto.
1535 * doc/gimple.texi: Ditto.
1536 * doc/implement-c.texi: Ditto.
1537 * doc/install.texi: Ditto.
1538 * doc/invoke.texi: Ditto.
1539 * doc/md.texi: Ditto.
1540 * doc/plugins.texi: Ditto.
1541 * doc/rtl.texi: Ditto.
1542 * doc/sourcebuild.texi: Ditto.
1543 * doc/tm.texi.in: Ditto.
1544 * doc/ux.texi: Ditto.
1545 * doc/tm.texi: Regenerate.
1547 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1549 * config/arm/arm-cpus.in (ares): New entry.
1550 * config/arm/arm-tables.opt: Regenerate.
1551 * config/arm/arm-tune.md: Likewise.
1552 * doc/invoke.texi (ARM Options): Document ares.
1554 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1556 * config/aarch64/aarch64-cores.def (ares): Define.
1557 * config/aarch64/aarch64-tune.md: Regenerate.
1558 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1560 2018-11-07 Jan Hubicka <jh@suse.cz>
1562 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1563 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1564 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1567 2018-11-07 Richard Biener <rguenther@suse.de>
1569 PR tree-optimization/87914
1570 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1572 (vectorizable_reduction): Handle shifts and rotates by dispatching
1573 to vectorizable_shift.
1574 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1575 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
1577 (vectorizable_shift): Export and handle being called as
1579 (vect_analyze_stmt): Call vectorizable_shift after
1580 vectorizable_reduction.
1581 * tree-vectorizer.h (vectorizable_shift): Declare.
1583 2018-11-07 Jan Hubicka <jh@suse.cz>
1585 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1586 than const decls in TREE_VALUE of enum.
1587 (dump_type_inheritance_graph): Improve duplicate dumping.
1588 (free_enum_values): New.
1589 (build_type_inheritance_graph): Use it.
1590 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
1591 which are not main variants or not ODR types.
1592 (verify_type_variant): Expect variants to have no TYPE_VALUES.
1594 2018-11-07 Richard Biener <rguenther@suse.de>
1596 * ipa-inline.c (want_inline_small_function_p): Compute
1597 big_speedup_p lazily and last.
1599 2018-11-07 Jan Hubicka <jh@suse.cz>
1601 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
1602 building incomplete variant of complete type.
1603 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
1604 variant of complete type.
1606 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1608 * config/mips/mips.c: Fix typo in documentation of
1609 mips_loongson_ext2_prefetch_cookie.
1610 (mips_option_override): fix brain twister logical.
1611 * config/mips/mips.h: Fix typo in documentation of
1612 ISA_HAS_CTZ_CTO and define pattern.
1613 * config/mips/mips.md (prefetch): Hoist EXT2 above
1615 (prefetch_indexed): Hoist EXT2 above the EXT block.
1617 2018-11-07 Jan Hubicka <jh@suse.cz>
1619 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
1620 return and parameter types of function and method types.
1621 (free_lang_data_in_cgraph): Update.
1623 2018-11-07 Martin Liska <mliska@suse.cz>
1625 PR rtl-optimization/87868
1626 * postreload-gcse.c (eliminate_partially_redundant_load): Set
1627 threshold to max_count if we would overflow.
1628 * profile-count.h: Make max_count a public constant.
1630 2018-11-07 Martin Liska <mliska@suse.cz>
1632 * mem-stats.h: Fix GNU coding style.
1634 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1636 * config/mips/gs264e.md: New.
1637 * config/mips/mips-cpus.def: Define gs264e.
1638 * config/mips/mips-tables.opt: Regenerate.
1639 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1641 (mips_issue_rate): Add support for gs264e.
1642 (mips_multipass_dfa_lookahead): Likewise.
1643 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
1644 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
1645 (MIPS_ASE_MSA_SPEC): New.
1646 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
1647 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
1648 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
1649 * config/mips/mips.md: Include gs264e.md.
1650 (processor): Add gs264e.
1651 * config/mips/mips.opt (MSA): Use Mask instead of Var.
1652 * doc/invoke.texi: Add gs264e to supported architectures.
1654 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1656 * config/mips/gs464e.md: New.
1657 * config/mips/mips-cpus.def: Define gs464e.
1658 * config/mips/mips-tables.opt: Regenerate.
1659 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
1661 (mips_issue_rate): Add support for gs464e.
1662 (mips_multipass_dfa_lookahead): Likewise.
1663 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
1664 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
1665 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
1666 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
1667 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
1668 * config/mips/mips.md: Include gs464e.md.
1669 (processor): Add gs464e.
1670 * doc/invoke.texi: Add gs464e to supported architectures.
1672 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1674 * config/mips/loongson3a.md: Rename to ...
1675 * config/mips/gs464.md: ... here.
1676 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
1677 as an alias of gs464 processor.
1678 * config/mips/mips-tables.opt: Regenerate.
1679 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
1680 instead of PROCESSOR_LOONGSON_3A.
1681 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
1683 (mips_option_override): Enable MMI and EXT for gs464.
1684 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
1685 Rename TUNE_LOONGSON_3A to TUNE_GS464.
1686 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
1687 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
1688 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
1690 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
1691 (processor): Add gs464;
1692 * doc/invoke.texi: Add gs464 to supported architectures.
1694 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1696 * config/mips/mips-protos.h
1697 (mips_loongson_ext2_prefetch_cookie): New prototype.
1698 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
1699 (mips_option_override): Enable TARGET_LOONGSON_EXT when
1700 TARGET_LOONGSON_EXT2 is true.
1701 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
1702 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
1703 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
1704 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
1705 TARGET_LOONGSON_EXT2.
1706 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
1707 (define_insn "ctz<mode>2"): New insn pattern.
1708 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
1709 (define_insn "prefetch_indexed_<mode>"): Include
1710 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
1711 * config/mips/mips.opt (-mloongson-ext2): Add option.
1712 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
1714 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1716 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
1717 __mips_loongson_ext.
1718 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
1719 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
1721 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
1722 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
1723 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
1724 instead of TARGET_LOONGSON_3A.
1725 * config/mips/mips.opt (-mloongson-ext): Add option.
1726 * gcc/doc/invoke.texi (-mloongson-ext): Document.
1728 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
1730 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
1731 * config/mips/loongson.md: Move to ...
1732 * config/mips/loongson-mmi.md: here; Adjustment.
1733 * config/mips/loongson.h: Move to ...
1734 State as deprecated. Include loongson-mmiintrin.h for back
1735 compatibility and warning.
1736 * config/mips/loongson-mmiintrin.h: ... here.
1737 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
1738 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
1739 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1740 (mips_option_override): Make sure MMI use hard float;
1741 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
1742 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
1743 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
1744 TARGET_LOONGSON_VECTORS.
1745 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
1746 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
1747 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
1748 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
1750 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
1751 TARGET_LOONGSON_VECTORS.
1752 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
1753 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
1754 (Loongson MMI patterns): Include loongson-mmi.md instead of
1756 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
1757 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
1759 2018-11-07 Richard Biener <rguenther@suse.de>
1762 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
1763 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
1765 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
1767 PR rtl-optimization/87874
1768 * lra.c (lra_substitute_pseudo): Do not create a subreg for
1771 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
1773 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
1774 if not in indexed or indirect form.
1775 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
1776 (bswap<mode>2_store): Ditto.
1778 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
1780 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
1783 2018-11-06 Richard Biener <rguenther@suse.de>
1785 PR tree-optimization/86850
1786 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
1787 instead of src.m_vec.
1789 2018-11-06 Jan Hubicka <jh@suse.cz>
1791 * tree.c (fld_simplified_type_name): Break out form ...
1792 (free_lang_data_in_type): ... here.
1793 (fld_type_variant_equal_p): Use it.
1795 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1797 * config/default-d.c: Include memmodel.h.
1799 * config/sol2-d.c: New file.
1800 * config/t-sol2 (sol2-d.o): New rule.
1801 * config.gcc <*-*-solaris2*>: Set d_target_objs,
1802 target_has_targetdm.
1804 2018-11-06 Jan Hubicka <jh@suse.cz>
1806 * tree.c (fld_type_variant): Also copy alignment; be sure that
1807 new variant is equal.
1809 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
1812 * config/s390/s390.md: Add relative_long attribute.
1814 2018-11-06 Jan Hubicka <jh@suse.cz>
1816 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
1817 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
1818 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
1819 stream TYPE_NEEDS_CONSTRUCTING.
1820 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
1821 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
1823 2018-11-06 Richard Biener <rguenther@suse.de>
1825 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
1827 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
1829 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1831 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
1832 REG_CLASS_CONTENTS[GEN_REGS].
1833 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
1835 2018-11-06 Jan Hubicka <jh@suse.cz>
1837 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
1839 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1842 * config/sol2.h (ASAN_CC1_SPEC): Define.
1843 (LD_WHOLE_ARCHIVE_OPTION): Define.
1844 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
1845 (ASAN_REJECT_SPEC): Provide default.
1846 (LIBASAN_EARLY_SPEC): Define.
1847 (LIBTSAN_EARLY_SPEC): Define.
1848 (LIBLSAN_EARLY_SPEC): Define.
1849 * config/i386/sol2.h (CC1_SPEC): Redefine.
1850 (ASAN_REJECT_SPEC): Define.
1852 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
1853 (TARGET_ASAN_SHADOW_OFFSET): Define.
1854 (sparc_asan_shadow_offset): New function.
1855 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
1856 (ASAN_REJECT_SPEC): Define.
1858 2018-11-06 Jan Hubicka <jh@suse.cz>
1860 * tree.c (fld_type_variant): Copy canonical type.
1861 (fld_incomplete_type_of): Check that canonical types looks sane;
1862 copy canonical type.
1863 (verify_type): Accept when incomplete type has complete canonical type.
1865 2018-11-06 Jan Hubicka <jh@suse.cz>
1867 * tree.c (free_lang_data): Reset overwite_assembler_name,
1868 print_xnode, print_decl, print_type and print_identifier of
1871 2018-11-06 Richard Biener <rguenther@suse.de>
1873 PR tree-optimization/87889
1874 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1875 Do nothing if old and new arg are the same
1877 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
1880 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
1881 attributes for operands 3 and 4.
1883 2018-11-06 Richard Biener <rguenther@suse.de>
1886 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
1887 matching bitfield insertion.
1889 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
1891 * auto-inc-dec.c: Include valtrack.h. Improve comments.
1892 (reg_next_debug_use): New.
1893 (attempt_change): Propagate adjusted expression into affected
1895 (merge_in_block): Track uses in debug insns.
1896 (pass_inc_dec::execute): Allocate and release
1899 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
1901 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
1902 (_mm512_fixupimm_round_pd): Update parameters and builtin.
1903 (_mm512_maskz_fixupimm_round_pd): Ditto.
1904 (_mm512_fixupimm_round_ps): Ditto.
1905 (_mm512_maskz_fixupimm_round_ps): Ditto.
1906 (_mm_fixupimm_round_sd): Ditto.
1907 (_mm_maskz_fixupimm_round_sd): Ditto.
1908 (_mm_fixupimm_round_ss): Ditto.
1909 (_mm_maskz_fixupimm_round_ss): Ditto.
1910 (_mm512_fixupimm_pd): Ditto.
1911 (_mm512_maskz_fixupimm_pd): Ditto.
1912 (_mm512_fixupimm_ps): Ditto.
1913 (_mm512_maskz_fixupimm_ps): Ditto.
1914 (_mm_fixupimm_sd): Ditto.
1915 (_mm_maskz_fixupimm_sd): Ditto.
1916 (_mm_fixupimm_ss): Ditto.
1917 (_mm_maskz_fixupimm_ss): Ditto.
1918 (_mm512_mask_fixupimm_round_pd): Update builtin.
1919 (_mm512_mask_fixupimm_round_ps): Ditto.
1920 (_mm_mask_fixupimm_round_sd): Ditto.
1921 (_mm_mask_fixupimm_round_ss): Ditto.
1922 (_mm512_mask_fixupimm_pd): Ditto.
1923 (_mm512_mask_fixupimm_ps): Ditto.
1924 (_mm_mask_fixupimm_sd): Ditto.
1925 (_mm_mask_fixupimm_ss): Ditto.
1926 * config/i386/avx512vlintrin.h:
1927 (_mm256_fixupimm_pd): Update parameters and builtin.
1928 (_mm256_maskz_fixupimm_pd): Ditto.
1929 (_mm256_fixupimm_ps): Ditto.
1930 (_mm256_maskz_fixupimm_ps): Ditto.
1931 (_mm_fixupimm_pd): Ditto.
1932 (_mm_maskz_fixupimm_pd): Ditto.
1933 (_mm_fixupimm_ps): Ditto.
1934 (_mm_maskz_fixupimm_ps): Ditto.
1935 (_mm256_mask_fixupimm_pd): Update builtin.
1936 (_mm256_mask_fixupimm_ps): Ditto.
1937 (_mm_mask_fixupimm_pd): Ditto.
1938 (_mm_mask_fixupimm_ps): Ditto.
1939 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
1940 * config/i386/i386-builtin.def: Update builtin definitions.
1941 * config/i386/i386.c: Handle new builtin types and remove useless ones.
1942 * config/i386/sse.md: Update VFIXUPIMM* patterns.
1943 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1944 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1945 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
1946 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1947 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1948 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
1949 * config/i386/subst.md:
1950 (round_saeonly_sd_mask_operand4): Add new subst_attr.
1951 (round_saeonly_sd_mask_op4): Ditto.
1952 (round_saeonly_expand_operand5): Ditto.
1953 (round_saeonly_expand): Update.
1955 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
1957 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
1959 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
1961 PR rtl-optimization/87871
1962 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
1964 2018-11-05 Paul Koning <ni1d@arrl.net>
1966 * doc/sourcebuild.texi (target attributes): Document new "inf"
1967 effective target keyword.
1969 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1971 * config/s390/s390.c (s390_register_move_cost): Increase costs for
1972 moves involving the CC reg.
1974 2018-11-05 Richard Biener <rguenther@suse.de>
1976 PR tree-optimization/87873
1977 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
1979 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
1980 * tree-vect-loop.c (vect_transform_loop): When splitting the
1981 loop exit also create forwarder PHIs for constants.
1982 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
1983 Handle constant to_arg, add extra checking we match up the correct
1986 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1988 * config/s390/s390.md: QImode and HImode for load on condition.
1990 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
1992 * config/s390/predicates.md: Fix typo.
1993 * config/s390/s390.md: Allow immediates for load on condition.
1995 2018-11-05 Martin Liska <mliska@suse.cz>
1997 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
1998 * bitmap.h (struct bitmap_usage): Likewise.
1999 * ggc-common.c (SCALE): Remove.
2001 (struct ggc_usage): Use SIZE_AMOUNT. And update
2003 * ggc-page.c (SCALE): Remove.
2004 (STAT_LABEL): Remove.
2005 (ggc_print_statistics): Use SIZE_AMOUNT.
2006 * gimple.h (SCALE): Remove.
2008 * input.c (ONE_K): Remove.
2011 (STAT_LABEL): Likewise.
2012 (FORMAT_AMOUNT): Likewise.
2013 (dump_line_table_statistics): Use SIZE_AMOUNT.
2014 * mem-stats.h (struct mem_usage): Likewise.
2015 * rtl.c (dump_rtx_statistics): Likewise.
2016 (rtx_alloc_counts): Change type to size_t.
2017 (rtx_alloc_sizes): Likewise.
2018 (rtx_count_cmp): New.
2019 (dump_rtx_statistics): Sort first based on counts.
2020 * tree.c (tree_nodes_cmp): New.
2021 (tree_codes_cmp): New.
2022 (dump_tree_statistics): Sort first based on counts.
2023 * system.h (ONE_K): New.
2025 (SIZE_SCALE): Likewise.
2026 (SIZE_LABEL): Likewise.
2027 (SIZE_AMOUNT): Likewise.
2028 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2029 * tree-dfa.c (dump_dfa_stats): Likewise.
2030 * tree-phinodes.c (phinodes_print_statistics): Likewise.
2031 * tree-ssanames.c (ssanames_print_statistics): Likewise.
2032 * tree.c (dump_tree_statistics): Likewise.
2033 * vec.c (struct vec_usage): Likewise.
2034 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2035 trigger a -Werror=format-overflow with
2036 --enable-gather-detailed-stats.
2038 2018-11-05 Martin Liska <mliska@suse.cz>
2040 * mem-stats.h (mem_alloc_description::release_instance_overhead):
2042 * vec.c (struct vec_usage): Register m_element_size.
2043 (vec_prefix::register_overhead): New arguments: elements and
2045 (vec_prefix::release_overhead): Subtract elements.
2046 * vec.h (struct vec_prefix): Change signature.
2047 (va_heap::reserve): Pass proper arguments.
2048 (va_heap::release): Likewise.
2050 2018-11-05 Martin Liska <mliska@suse.cz>
2052 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2056 2018-11-05 Richard Biener <rguenther@suse.de>
2058 * tree-scalar-evolution.h (final_value_replacement_loop): Update
2060 * tree-scalar-evolution.c (final_value_replacement_loop): Return
2061 whether anything was done.
2062 (scev_const_prop): Remove constant propagation part, fold
2064 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2065 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2068 2018-11-05 Jakub Jelinek <jakub@redhat.com>
2070 PR tree-optimization/87859
2071 * gimple-ssa-store-merging.c (struct merged_store_group): Add
2072 only_constants and first_nonmergeable_order members.
2073 (merged_store_group::merged_store_group): Initialize them.
2074 (merged_store_group::do_merge): Clear only_constants member if
2075 adding something other than INTEGER_CST store.
2076 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2077 stores with order >= first_nonmergeable_order. Use
2078 merged_store->only_constants instead of always recomputing it.
2079 Set merged_store->first_nonmergeable_order if we've skipped any
2080 stores. Attempt to merge overlapping INTEGER_CST stores that
2081 we would otherwise skip.
2084 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2086 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
2089 * config/i386/emmintrin.h (__v16qs): New to cope with option
2091 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2092 (_mm_cmplt_epi8): Likewise.
2093 (_mm_cmpgt_epi8): Likewise.
2095 2018-11-05 Richard Biener <rguenther@suse.de>
2097 PR rtl-optimization/87852
2098 * fwprop.c (use_killed_between): Only consider single-defs of the
2099 use whose definition statement dominates the use.
2101 2018-11-05 Martin Liska <mliska@suse.cz>
2104 * doc/invoke.texi: Remove options that are
2105 not disabled with -Os.
2107 2018-11-05 Martin Liska <mliska@suse.cz>
2110 * doc/extend.texi: Update constrain about the last argument
2111 of __builtin_expect_with_probability.
2113 2018-11-05 Martin Liska <mliska@suse.cz>
2116 * predict.c (expr_expected_value_1): Verify
2117 that last argument is a real constants and emit
2120 2018-11-05 Martin Liska <mliska@suse.cz>
2122 PR gcov-profile/77698
2123 * ipa-profile.c (ipa_profile): Adjust hotness threshold
2126 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2128 PR tree-optimization/86572
2129 * builtins.c (c_strlen): Handle negative offsets in a safe way.
2131 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2133 PR tree-optimization/87672
2134 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2135 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2137 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
2140 * cfgexpand.c (pass_expand::execute): Move the call to
2141 finish_eh_generation in front of the call to expand_stack_alignment.
2143 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2145 * common/config/i386/i386-common.c (processor_alias_table): Add
2147 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2148 (case ${target}): Add znver2.
2149 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2150 -march=native recognize znver2 processors.
2151 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2152 * config/i386/i386.c (m_znver2): New definition.
2153 (m_ZNVER): New definition.
2154 (m_AMD_MULTIPLE): Includes m_znver2.
2155 (processor_cost_table): Add znver2 entry.
2156 (processor_target_table): Add znver2 entry.
2157 (get_builtin_code_for_version): Set priority for
2159 (processor_model): Add M_AMDFAM17H_ZNVER2.
2160 (arch_names_table): Ditto.
2161 (ix86_reassociation_width): Include znver2.
2162 * config/i386/i386.h (TARGET_znver2): New definition.
2163 (struct ix86_size_cost): Add TARGET_ZNVER2.
2164 (enum processor_type): Add PROCESSOR_ZNVER2.
2165 * config/i386/i386.md (define_attr "cpu"): Add znver2.
2166 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2167 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2168 (ix86_adjust_cost): Add znver2.
2169 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
2170 * gcc/doc/extend.texi: Add details about znver2.
2171 * gcc/doc/invoke.texi: Add details about znver2.
2173 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
2177 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2180 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2182 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2183 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2185 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
2187 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2188 shrink-wrapping when -mtrack-speculation.
2190 2018-11-02 Richard Biener <rguenther@suse.de>
2192 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2193 (pop_cost_one_pair): Do not free pair.
2194 (pop_best_coalesce): Likewise.
2195 (create_coalesce_list): Initialize obstack.
2196 (delete_coalesce_list): Free obstack.
2197 (find_coalesce_pair): Obstack-allocate coalesce pairs.
2198 (add_cost_one_coalesce): Likewise.
2199 (struct live_track): Remove bitmap pointer indirections.
2200 (new_live_track): Adjust.
2201 (delete_live_track): Likewise.
2202 (live_track_remove_partition): Likewise.
2203 (live_track_add_partition): Likewise.
2204 (live_track_live_p): Likewise.
2205 (live_track_process_def): Likewise.
2206 (live_track_clear_base_vars): Likewise.
2208 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2210 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2211 * configure: Regenerate.
2212 * config.in: Regenerate.
2213 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2214 HAVE_AS_REGISTER_PSEUDO_OP guard.
2215 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2216 !HAVE_AS_REGISTER_PSEUDO_OP support.
2217 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2220 2018-11-02 Richard Biener <rguenther@suse.de>
2222 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2225 2018-11-02 Richard Biener <rguenther@suse.de>
2227 PR tree-optimization/87776
2228 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2229 executable when iterating but running into rpo-vn-max-loop-depth
2230 and not eliding the iteration.
2232 2018-11-30 Jan Hubicka <jh@suse.cz>
2234 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2237 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
2239 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2241 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2242 Combine with rs6000_address_for_fpconvert.
2243 (rs6000_address_for_fpconvert): Combine with
2244 rs6000_force_indexed_or_indirect_mem.
2245 (rs6000_expand_vector_init): Change function call from
2246 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2247 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2248 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2249 (floatsi<mode>2_lfiwax_mem): Ditto.
2250 (floatunssi<mode>2_lfiwzx): Ditto.
2251 (floatunssi<mode>2_lfiwzx_mem): Ditto.
2252 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2253 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2254 (fix_trunc<mode>si2_stfiwx): Ditto.
2255 (fixuns_trunc<mode>si2_stfiwx): Ditto.
2256 (float_<mode>si2_hw): Ditto.
2257 (floatuns_<mode>si2_hw): Ditto.
2258 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2259 (vsx_splat_<mode>): Ditto.
2261 2018-11-01 Joseph Myers <joseph@codesourcery.com>
2263 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2264 at end of assembler input text.
2265 * configure: Regenerate.
2267 2018-11-01 Jakub Jelinek <jakub@redhat.com>
2269 PR tree-optimization/87826
2270 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2271 negative or larger or equal to type's precision.
2273 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
2275 * opts.c (default_options_table): Do not enable
2276 OPT_fdelayed_branch at -Og.
2277 * doc/invoke.texi (-fdelayed-branch): Document it.
2279 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
2281 * optabs-libfuncs.c (build_libfunc_function_visibility):
2282 New, split out from...
2283 (build_libfunc_function): ... here.
2284 (init_one_libfunc_visibility): New, split out from ...
2285 (init_one_libfunc): ... here.
2287 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2288 scratch register need not be early-clobber. Document the reason
2289 why we cannot use ST<OP>.
2291 2018-10-31 Joseph Myers <joseph@codesourcery.com>
2294 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
2295 line for second argument of AC_DEFINE_UNQUOTED.
2296 * doc/install.texi (Tools/packages necessary for modifying GCC):
2297 Update to autoconf 2.69 and automake 1.15.1.
2298 * aclocal.m4, config.in, configure: Regenerate.
2300 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
2302 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2305 2018-10-31 Martin Liska <mliska@suse.cz>
2308 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2309 Add new argument invalid_extension.
2310 (aarch64_get_all_extension_candidates): New function.
2311 (aarch64_rewrite_selected_cpu): Add NULL to function call.
2312 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2314 (aarch64_get_all_extension_candidates): New function.
2315 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2316 argument invalid_extension.
2317 (aarch64_parse_cpu): Likewise.
2318 (aarch64_print_hint_for_extensions): New function.
2319 (aarch64_validate_mcpu): Provide hint about invalid extension.
2320 (aarch64_validate_march): Likewise.
2321 (aarch64_handle_attr_arch): Pass new argument.
2322 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2323 (aarch64_handle_attr_isa_flags): Likewise.
2325 2018-10-31 Richard Biener <rguenther@suse.de>
2329 * tree-outof-ssa.c (insert_backedge_copies): Restrict
2330 copy generation to useful cases. Place the copy before
2331 the definition of the backedge value when possible.
2333 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2335 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2336 * config/arc/arc.c (arc_active_insn): New function.
2337 (check_store_cacheline_hazard): Likewise.
2338 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2339 (arc_override_options): Disable delay slot scheduler for older
2341 (arc_store_addr_hazard_p): New implementation, old one renamed to
2343 (arc_store_addr_hazard_internal_p): Renamed.
2344 (arc_reorg): Don't combine into brcc instructions which are part
2345 of hardware hazard solution.
2346 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2347 (tune_arc700): Likewise.
2348 * config/arc/arc.opt (arc7xx): New tune value.
2349 * config/arc/arc700.md: Improve A7 scheduler.
2351 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2353 * config/arc/arc.c (arc_override_options): Remove
2354 TARGET_COMPACT_CASESI.
2355 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2356 (CASE_VECTOR_MODE): Likewise.
2357 (CASE_VECTOR_PC_RELATIVE): Likewise.
2358 (CASE_VECTOR_SHORTEN_MODE): Likewise.
2359 (CASE_VECTOR_SHORTEN_MODE1): Delete.
2360 (ADDR_VEC_ALIGN): Update.
2361 (ASM_OUTPUT_CASE_LABEL): Undefine.
2362 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2363 (TARGET_BI_BIH): Define.
2364 (DEFAULT_BRANCH_INDEX): Likewise.
2365 * config/arc/arc.md (casesi): Rework to accept BI/BIH
2366 instructions, remove compact_casesi use case.
2367 (casesi_compact_jump): Remove.
2368 (casesi_dispatch): New pattern.
2369 * config/arc/arc.opt: Add mbranch-index option. Deprecate
2370 compact_casesi option.
2371 * doc/invoke.texi: Document mbranch-index option.
2373 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2375 * config/arc/arc.c (arc_get_tp): Remove function.
2376 (arc_emit_call_tls_get_addr): Likewise.
2377 (arc_call_tls_get_addr): New function.
2378 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2379 * config/arc/arc.md (tls_load_tp_soft): Remove.
2380 (tls_gd_get_addr): Likewise.
2382 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2384 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2385 (mulsi3_600_lib): Remove pattern.
2386 (umulsi3_highpart_600_lib_le): Likewise.
2387 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2388 (umulsidi3): Remove call to umulsidi3_600_lib.
2389 (umulsidi3_600_lib): Remove pattern.
2390 (peephole2): Remove peephole using the above deprecated patterns.
2392 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2395 * config/arm/arm.c (arm_option_check_internal): Disable the combined
2396 use of -mslow-flash-data and -mword-relocations.
2397 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2398 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2400 * doc/invoke.texi (-mword-relocations): Mention conflict with
2402 (-mslow-flash-data): Reciprocally.
2404 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
2406 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2407 16-byte modes held in GP registers to use an even regno.
2409 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2410 (aarch64_atomic_ldop_supported_p): Remove.
2411 (aarch64_gen_atomic_ldop): Remove.
2412 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2413 Fully expand LSE operations here.
2414 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2415 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2416 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2417 and use ATOMIC_LDOP instead; use register_operand for the input;
2418 drop the split and emit insns directly.
2419 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2420 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2421 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2423 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2424 (aarch64_gen_atomic_ldop): Don't call it.
2425 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2426 Use aarch64_reg_or_zero.
2427 (aarch64_atomic_exchange<ALLI>): Likewise.
2428 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2429 operand 0; use aarch64_reg_or_zero for input; merge ...
2430 (@aarch64_atomic_swp<ALLI>): ... this and remove.
2432 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2433 (aarch64_split_compare_and_swap): Use it.
2434 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
2435 test oldval against the proper predicate.
2436 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2437 Use nonmemory_operand for expected.
2438 (cas_short_expected_pred): New.
2439 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2440 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2441 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2442 (aarch64_plushi_operand): New.
2444 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2445 Force oldval into the rval register for TARGET_LSE; emit the compare
2446 during initial expansion so that it may be deleted if unused.
2447 (aarch64_gen_atomic_cas): Remove.
2448 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2449 Change =&r to +r for operand 0; use match_dup for operand 2;
2450 remove is_weak and mod_f operands as unused. Drop the split
2452 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2453 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2454 (@aarch64_atomic_cas<GPI>): Similarly.
2456 2018-10-31 Richard Biener <rguenther@suse.de>
2458 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2461 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2463 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2464 * config/aarch64/aarch64.md: Include saphira.md
2465 * config/aarch64/saphira.md: New file for pipeline description.
2467 2018-10-30 Martin Sebor <msebor@redhat.com>
2470 * gimple-ssa-sprintf.c (format_directive): Use %G to include
2472 (sprintf_dom_walker::compute_format_length):
2473 Avoid setting POSUNDER4K here.
2474 (get_destination_size): Handle null argument values.
2475 (get_user_idx_format): New function.
2476 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2477 functions, including user-defined with attribute format printf.
2478 Use %G to include inlining context.
2479 Set POSUNDER4K here.
2481 2018-10-30 Jan Hubicka <jh@suse.cz>
2483 * params.def (lto-partitions): Bump from 32 to 128.
2485 2018-10-30 Jan Hubicka <jh@suse.cz>
2488 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2490 (free_lang_data_in_type): Forward declare.
2491 (fld_type_variant_equal_p): New function.
2492 (fld_type_variant): New function
2493 (fld_incomplete_types): New hash.
2494 (fld_incomplete_type_of): New function
2495 (fld_simplfied-type): New function.
2496 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2497 (free_lang_data): Allocate and free fld_incomplete_type; update call
2498 of free_lang_data_in_decl.
2500 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
2502 * gcov.c (output_lines): Remove duplicate line.
2504 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
2506 * config/rs6000/rs6000.md (bswapdi2): Force address into register
2507 if not in indexed or indirect form.
2508 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2509 (bswapdi2_store): Ditto.
2510 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2512 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2513 Prototype for helper function.
2515 2018-10-30 Martin Sebor <msebor@redhat.com>
2517 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2518 (target, pragma GCC optimize, pragma GCC target): Ditto.
2520 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
2522 * doc/extend.texi: Fix prototype and description of
2523 __builtin_expect_with_probability.
2525 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
2527 * cgraph.h (clone_function_name_1): Replaced by new
2528 clone_function_name_numbered that takes name as string; for
2529 privatize_symbol_name_1 use only.
2530 (clone_function_name): Renamed to
2531 clone_function_name_numbered to be explicit about numbering.
2532 (clone_function_name): New two-argument function that does
2533 not number its output.
2534 (clone_function_name): New three-argument function that
2535 takes a number to append to its output.
2536 * cgraphclones.c (duplicate_thunk_for_node):
2537 (clone_function_name_1): Renamed.
2538 (clone_function_name_numbered): Two new functions.
2539 (clone_function_name): Improved documentation.
2540 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2541 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2542 * final.c (final_scan_insn_1): Use the new clone_function_name
2544 * multiple_target.c (create_dispatcher_calls): Ditto.
2545 (create_target_clone): Ditto.
2546 * omp-expand.c (grid_expand_target_grid_body): Ditto.
2547 * omp-low.c (create_omp_child_function_name): Ditto.
2548 * omp-simd-clone.c (simd_clone_create): Ditto.
2549 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2550 new clone_function_name without numbering.
2552 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
2554 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2555 Assert that the allocation size is not zero.
2557 2018-10-30 Richard Biener <rguenther@suse.de>
2559 PR tree-optimization/87800
2560 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2561 non-induction or reduction PHIs.
2563 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2565 * config/aarch64/falkor-tag-collision-avoidance.c
2566 (execute_tag_collision_avoidance): Call df_note_add_problem.
2568 2018-10-30 Martin Liska <mliska@suse.cz>
2570 * doc/extend.texi: Fix typo in documentation
2571 of __builtin_expect_with_probability.
2573 2018-10-29 David Malcolm <dmalcolm@redhat.com>
2576 * input.c (get_substring_ranges_for_loc): Detect if
2577 linemap_resolve_location gives us a NULL map, and reject
2580 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
2582 * config.gcc (xstormy16-*-elf): Set tm_d_file.
2584 2018-10-29 David Malcolm <dmalcolm@redhat.com>
2585 Martin Sebor <msebor@redhat.com>
2586 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
2588 * doc/ux.texi (Quoting): New subsection, adapted from material at
2589 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
2590 MartinSebor and ManuelLopezIbanez.
2591 (Fix-it hints): Note that fix-it hints shouldn't be marked for
2594 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
2597 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
2600 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
2602 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
2604 2018-10-29 Paul Koning <ni1d@arrl.net>
2606 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
2608 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
2610 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
2611 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
2612 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
2613 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
2614 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
2615 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
2616 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
2617 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
2618 Change 'vector' to '__vector'.
2619 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
2620 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
2621 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
2622 _mm_avg_pu16): Likewise. And, whitespace corrections.
2624 2018-10-29 Richard Biener <rguenther@suse.de>
2626 PR tree-optimization/87785
2627 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
2630 2018-10-29 Olivier Hainque <hainque@adacore.com>
2632 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
2634 2018-10-29 Olivier Hainque <hainque@adacore.com>
2636 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
2637 ports configuration macro, defaults to "gnu".
2638 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
2639 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
2641 2018-10-29 Olivier Hainque <hainque@adacore.com>
2643 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
2645 2018-10-29 Richard Biener <rguenther@suse.de>
2647 PR tree-optimization/87790
2648 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
2649 (vect_make_slp_decision): Adjust.
2650 (vect_slp_analyze_bb_1): Likewise.
2651 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
2654 2018-10-29 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/87785
2657 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
2659 (vect_build_slp_tree): Likewise.
2660 (vect_gather_slp_loads): New function.
2661 (vect_analyze_slp_instance): Gather loads separately from the
2664 2018-10-29 Martin Liska <mliska@suse.cz>
2666 * Makefile.in: Make dependency to json.o.
2667 * doc/gcov.texi: Document new JSON format, remove
2668 old intermediate format documentation.
2669 * gcov.c (struct function_info): Come up with m_name and
2671 (function_info::function_info): Initialize it.
2672 (function_info::~function_info): Release it.
2673 (main): Rename flag_intermediate_format to flag_json_format.
2674 (print_usage): Describe --json-format.
2675 (process_args): Set flag_json_format.
2676 (output_intermediate_line): Remove.
2677 (output_intermediate_json_line): Likewise.
2678 (get_gcov_intermediate_filename): Return new extension
2680 (output_intermediate_file): Implement JSON emission.
2681 (output_json_intermediate_file): Implement JSON emission.
2682 (generate_results): Use ::get_name for function name.
2683 Handle JSON output file.
2684 (read_graph_file): Use ::get_name instead of cplus_demangle.
2685 (read_count_file): Likewise.
2686 (solve_flow_graph): Likewise.
2687 (add_line_counts): Likewise.
2688 (accumulate_line_counts): Use new flag_json_format.
2689 (output_function_details): Use ::get_name instead of cplus_demangle.
2690 (output_lines): Likewise.
2691 * json.cc (test_writing_literals): Add new tests.
2692 * json.h (class literal): Add new boolean constructor.
2694 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
2696 PR rtl-optimization/87701
2697 PR rtl-optimization/87780
2698 * combine.c (make_more_copies): Rewrite.
2700 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
2702 * doc/generic.texi (ABSU_EXPR): Document.
2703 * match.pd (absu(x)*absu(x) -> x*x): Handle.
2704 (absu(absu(X)) -> absu(X)): Likewise.
2705 (absu(-X) -> absu(X)): Likewise.
2706 (absu(X) where X is nonnegative -> X): Likewise.
2708 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
2710 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
2711 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
2712 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
2713 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
2714 (s-tm-texi): Also check timestamp on d-target.def.
2715 (generated_files): Add TM_D_H and d-target-hooks-def.h.
2716 (build/genhooks.o): Also depend on D_TARGET_DEF.
2717 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
2719 * config/aarch64/aarch64-d.c: New file.
2720 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
2722 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
2724 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
2725 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
2726 * config/arm/arm-d.c: New file.
2727 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
2728 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
2729 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2730 * config/arm/t-arm (arm-d.o): New rule.
2731 * config/default-d.c: New file.
2732 * config/glibc-d.c: New file.
2733 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2734 * config/i386/i386-d.c: New file.
2735 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
2736 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
2737 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2738 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
2739 * config/i386/t-i386 (i386-d.o): New rule.
2740 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2741 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2742 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
2743 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2744 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
2745 * config/mips/mips-d.c: New file.
2746 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
2747 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
2748 * config/mips/t-mips (mips-d.o): New rule.
2749 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2750 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2751 * config/powerpcspe/powerpcspe-d.c: New file.
2752 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
2754 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
2755 Support GNU D by using 0 as the language type.
2756 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
2757 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
2758 * config/riscv/riscv-d.c: New file.
2759 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
2761 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
2762 * config/riscv/t-riscv (riscv-d.o): New rule.
2763 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2764 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
2765 * config/rs6000/rs6000-d.c: New file.
2766 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
2768 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
2769 Support GNU D by using 0 as the language type.
2770 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
2771 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
2772 * config/s390/s390-d.c: New file.
2773 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
2774 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
2775 * config/s390/t-s390 (s390-d.o): New rule.
2776 * config/sparc/sparc-d.c: New file.
2777 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
2779 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
2780 * config/sparc/t-sparc (sparc-d.o): New rule.
2781 * config/t-glibc (glibc-d.o): New rule.
2782 * configure: Regenerated.
2783 * configure.ac (tm_d_file): New variable.
2784 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
2785 * doc/contrib.texi (Contributors): Add self for the D frontend.
2786 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
2787 * doc/install.texi (Configuration): Mention libphobos as an option for
2788 --enable-shared. Mention d as an option for --enable-languages.
2789 (Testing): Mention check-d as a target.
2790 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
2791 name suffixes. Mention d as a -x option.
2792 * doc/sourcebuild.texi (Top Level): Mention libphobos.
2793 * doc/standards.texi (Standards): Add section on D language.
2794 * doc/tm.texi: Regenerated.
2795 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
2796 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
2797 * dwarf2out.c (is_dlang): New function.
2798 (gen_compile_unit_die): Use DW_LANG_D for D.
2799 (declare_in_namespace): Return module die for D, instead of adding
2800 extra declarations into the namespace.
2801 (gen_namespace_die): Generate DW_TAG_module for D.
2802 (gen_decl_die): Handle CONST_DECLSs for D.
2803 (dwarf2out_decl): Likewise.
2804 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
2805 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
2806 kinds of aggregates.
2807 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
2808 * genhooks.c: Include d/d-target.def.
2810 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
2813 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
2814 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
2815 (STACK_DYNAMIC_OFFSET): Likewise.
2817 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
2820 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
2823 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
2825 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
2826 return type and other typos.
2828 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2830 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
2831 a shorter sequence with fewer branches.
2832 (emit_final_str_compare_gpr): Ditto.
2834 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
2836 * config/rs6000/tmmintrin.h: New file.
2837 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
2839 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
2841 * config/rs6000/mmintrin.h: Enable 32bit compilation.
2842 * config/rs6000/xmmintrin.h: Likewise.
2844 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
2846 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
2848 2018-10-26 Richard Biener <rguenther@suse.de>
2850 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
2852 (vect_mark_slp_stmts_relevant): Likewise.
2853 (vect_detect_hybrid_slp_stmts): Likewise.
2854 (vect_bb_slp_scalar_cost): Likewise.
2855 (vect_remove_slp_scalar_calls): Likewise.
2857 2018-10-26 Jan Hubicka <jh@suse.cz>
2859 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
2860 (warn_types_mismatch): Fix walk of DECL_NAME.
2861 (odr_types_equivalent_p): Fix overactive assert.
2863 2018-10-26 Richard Biener <rguenther@suse.de>
2865 PR tree-optimization/87105
2866 * tree-vectorizer.h (_slp_tree::refcnt): New member.
2867 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
2869 (vect_create_new_slp_node): Initialize refcnt to one.
2871 (scalar_stmts_set_t, bst_fail): Remove.
2872 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
2873 (vect_build_slp_tree): Add bst_map argument and lookup
2874 already created SLP nodes.
2875 (vect_print_slp_tree): Handle a SLP graph, print SLP node
2877 (vect_slp_rearrange_stmts): Handle a SLP graph.
2878 (vect_analyze_slp_instance): Adjust and free SLP nodes from
2879 the CSE map. Fix indenting.
2880 (vect_schedule_slp_instance): Add short-cut.
2882 2018-10-26 Martin Liska <mliska@suse.cz>
2885 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
2886 addr_expr and not with pointers.
2888 2018-10-26 Jan Hubicka <jh@suse.cz>
2890 * tree.c (free_lang_data_in_type): Only check main variants.
2891 * ipa-devirt.c (warn_odr): Make static.
2892 (types_same_for_odr): Drop strict variant.
2893 (types_odr_comparable): Likewise.
2894 (odr_or_derived_type_p): Look for main variants.
2895 (odr_name_hasher::equal): Cleanup comment.
2896 (odr_subtypes_equivalent): Add warn and warned arguments; check main
2898 (type_variants_equivalent_p): break out from ...
2899 (odr_types_equivalent): ... here; go for main variants where needed.
2900 (warn_odr): ... here; turn static.
2901 (warn_types_mismatch): Compare mangled names of main variants.
2902 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
2903 (type_with_linkage_p): Sanity check that we look at main variant.
2904 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
2905 * tree.h (types_same_for_odr): Drop strict argument.
2907 2018-10-26 Richard Biener <rguenther@suse.de>
2909 PR tree-optimization/87746
2910 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
2911 Simplify and fix WRT strided store groups with size not
2912 equal to step in element count.
2913 (vect_analyze_group_access_1): Dump the whole group.
2915 2018-10-25 Carl Love <cel@us.ibm.com>
2917 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
2918 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
2919 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
2920 precicion entry for each overloaded builtin.
2921 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
2922 VSCEDPUO): Rename overloaded name.
2923 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
2924 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
2925 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
2926 define_expand for xscmpexqp instruction.
2927 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
2929 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2930 Jinsong Ji <jji@us.ibm.com>
2932 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
2933 function with vec_sl.
2934 (_mm_slli_epi32): Likewise.
2935 (_mm_slli_epi64): Likewise.
2936 (_mm_srai_epi16): Replace deprecated function with vec_sra.
2937 (_mm_srai_epi32): Likewise.
2938 (_mm_srli_epi16): Replace deprecated function with vec_sr.
2939 (_mm_srli_epi32): Likewise.
2940 (_mm_srli_epi64): Likewise.
2941 (_mm_sll_epi16): Replace deprecated function with vec_sl.
2942 (_mm_sll_epi32): Likewise.
2943 (_mm_sll_epi64): Likewise.
2944 (_mm_sra_epi16): Replace deprecated function with vec_sra.
2945 (_mm_sra_epi32): Likewise.
2946 (_mm_srl_epi16): Replace deprecated function with vec_sr.
2947 (_mm_srl_epi32): Likewise.
2948 (_mm_srl_epi64): Likewise.
2950 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2951 Jinsong Ji <jji@us.ibm.com>
2953 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
2954 comparison operators with vec_cmp* for compatibility due to
2955 unfortunate history; clean up formatting and use types more
2957 (_mm_sll_epi32): Likewise.
2958 (_mm_sll_epi64): Likewise.
2959 (_mm_srl_epi16): Likewise.
2960 (_mm_srl_epi32): Likewise.
2961 (_mm_srl_epi64): Likewise.
2963 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
2964 Jinsong Ji <jji@us.ibm.com>
2966 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
2967 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
2968 __vector __bool int. Use vec_cmpgt in preference to deprecated
2969 function vec_vcmpgtfp.
2970 (_mm_max_ps): Likewise.
2972 2018-10-25 Jeff Law <law@redhat.com>
2974 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
2977 2018-10-25 Martin Sebor <msebor@redhat.com>
2979 * doc/extend.texi (aligned): Expand attribute description.
2980 (Alignment): Rename section. Discuss function arguments.
2982 2018-10-25 Jan Hubicka <jh@suse.cz>
2984 * ipa-devirt.c (main_odr_variant): Remove.
2985 (hash_odr_name, types_same_for_odr, types_odr_comparable,
2986 odr_name_hasher::equal, odr_subtypes_equivalent_p):
2987 Drop use of main_odr_variant.
2988 (add_type_duplicate): Silence confused warnings on integer types.
2989 (get_odr_type): Always look for main variant.
2990 (register_odr_type): Simplify.
2992 2018-10-25 Richard Biener <rguenther@suse.de>
2994 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
2995 Initialize ng to silence error with release checking bootstrap.
2997 2018-10-25 Richard Biener <rguenther@suse.de>
2999 * tree-if-conv.c: Include tree-ssa-sccvn.h.
3000 (tree_if_conversion): Run CSE on the if-converted loop body.
3002 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
3004 * config/s390/constraints.md (ZL): New constraint.
3005 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3007 * config/s390/s390.md (movdi_larl): Remove.
3008 (movdi_64): Add the LARL alternative.
3010 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
3013 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3014 (RTX_CODE_HWINT_P): New macro.
3015 (rtx_code_size): Use RTX_CODE_HWINT_P ().
3017 2018-10-25 Jan Hubicka <jh@suse.cz>
3019 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3022 2018-10-25 Richard Biener <rguenther@suse.de>
3024 PR tree-optimization/87665
3025 PR tree-optimization/87745
3026 * tree-vectorizer.h (get_earlier_stmt): Remove.
3027 (get_later_stmt): Pick up UID from the original non-pattern stmt.
3029 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
3031 * options.texi (Deprecated): Move list to Var section.
3033 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
3034 Jinsong Ji <jji@us.ibm.com>
3036 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3037 __vector long to __vector long long.
3038 (_mm_cvtpd_ps): Likewise.
3039 (_mm_cvttpd_epi32): Likewise.
3040 (_mm_cvtpi32_pd): Likewise.
3041 (_mm_unpackhi_epi64): Likewise.
3042 (_mm_unpacklo_epi64): Likewise.
3044 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
3046 PR rtl-optimization/87720
3047 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3049 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
3051 * gimple-ssa-isolate-paths.c
3052 (find_implicit_erroneous_behavior): Do not change code if the
3053 pass is running for warnings only.
3054 (find_explicit_erroneous_behavior): Likewise.
3056 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
3058 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3059 Define as rs6000_mangle_decl_assembler_name.
3060 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3061 long double to IEEE long double, switch the names of the long
3062 double built-in functions to be <func>f128 instead of <func>l.
3064 2018-10-24 Martin Sebor <msebor@redhat.com>
3066 * doc/extend.texi (nonnull): List no-argument form. Reference
3067 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3069 2018-10-24 Richard Biener <rguenther@suse.de>
3071 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3073 2018-10-24 Martin Liska <mliska@suse.cz>
3075 PR tree-optimization/84436
3076 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3078 (switch_conversion::contains_linear_function_p): New.
3079 (switch_conversion::build_one_array): Support linear
3080 transformation on input.
3081 * tree-switch-conversion.h (struct switch_conversion): Add
3082 contains_linear_function_p declaration.
3084 2018-10-24 Richard Biener <rguenther@suse.de>
3086 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3087 if its argument is CONSTANT_CLASS_P.
3089 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
3091 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3092 it is wrong for forward declarations.
3094 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
3096 * config/s390/s390.c (s390_check_qrst_address): Add the missing
3097 SYMBOL_REF_P () check.
3099 2018-10-24 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/87105
3102 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3103 dump classification.
3104 (vect_analyze_data_ref_accesses): Handle duplicate loads and
3105 stores by splitting the affected group after the fact.
3106 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3107 fail the SLP build because of size constraints.
3109 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3111 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3112 * configure: Regenerate.
3113 * config.in: Regenerate.
3114 * varasm.c (mergeable_string_section): Use readonly_data_section
3115 if linker doesn't support SHF_MERGE with alignment > 8.
3116 (mergeable_constant_section): Likewise.
3118 2018-10-24 Richard Biener <rguenther@suse.de>
3120 PR tree-optimization/84013
3121 * tree-ssa-structalias.c (struct msdi_data): New struct for
3122 marshalling data to walk_stmt_load_store_ops.
3123 (maybe_set_dependence_info): Refactor as callback for
3124 walk_stmt_load_store_ops.
3125 (compute_dependence_clique): Set restrict info on all stmt kinds.
3127 2018-10-24 Martin Liska <mliska@suse.cz>
3129 * cgraph.c (cgraph_node::dump):
3130 Remove reduntant dumps and make tp_first_run dump more compact.
3132 2018-10-24 Richard Biener <rguenther@suse.de>
3134 PR tree-optimization/87665
3135 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3138 2018-10-12 Jeff Law <law@redhat.com>
3140 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3143 2018-10-23 Richard Biener <rguenther@suse.de>
3145 * tree-vrp.c (add_assert_info): Guard dump_printf with
3147 * gimple-ssa-evrp-analyze.c
3148 (evrp_range_analyzer::record_ranges_from_incoming_edge):
3149 Use value_range::ignore_equivs_equal_p.
3151 2018-10-23 Richard Biener <rguenther@suse.de>
3153 PR tree-optimization/87105
3154 PR tree-optimization/87608
3155 * passes.def (pass_all_early_optimizations): Add early phi-opt
3157 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3158 addition to debug stmts.
3159 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3160 and abs replacement early.
3161 * tree-cfg.c (gimple_empty_block_p): Likewise.
3163 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
3166 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3167 specified to configure.
3168 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3170 2018-10-23 Richard Biener <rguenther@suse.de>
3172 PR tree-optimization/87700
3173 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3175 2018-10-23 Jakub Jelinek <jakub@redhat.com>
3178 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3179 second argument from __mmask16 to __mmask8.
3180 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3181 _mm_mask_packs_epi32): Likewise.
3182 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3184 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
3186 2018-10-23 Richard Biener <rguenther@suse.de>
3188 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3190 2018-10-23 Richard Biener <rguenther@suse.de>
3192 PR tree-optimization/86144
3193 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3194 over simd attribute.
3196 2018-10-23 Richard Biener <rguenther@suse.de>
3198 PR tree-optimization/87693
3199 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3200 the case we do not find the taken edge.
3202 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
3203 Jinsong Ji <jji@us.ibm.com>
3205 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3206 (_mm_store_pd): Use unaligned vector type for pointer cast.
3207 (_mm_maskmoveu_si128): Likewise.
3208 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3209 (_mm_store_ps): Use unaligned vector type for pointer cast.
3211 2018-10-22 Paul Koning <ni1d@arrl.net>
3213 * symtab.c (symtab_node::increase_alignment): Correct max
3216 2018-10-22 Yury Gribov <tetra2005@gmail.com>
3218 PR tree-optimization/87633
3219 * match.pd: Do not generate unordered integer comparisons.
3221 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
3223 PR rtl-optimization/87600
3224 * combine.c: Add include of expr.h.
3225 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3226 register to a pseudo.
3227 (make_more_copies): New function, add a copy to a new pseudo after
3228 the moves from hard registers into pseudos.
3229 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3230 later. Call make_more_copies.
3232 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
3234 * lra-constraints.c (process_alt_operands): New local array,
3235 matching_early_clobber. Check matching_early_clobber before
3236 decrementing reject, and set matching_early_clobber after.
3238 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
3241 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3242 call output_addr_const and hope for the best.
3244 2018-10-22 Richard Biener <rguenther@suse.de>
3246 * gimple-ssa-evrp-analyze.c
3247 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3248 smarter about what ranges to use.
3249 * tree-vrp.c (add_assert_info): Dump here.
3250 (register_edge_assert_for_2): Instead of here at multiple but
3253 * gcc.dg/tree-ssa/evrp12.c: New testcase.
3254 * gcc.dg/predict-6.c: Adjust.
3255 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3256 * gcc.dg/tree-ssa/vrp02.c: Likewise.
3257 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3259 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
3260 Richard Biener <rguenther@suse.de>
3262 * bitmap.h: Update data structure documentation, including a
3263 description of bitmap views as either linked-lists or splay trees.
3264 (struct bitmap_element_def): Update comments for splay tree bitmaps.
3265 (struct bitmap_head_def): Likewise.
3266 (bitmap_list_view, bitmap_tree_view): New prototypes.
3267 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3269 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3270 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3271 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3272 released bitmap element here.
3273 (bitmap_element_free): Remove.
3274 (bitmap_elt_clear_from): Work on splay tree bitmaps.
3275 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
3276 this function similar ones such that linked-list bitmap implementation
3277 functions are grouped.
3278 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3279 and moved for grouping.
3280 (bitmap_list_insert_element_after): Renamed from
3281 bitmap_elt_insert_after, and moved for grouping.
3282 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3283 (bitmap_tree_link_left, bitmap_tree_link_right,
3284 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3285 bitmap_tree_link_element, bitmap_tree_unlink_element,
3286 bitmap_tree_find_element): New functions for splay-tree bitmap
3288 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3289 Renamed and moved, see above entries.
3290 (bitmap_tree_listify_from): New function to convert part of a splay
3291 tree bitmap to a linked-list bitmap.
3292 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3293 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3294 (bitmap_find_bit): Remove.
3295 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3296 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3297 Handle splay tree bitmaps.
3298 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3299 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3300 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3301 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3302 bitmap_intersect_compl_p, bitmap_ior_and_compl,
3303 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3304 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
3305 corresponding changes to use linked-list specific bitmap_element
3306 manipulation functions as applicable for efficiency.
3307 (bitmap_tree_to_vec): New function.
3308 (debug_bitmap_elt_file): New function split out from ...
3309 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
3310 (bitmap_print): Likewise.
3312 PR tree-optimization/63155
3313 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3315 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3318 2018-10-22 Martin Liska <mliska@suse.cz>
3320 PR tree-optimization/87686
3322 2018-08-29 Martin Liska <mliska@suse.cz>
3324 * tree-switch-conversion.c (switch_conversion::expand):
3325 Strenghten assumption about gswitch statements.
3327 2018-10-22 Martin Liska <mliska@suse.cz>
3329 * ipa-icf.c (sem_item::compare_attributes): Remove.
3330 (sem_item::compare_referenced_symbol_properties): Use
3331 attribute_list_equal instead.
3332 (sem_function::equals_wpa): Likewise.
3333 * ipa-icf.h: Remove compare_attributes.
3335 2018-10-22 Richard Biener <rguenther@suse.de>
3338 * mem-stats.h (mem_usage::operator==): Fix pasto.
3340 2018-10-22 Richard Biener <rguenther@suse.de>
3342 PR tree-optimization/87640
3343 * tree-vrp.c (set_value_range_with_overflow): Decompose
3345 (extract_range_from_binary_expr_1): Adjust.
3347 2018-10-22 Martin Jambor <mjambor@suse.cz>
3349 * tree-eh.h (stmt_could_throw_p): Add function parameter.
3350 (stmt_can_throw_external): Likewise.
3351 (stmt_can_throw_internal): Likewise.
3352 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3353 (lower_eh_constructs_2): Likewise.
3354 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3355 (stmt_can_throw_external): Likewise.
3356 (stmt_can_throw_internal): Likewise.
3357 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3358 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3359 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3360 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3361 (pass_lower_eh_dispatch::execute): Pass cfun to
3362 stmt_can_throw_external.
3363 (cleanup_empty_eh): Likewise.
3364 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3365 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3366 stmt_can_throw_external instead of pushing it to cfun.
3367 (symbol_table::create_edge): Likewise.
3368 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3369 stmt_can_throw_internal.
3370 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3371 to stmt_could_throw_p.
3372 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3373 stmt_can_throw_internal.
3374 (pass_store_merging::execute): Likewise.
3375 * gimple-ssa-strength-reduction.c
3376 (find_candidates_dom_walker::before_dom_children): Pass cfun to
3378 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3379 stmt_can_throw_internal.
3380 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3381 to stmt_can_throw_external.
3382 (check_stmt): Pass cfun to stmt_could_throw_p.
3383 (check_stmt): Pass cfun to stmt_can_throw_external.
3384 (pass_nothrow::execute): Likewise.
3385 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3386 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3387 stmt_can_throw_internal.
3388 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3389 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3390 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3391 * tree-complex.c (expand_complex_libcall): Pass cfun to
3392 stmt_could_throw_p and to stmt_can_throw_internal.
3393 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3394 * tree-inline.c (copy_edges_for_bb): Likewise.
3395 (maybe_move_debug_stmts_to_successors): Likewise.
3396 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3398 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3399 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3400 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3401 stmt_can_throw_internal.
3402 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3403 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3405 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3406 stmt_can_throw_internal.
3407 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3408 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3410 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3411 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3412 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3413 (convert_mult_to_fma_1): Likewise.
3414 (convert_to_divmod): Likewise.
3415 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3416 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3417 * tree-ssa-propagate.c
3418 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3419 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3420 (maybe_optimize_range_tests): Likewise.
3421 (linearize_expr_tree): Likewise.
3422 (reassociate_bb): Likewise.
3423 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3424 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3425 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3426 (handle_char_store): Likewise.
3427 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3428 stmt_can_throw_internal.
3429 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3431 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3432 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3433 (vectorizable_simd_clone_call): Likewise.
3434 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3435 (gimple_stringop_fixed_value): Likewise.
3437 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
3439 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3440 literal pool references.
3441 (s390_check_qrst_address): Adapt to the new behavior of
3442 s390_loadrelative_operand_p ().
3444 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3447 * config/i386/sse.md (*andnot<mode>3_bcst): New.
3449 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3452 * config/i386/sse.md (*<code><mode>3_bcst): New.
3454 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3457 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3458 V4DI, V16SI and V8DI.
3459 (*sub<mode>3<mask_name>_bcst): New.
3460 (*add<mode>3<mask_name>_bcst): Likewise.
3462 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
3463 Jinsong Ji <jji@us.ibm.com>
3465 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3466 __m64 with __vector unsigned long long for compatibility.
3467 (_mm_movemask_epi8): Likewise.
3468 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3469 (_mm_cvttps_pi32): Likewise.
3470 (_mm_cvtpi32_ps): Likewise.
3471 (_mm_cvtps_pi16): Likewise.
3472 (_mm_loadh_pi): Likewise.
3473 (_mm_storeh_pi): Likewise.
3474 (_mm_movehl_ps): Likewise.
3475 (_mm_movelh_ps): Likewise.
3476 (_mm_loadl_pi): Likewise.
3477 (_mm_storel_pi): Likewise.
3478 (_mm_movemask_ps): Likewise.
3479 (_mm_shuffle_pi16): Likewise.
3481 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3484 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3485 __builtin_ia32_vfnmsubpd512_mask.
3486 (_mm512_mask_fnmsub_round_pd): Likewise.
3487 (_mm512_fnmsub_pd): Likewise.
3488 (_mm512_mask_fnmsub_pd): Likewise.
3489 (_mm512_maskz_fnmsub_round_pd): Use
3490 __builtin_ia32_vfnmsubpd512_maskz.
3491 (_mm512_maskz_fnmsub_pd): Likewise.
3492 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3493 (_mm512_mask_fnmsub_round_ps): Likewise.
3494 (_mm512_fnmsub_ps): Likewise.
3495 (_mm512_mask_fnmsub_ps): Likewise.
3496 (_mm512_maskz_fnmsub_round_ps): Use
3497 __builtin_ia32_vfnmsubps512_maskz.
3498 (_mm512_maskz_fnmsub_ps): Likewise.
3499 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3500 __builtin_ia32_vfnmsubpd256_mask.
3501 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3502 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3503 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3504 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3505 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3506 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3507 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3508 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3509 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3510 __builtin_ia32_vfnmsubpd.
3511 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3512 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3513 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3514 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3515 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3516 * config/i386/i386-builtin.def: Add
3517 __builtin_ia32_vfnmsubpd256_mask,
3518 __builtin_ia32_vfnmsubpd256_maskz,
3519 __builtin_ia32_vfnmsubpd128_mask,
3520 __builtin_ia32_vfnmsubpd128_maskz,
3521 __builtin_ia32_vfnmsubps256_mask,
3522 __builtin_ia32_vfnmsubps256_maskz,
3523 __builtin_ia32_vfnmsubps128_mask,
3524 __builtin_ia32_vfnmsubps128_maskz,
3525 __builtin_ia32_vfnmsubpd512_mask,
3526 __builtin_ia32_vfnmsubpd512_maskz,
3527 __builtin_ia32_vfnmsubps512_mask,
3528 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3529 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3530 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3531 __builtin_ia32_vfnmsubpd256.
3532 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3533 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3534 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3536 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3538 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3540 (fmai_vmfnmsub_<mode><round_name>): Likewise.
3542 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3545 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3546 __builtin_ia32_vfnmaddpd512_mask.
3547 (_mm512_mask_fnmadd_round_pd): Likewise.
3548 (_mm512_fnmadd_pd): Likewise.
3549 (_mm512_mask_fnmadd_pd): Likewise.
3550 (_mm512_maskz_fnmadd_round_pd): Use
3551 __builtin_ia32_vfnmaddpd512_maskz.
3552 (_mm512_maskz_fnmadd_pd): Likewise.
3553 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3554 (_mm512_mask_fnmadd_round_ps): Likewise.
3555 (_mm512_fnmadd_ps): Likewise.
3556 (_mm512_mask_fnmadd_ps): Likewise.
3557 (_mm512_maskz_fnmadd_round_ps): Use
3558 __builtin_ia32_vfnmaddps512_maskz.
3559 (_mm512_maskz_fnmadd_ps): Likewise.
3560 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3561 __builtin_ia32_vfnmaddpd256_mask.
3562 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3563 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3564 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3565 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3566 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3567 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3568 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3569 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3570 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3571 __builtin_ia32_vfnmaddpd.
3572 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3573 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3574 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3575 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3576 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3577 * config/i386/i386-builtin.def: Add
3578 __builtin_ia32_vfnmaddpd256_mask,
3579 __builtin_ia32_vfnmaddpd256_maskz,
3580 __builtin_ia32_vfnmaddpd128_mask,
3581 __builtin_ia32_vfnmaddpd128_maskz,
3582 __builtin_ia32_vfnmaddps256_mask,
3583 __builtin_ia32_vfnmaddps256_maskz,
3584 __builtin_ia32_vfnmaddps128_mask,
3585 __builtin_ia32_vfnmaddps128_maskz,
3586 __builtin_ia32_vfnmaddpd512_mask,
3587 __builtin_ia32_vfnmaddpd512_maskz,
3588 __builtin_ia32_vfnmaddps512_mask,
3589 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
3590 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
3591 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
3592 __builtin_ia32_vfnmaddpd256.
3593 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
3594 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
3595 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
3597 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
3599 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
3601 (fmai_vmfnmadd_<mode><round_name>): Likewise.
3603 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3606 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
3607 __builtin_ia32_vfmsubpd512_mask.
3608 (_mm512_mask_fmsub_round_pd): Likewise.
3609 (_mm512_fmsub_pd): Likewise.
3610 (_mm512_mask_fmsub_pd): Likewise.
3611 (_mm512_maskz_fmsub_round_pd): Use
3612 __builtin_ia32_vfmsubpd512_maskz.
3613 (_mm512_maskz_fmsub_pd): Likewise.
3614 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
3615 (_mm512_mask_fmsub_round_ps): Likewise.
3616 (_mm512_fmsub_ps): Likewise.
3617 (_mm512_mask_fmsub_ps): Likewise.
3618 (_mm512_maskz_fmsub_round_ps): Use
3619 __builtin_ia32_vfmsubps512_maskz.
3620 (_mm512_maskz_fmsub_ps): Likewise.
3621 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
3622 __builtin_ia32_vfmsubpd256_mask.
3623 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
3624 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3625 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
3626 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3627 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
3628 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
3629 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
3630 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
3631 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
3632 __builtin_ia32_vfmsubpd.
3633 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
3634 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
3635 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
3636 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
3637 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
3638 * config/i386/i386-builtin.def: Add
3639 __builtin_ia32_vfmsubpd256_mask,
3640 __builtin_ia32_vfmsubpd256_maskz,
3641 __builtin_ia32_vfmsubpd128_mask,
3642 __builtin_ia32_vfmsubpd128_maskz,
3643 __builtin_ia32_vfmsubps256_mask,
3644 __builtin_ia32_vfmsubps256_maskz,
3645 __builtin_ia32_vfmsubps128_mask,
3646 __builtin_ia32_vfmsubps128_maskz,
3647 __builtin_ia32_vfmsubpd512_mask,
3648 __builtin_ia32_vfmsubpd512_maskz,
3649 __builtin_ia32_vfmsubps512_mask,
3650 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
3651 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
3652 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
3653 __builtin_ia32_vfmsubpd256.
3654 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
3655 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
3656 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
3658 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
3660 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
3662 (fmai_vmfmsub_<mode><round_name>): Likewise.
3664 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3666 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
3667 Remove plus. Renamed to ...
3668 (*sub<mode>3<mask_name>_bcst): This.
3669 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
3670 (*add<mode>3<mask_name>_bcst): This.
3672 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3675 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
3677 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3680 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
3681 (_mm_or_epi32): Likewise.
3682 (_mm256_xor_epi32): Likewise.
3683 (_mm_xor_epi32): Likewise.
3684 (_mm256_or_epi64): Likewise.
3685 (_mm_or_epi64): Likewise.
3686 (_mm256_xor_epi64): Likewise.
3687 (_mm_xor_epi64): Likewise.
3689 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
3692 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
3694 2018-10-20 Jakub Jelinek <jakub@redhat.com>
3697 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
3699 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
3701 * doc/ux.texi: Move @section directly after @node.
3703 2018-10-19 Jakub Jelinek <jakub@redhat.com>
3707 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
3708 depend closely nested inside of loop with ordered clause with
3711 2018-10-19 David Malcolm <dmalcolm@redhat.com>
3713 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
3714 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
3715 * doc/ux.texi: New file.
3717 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
3719 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
3720 be the first CR field allocated.
3722 2018-10-19 Richard Biener <rguenther@suse.de>
3725 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
3726 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
3728 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
3731 * config/i386/sse.md
3732 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
3733 (*add<mode>3<mask_name>_bcst_2): Likewise.
3735 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
3737 * config/i386/sse.md
3738 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3739 Replace nonimmediate_operand with register_operand.
3740 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3742 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3745 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
3747 PR rtl-optimization/87596
3748 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
3749 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
3750 for instructions in FROM..TO range.
3752 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
3754 * cfgexpand.c (expand_one_var): Use specific wording in error message
3755 for non-local frame variables.
3756 * stor-layout.c (layout_decl): Do not issue a warning for them.
3758 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
3760 * haifa-sched.c (priority): Add force_recompute parameter.
3761 (apply_replacement): Call priority () with force_recompute = true.
3762 (restore_pattern): Likewise.
3764 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
3766 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
3767 HOST_BITS_PER_WIDE_INT.
3768 (test_vector_ops_duplicate): Likewise.
3770 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
3773 * config/i386/sse.md (VF_AVX512): New.
3774 (avx512bcst): Likewise.
3775 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
3777 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
3779 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
3782 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
3784 * doc/invoke.texi (-dumpversion): Improve grammar.
3785 (-dumpfullversion): Make more consistent with -dumpversion.
3787 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
3789 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
3790 Set JUMP_LABEL to the jump insn.
3791 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
3792 Predict emitted jump and add label to jump insn.
3794 2018-10-18 David Malcolm <dmalcolm@redhat.com>
3796 PR tree-optimization/87562
3797 * input.c (get_substring_ranges_for_loc): Use
3798 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
3799 getting the linemap for the endpoint. Verify that it's either
3800 in the same linemap as the start point's spelling location, or
3801 at least in the same file.
3803 2018-10-18 Richard Biener <rguenther@suse.de>
3805 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
3806 feed width-specific load/store costs through ix86_vec_cost.
3807 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
3808 (k8_cost): Likewise.
3809 (bdver_cost): Likewise.
3810 (znver1_cost): Likewise.
3811 (btver1_cost): Likewise.
3812 (btver2_cost): Likewise.
3814 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
3816 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
3817 to simplify subreg of vec_merge.
3819 2018-10-18 Richard Biener <rguenther@suse.de>
3821 * config/i386/i386.c: Fix costing of vector FMA.
3823 2018-10-18 Richard Biener <rguenther@suse.de>
3825 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
3827 (ix86_builtin_vectorization_cost): For vec_construct properly
3828 cost insertion into SSE regs.
3829 (...): Adjust calls to ix86_vec_cost.
3831 2018-10-18 Richard Biener <rguenther@suse.de>
3835 2018-02-07 Richard Biener <rguenther@suse.de>
3837 PR tree-optimization/84204
3838 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
3841 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
3844 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
3846 (test_vector_ops_duplicate): Add test for a scalar subreg of a
3847 VEC_MERGE of a VEC_DUPLICATE.
3849 2018-10-17 Joseph Myers <joseph@codesourcery.com>
3851 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
3852 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
3853 * doc/standards.texi (C Language): Document C2X.
3854 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
3855 (rl78_option_override): Handle "GNU C2X" language name.
3857 2018-10-17 Joseph Myers <joseph@codesourcery.com>
3859 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
3860 Document C17 as published in 2018.
3862 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
3865 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
3866 bail out if both sides do not have the same storage order.
3868 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
3870 * bitmap.c (bitmap_head::dump): New.
3871 * bitmap.h (bitmap_head): Add dump().
3872 * gimple-ssa-evrp-analyze.c
3873 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
3874 (evrp_range_analyzer::set_ssa_range_info): Same.
3875 (evrp_range_analyzer::record_ranges_from_phis): Same.
3876 (evrp_range_analyzer::record_ranges_from_stmt): Same.
3877 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
3878 * gimple-ssa-sprintf.c (get_int_range): Same.
3879 (format_integer): Same.
3880 (sprintf_dom_walker::handle_gimple_call): Same.
3881 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
3882 (ipcp_vr_lattice::top_p): Same.
3883 (ipcp_vr_lattice::bottom_p): Same.
3884 (ipcp_vr_lattice::set_to_bottom): Same.
3885 (ipa_vr_operation_and_type_effects): Same.
3886 (propagate_vr_across_jump_function): Same.
3887 (ipcp_store_vr_results): Same.
3888 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
3889 (ipa_print_node_jump_functions_for_edge): Same.
3890 (ipa_get_value_range): Same.
3891 (ipa_compute_jump_functions_for_edge): Same.
3892 (ipa_write_jump_function): Same.
3893 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
3894 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
3896 * vr-values.c (set_value_range_to_nonnegative): Same.
3897 (set_value_range_to_truthvalue): Same.
3898 (vr_values::get_value_range): Same.
3899 (vr_values::set_defs_to_varying): Same.
3900 (vr_values::update_value_range): Same.
3901 (symbolic_range_based_on_p): Same.
3902 (vr_values::op_with_boolean_value_range_p): Same.
3903 (vr_values::extract_range_for_var_from_comparison_expr): Same.
3904 (vr_values::extract_range_from_ssa_name): Same.
3905 (vr_values::extract_range_from_binary_expr): Same.
3906 (vr_values::extract_range_from_unary_expr): Same.
3907 (vr_values::extract_range_from_cond_expr): Same.
3908 (vr_values::extract_range_from_comparison): Same.
3909 (vr_values::check_for_binary_op_overflow): Same.
3910 (vr_values::extract_range_basic): Same.
3911 (vr_values::extract_range_from_assignment): Same.
3912 (compare_ranges): Same.
3913 (compare_range_with_value): Same.
3914 (vr_values::adjust_range_with_scev): Same.
3915 (vrp_valueize): Same.
3916 (vrp_valueize_1): Same.
3917 (vr_values::get_vr_for_comparison): Same.
3918 (vr_values::compare_name_with_value): Same.
3919 (vr_values::compare_names): Same.
3920 (vr_values::vrp_evaluate_conditional): Same.
3921 (find_case_label_ranges): Same.
3922 (vr_values::vrp_visit_switch_stmt): Same.
3923 (vr_values::extract_range_from_phi_node): Same.
3924 (vr_values::simplify_div_or_mod_using_ranges): Same.
3925 (vr_values::simplify_bit_ops_using_ranges): Same.
3926 (test_for_singularity): Same.
3927 (range_fits_type_p): Same.
3928 (vr_values::simplify_cond_using_ranges_1): Same.
3929 (vr_values::simplify_switch_using_ranges): Same.
3930 (vr_values::simplify_float_conversion_using_ranges): Same.
3931 (vr_values::two_valued_val_range_p): Same.
3932 (vr_values::add_equivalence): Move to value_range::equiv_add.
3933 * vr-values.h (vr_values::add_equivalence): Remove.
3934 (VR_INITIALIZER): Remove.
3935 * tree-vrp.c (value_range::set): New.
3936 (value_range::equiv_add): New.
3937 (value_range::value_range): New.
3938 (value_range::deep_copy): New.
3939 (value_range::check): New.
3940 (value_range::equal_p): New.
3941 (value_range::ignore_equivs_equal_p): New.
3942 (value_range::operator==): New.
3943 (value_range::operator!=): New.
3944 (value_range::symbolic_p): New.
3945 (value_range::numeric_p): New.
3946 (value_range::set_undefined): New.
3947 (value_range::set_varying): New.
3948 (value_range::may_contain_p): New.
3949 (value_range::equiv_clear): New.
3950 (value_range::singleton_p): New.
3951 (value_range::intersect): New.
3952 (value_range::dump): New.
3953 (value_range::set_and_canonicalize): New.
3954 (set_value_range): Adjust for value_range API.
3955 (set_value_range_to_undefined): Same.
3956 (set_value_range_to_varying): Same.
3957 (set_and_canonicalize_value_range): Same.
3958 (set_value_range_to_nonnull): Same.
3959 (set_value_range_to_null): Same.
3960 (range_is_null): Same.
3961 (range_is_nonnull): Same.
3962 (range_int_cst_p): Same.
3963 (range_int_cst_singleton_p): Same.
3964 (symbolic_range_p): Same.
3965 (range_includes_zero_p): Same.
3966 (value_range_constant_singleton): Same.
3967 (vrp_set_zero_nonzero_bits): Same.
3968 (ranges_from_anti_range): Same.
3969 (extract_range_into_wide_ints): Same.
3970 (extract_range_from_multiplicative_op): Same.
3971 (set_value_range_with_overflow): Same.
3972 (extract_range_from_binary_expr_1): Same.
3973 (extract_range_from_unary_expr): Same.
3974 (dump_value_range): Same.
3975 (debug_value_range): Same.
3976 (vrp_prop::check_array_ref): Same.
3977 (vrp_prop::check_mem_ref): Same.
3978 (vrp_prop::vrp_initialize): Same.
3979 (vrp_prop::visit_stmt): Same.
3980 (intersect_ranges): Same.
3981 (vrp_prop::visit_phi): Same.
3982 (vrp_prop::vrp_finalize): Same.
3983 (determine_value_range_1): Same.
3984 (determine_value_range): Same.
3985 (vrp_intersect_ranges_1): Rename to...
3986 (vrp_intersect_1): this.
3987 (vrp_intersect_ranges): Rename to...
3988 (value_range::intersect_helper): ...this.
3989 (vrp_meet_1): Rename to...
3990 (value_range::union_helper): ...this.
3991 (vrp_meet): Rename to...
3992 (value_range::union_): ...this.
3993 (copy_value_range): Remove.
3994 * tree-vrp.h (struct value_range): Rewrite into a proper class.
3995 (value_range::vrtype): New.
3996 (value_range::type): New.
3997 (value_range::equiv): New.
3998 (value_range::min): New.
3999 (value_range::max): New.
4000 (value_range::varying_p): New.
4001 (value_range::undefined_p): New.
4002 (value_range::null_p): New.
4003 (value_range::equiv_add): New.
4004 (copy_value_range): Remove.
4006 2018-10-17 David Malcolm <dmalcolm@redhat.com>
4008 * Makefile.in (SELFTEST_TARGETS): New.
4009 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4010 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4011 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4013 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4014 (selftest-c++-gdb, selftest-c++-valgrind): Move to
4016 * configure: Regenerate.
4017 * configure.ac (selftest_languages): New.
4019 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
4021 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4022 overflow wraps argument.
4023 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4024 wide_int_range_multiplicative_op.
4025 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4026 overflow wraps argument.
4027 (wide_int_range_multiplicative_op): Same.
4028 (wide_int_range_lshift): Same.
4029 (wide_int_range_div): Same.
4030 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4031 (wide_int_range_lshift): Same.
4032 (wide_int_range_div): Same.
4034 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
4036 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4037 use sign as argument.
4038 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4039 wide_int_range_shift_undefined_p.
4041 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4043 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4045 (@despeculate_copy<ALLI_TI:mode>): ... This.
4046 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4049 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4051 * config.gcc: Obsolete *-*-solaris2.10*.
4052 * doc/install.texi (Specific, *-*-solaris2*): Document it.
4054 2018-10-12 Jeff Law <law@redhat.com>
4056 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4057 reg + sym +- const_int addressing modes.
4059 2018-10-15 David Malcolm <dmalcolm@redhat.com>
4061 * common.opt (fdiagnostics-minimum-margin-width=): New option.
4062 * diagnostic-show-locus.c (layout::layout): Apply the minimum
4064 (layout::start_annotation_line): Only print up to 3 of the
4065 margin character, to avoid touching the left-hand side.
4066 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4067 minimum margin width, as set by test_diagnostic_context's ctor.
4068 (selftest::test_fixit_insert_containing_newline): Likewise.
4069 (selftest::test_fixit_insert_containing_newline_2): Likewise.
4070 (selftest::test_line_numbers_multiline_range): Clear
4071 dc.min_margin_width.
4072 * diagnostic.c (diagnostic_initialize): Initialize
4074 * diagnostic.h (struct diagnostic_context): Add field
4076 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4077 * opts.c (common_handle_option): Handle
4078 OPT_fdiagnostics_minimum_margin_width_.
4079 * selftest-diagnostic.c
4080 (selftest::test_diagnostic_context::test_diagnostic_context):
4081 Initialize min_margin_width to 6.
4082 * toplev.c (general_init): Initialize global_dc->min_margin_width.
4084 2018-10-15 David Malcolm <dmalcolm@redhat.com>
4086 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4087 Fix usage of "error_at_rich_loc" in the comment.
4089 2018-10-15 Renlin Li <renlin.li@arm.com>
4092 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4093 if-conversioned loop when it contains ifn with types not
4094 supported by backend.
4095 * internal-fn.c (expand_direct_optab_fn): Add an assert.
4096 (direct_internal_fn_supported_p): New helper function.
4097 * internal-fn.h (direct_internal_fn_supported_p): Declare.
4099 2018-10-15 Jakub Jelinek <jakub@redhat.com>
4102 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4103 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4104 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4106 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
4108 PR tree-optimization/87022
4109 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4110 bits in dist vector rather than the first one.
4112 2018-10-15 Richard Biener <rguenther@suse.de>
4115 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4116 (visit_loadstore): When a used restrict tag escaped verify that
4117 the points-to solution of "other" pointers do not include
4119 (compute_dependence_clique): If a used restrict tag escaped
4120 communicated that down to visit_loadstore.
4122 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
4124 * config/s390/s390.c (s390_expand_vec_init): Force vector element
4125 into reg if it isn't a general operand.
4127 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
4130 * config/i386/sse.md (*vec_dupv2di): Add register source to
4133 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
4136 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4137 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4138 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4139 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4141 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
4143 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4144 (notice_args_size): Set it in the current trace if no insn that can
4145 throw internally has been seen yet.
4146 (connect_traces): When connecting args_size between traces, allow the
4147 incoming values not to match if there is an insn setting it before the
4148 first insn that can throw internally; in that case, force the creation
4149 of a CFI note on this latter insn.
4151 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
4153 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4154 the base class, because using the injected-class-name was not clearly
4155 specified until DR 176.
4157 2018-10-12 Paul Koning <ni1d@arrl.net>
4159 * config/pdp11/pdp11.md (doloop_end): New expander.
4160 (doloop_end_insn): renamed from "doloop_end".
4161 (addqi3): New pattern.
4162 (subqi3): New pattern.
4163 * config/pdp11/predicates.md (incdec_operand): New predicate.
4165 2018-10-12 Yury Gribov <tetra2005@gmail.com>
4168 * real.c (format_helper::can_represent_integral_type_p): New function
4169 * real.h (format_helper::can_represent_integral_type_p): Ditto.
4170 * match.pd: New pattern.
4172 2018-10-12 Alexandre Oliva <oliva@adacore.com>
4174 * configure.ac: Introduce --enable-large-address-aware
4175 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4176 * doc/install.texi: Document it.
4177 * configure, config.in: Rebuilt.
4178 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4179 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4180 (LINK_SPEC): Insert it.
4181 * config/i386/mingw-w64.h: Likewise.
4183 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4185 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
4187 PR rtl-optimization/87600
4188 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4190 2018-10-12 Paul Koning <ni1d@arrl.net>
4192 * doc/md.texi (doloop_end): Document that the pattern code may
4193 need to check operand mode.
4195 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
4197 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4198 to zero-extend between int and floating-point registers.
4199 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4200 ldp into floating-point registers. Add type and arch attributes.
4201 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4202 Use f_loads for type attribute.
4204 2018-10-11 Martin Sebor <msebor@redhat.com>
4206 * doc/extend.texi (attribute packed): Correct typos.
4208 2018-10-11 Martin Sebor <msebor@redhat.com>
4210 * doc/extend.texi (attribute flatten): Mention interaction with
4213 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
4216 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4217 Set new_decl virtual flag to zero.
4219 2018-10-11 Martin Sebor <msebor@redhat.com>
4222 * doc/extend.texi (attribute format_arg): Discuss using multiple
4223 attributes on a single function.
4225 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
4227 PR tree-optimization/86829
4228 * match.pd (sin (atan (x))): New simplification rules.
4229 (cos (atan (x))): Likewise.
4230 * real.c (build_sinatan_real): New function.
4231 * real.h (build_sinatan_real): Prototype.
4233 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
4235 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4237 (fold_mergeeo_helper): New helper function.
4238 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4239 intrinsics. Correct some whitespace indentation issues.
4241 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
4244 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4245 Use HOST_WIDE_INT_1U for shift.
4247 2018-10-11 Doug Rupp <rupp@adacore.com>
4248 Olivier Hainque <hainque@adacore.com>
4250 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4251 Pass --relax to the linker for RTPs.
4252 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4254 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
4255 Jan Hubicka <jh@suse.cz>
4256 Martin Jambor <mjambor@suse.cz>
4258 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4259 the same elements are repeated rather than printing all of them.
4260 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4261 "repeated" elements.
4262 * read-rtl-function.c (test_loading_repeat): New function.
4263 (read_rtl_function_c_tests): Call test_loading_repeat.
4264 * rtl-tests.c (test_dumping_repeat): New function.
4265 (rtl_tests_c_tests): Call test_dumping_repeat.
4267 2018-10-11 Richard Biener <rguenther@suse.de>
4269 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4270 bdver?_cost): Unify to ...
4271 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4272 * config/i386/i386.c (processor_cost_table): Adjust.
4274 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
4277 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4278 the thunk when expanding to GIMPLE.
4280 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
4282 * varasm.c (mergeable_string_section): Don't try to move zero-length
4283 strings to the merge section.
4285 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
4288 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4290 2018-10-10 Jakub Jelinek <jakub@redhat.com>
4293 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4294 to special_args set.
4296 2018-10-10 Richard Biener <rguenther@suse.de>
4298 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4299 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4300 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4301 and recursing and pattern terminating the recursion on SSE
4302 vector width using ix86_expand_reduc.
4303 (reduc_sminmax_scal_<mode>): Split into part reducing to half
4304 width and recursing and SSE2 vector variant doing the final
4305 reduction with ix86_expand_reduc.
4306 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4307 with terminating the recursion at AVX level, splitting that
4310 2018-10-09 David Malcolm <dmalcolm@redhat.com>
4312 * genmatch.c (error_cb): Rename to...
4313 (diagnostic_cb): ...this, converting int params to enums.
4314 (fatal_at): Update for renaming.
4315 (warning_at): Likewise.
4317 * input.c (selftest::ebcdic_execution_charset::apply):
4318 Update for renaming of...
4319 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4321 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4322 converting level and reason to enums.
4323 (class selftest::lexer_error_sink): Rename to...
4324 (class selftest::lexer_test_options): ...this, renaming field
4325 "m_errors" to "m_diagnostics".
4326 (selftest::lexer_test_options::apply): Update for renaming of...
4327 (selftest::lexer_test_options::on_error): ...this, renaming to...
4328 (selftest::lexer_test_options::on_diagnostic): ...this
4329 converting level and reason to enums.
4330 (selftest::test_lexer_string_locations_raw_string_unterminated):
4331 Update for renamings.
4332 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4335 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
4337 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4338 * config/rs6000/pmmintrin.h: New file.
4340 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
4342 PR tree-optimization/86659
4343 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4345 2018-10-09 Richard Biener <rguenther@suse.de>
4347 PR tree-optimization/63155
4348 * tree-ssa-structalias.c: Include tree-ssa.h.
4349 (get_constraint_for_ssa_var): For undefs return nothing_id.
4350 (find_func_aliases): Cleanup PHI handling.
4352 2018-10-09 Richard Biener <rguenther@suse.de>
4354 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4357 2018-10-09 Martin Liska <mliska@suse.cz>
4359 * asan.c (asan_emit_stack_protection): If a stack variable
4360 is located in a same file as current function, then emit
4361 line info into variable definition string.
4363 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4365 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4368 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4370 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4373 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4375 PR tree-optimization/86659
4376 * gimple-match.h (struct gimple_match_op): Add reverse field.
4377 (gimple_match_op::set_op): New overloaded method.
4378 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4379 the REF_REVERSE_STORAGE_ORDER flag on the value.
4380 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4381 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4383 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
4386 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4387 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4389 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
4392 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4393 Defined with __builtin_ia32_vfmaddsubpd512_mask.
4395 2018-10-08 Richard Biener <rguenther@suse.de>
4397 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4398 cost the same as AVX128 ones.
4400 2018-10-08 Paul Koning <ni1d@arrl.net>
4402 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4403 (expand_block_move): New function.
4404 * config/pdp11/pdp11.c (output_block_move): Remove.
4405 (expand_block_move): New function.
4406 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4407 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4408 (*movmemhi1): Remove.
4410 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
4412 * config/s390/2827.md: Increase latencies for some FP instructions.
4414 2018-10-08 Richard Biener <rguenther@suse.de>
4416 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4418 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4419 * tree-vectorizer.h (dump_stmt_cost): Adjust.
4420 (add_stmt_cost): Dump return value of the hook.
4422 2018-10-08 Richard Biener <rguenther@suse.de>
4424 PR tree-optimization/63155
4425 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4426 (ssa_propagation_engine::ssa_propagate): Remove redundant
4427 bitmap bit clearing.
4429 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
4431 PR rtl-optimization/86939
4432 PR rtl-optimization/87479
4433 * ira.h (non_conflicting_reg_copy_p): New prototype.
4434 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4435 (make_hard_regno_dead): Don't add conflicts for register
4436 ignore_reg_for_conflicts.
4437 (make_object_dead): Likewise.
4438 (non_conflicting_reg_copy_p): New function.
4439 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4440 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4441 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4442 (make_hard_regno_dead): Don't add conflicts for register
4443 ignore_reg_for_conflicts. Remove special conflict handling of
4444 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
4445 check_pic_pseudo_p and update callers.
4446 (mark_pseudo_dead): Don't add conflicts for register
4447 ignore_reg_for_conflicts.
4448 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4450 2018-10-05 Andrew Waterman <andrew@sifive.com>
4451 Jim Wilson <jimw@sifive.com>
4453 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4454 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
4455 new pattern using HONOR_SNANS that emits one extra instruction.
4457 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
4459 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4460 patterns): Merge SI and DI patterns to a GPR pattern.
4461 (unnamed define_insn and define_split for record form of that): Merge
4462 to a single define_insn_and_split pattern.
4464 2018-10-05 David Malcolm <dmalcolm@redhat.com>
4467 * input.c (expand_location_to_spelling_point): Add param "aspect"
4468 and use rather than hardcoding LOCATION_ASPECT_CARET.
4469 (get_substring_ranges_for_loc): Handle the case of a single token
4470 within a macro expansion.
4471 * input.h (expand_location_to_spelling_point): Add "aspect" param,
4472 defaulting to LOCATION_ASPECT_CARET.
4474 2018-10-05 Paul Koning <ni1d@arrl.net>
4476 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4477 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4478 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4479 (pdp11_guard_type): New function.
4481 2018-10-05 Paul Koning <ni1d@arrl.net>
4483 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4484 * config/pdp11/pdp11.opt (mfloat32): Remove.
4486 * doc/invoke.texi (pdp11 -mfloat32): Remove:
4487 (pdp11 -mfloat64): Remove.
4489 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4491 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4492 (*cmp<mode>_cc_i387): Ditto.
4493 (*cmpu<mode>_cc_i387): Ditto.
4494 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4495 * config/i386/i386.c (ix86_expand_fp_compare): Remove
4497 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4498 Emit x86_sahf_1 pattern.
4499 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4500 (ix86_expand_carry_flag_compare): Ditto.
4502 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4504 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4505 to reg_or_0_operand. Add "C" constraint.
4506 (*cmpxf_cc_i387): Ditto.
4507 (*cmp<mode>_i387): Change operand 2 predicate
4508 to nonimm_or_0_operand. Add "C" constraint.
4509 (*cmp<mode>_cc_i387): Ditto.
4510 (*cmp<mode>_0_i387): Remove insn pattern.
4511 (*cmp<mode>_0_cc_i387): Ditto.
4513 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4515 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4516 * config/i386/predicates.md (nonimm_or_0_operand): Rename
4517 from vector_move_operand. Update all uses.
4519 2018-10-05 Martin Sebor <msebor@redhat.com>
4521 PR tree-optimization/87490
4522 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4525 2018-10-05 Richard Biener <rguenther@suse.de>
4527 PR tree-optimization/63155
4528 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4529 vertical space in dumpfiles.
4530 * tree-ssa-propagate.h
4531 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4532 * tree-ssa-propagate.c (cfg_blocks_back): New global.
4533 (ssa_edge_worklist_back): Likewise.
4534 (curr_order): Likewise.
4535 (cfg_blocks_get): Remove abstraction.
4536 (cfg_blocks_add): Likewise.
4537 (cfg_blocks_empty_p): Likewise.
4538 (add_ssa_edge): Add to current or next worklist based on
4540 (add_control_edge): Likewise.
4541 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4543 (ssa_propagation_engine::ssa_propagate): ... here. Unify
4544 iteration from CFG and SSA edge worklist so we process
4545 everything in RPO order, prioritizing forward progress
4547 (ssa_prop_init): Allocate new worklists, do not dump
4549 (ssa_prop_fini): Free new worklists.
4551 2018-10-05 Richard Biener <rguenther@suse.de>
4553 * tree-core.h (tree_block::abstract_flag): Remove.
4554 (tree_block::block_num): Make full 32bits.
4555 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4556 * tree.h (BLOCK_ABSTRACT): Remove.
4557 * dwarf2out.c (gen_lexical_block_die): Remove dead code
4558 resulting from BLOCK_ABSTRACT being always false.
4559 (gen_inlined_subroutine_die): Likewise.
4560 (gen_block_die): Likewise.
4561 * tree.c (block_ultimate_origin): Likewise.
4562 * tree-pretty-print.c (dump_block_node): Remove code dealing
4563 with BLOCK_ABSTRACT.
4564 * tree-ssa-live.c (dump_scope_block): Likewise.
4565 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4566 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4568 2018-10-05 Richard Biener <rguenther@suse.de>
4570 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4571 is asked for initialize mode to the component mode of the
4574 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
4577 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4578 assembler for -mavx.
4579 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4581 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
4584 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4586 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4587 to be DImode. When using mffscrn, force the operand to a register.
4589 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
4591 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
4592 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
4593 X87MODEF mode iterator.
4594 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
4595 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
4596 X87MODEF mode iterator.
4598 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
4600 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
4601 -Wno-prio-ctor-dtor.
4603 2018-10-04 David Malcolm <dmalcolm@redhat.com>
4605 * Makefile.in (OBJS): Add opt-problem.o.
4606 * dump-context.h: Include "selftest.h.
4607 (selftest::temp_dump_context): New forward decl.
4608 (class dump_context): Make friend of class
4609 selftest::temp_dump_context.
4610 (dump_context::dump_loc_immediate): New decl.
4611 (class dump_pretty_printer): Move here from dumpfile.c.
4612 (class temp_dump_context): Move to namespace selftest.
4613 (temp_dump_context::temp_dump_context): Add param
4614 "forcibly_enable_dumping".
4615 (selftest::verify_dumped_text):
4616 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
4617 (selftest::verify_item):
4618 (ASSERT_IS_TEXT): Move here from dumpfile.c.
4619 (ASSERT_IS_TREE): Likewise.
4620 (ASSERT_IS_GIMPLE): Likewise.
4621 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
4623 (dump_context::dump_loc_immediate): ...this new function.
4624 (class dump_pretty_printer): Move to dump-context.h.
4625 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
4626 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
4627 (temp_dump_context::temp_dump_context): Move to "selftest"
4628 namespace. Add param "forcibly_enable_dumping", and use it to
4629 conditionalize the use of m_pp;
4630 (selftest::verify_dumped_text): Make non-static.
4631 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
4632 (selftest::verify_item): Make non-static.
4633 (ASSERT_IS_TEXT): Move to dump-context.h.
4634 (ASSERT_IS_TREE): Likewise.
4635 (ASSERT_IS_GIMPLE): Likewise.
4636 (selftest::test_capture_of_dump_calls): Pass "true" for new
4637 param of temp_dump_context.
4638 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
4639 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
4641 * opt-problem.cc: New file.
4642 * opt-problem.h: New file.
4643 * optinfo-emit-json.cc
4644 (selftest::test_building_json_from_dump_calls): Pass "true" for
4645 new param of temp_dump_context.
4646 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
4647 (optinfo::emit_for_opt_problem): New function.
4648 (optinfo::emit): Clarity which emit_item is used.
4649 * optinfo.h (optinfo::get_dump_location): New accessor.
4650 (optinfo::emit_for_opt_problem): New decl.
4651 (optinfo::emit): Make const.
4652 * selftest-run-tests.c (selftest::run_tests): Call
4653 selftest::opt_problem_cc_tests.
4654 * selftest.h (selftest::opt_problem_cc_tests): New decl.
4655 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
4656 bool to opt_result, converting fprintf messages to
4657 opt_result::failure_at calls. Add "stmt" param for use by the
4659 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
4660 (runtime_alias_check_p): Convert return type from bool to
4661 opt_result, converting dump_printf calls to
4662 opt_result::failure_at, using the statement DDR_A for their
4664 (find_data_references_in_stmt): Convert return type from bool to
4665 opt_result, converting "return false" to opt_result::failure_at
4667 * tree-data-ref.h: Include "opt-problem.h".
4668 (dr_analyze_innermost): Convert return type from bool to opt_result,
4669 and add a const gimple * param.
4670 (find_data_references_in_stmt): Convert return type from bool to
4672 (runtime_alias_check_p): Likewise.
4673 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
4674 dr_analyze_innermost.
4675 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
4676 Convert return type from bool to opt_result, adding a message for
4677 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
4678 (vect_analyze_data_ref_dependence): Convert return type from bool
4679 to opt_result. Change sense of return type from "false"
4680 effectively meaning "no problems" to "false" meaning a problem,
4681 so that "return false" becomes "return opt_result::success".
4682 Convert "return true" calls to opt_result::failure_at, using
4683 the location of statement A rather than vect_location.
4684 (vect_analyze_data_ref_dependences): Convert return type from bool
4686 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
4687 calls to opt_result::failure_at, using the stmt location rather
4689 (vect_verify_datarefs_alignment): Convert return type from bool
4691 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
4692 into multiple more-tightly-scoped copies.
4693 (vect_analyze_data_refs_alignment): Convert return type from bool
4695 (vect_analyze_data_ref_accesses): Likewise, converting a
4696 "return false" to a "return opt_result::failure_at", adding a
4698 (vect_prune_runtime_alias_test_list): Convert return type from
4699 bool to opt_result, converting dump_printf_loc to
4700 opt_result::failure_at. Add a %G to show the pertinent statement,
4701 and use the stmt's location rather than vect_location.
4702 (vect_find_stmt_data_reference): Convert return type from
4703 bool to opt_result, converting dump_printf_loc to
4704 opt_result::failure_at, using stmt's location.
4705 (vect_analyze_data_refs): Convert return type from bool to
4706 opt_result. Convert "return false" to "return
4707 opt_result::failure_at", adding messages as needed.
4708 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
4709 type from bool to opt_result.
4710 (vect_determine_vf_for_stmt): Likewise.
4711 (vect_determine_vectorization_factor): Likewise, converting
4712 dump_printf_loc to opt_result::failure_at, using location of phi
4713 rather than vect_location.
4714 (vect_analyze_loop_form_1): Convert return type from bool to
4715 opt_result, converting dump_printf_loc calls, retaining the use of
4717 (vect_analyze_loop_form): Convert return type from loop_vec_info
4718 to opt_loop_vec_info.
4719 (vect_analyze_loop_operations): Convert return type from bool to
4720 opt_result, converting dump_printf_loc calls, using the location
4721 of phi/stmt rather than vect_location where available. Convert
4722 various "return false" to "return opt_result::failure_at" with
4723 "unsupported phi" messages.
4724 (vect_get_datarefs_in_loop): Convert return type from bool to
4725 opt_result. Add a message for the
4726 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
4727 (vect_analyze_loop_2): Convert return type from bool to
4728 opt_result. Ensure "ok" is set to a opt_result::failure_at before
4729 each "goto again;", adding new messages where needed.
4730 Add "unsupported grouped {store|load}" messages.
4731 (vect_analyze_loop): Convert return type from loop_vec_info to
4733 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
4735 * tree-vect-stmts.c (process_use): Likewise, converting
4736 dump_printf_loc call and using stmt location, rather than
4738 (vect_mark_stmts_to_be_vectorized): Likeise.
4739 (vect_analyze_stmt): Likewise, adding a %G.
4740 (vect_get_vector_types_for_stmt): Convert return type from bool to
4741 opt_result, converting dump_printf_loc calls and using stmt
4742 location, rather than vect_location.
4743 (vect_get_mask_type_for_stmt): Convert return type from tree to
4744 opt_tree, converting dump_printf_loc calls and using stmt location.
4745 * tree-vectorizer.c: Include "opt-problem.h.
4746 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
4747 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
4748 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
4749 enabled, use it to report at the top level "couldn't vectorize
4750 loop" followed by the problem.
4751 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
4752 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
4754 (vect_analyze_stmt): Likewise.
4755 (vect_get_vector_types_for_stmt): Likewise.
4756 (tree vect_get_mask_type_for_stmt): Likewise.
4757 (vect_analyze_data_ref_dependences): Likewise.
4758 (vect_enhance_data_refs_alignment): Likewise.
4759 (vect_analyze_data_refs_alignment): Likewise.
4760 (vect_verify_datarefs_alignment): Likewise.
4761 (vect_analyze_data_ref_accesses): Likewise.
4762 (vect_prune_runtime_alias_test_list): Likewise.
4763 (vect_find_stmt_data_reference): Likewise.
4764 (vect_analyze_data_refs): Likewise.
4765 (vect_analyze_loop): Convert return type from loop_vec_info to
4767 (vect_analyze_loop_form): Likewise.
4768 (vect_analyze_slp): Convert return type from bool to opt_result.
4770 2018-10-04 David Malcolm <dmalcolm@redhat.com>
4772 * doc/invoke.texi (-fopt-info): Document new "internals"
4774 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
4775 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
4777 (optinfo_verbosity_options): Add "internals".
4778 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
4779 (dump_context::apply_dump_filter_p): New member function.
4780 (dump_context::dump_loc): Use apply_dump_filter_p rather than
4781 explicitly masking the dump_kind.
4782 (dump_context::begin_scope): Increment the scope depth first. Use
4783 apply_dump_filter_p rather than explicitly masking the dump_kind.
4784 (dump_context::emit_item): Use apply_dump_filter_p rather than
4785 explicitly masking the dump_kind.
4786 (dump_dec): Likewise.
4787 (dump_hex): Likewise.
4788 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
4789 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
4790 (opt_info_switch_p): Update handling of default
4791 MSG_OPTIMIZED_LOCATIONS to cope with default of
4792 MSG_PRIORITY_USER_FACING.
4793 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
4794 masking the dump_kind.
4795 (selftest::test_capture_of_dump_calls): Update test_dump_context
4796 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
4797 than MSG_ALL. Generalize scope test to be run at all four
4798 combinations of with/without MSG_PRIORITY_USER_FACING and
4799 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
4800 for each of the two values.
4801 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
4802 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
4803 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
4804 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
4805 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
4806 with MSG_PRIORITY_*.
4807 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
4808 dump messages as MSG_PRIORITY_USER_FACING.
4809 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
4810 about the interaction with MSG_PRIORITY_*.
4812 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
4814 * varasm.c (output_constant): Add new parameter merge_strings.
4815 Make strings properly zero terminated in merge string sections.
4816 (mergeable_string_section): Don't fail if the last char is non-zero.
4817 (assemble_variable_contents): Handle merge string sections.
4818 (assemble_variable): Likewise.
4819 (assemble_constant_contents): Likewise.
4820 (output_constant_def_contents): Likewise.
4821 (output_constructor_array_range,
4822 output_constructor_regular_field): Adjust call to output_constant.
4823 (output_object_block): Adjust call to assemble_constant_contents
4824 and assemble_variable_contents.
4826 2018-10-04 Martin Liska <mliska@suse.cz>
4829 * cgraphunit.c (process_function_and_variable_attributes):
4830 Warn about a function with alias attribute and a body.
4832 2018-10-04 Martin Liska <mliska@suse.cz>
4835 * multiple_target.c (redirect_to_specific_clone): New function.
4836 (ipa_target_clone): Use it.
4837 * tree-inline.c: Fix comment.
4839 2018-10-04 David Malcolm <dmalcolm@redhat.com>
4841 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
4843 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
4844 (gcc::dump_manager::register_pass): New member function, adapted
4845 from loop body in gcc::pass_manager::register_pass, adding a
4846 call to update_dfi_for_opt_info.
4847 (gcc::dump_manager::opt_info_enable_passes): Store the
4848 -fopt-info options into the new fields. Move the loop
4850 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
4852 * dumpfile.h (struct opt_pass): New forward decl.
4853 (gcc::dump_manager::register_pass): New decl.
4854 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
4855 (class gcc::dump_manager): Add fields "m_optgroup_flags",
4856 "m_optinfo_flags", and "m_optinfo_filename".
4857 * passes.c (gcc::pass_manager::register_pass): Move all of the
4858 dump-handling code to gcc::dump_manager::register_pass.
4860 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
4862 PR rtl-optimization/87466
4863 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
4864 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
4865 * doc/tm.texi: Regenerate.
4866 * ira-lives.c (process_bb_node_lives): Use the new target hook.
4867 * lra-lives.c (process_bb_lives): Likewise.
4868 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
4871 2018-10-04 Tamar Christina <tamar.christina@arm.com>
4873 * params.c (add_params): Fix initialization.
4875 2018-10-04 Martin Liska <mliska@suse.cz>
4877 PR gcov-profile/84107
4878 * tree-profile.c (init_ic_make_global_vars):
4879 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
4880 Come up with new ic_tuple* variables. Emit
4881 __gcov_indirect_call{,_topn} variables.
4882 (gimple_gen_ic_profiler): Access the variable
4884 (gimple_gen_ic_func_profiler): Access
4885 __gcov_indirect_call.callee field.
4886 (gimple_init_gcov_profiler): Use ptr_type_node.
4887 * value-prof.c (gimple_ic): Use ptr_type_node.
4889 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4891 PR tree-optimization/85787
4892 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
4893 into this function and add support for detecting multiple phis.
4894 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
4896 2018-10-04 Martin Liska <mliska@suse.cz>
4899 * ipa-inline.c (inline_to_all_callers_1):
4900 Call ultimate_alias_target for node being inlined.
4902 2018-10-03 Jeff Law <law@redhat.com>
4904 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
4906 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
4907 * tree.h (get_typenode_from_name): Prototype.
4909 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
4911 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
4912 Change operand 2 predicate to nonimmediate_operand.
4913 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4915 2018-10-03 Martin Sebor <msebor@redhat.com>
4916 Jeff Law <law@redhat.com>
4918 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
4920 (get_string_length): Detect unterminated arrays.
4921 (format_string): Same.
4922 (format_directive): Warn about unterminated arrays.
4923 (handle_gimple_call): Mark statements with no_warning as needed.
4925 2018-10-03 Jim Wilson <jimw@sifive.com>
4927 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
4928 also define __riscv_abi_rve. Delete trailing white space.
4930 2018-10-03 Paul Koning <ni1d@arrl.net>
4932 Enable LRA register allocator for PDP11.
4933 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
4936 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
4937 * config/pdp11/pdp11.opt (-mlra): New option.
4938 * doc/invoke.texi (PDP-11 Options): Document -mlra.
4940 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
4942 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
4943 (*<absneg:code>extend<mode>xf2): Ditto.
4945 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
4947 PR tree-optimization/87415
4948 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
4951 2018-10-02 Jeff Law <law@redhat.com>
4953 * gimple-fold.c (get_range_strlen): Only set *nonstr when
4954 an unterminated string is discovered. Bubble up range
4955 even for unterminated strings.
4956 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
4957 indicates the string was not terminated via NONSTR.
4959 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
4961 * tree-vrp.c (extract_range_from_unary_expr): Special case all
4962 pointer conversions.
4963 Do not do anything special for anti-ranges.
4965 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
4967 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
4968 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
4970 2018-10-03 Martin Liska <mliska@suse.cz>
4972 PR gcov-profile/86109
4973 * coverage.c (coverage_begin_function): Do not
4974 mark lambdas as artificial.
4975 * tree-core.h (struct GTY): Remove tm_clone_flag
4976 and introduce new lambda_function.
4977 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
4979 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4982 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
4983 P8_VECTOR and VSX are enabled.
4985 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
4987 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
4988 0x3907 as CPU model number.
4990 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
4992 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
4993 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
4994 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
4995 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
4996 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
4997 * config/s390/s390.md: Likewise. Rename also the cpu attribute
4998 value from arch12 to z14.
5000 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
5002 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5004 (isinf<mode>2): Ditto.
5006 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
5008 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5009 before emitting fxam. Perform calculations in XFmode.
5011 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
5013 * match.pd (((X /[ex] A) +- B) * A): New transformation.
5015 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
5018 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5019 * tree.c (signed_or_unsigned_type_for): Handle complex.
5021 2018-10-02 Jeff Law <law@redhat.com>
5023 * gimple-fold.c (get_range_strlen): Remove dead code.
5025 2018-10-02 Martin Sebor <msebor@redhat.com>
5026 Jeff Law <law@redhat.com>
5028 * builtins.c (unterminated_array): Add new arguments.
5029 If argument is not terminated, bubble up size and exact
5031 (expand_builtin_strnlen): Detect, avoid expanding
5032 and diagnose unterminated arrays.
5033 (c_strlen): Fill in offset of start of unterminated strings.
5034 * builtins.h (unterminated_array): Update prototype.
5036 2018-10-02 Richard Biener <rguenther@suse.de>
5038 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5039 of haddv4df, first reduce to SSE width and exploit the fact
5040 that we only need element zero with the reduction result.
5041 (reduc_plus_scal_v2df): Likewise.
5043 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
5045 * dojump.h (do_jump): Delete.
5046 (do_jump_1): Likewise.
5047 (split_comparison): Move around.
5048 * dojump.c (do_jump): Make static.
5049 (do_jump_1): Likewise.
5050 (jumpifnot): Move around.
5051 (jumpifnot_1): Likewise.
5053 (jumpif_1): Likewise.
5054 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5056 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
5058 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5059 insns in the delay slot and add_insn_after for the jump insn.
5061 2018-10-02 Richard Biener <rguenther@suse.de>
5063 * tree-inline.c (expand_call_inline): Use the location of
5064 the callee declaration for the inline-entry marker.
5065 * final.c (notice_source_line): Remove special-casing of
5066 NOTE_INSN_INLINE_ENTRY.
5068 2018-10-01 Carl Love <cel@us.ibm.com>
5071 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5072 (__builtin_mtfsb0): New.
5073 (__builtin_mtfsb1): New.
5074 ( __builtin_set_fpscr_rn): New.
5075 (__builtin_set_fpscr_drn): New.
5076 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5077 (rs6000_expand_set_fpscr_rn_builtin): Add.
5078 (rs6000_expand_set_fpscr_drn_builtin): Add.
5079 (rs6000_expand_builtin): Add case statement entries for
5080 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5081 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5082 RS6000_BUILTIN_MFFSL.
5083 (rs6000_init_builtins): Add ftype initialization and def_builtin
5084 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5085 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5086 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5087 rs6000_mffscdrn): Add define_insn.
5088 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5089 * doc/extend.texi: Add documentation for the builtins.
5091 2018-10-01 Richard Biener <rguenther@suse.de>
5093 PR tree-optimization/87465
5094 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5095 causing branch miscounts.
5097 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5099 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5100 aarch64_option_default_param): New.
5101 (params.h): Include.
5102 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5103 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5104 stack-clash protection validation code.
5106 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5108 * params.c (validate_param): New.
5109 (add_params): Use it.
5110 (set_param_value): Refactor param validation into validate_param.
5111 (diagnostic.h): Include.
5112 * diagnostic.h (diagnostic_ready_p): New.
5114 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5116 * params.c (set_param_value):
5117 Add index of parameter being validated.
5118 * common/common-target.def (option_validate_param): New.
5119 * common/common-targhooks.h (default_option_validate_param): New.
5120 * common/common-targhooks.c (default_option_validate_param): New.
5121 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5122 * doc/tm.texi: Regenerate.
5124 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5127 * config/aarch64/aarch64.c (aarch64_override_options_internal):
5128 Add validation for stack-clash parameters and set defaults.
5130 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5133 * configure.ac: Add stack-clash-protection-guard-size.
5134 * doc/install.texi: Document it.
5135 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5136 * params.def: Update comment for guard-size.
5137 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5138 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5139 * configure: Regenerate.
5141 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5144 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5145 STACK_DYNAMIC_OFFSET): New.
5146 * config/aarch64/aarch64.c (aarch64_layout_frame):
5147 Update outgoing args size.
5148 (aarch64_stack_clash_protection_alloca_probe_range,
5149 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5151 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5154 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5156 * target.def (stack_clash_protection_alloca_probe_range): New.
5157 (stack_clash_protection_final_dynamic_probe): Remove.
5158 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5159 (default_stack_clash_protection_final_dynamic_probe): Remove.
5160 * targhooks.c: Likewise.
5161 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5162 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5163 * doc/tm.texi: Regenerate.
5165 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5168 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5169 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5170 aarch64_clamp_to_uimm12_shift): New.
5171 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5172 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5174 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5177 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5179 2018-10-01 Jeff Law <law@redhat.com>
5180 Richard Sandiford <richard.sandiford@linaro.org>
5181 Tamar Christina <tamar.christina@arm.com>
5184 * config/aarch64/aarch64.md
5185 (probe_stack_range): Add k (SP) constraint.
5186 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5187 STACK_CLASH_MAX_UNROLL_PAGES): New.
5188 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5189 stack probes for stack clash.
5190 (aarch64_allocate_and_probe_stack_space): New.
5191 (aarch64_expand_prologue): Use it.
5192 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5193 (aarch64_sub_sp): Add emit_move_imm optional param.
5195 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
5197 PR tree-optimization/87261
5198 * match.pd: Remove trailing whitespace.
5199 Add (x & y) | ~(x | y) -> ~(x ^ y),
5200 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5202 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
5204 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5206 (ashlsi3_insn): Update instruction constraints.
5207 (ashrsi3_insn): Likewise.
5208 (rotrsi3): Likewise.
5209 (add_shift): Likewise.
5210 * config/arc/constraints.md (Csz): New 32 bit constraint. It
5211 avoids placing in the limm field small constants which, otherwise,
5212 could end into a small instruction.
5214 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
5216 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5217 destination register is not odd-even.
5218 (umaddsidi4_split): Likewise.
5220 2018-10-01 Richard Biener <rguenther@suse.de>
5222 * tree-inline.c (expand_call_inline): Store origin of fn
5223 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5224 * tree.c (block_ultimate_origin): Simplify and do some
5227 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
5229 * config/i386/mmx.md (EMMS): New int iterator.
5230 (emms): New int attribute.
5231 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5232 EMMS int iterator. Explicitly declare clobbers.
5233 (mmx_emms): Remove expander.
5235 * config/i386/predicates.md (emms_operation): Remove predicate.
5236 (vzeroall_pattern): New predicate.
5237 (vzeroupper_pattern): Rename from vzeroupper_operation.
5238 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5239 vzeroupper_pattern and vzeroall_pattern predicates.
5241 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
5243 PR rtl-optimization/86939
5244 * ira-lives.c (make_hard_regno_born): Rename from this...
5245 (make_hard_regno_live): ... to this. Remove update to conflict
5246 information. Update function comment.
5247 (make_hard_regno_dead): Add conflict information update. Update
5249 (make_object_born): Rename from this...
5250 (make_object_live): ... to this. Remove update to conflict information.
5251 Update function comment.
5252 (make_object_dead): Add conflict information update. Update function
5254 (mark_pseudo_regno_live): Call make_object_live.
5255 (mark_pseudo_regno_subword_live): Likewise.
5256 (mark_hard_reg_dead): Update function comment.
5257 (mark_hard_reg_live): Call make_hard_regno_live.
5258 (process_bb_node_lives): Likewise.
5259 * lra-lives.c (make_hard_regno_born): Rename from this...
5260 (make_hard_regno_live): ... to this. Remove update to conflict
5261 information. Remove now uneeded check_pic_pseudo_p argument.
5262 Update function comment.
5263 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5264 to conflict information. Update function comment.
5265 (mark_pseudo_live): Remove update to conflict information. Update
5267 (mark_pseudo_dead): Add conflict information update.
5268 (mark_regno_live): Call make_hard_regno_live.
5269 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5270 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5272 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
5275 * config/i386/i386.c (construct_container): Use TImode for
5276 BLKmode values in 2 integer registers.
5278 2018-09-29 Jeff Law <law@redhat.com>
5280 * builtins.c (unterminated_array): Pass in c_strlen_data * to
5281 c_strlen rather than just a tree *.
5282 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5283 Update recursive calls appropriately. If caller did not provide a
5284 suitable data pointer, create a local one. When a non-terminated
5285 string is discovered, bubble up information about the string via the
5286 c_strlen_data object.
5287 * builtins.h (c_strlen): Update prototype.
5288 (c_strlen_data): New structure.
5289 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5290 For a type 2 call, if c_strlen indicates a non-terminated string
5291 use the length of the non-terminated string.
5292 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5294 2018-09-29 Jakub Jelinek <jakub@redhat.com>
5297 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5298 __m512d type for __A argument rather than __m512.
5300 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
5302 * match.pd (simple_comparison): Don't optimize if either operand is
5303 a function pointer when target needs function pointer canonicalization.
5305 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5307 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5308 power5 .. power9 to remove indirection.
5309 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5310 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5311 ASM_CPU_476_SPEC): Delete.
5312 (ASM_CPU_SPEC): Adjust.
5313 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5314 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5316 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5318 * config.in: Delete HAVE_AS_DCI.
5319 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5320 * config/rs6000/rs6000.h: Ditto.
5321 * configure.ac: Delete HAVE_AS_DCI.
5322 * configure: Regenerate.
5324 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5326 * config.in (HAVE_AS_LWSYNC): Delete.
5327 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5328 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5330 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5331 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5333 * configure.ac: Delete HAVE_AS_LWSYNC.
5334 * configure: Regenerate.
5336 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
5337 Pierre-Marie de Rodat <derodat@adacore.com>
5339 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5340 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5341 (cgraph_node::create_thunk): Add indirect_offset parameter.
5342 (thunk_adjust): Likewise.
5343 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5344 and initialize the corresponding field with it.
5345 (cgraph_node::dump): Dump indirect_offset field.
5346 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5347 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5348 (thunk_adjust): Add indirect_offset parameter and deal with it.
5349 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5350 pass it to thunk_adjust. Do not call the target hook if it's non-zero
5351 or if the thunk is external or local. Fix formatting. Do not chain
5352 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
5353 if any, in the GIMPLE representation.
5354 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5355 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5356 (input_node): Read indirect_offset field.
5357 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5358 call to thunk_adjust.
5359 * tree-nested.c (struct nesting_info): Add thunk_p field.
5360 (create_nesting_tree): Set it.
5361 (convert_all_function_calls): Copy static chain from targets to thunks.
5362 (finalize_nesting_tree_1): Return early for thunks.
5363 (unnest_nesting_tree_1): Do not finalize thunks.
5364 (gimplify_all_functions): Do not gimplify thunks.
5366 2018-09-28 David Malcolm <dmalcolm@redhat.com>
5368 * opt-suggestions.c (option_proposer::build_option_suggestions):
5369 Release "option_values".
5371 2018-09-28 David Malcolm <dmalcolm@redhat.com>
5373 * coverage.c (get_coverage_counts): Convert problem-reporting dump
5374 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5375 * dumpfile.c (kind_as_string): New function.
5376 (dump_loc): Rather than a hardcoded prefix of "note: ", use
5377 kind_as_string to vary the prefix based on dump_kind.
5378 (selftest::test_capture_of_dump_calls): Update for above.
5380 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
5382 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5383 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
5385 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
5387 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5388 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5389 INVALID_REGNUM instead of FPSR_REG.
5390 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5391 * config/i386/i386.md: Update comment of FP compares.
5392 (fldenv): Do not clobber FPSR_REG.
5394 2018-09-28 Richard Biener <rguenther@suse.de>
5396 * tree.h (BLOCK_ORIGIN): New.
5397 * omp-expand.c (grid_expand_target_grid_body): Assign
5398 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5399 * tree-inline.c (remap_block): Likewise.
5400 * auto-profile.c (get_function_decl_from_block): Simplify
5401 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5402 * langhooks.c (lhd_print_error_function): Likewise.
5403 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5405 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5406 * tree.c (block_nonartificial_location): Likewise.
5407 (block_ultimate_origin): Likewise.
5408 * tree-pretty-print.c (percent_K_format): Likewise. Remove
5409 no longer needed LTO case.
5411 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
5412 Jan Hubicka <jh@suse.cz>
5413 Martin Jambor <mjambor@suse.cz>
5415 * simplify-rtx.c (simplify_merge_mask): New function.
5416 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5417 same masks are used in op1 or op2.
5418 (test_vec_merge): New function.
5419 (test_vector_ops): Call test_vec_merge.
5421 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
5423 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5424 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5425 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5426 (cypress_costs): Set it.
5427 (supersparc_costs): Likewise.
5428 (hypersparc_costs): Likewise.
5429 (leon_cost): Likewise.
5430 (leon3_costs): Likewise.
5431 (sparclet_costs): Likewise.
5432 (ultrasparc_costs): Likewise.
5433 (ultrasparc_costs): Likewise.
5434 (niagara_costs): Likewise.
5435 (niagara2_costs): Likewise.
5436 (niagara3_costs): Likewise.
5437 (niagara4_costs): Likewise.
5438 (niagara7_costs): Likewise.
5439 (m8_costs): Likewise.
5440 (TARGET_CAN_FOLLOW_JUMP): Define.
5441 (pass_work_around_errata::gate): Minor tweak.
5442 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5443 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5444 Automaitcally clear MASK_FSMULD mask for V7 processors.
5445 (sparc_can_follow_jump): New static function.
5446 (output_ubranch): Deal with CROSSING_JUMP_P.
5447 (sparc_use_sched_lookahead): Rewrite using switch statement.
5448 (sparc_issue_rate): Reorder.
5449 (sparc_branch_cost): New function.
5451 2018-09-27 Martin Sebor <msebor@redhat.com>
5453 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5454 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5455 (int_fits_type_p): Same.
5457 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
5459 * config/i386/i386.md (FPCR_REG): Remove.
5460 (UNSPEC_FLDCW): Remove.
5461 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5462 (x86_fldcw_1): Remove insn pattern.
5463 (fnstenv): Do not clobber FPCR_REG.
5465 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5466 (CALL_USED_REGISTERS): Ditto.
5467 (REG_ALLOC_ORDER): Ditto.
5468 (REG_CLASS_CONTENTS): Ditto.
5469 (HI_REGISTER_NAMES): Ditto.
5470 (ADDITIONAL_REGISTER_NAMES): Use defines instead
5471 of numerical constants.
5472 * config/i386/i386.c (regclass_map): Remove fpsr register.
5473 (dbx_register_map): Ditto.
5474 (dbx64_register_map): Ditto.
5475 (svr4_dbx_register_map): Ditto.
5476 (print_reg): Do not handle FPCR_REG.
5478 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
5481 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5482 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5483 Delete, always treat as true.
5484 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5485 Ditto. Simplify remaining code.
5486 * config/powerpcspe/powerpcspe.h: Ditto.
5487 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5488 Simplify remaining code.
5489 (rs6000_expand_builtin): Ditto.
5490 * config/rs6000/rs6000.h: Ditto.
5491 * configure.ac: Ditto.
5492 * configure: Regenerate.
5494 2018-09-27 Martin Liska <mliska@suse.cz>
5496 * coverage.c (get_coverage_counts): Revert the formatting
5497 of missing profile opt info.
5499 2018-09-27 Richard Biener <rguenther@suse.de>
5503 * dwarf2out.c (set_block_origin_self): Do not mark outermost
5504 block as we do not output that.
5505 (gen_inlined_subroutine_die): Elide the originally outermost
5506 block, matching what we do for concrete instances.
5507 (decls_for_scope): Add parameter specifying whether to recurse
5510 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
5511 Tom de Vries <tom@codesourcery.com>
5515 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5516 STORE_FLAG_VALUE == 1.
5518 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
5520 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5521 constant definitions.
5522 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5523 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5524 ("speculation_barrier"): New expander definition.
5526 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
5528 PR gcov-profile/86957
5529 * common.opt: New warning option -Wmissing-profile.
5530 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5531 * doc/invoke.texi: Document -Wmissing-profile.
5533 2018-09-26 Jim Wilson <jimw@sifive.com>
5535 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5536 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5538 2018-09-26 Martin Sebor <msebor@redhat.com>
5540 * tree.c (zerop): Change return type to bool.
5541 (integer_zerop, integer_onep, integer_each_onep): Same.
5542 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5543 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5544 (real_onep, real_minus_onep, chain_index): Same.
5545 (print_type_hash_statistics, type_list_equal): Same.
5546 * tree.h (zerop): Same.
5547 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5548 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5549 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5550 (real_onep, real_minus_onep, chain_index): Same.
5551 (print_type_hash_statistics, type_list_equal): Same.
5553 2018-09-26 Jim Wilson <jimw@sifive.com>
5555 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5557 2018-09-26 Jakub Jelinek <jakub@redhat.com>
5560 * config/i386/i386.c: Include debug.h and dwarf2out.h.
5561 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5564 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
5566 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5568 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
5570 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5571 and FP_SECOND_SSE_REGS.
5572 (REG_CLASS_NAMES): Ditto.
5573 (REG_CLASS_CONTENTS): Ditto.
5574 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5575 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5576 (ix86_preferred_output_reload_class): Ditto.
5577 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5578 clobber constraint to "=&f".
5579 (fix_truncdi_i387): Ditto.
5580 (lrintxfdi2): Ditto.
5581 (fistdi2_<rounding>): Ditto.
5582 (fpremxf4_i387): Change "=u" constraint to "=f".
5583 (fprem1xf4_i387): Ditto.
5585 (fptanxf4_i387): Ditto.
5586 (fxtractxf3_i387): Ditto.
5587 (fscalexf4_i387): Ditto.
5588 (atan2xf3): Change "u" constraint to "f".
5589 (fyl2xxf3_i387): Ditto.
5590 (fyl2xp1xf3_i387): Ditto.
5592 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
5595 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
5596 for removed I387_MASK_PM entity.
5599 2018-09-26 Jeff Law <law@redhat.com>
5601 2018-09-26 Alexey Neyman <stilor@att.net>
5603 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5604 headers are no longer pulled in by <isl/val.h>.
5606 2018-09-26 Richard Biener <rguenther@suse.de>
5609 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
5610 or concrete instance DIE to the tree. Create abstract origin
5611 attributes also for concrete instances.
5613 2018-09-26 Alexey Neyman <stilor@att.net>
5615 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
5616 headers are no longer pulled in by <isl/val.h>.
5618 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
5620 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
5621 Use new helper functions.
5622 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
5623 Use new helper functions.
5624 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
5625 aarch_mm_needs_release): New declarations.
5626 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
5627 aarch_mm_needs_release): New.
5629 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
5631 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
5632 (arm32_output_mi_thunk): Deal with long calls.
5634 2018-09-26 Richard Biener <rguenther@suse.de>
5638 * tree-inline.c (expand_call_inline): When the location
5639 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
5640 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
5641 the inserted BLOCK to make inlined_function_outer_scope_p
5643 * dwarf2out.c (add_call_src_coords_attributes): Do not add
5644 coords for reserved locations.
5646 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
5648 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
5649 (*call_indirect_nonlocal_sysv<mode>): Ditto.
5650 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
5651 (*sibcall_nonlocal_sysv<mode>): Ditto.
5652 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
5653 (<bd>_<mode>): Ditto.
5654 (<bd>tf_<mode>): Ditto.
5656 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
5658 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
5659 control string as a list of templates instead of as C code.
5660 (*altivec_movti): Ditto.
5661 * config/rs6000/darwin.md (movdf_low_di): Ditto.
5663 2018-09-25 Jim Wilson <jimw@sifive.com>
5665 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
5666 when target symbol is weak.
5668 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
5671 * builtins.c (unterminated_array): Simplify.
5672 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
5673 where pointer arithmetic is safe.
5675 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
5678 * config/rs6000/altivec.md (altivec_vspltb): Use
5679 const_0_to_15_operand instead of u5bit_cint_operand.
5680 (*altivec_vspltb_internal): Ditto.
5681 (altivec_vspltb_direct): Ditto.
5682 (altivec_vsplth): Use const_0_to_7_operand instead of
5684 (*altivec_vsplth_internal): Ditto.
5685 (altivec_vsplth_direct): Ditto.
5686 (altivec_vspltw): Use const_0_to_3_operand instead of
5688 (*altivec_vspltw_internal): Ditto.
5689 (altivec_vspltw_direct): Ditto.
5690 (altivec_vspltsf): Ditto.
5691 (*altivec_vspltsf_internal): Ditto.
5692 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
5693 various splats with the proper size immediate. Reorder the various
5694 cases by ascending size of immediate, and put all such together.
5696 2018-09-25 Richard Biener <rguenther@suse.de>
5699 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
5701 (maybe_create_die_with_external_ref): Do not create
5702 DW_TAG_imported_unit here.
5703 (add_abstract_origin_attribute): Handle external BLOCK refs.
5704 (dwarf2out_abstract_function): Simplify LTO case.
5705 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
5706 rather than using maybe_create_die_with_external_ref.
5708 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
5711 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
5712 (frndintxf2_mask_pm_i387): Ditto.
5713 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
5714 Enable for !flag_trapping_math.
5715 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
5716 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
5717 Change operand 1 predicate to nonimmediate_operand.
5718 (attr "i387_cw"): Remove mask_pm.
5719 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
5720 (enum ix86_entity): Remove I387_MASK_PM.
5721 * config/i386/i386.c (ix86_i387_mode_needed): Do not
5722 handle I387_MASK_PM.
5723 (ix86_mode_needed): Ditto.
5724 (ix86_mode_after): Ditto.
5725 (ix86_mode_entry): Ditto.
5726 (ix86_mode_exit): Ditto.
5727 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
5729 2018-09-25 Jakub Jelinek <jakub@redhat.com>
5731 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
5732 to_update_switch_stmts to vNULL instead of calling create on them
5735 2018-09-25 Richard Biener <rguenther@suse.de>
5737 PR tree-optimization/87402
5738 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
5739 (visit_phi): Re-instantiate handling of supposed to be VARYING
5740 but non-VARYING backedge value.
5742 2018-09-25 Richard Biener <rguenther@suse.de>
5745 * dwarf2out.c (struct sym_off_pair): New.
5746 (external_die_map): New global.
5747 (lookup_decl_die): When in LTO create DIEs lazily from the
5749 (lookup_block_die): New function, create DIEs lazily in LTO.
5750 (equate_block_to_die): New function.
5751 (dwarf2out_die_ref_for_decl): During WPA get the association
5752 from the external DIE map.
5753 (dwarf2out_register_external_die): Record mapping into the
5755 (maybe_create_die_with_external_ref): New function split out from
5756 DIE generation part of old dwarf2out_register_external_die.
5757 (add_abstract_origin_attribute): Do not return the DIE. When
5758 in LTO reference externals directly.
5759 (dwarf2out_abstract_function): When in LTO ignore calls for
5760 decls with external DIEs (already present abstract instances).
5761 (gen_call_site_die): Adjust.
5762 (add_high_low_attributes): Likewise.
5763 (gen_lexical_block_die): Likewise.
5764 (gen_inlined_subroutine_die): Likewie.
5765 (gen_block_die): Likewise.
5766 (dwarf2out_inline_entry): Likewise.
5767 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
5770 2018-09-25 Martin Liska <mliska@suse.cz>
5772 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
5773 integers and not by a float value.
5775 2018-09-25 Martin Liska <mliska@suse.cz>
5778 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
5780 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
5781 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
5782 * godump.c (go_format_type): Remove extra parenthesis.
5784 2018-09-25 Martin Liska <mliska@suse.cz>
5786 * alias.c (set_dest_equal_p): Remove unused function.
5787 * config/i386/i386.c (def_builtin_pure2): Likewise.
5788 * diagnostic-show-locus.c (class layout): Remove
5790 (layout::layout): Likewise here.
5791 * dump-context.h (class temp_dump_context): Likewise.
5792 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
5793 (add_AT_loclistsptr): Likewise.
5794 (add_AT_offset): Likewise.
5795 (get_AT_hi_pc): Likewise.
5796 (is_comdat_die): Likewise.
5797 (type_is_enum): Likewise.
5798 (ceiling): Likewise.
5799 (add_AT_vms_delta): Likewise.
5800 (is_class_die): Likewise.
5801 * edit-context.c (class line_event): Remove unused field.
5802 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
5804 * ipa-cp.c (ipa_get_vr_lat): Likewise.
5805 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
5806 (ok_for_base_p_nonstrict): Likewise.
5807 * tree-chrec.c (is_not_constant_evolution): Likewise.
5808 (chrec_fold_poly_cst): Likewise.
5809 * tree-if-conv.c (has_pred_critical_p): Likewise.
5810 * tree-ssa-coalesce.c (print_exprs): Likewise.
5811 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
5812 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
5813 * tree-vrp.c (value_ranges_intersect_p): Likewise.
5814 (value_range_nonnegative_p): Likewise.
5816 2018-09-25 Martin Liska <mliska@suse.cz>
5818 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
5819 Do not handle "GNU Pascal".
5820 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
5822 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
5823 from documentation. Likewise.
5824 * dbxout.c (dbxout_range_type): Likewise.
5825 * doc/cpp.texi: Likewise.
5826 * doc/extend.texi: Likewise.
5827 * doc/frontends.texi: Likewise.
5828 * doc/invoke.texi: Remove Pascal entry.
5829 * tree.def (CLEANUP_POINT_EXPR): Likewise.
5830 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
5832 2018-09-25 Martin Liska <mliska@suse.cz>
5835 * doc/invoke.texi: Document all parameters and remove default
5838 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
5841 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
5842 contains HOST_WIDE_INTs when computing its size.
5844 2018-09-24 Jim Wilson <jimw@sifive.com>
5847 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
5849 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
5851 2018-09-24 Andrew Pinski <apinski@marvell.com>
5853 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
5854 access prev before checking it for NULLness in the
5855 AARCH64_FUSE_CMP_BRANCH case.
5857 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
5860 * config/i386/i386.c (rest_of_insert_endbranch): Set
5861 endbr_queued_at_entrance to true and don't insert ENDBR if
5862 x86_function_profiler will be called.
5863 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
5865 * config/i386/i386.h (machine_function): Add
5866 endbr_queued_at_entrance.
5868 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
5870 * genattrtab.c (mk_attr_alt): Use alternative_mask.
5871 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
5873 (check_attr_test): Use alternative_mask.
5874 (get_attr_value): Likewise.
5875 (compute_alternative_mask): Use alternative_mask and XWINT.
5876 (make_alternative_compare): Use alternative_mask.
5877 (attr_alt_subset_p): Use XWINT.
5878 (attr_alt_subset_of_compl_p): Likewise.
5879 (attr_alt_intersection): Use alternative_mask and XWINT.
5880 (attr_alt_union): Likewise.
5881 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
5882 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
5883 (simplify_test_exp): Use alternative_mask and XWINT.
5884 (write_test_expr): Use alternative_mask and XWINT, adjust bit
5885 number calculation to support 64 bits. Generate code that
5886 checks 64-bit masks.
5887 (main): Use alternative_mask.
5888 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
5890 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
5893 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
5894 RETURN+USE when returning via %r14.
5896 2018-09-24 Martin Liska <mliska@suse.cz>
5898 * gcov.c (output_lines): Print colorization legend
5899 for both flag_use_colors and flag_use_hotness_colors.
5902 2018-09-24 Martin Liska <mliska@suse.cz>
5904 * coverage.c (get_coverage_counts): Use warning_at
5905 with current_function_decl location. Use %qD in warning
5908 2018-09-24 Martin Liska <mliska@suse.cz>
5910 * memory-block.h (memory_block_pool::release): Annotate with
5911 valgrind that the memory is not accessible.
5913 2018-09-24 Martin Liska <mliska@suse.cz>
5916 * asan.c: Make asan_handled_variables extern.
5918 * cfgexpand.c (expand_stack_vars): Make sure
5919 a representative is unpoison if another
5920 variable in the partition is handled by
5921 use-after-scope sanitization.
5923 2018-09-24 Richard Biener <rguenther@suse.de>
5925 PR tree-optimization/63155
5926 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
5927 the worklist when the edge of the respective argument isn't
5930 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5932 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
5933 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
5934 (MASK_CLASS_P): Update for rename.
5935 (MAYBE_MASK_CLASS_P): Ditto.
5936 (REG_CLASS_NAMES): Update.
5937 (REG_CLASS_CONTENT): Update.
5938 * config/i386/i386.c (regclass_map): Update for MASK_REG
5939 and ALL_MASK_REGS rename.
5940 * config/i386/constraints.md (Yk): Update for rename.
5943 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5945 * config/i386/i386.h (enum reg_class): Remove
5946 EVEX_SSE_REGS and MOD4_SSE_REGS.
5947 (REG_CLASS_NAMES): Update.
5948 (REG_CLASS_CONTENT): Update.
5949 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
5950 registers as ALL_SSE_REGS.
5951 (ix86_additional_allocno_class_p): Remove.
5952 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
5953 (ix86_register_priority): Lower priority of EVEX SSE registers.
5954 Use IN_RANGE macro where appropriate.
5955 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
5956 AVX-5124VNNIW checks.
5957 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
5958 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
5959 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
5960 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
5961 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
5962 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
5963 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
5964 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
5965 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
5966 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
5967 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
5968 * config/i386/constraints.md (Yh): Remove.
5970 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
5972 * config/i386/i386.c (regclass_map): Declare integer REX registers
5975 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
5977 * doc/service.texi (Service): Switch the fsf.org link to https.
5979 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
5982 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5983 Define to speculation_safe_value_not_needed.
5985 2018-09-21 Florian Weimer <fweimer@redhat.com>
5988 * doc/extend.texi (Common Function Attributes): Mention that
5989 noreturn suppresses tail call optimization.
5991 2018-09-21 Jeff Law <law@redhat.com>
5993 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
5994 vr_values::cleanup_edges_and_switches.
5995 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
5997 (identify_jump_threads): Remove EDGE_IGNORE handling.
5998 (execute_vrp): Move handling of to_remove_edges and
5999 to_update_switch_stmts into vr_values class member functions.
6000 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6001 (to_update_switch_stmts): Likewise.
6002 * vr-values.c: Include cfghooks.h.
6003 (vr_values::vr_values): Initialize to_remove_edges and
6004 to_update_switch_stmts.
6005 (vr_values::~vr_values): Verify to_remove_edges and
6006 to_update_switch_stmts are empty.
6007 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6008 (vr_values::cleanup_edges_and_switches): New member function.
6009 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6010 function. Add new data members.
6012 2018-09-21 David Malcolm <dmalcolm@redhat.com>
6014 PR tree-optimization/87309
6015 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6016 calls with pflags and alt_flags.
6017 (selftest::test_capture_of_dump_calls): Add test of interaction of
6018 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6020 2018-09-21 Olivier Hainque <hainque@adacore.com>
6022 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6024 2018-09-21 Olivier Hainque <hainque@adacore.com>
6026 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6028 2018-09-21 Olivier Hainque <hainque@adacore.com>
6030 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6031 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6032 (VXWORKS_LIBS_RTP): Minor reordering.
6034 2018-09-21 Olivier Hainque <hainque@adacore.com>
6036 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6037 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6039 2018-09-21 Olivier Hainque <hainque@adacore.com>
6041 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6042 (PTRDIFF_TYPE): Likewise.
6044 2018-09-21 Olivier Hainque <hainque@adacore.com>
6046 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6047 triplet, similar to support for VxWorks7.
6048 * config/vxworks-dummy.h: Provide a default definition
6049 of TARGET_VXWORKS64 to 0.
6051 2018-09-21 Olivier Hainque <hainque@adacore.com>
6053 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6054 * config/vxworks-dummy.h: here.
6056 2018-09-21 Olivier Hainque <hainque@adacore.com>
6058 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6060 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
6061 Bo Zhou <zbo.zhou@hisilicon.com>
6063 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6064 * config/aarch64/aarch64-tune.md: Regenerated.
6065 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6066 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6067 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6069 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
6070 Julian Brown <julian@codesourcery.com>
6072 * builtins.c (get_builtin_sync_mem): Handle address spaces.
6074 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
6076 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6077 if the call takes a static chain.
6079 2018-09-21 Martin Liska <mliska@suse.cz>
6081 * auto-profile.c (autofdo_source_profile::read): Do not
6083 (read_profile): Do not add working sets.
6084 (read_autofdo_file): Remove sum_all.
6085 (afdo_callsite_hot_enough_for_early_inline): Remove const
6087 * coverage.c (struct counts_entry): Remove gcov_summary.
6088 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6089 do not support GCOV_TAG_PROGRAM_SUMMARY.
6090 (get_coverage_counts): Remove summary and expected
6092 * coverage.h (get_coverage_counts): Likewise.
6093 * doc/gcov-dump.texi: Remove -w option.
6094 * gcov-dump.c (dump_working_sets): Remove.
6095 (main): Do not support '-w' option.
6096 (print_usage): Likewise.
6097 (tag_summary): Likewise.
6098 * gcov-io.c (gcov_write_summary): Do not dump
6100 (gcov_read_summary): Likewise.
6101 (gcov_histo_index): Remove.
6102 (gcov_histogram_merge): Likewise.
6103 (compute_working_sets): Likewise.
6104 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6106 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6107 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6108 (GCOV_HISTOGRAM_SIZE): Remove.
6109 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6110 (struct gcov_summary): Simplify rapidly just
6111 to runs and sum_max fields.
6112 (gcov_histo_index): Remove.
6113 (NUM_GCOV_WORKING_SETS): Likewise.
6114 (compute_working_sets): Likewise.
6115 * gcov-tool.c (print_overlap_usage_message): Remove
6116 trailing empty line.
6117 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6118 (output_lines): Remove program related line.
6119 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6120 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6122 (input_profile_summary): Do not read it.
6123 (merge_profile_summaries): And do not merge it.
6124 (input_symtab): Do not call removed function.
6125 * modulo-sched.c (sms_schedule): Do not print sum_max.
6126 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6127 removed when histogram method was invented.
6128 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6130 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6132 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6133 and dump selected value.
6134 * profile.c (add_working_set): Remove.
6135 (get_working_sets): Likewise.
6136 (find_working_set): Likewise.
6137 (get_exec_counts): Do not work with working sets.
6138 (read_profile_edge_counts): Do not inform as sum_max is removed.
6139 (compute_branch_probabilities): Likewise.
6140 (compute_value_histograms): Remove argument for call of
6141 get_coverage_counts.
6142 * profile.h: Do not make gcov_summary const.
6144 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
6146 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6148 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
6150 PR tree-optimization/86990
6151 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6152 Check that the entire merged store group is made of constants only for
6155 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
6157 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6158 (VTABLE_VERIFICATION_SPEC): Likewise.
6159 (SANITIZER_EARLY_SPEC): Likewise.
6160 (SANITIZER_SPEC): Likewise.
6161 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6162 * doc/invoke.texi (Link Options): Document -r.
6164 2018-09-20 Richard Biener <rguenther@suse.de>
6167 * gimplify.c (gimplify_expr): Retain alignment of
6168 addressable lvalue in dereference.
6170 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
6173 * configure.ac: Check for .loc is_stmt support.
6174 * configure, config.in: Rebuilt.
6175 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6178 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
6180 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6181 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6184 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
6186 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6188 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6191 2018-09-20 Marek Polacek <polacek@redhat.com>
6193 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6195 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
6197 PR tree-optimization/87288
6198 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6199 into account when determining PEELING_FOR_NITERS.
6201 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
6203 PR tree-optimization/86877
6204 * tree-vect-loop.c (vect_analyze_loop_2): Call
6205 vect_verify_datarefs_alignment.
6207 2018-09-19 Marek Polacek <polacek@redhat.com>
6209 * doc/invoke.texi: Document -Wclass-conversion.
6211 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
6213 * config/pa/pa.c (pa_adjust_priority): Delete.
6214 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6216 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6217 (atomic_storehi): Likewise.
6218 (atomic_storesi): Likewise.
6219 (atomic_loaddi): Restore compare and swap exchange loop code.
6221 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
6223 PR rtl-optimization/86902
6224 * combine.c (try_combine): When changing the CC mode used, don't change
6225 an unrelated mode in other_insn to that new CC mode.
6227 2018-09-19 David Malcolm <dmalcolm@redhat.com>
6229 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6230 with %T in place of calls to dump_generic_expr.
6231 (prune_runtime_alias_test_list): Likewise.
6232 (create_runtime_alias_checks): Likewise.
6233 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6234 (vect_analyze_data_ref_dependence): Likewise.
6235 (vect_slp_analyze_data_ref_dependence): Likewise.
6236 (vect_record_base_alignment): Likewise. Use %G in place of call
6237 to dump_gimple_stmt.
6238 (vect_compute_data_ref_alignment): Likewise.
6239 (verify_data_ref_alignment): Likewise.
6240 (vect_find_same_alignment_drs): Likewise.
6241 (vect_analyze_group_access_1): Likewise.
6242 (vect_analyze_data_ref_accesses): Likewise.
6243 (dependence_distance_ge_vf): Likewise.
6244 (dump_lower_bound): Likewise.
6245 (vect_prune_runtime_alias_test_list): Likewise.
6246 (vect_find_stmt_data_reference): Likewise.
6247 (vect_analyze_data_refs): Likewise.
6248 (vect_create_addr_base_for_vector_ref): Likewise.
6249 (vect_create_data_ref_ptr): Likewise.
6250 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6251 (vect_can_advance_ivs_p): Likewise.
6252 (vect_update_ivs_after_vectorizer): Likewise.
6253 (vect_gen_prolog_loop_niters): Likewise.
6254 (vect_prepare_for_masked_peels): Likewise.
6255 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6256 (vect_determine_vectorization_factor): Likewise.
6257 (vect_is_simple_iv_evolution): Likewise.
6258 (vect_analyze_scalar_cycles_1): Likewise.
6259 (vect_analyze_loop_operations): Likewise.
6260 (report_vect_op): Likewise.
6261 (vect_is_slp_reduction): Likewise.
6262 (check_reduction_path): Likewise.
6263 (vect_is_simple_reduction): Likewise.
6264 (vect_create_epilog_for_reduction): Likewise.
6265 (vect_finalize_reduction:): Likewise.
6266 (vectorizable_induction): Likewise.
6267 (vect_transform_loop_stmt): Likewise.
6268 (vect_transform_loop): Likewise.
6269 (optimize_mask_stores): Likewise.
6270 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6271 (vect_split_statement): Likewise.
6272 (vect_recog_over_widening_pattern): Likewise.
6273 (vect_recog_average_pattern): Likewise.
6274 (vect_determine_min_output_precision_1): Likewise.
6275 (vect_determine_precisions_from_range): Likewise.
6276 (vect_determine_precisions_from_users): Likewise.
6277 (vect_mark_pattern_stmts): Likewise.
6278 (vect_pattern_recog_1): Likewise.
6279 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6280 (vect_record_max_nunits): Likewise.
6281 (vect_build_slp_tree_1): Likewise.
6282 (vect_build_slp_tree_2): Likewise.
6283 (vect_print_slp_tree): Likewise.
6284 (vect_analyze_slp_instance): Likewise.
6285 (vect_detect_hybrid_slp_stmts): Likewise.
6286 (vect_detect_hybrid_slp_1): Likewise.
6287 (vect_slp_analyze_operations): Likewise.
6288 (vect_slp_analyze_bb_1): Likewise.
6289 (vect_transform_slp_perm_load): Likewise.
6290 (vect_schedule_slp_instance): Likewise.
6291 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6292 (vect_mark_stmts_to_be_vectorized): Likewise.
6293 (vect_init_vector_1): Likewise.
6294 (vect_get_vec_def_for_operand): Likewise.
6295 (vect_finish_stmt_generation_1): Likewise.
6296 (vect_check_load_store_mask): Likewise.
6297 (vectorizable_call): Likewise.
6298 (vectorizable_conversion): Likewise.
6299 (vectorizable_operation): Likewise.
6300 (vectorizable_load): Likewise.
6301 (vect_analyze_stmt): Likewise.
6302 (vect_is_simple_use): Likewise.
6303 (vect_get_vector_types_for_stmt): Likewise.
6304 (vect_get_mask_type_for_stmt): Likewise.
6305 * tree-vectorizer.c (increase_alignment): Likewise.
6307 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
6309 * doc/rtl.texi: Adjust vec_select description.
6310 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6311 non-constant selectors.
6313 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
6315 * config/aarch64/aarch64-protos.h
6316 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6317 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6318 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6319 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6320 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6321 (AARCH64_FL_PROFILE): Move index so flags are ordered.
6322 (AARCH64_ISA_RCPC8_4): New flag.
6323 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6324 to aarch64_offset_9bit_signed_unscaled_p.
6325 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6327 * config/aarch64/constraints.md (Ust): New constraint.
6328 * config/aarch64/predicates.md.
6329 (aarch64_9bit_offset_memory_operand): New predicate.
6330 (aarch64_rcpc_memory_operand): New predicate.
6332 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
6334 PR rtl-optimization/87361
6335 * rtlanal.c (nonzero_bits1): Revert accidental change.
6337 2018-09-19 Richard Biener <rguenther@suse.de>
6339 PR tree-optimization/87349
6340 PR tree-optimization/87342
6341 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6343 2018-09-18 Marek Polacek <polacek@redhat.com>
6345 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6346 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6348 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
6350 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6352 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
6354 PR rtl-optimization/86882
6355 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6357 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
6359 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6360 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6362 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
6365 * doc/invoke.texi (Link Options): Fix formatting and grammar.
6367 2018-09-18 Richard Biener <rguenther@suse.de>
6370 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6371 (compute_samebase_partition_bases): Likewise.
6372 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6373 (gimple_can_coalesce_p): Simplify.
6375 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
6377 Handle a library implementation of ffs calling __builtin_ffs.
6378 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6379 (mmix_init_libfuncs): New function: make __builtin_ffs expand
6382 2018-09-17 David Malcolm <dmalcolm@redhat.com>
6384 * diagnostic-show-locus.c (class layout_range): Add field
6386 (layout_range::layout_range): Add "original_idx" param and use it
6387 to initialize new field.
6388 (make_range): Use 0 for original_idx.
6389 (layout::layout): Pass in index to calls to
6390 maybe_add_location_range.
6391 (layout::maybe_add_location_range): Add param "original_idx" and
6392 pass it on to layout_range.
6393 (layout::print_any_labels): Pass on range->m_original_idx to
6395 (gcc_rich_location::add_location_if_nearby): Use 0 for
6397 * gcc-rich-location.h (text_range_label::get_text): Update for new
6399 (range_label_for_type_mismatch::get_text): Likewise.
6401 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
6403 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6405 2018-09-17 David Malcolm <dmalcolm@redhat.com>
6407 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6408 format_string_diagnostic_t.
6409 (fmtwarn_n): Likewise.
6410 * substring-locations.c
6411 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6412 (format_warning_n_va): Convert to...
6413 (format_string_diagnostic_t::emit_warning_n_va): ...this.
6414 (format_warning_va): Convert to...
6415 (format_string_diagnostic_t::emit_warning_va): ...this.
6416 (format_warning_at_substring): Convert to...
6417 (format_string_diagnostic_t::emit_warning): ...this.
6418 (format_warning_at_substring_n): Convert to...
6419 (format_string_diagnostic_t::emit_warning_n): ...this.
6420 * substring-locations.h (class format_string_diagnostic_t): New
6422 (format_warning_va): Convert to
6423 format_string_diagnostic_t::emit_warning_va.
6424 (format_warning_n_va): Convert to
6425 format_string_diagnostic_t::emit_warning_n_va.
6426 (format_warning_at_substring): Convert to
6427 format_string_diagnostic_t::emit_warning.
6428 (format_warning_at_substring_n): Convert to
6429 format_string_diagnostic_t::emit_warning_n.
6431 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
6432 Bernd Schmidt <bernds_cb1@t-online.de>
6434 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6437 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
6439 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6440 operand 0 predicate to nonimmediate operand.
6441 (rint<mode>2_frndint): Remove insn pattern.
6442 (rint<mode>2): Change operand 1 predicate to general_operand.
6443 Extend operand 1 to XFmode and generate rintxf2 insn.
6444 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6445 Do not use X87MODEF mode macro.
6446 (frndintxf2_<rounding>_i387): Rename from
6447 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
6448 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6449 to XFmode and generate significandxf3 insn.
6451 2018-09-17 Richard Biener <rguenther@suse.de>
6453 PR tree-optimization/87328
6454 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6455 visiting unexecutable backedges when not iterating.
6456 (do_rpo_vn): Mark all edges not executable even when not
6459 2018-09-17 Martin Jambor <mjambor@suse.cz>
6462 * doc/invoke.texi (Warning Options): Likewise.
6464 2018-09-17 Richard Biener <rguenther@suse.de>
6466 PR tree-optimization/87301
6467 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6468 clean EH info from leftover copy assignments.
6470 2018-09-17 Martin Liska <mliska@suse.cz>
6472 PR gcov-profile/85871
6473 * gcov.c (output_intermediate_file): Fix out of bounds
6476 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
6478 * config/arc/arc.c: Object attributes for core4 not reflected
6480 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6483 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
6485 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6487 2018-09-17 Martin Liska <mliska@suse.cz>
6489 * doc/gcov.texi: Document new option --use-hotness-colors.
6490 * gcov.c (struct source_info): Declare new field.
6491 (source_info::source_info): Set default for maximum_count.
6492 (print_usage): Add new -q option.
6493 (process_args): Process it.
6494 (accumulate_line_info): Save src->maximum_count.
6495 (output_line_beginning): Make color line number if
6496 flag_use_hotness_colors is set.
6497 (output_line_details): Pass default argument value.
6498 (output_lines): Pass src->maximum_count.
6500 2018-09-17 Martin Liska <mliska@suse.cz>
6502 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6503 Use processor_names table.
6504 * config/i386/i386.c (ix86_default_align): Use
6505 processor_cost_table for alignment values.
6506 (ix86_option_override_internal): Use processor_names.
6507 (ix86_function_specific_print): Likewise.
6508 * config/i386/i386.h (struct processor_costs):
6509 Add alignment values.
6510 (struct ptt): Remove and replace with const char *.
6511 * config/i386/x86-tune-costs.h (struct processor_costs):
6512 Declare default alignments for all costs.
6514 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
6516 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6517 symbolics or VR_VARYING ranges for ABS_EXPR.
6518 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6519 when range will wrap.
6521 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
6524 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6525 before and after a JUMP_TABLE_DATA.
6527 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
6530 * dojump.c (do_compare_and_jump): Canonicalize function pointers
6531 when one operand is a function pointer. Use POINTER_TYPE_P and
6532 FUNC_OR_METHOD_TYPE_P.
6533 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6534 * fold-const.c (build_range_check): Likewise.
6535 * match.pd (simple_comparison): Likewise.
6537 2018-09-14 David Malcolm <dmalcolm@redhat.com>
6540 * spellcheck.c (get_edit_distance_cutoff): New function.
6541 (selftest::test_edit_distance_unit_test_oneway): Rename to...
6542 (selftest::test_get_edit_distance_one_way): ...this.
6543 (selftest::test_get_edit_distance_unit): Rename to...
6544 (selftest::test_get_edit_distance_both_ways): ...this.
6545 (selftest::test_edit_distances): Move tests to this new function,
6546 and test some more pairs of strings. Update for above renaming.
6547 (selftest::get_old_cutoff): New function.
6548 (selftest::test_get_edit_distance_cutoff): New function.
6549 (selftest::assert_suggested_for): New function.
6550 (ASSERT_SUGGESTED_FOR): New macro.
6551 (selftest::assert_not_suggested_for): New function.
6552 (ASSERT_NOT_SUGGESTED_FOR): New macro.
6553 (selftest::test_suggestions): New function.
6554 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6555 tests to selftest::test_edit_distances and call it. Add calls to
6556 selftest::test_get_edit_distance_cutoff and
6557 selftest::test_suggestions.
6558 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6559 (best_match::consider): Replace hard-coded cutoff calculation with
6561 (best_match::get_cutoff): New declaration.
6562 (best_match::get_best_meaningful_candidate): Likewise.
6564 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
6566 * builtins.c (fold_builtin_strlen): Remove TODO comment.
6568 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
6571 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6573 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6574 terminated string literal.
6576 2018-09-14 Martin Sebor <msebor@redhat.com>
6578 * builtins.c (unterminated_array): Handle ARRAY_REF.
6579 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6580 * builtins.h (unterminated_array): Declare extern.
6581 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6583 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6586 2018-09-14 Martin Sebor <msebor@redhat.com>
6587 Jeff Law <law@redhat.com>
6589 * builtins.c (unterminated_array): New.
6590 (expand_builtin_strcpy): Adjust.
6591 (expand_builtin_strcpy_args): Detect unterminated arrays.
6592 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
6593 unterminated arrays.
6594 * gimple-fold.h (get_maxval_strlen): Add argument.
6595 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
6597 * gimple-fold.c (get_range_strlen): Add argument.
6598 (get_maxval_strlen): Adjust.
6599 * gimple-fold.h (get_range_strlen): Add argument.
6601 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
6603 * config/i386/movdirintrin.h: Fix copyright year.
6605 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
6607 * reg-stack.c: Include regs.h.
6608 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
6609 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
6610 FIRST_STACK_REG, not DFmode.
6611 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
6612 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
6613 (change stack): Default register mode to the reg_raw_mode of
6614 FIRST_STACK_REG, not DFmode.
6615 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
6616 (*swapxf): Rename from swapxf.
6618 2018-09-14 Carl Love <cel@us.ibm.com>
6620 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
6621 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
6623 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
6626 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
6629 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
6632 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
6634 2018-09-14 Jason Merrill <jason@redhat.com>
6636 Fix --enable-gather-detailed-mem-stats.
6637 * hash-table.c (hash_table_usage): Change from variable to function.
6638 * hash-table.h: Adjust.
6639 * Makefile.in: Add missing dependencies on hash-table.h.
6641 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6643 PR tree-optimization/87259
6645 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
6646 execute_cse_reciprocals_1 has tried transforming.
6648 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
6650 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
6651 VR_VARYING for PLUS/MINUS_EXPR.
6653 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
6655 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
6658 2018-09-14 Richard Biener <rguenther@suse.de>
6661 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
6662 bits for the merged partition.
6664 2018-09-13 Martin Sebor <msebor@redhat.com>
6665 Bernd Edlinger <bernd.edlinger@hotmail.de>
6667 * builtins.h (c_srlen): Add argument.
6668 * builtins.c (warn_string_no_nul): New function.
6669 (c_strlen): Add argument and use it. Update recursive calls.
6670 Pass DECL argument to string_constant to get info on non
6671 terminated strings. Update *NONSTR as needed.
6672 (fold_builtin_strlen): Add argument to calls to c_strlen.
6673 Warn for unterminated arrays.
6674 (warn_string_no_null): Add prototype.
6675 * expr.c (string_constant): Update arguments. Update recursive
6676 calls appropriately. Detect missing NUL terminator and outermost
6677 declaration its missing in.
6678 Improve checks for arrays with nonzero lower bound or elements
6679 that are not a single byte. Simplify offset computation.
6680 Simplify checks for non-NUL terminated strings.
6681 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
6682 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
6684 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
6686 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
6688 * fold-const.c (c_getstr): Fix function comment. Remove unused third
6689 argument. Fix range checks.
6690 * fold-const.h (c_getstr): Adjust protoype.
6691 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
6692 string is constant but contains no NUL byte.
6694 * expr.c (string_constant): Adjust function comment.
6695 Remove bogus check for zero termination.
6697 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
6699 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
6700 (get_constant_size): Don't make STRING_CSTs larger than they are.
6701 (check_string_literal): New check function for STRING_CSTs.
6702 (output_constant): Use it.
6704 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
6707 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
6709 2018-09-13 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/87263
6712 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
6713 (struct unwind_state): Add max_rpo field.
6714 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
6715 Compute max_rpo, the max RPO number a block can be backwards reached
6716 from. Re-write non-iterating mode to a RPO ordered worklist approach,
6717 separating it from the iterating mode.
6719 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
6721 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
6722 (rfs_decision): New scheduling decision.
6724 2018-09-13 Richard Biener <rguenther@suse.de>
6727 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
6728 (vn_nary_op_insert_pieces_predicated): Do not write useless
6729 valid_dominated_by_p entry outside of the allocated storage.
6731 2018-09-13 Omar Sandoval <osandov@osandov.com>
6732 Tom de Vries <tdevries@suse.de>
6735 * dwarf2out.c (is_c): New function.
6736 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
6738 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
6741 * config/aarch64/aarch64.md (*aarch64_bfxil):
6743 * config/aarch64/constraints.md (Ulc): Define.
6744 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
6746 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
6749 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
6751 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
6752 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
6753 aarch64_layout_frame call.
6754 (aarch64_expand_epilogue): Likewise.
6755 (aarch64_initial_elimination_offset): Likewise.
6756 (aarch64_get_separate_components): Likewise.
6757 (aarch64_use_return_insn_p): Likewise.
6758 (aarch64_layout_frame): Remove unneeded check.
6760 2018-09-13 Jakub Jelinek <jakub@redhat.com>
6762 * configure.ac: Only append
6763 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
6764 gcc_config_arguments if it was never reconfigured or last reconfigure
6765 was with different arguments.
6766 * configure: Regenerated.
6768 2018-09-13 Jakub Jelinek <jakub@redhat.com>
6769 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6772 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
6773 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
6775 2018-09-13 Jakub Jelinek <jakub@redhat.com>
6777 PR tree-optimization/87287
6778 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
6779 X % C == 0 to X % (unsigned) C == 0 optimization to ...
6780 * match.pd (X % C == 0): ... here. New optimization.
6782 2018-09-12 Jakub Jelinek <jakub@redhat.com>
6785 * expr.h (maybe_optimize_mod_cmp): Declare.
6786 * expr.c (mod_inv): New function.
6787 (maybe_optimize_mod_cmp): New function.
6788 (do_store_flag): Use it.
6789 * cfgexpand.c (expand_gimple_cond): Likewise.
6791 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
6792 Julian Brown <julian@codesourcery.com>
6795 * gimplify.c (gimplify_scan_omp_clauses): Set
6796 target_firstprivatize_array_bases in OpenACC parallel and kernels
6797 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
6798 OpenACC data regions.
6800 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
6802 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
6803 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
6805 2018-09-12 Richard Biener <rguenther@suse.de>
6807 PR tree-optimization/87280
6808 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
6809 edge but unreachable target.
6810 (do_rpo_vn): For conservatively handling a PHI only mark
6811 the backedge executable but not the block reachable.
6813 2018-09-12 Richard Biener <rguenther@suse.de>
6815 PR tree-optimization/87266
6816 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
6819 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
6821 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
6823 ("trunc<BFP:mode><DFP_ALL:mode>2")
6824 ("trunc<DFP_ALL:mode><BFP:mode>2")
6825 ("extend<BFP:mode><DFP_ALL:mode>2")
6826 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
6827 according to the target operand type.
6829 2018-09-12 Jakub Jelinek <jakub@redhat.com>
6830 Andreas Krebbel <krebbel@linux.ibm.com>
6832 PR tree-optimization/86844
6833 * gimple-ssa-store-merging.c
6834 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
6835 there are any overlapping stores in between them, make sure they are
6836 also coalesced or we give up completely.
6838 2018-09-12 Jakub Jelinek <jakub@redhat.com>
6841 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
6842 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
6844 2018-09-12 Tom de Vries <tdevries@suse.de>
6846 * common.opt (gdescribe-dies): Add option.
6847 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
6848 attribute for artifical and nameless decls.
6849 (dwarf2out_register_external_die): Add description attribute to
6850 external reference die.
6851 (add_desc_attribute): New functions.
6852 (gen_subprogram_die): Add description attribute to
6853 DW_TAG_call_site_parameter.
6854 * tree-pretty-print.c (print_generic_expr_to_str): New function.
6855 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
6856 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
6858 (@item -gdescribe-dies): Add.
6860 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
6862 * tree-vrp.c (vrp_shift_undefined_p): Remove.
6863 (extract_range_from_binary_expr_1: Call
6864 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
6865 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
6868 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
6870 * gimple-ssa-warn-alloca.c
6871 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
6872 field for ALLOCA_BOUND_*_LARGE.
6874 2018-09-11 Nathan Sidwell <nathan@acm.org>
6876 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
6878 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
6880 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
6881 for clobbers. Remove obsolete comment.
6883 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
6885 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
6886 mpxchk, mpxld and mpxst types.
6887 (define_attr length_immediate): Remove all processing of mpx types.
6888 (define_attr prefix_0f): Ditto.
6889 (define_attr memory): Ditto.
6891 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
6893 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
6894 (log<mode>2): Change operand 1 predicate to general_operand.
6895 Extend operand 1 to XFmode and generate logxf3 insn.
6896 (log10<mode>2): Change operand 1 predicate to general_operand.
6897 Extend operand 1 to XFmode and generate log10xf3 insn.
6898 (log2<mode>2): Change operand 1 predicate to general_operand.
6899 Extend operand 1 to XFmode and generate log2xf3 insn.
6900 (fyl2xp1_extend<mode>xf3_i387): Remove.
6901 (log1p<mode>2): Change operand 1 predicate to general_operand.
6902 Extend operand 1 to XFmode and generate log1pxf3 insn.
6903 (fxtract_extend<mode>xf3_i387): Remove.
6904 (logb<mode>2): Change operand 1 predicate to general_operand.
6905 Extend operand 1 to XFmode and generate logbxf3 insn.
6906 (ilogb<mode>2): Change operand 1 predicate to general_operand.
6907 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
6908 (significand<mode>2): Change operand 1 predicate to general_operand.
6909 Extend operand 1 to XFmode and generate significandxf3 insn.
6911 2018-09-11 Nathan Sidwell <nathan@acm.org>
6913 * gcc.c (perror_with_name, pfatal_with_name): Delete.
6914 (load_specs): Use fatal_error.
6915 (DELETE_IF_ORDINARY, process_command): Use error.
6916 (execute, run_attempt): Use fatal_error.
6918 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
6920 * diagnostic-core.h (sorry_at): New prototype.
6921 * diagnostic.c (sorry_at): New function.
6923 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
6925 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
6926 by zero as VR_UNDEFINED.
6928 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
6930 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
6931 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
6932 (<sincos>mode2): New expander.
6933 (sincos_extend<mode>xf3_i387): Remove insn pattern.
6934 (sincos -> sin, cos splitters): Remove splitter patterns.
6935 (sincos<mode>3): Change operand 2 predicate to general_operand.
6936 Extend operand 2 to XFmode and generate sincosxf3 insn.
6937 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
6938 Change operand 3 predicate to const1_operand.
6939 (fptan_extend<mode>xf4_i387): Remove insn pattern.
6940 (tanxf2): Update operands in the call to fptanxf4_i387.
6941 (tan<mode>2): Change operand 1 predicate to general_operand.
6942 Extend operand 1 to XFmode and generate tanxf3 insn.
6943 (atan2xf3): Rename from *fpatanxf3_i387.
6944 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
6945 (atan2xf3): Remove expander.
6946 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
6947 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
6948 (atan<mode>2): Change operand 1 predicate to general_operand.
6949 Extend operand 1 to XFmode and generate atanxf3 insn.
6951 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
6953 * config/i386/i386.md (x87/SSE constant load splitter): Use
6954 memory_operand instead of nonimmediate_operand for input operand
6957 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
6959 * config/i386/i386.md (float partial SSE register stall splitter): Move
6960 splitter near its instruction pattern.
6961 (float_extend partial SSE register stall splitter): Ditto.
6962 (float_truncate partial SSE register stall splitter): Ditto.
6964 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
6967 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6968 to speculation_safe_value_not_needed.
6971 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
6972 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
6973 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
6974 leaf_function_p, instead use has_hard_reg_initial_val.
6976 2018-09-09 Nathan Sidwell <nathan@acm.org>
6978 * gcc.h (pfatal_with_name): Don't declare here.
6979 * gcc.c (pfatal_with_name): Make static.
6981 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
6983 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
6986 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
6988 PR rtl-optimization/85458
6989 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
6990 priority hook to reduce the priority of EXPR.
6992 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
6994 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
6995 DImode for x87 on 32bit targets. Conditionally disable x87 modes
6996 with X87_ENABLE_FLOAT. Remove preparation code.
6997 (*float<SWI48:mode><MODEF:mode>2): Rename from
6998 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
6999 math using "enabled" attribute.
7000 (*floatdi<MODEF:mode>2_i387): Rename from
7001 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
7002 enable for 32bit targets only.
7003 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7005 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7006 as operand 1 predicate. Rewrite as define_insn_and_split.
7007 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7009 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
7011 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7012 to fallthru to FLOAT case.
7014 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7017 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7018 around folding of vec_sl to handle out of range shift values.
7020 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
7022 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7023 Update callers to gen_fix_trunc<mode>_i387_fisttp
7024 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7025 nonimmediate_operand.
7026 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7027 and corresponding splitters.
7028 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7029 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7030 (fix_truncdi_i387_with_temp): Remove insn pattern
7031 and corresponding splitters.
7032 (fix_trunc<mode>_i387): Change operand 0 predicate to
7033 nonimmediate_operand.
7034 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7035 and corresponding splitters.
7036 (*fistdi2_1): Remove.
7038 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7039 (lrintxfdi2): Remove expander. Reimplement as define_insn.
7040 (*fist<mode>2_1): Remove.
7041 (fist<mode>2): Ditto.
7042 (fist<mode>2_with_temp): Remove insn pattern and corresponding
7044 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
7045 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7046 (fistdi2_<rounding>): Change operand 0 predicate to
7047 nonimmediate_operand.
7048 (fistdi2_<rounding>_with_temp): Remove insn pattern
7049 and corresponding splitters.
7050 (fist<mode>2_<rounding>): Change operand 0 predicate to
7051 nonimmediate_operand.
7052 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7053 and corresponding splitters.
7055 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7057 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
7059 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7062 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7064 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7065 early gimple folding of vec_splat().
7066 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7067 * gimple-fold.h: Add an extern define for tree_vec_extract().
7069 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7071 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7072 wrappers around TREE_TYPE comparisons.
7074 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7077 * config/s390/predicates.md: Add nonsym_memory_operand.
7078 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7079 contains a SYMBOL_REF, load it into an intermediate pseudo.
7080 (s390_emit_compare_and_swap): Legitimize operand.
7081 * config/s390/s390.md: Use the new nonsym_memory_operand
7082 with UNSPECV_CAS patterns.
7084 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7087 * config/s390/s390-passes.def: New file.
7088 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7090 (make_pass_s390_early_mach): Add declaration.
7091 * config/s390/s390.c (make_pass_s390_early_mach):
7092 (s390_option_override): Remove dynamic registration.
7093 * config/s390/t-s390: Add s390-passes.def.
7095 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7097 * config/s390/s390.c (s390_decompose_constant_pool_ref):
7098 Remove UNSPEC_LTREL_BASE check.
7099 (annotate_constant_pool_refs): Likewise.
7100 (find_constant_pool_ref): Likewise.
7101 (find_ltrel_base): Removed.
7102 (replace_ltrel_base): Removed.
7103 (s390_mainpool_finish): Remove replace_ltrel_base call.
7104 (s390_chunkify_start): Remove pending LTREL_BASE logic.
7105 (s390_chunkify_finish): Remove replace_ltrel_base call.
7106 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7108 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
7111 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7112 to speculation_safe_value_not_needed.
7114 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
7115 Bernd Schmidt <bernds_cb1@t-online.de>
7117 * config/nvptx/nvptx-opts.h: New file.
7118 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7119 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7121 (TARGET_SM35): New macro.
7122 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7124 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7126 (misa=): New option.
7127 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7129 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
7131 * config/i386/i386.md (truncdfsf2): Remove expander.
7132 (truncdfsf2_with_temp): Ditto.
7133 (truncxf<mode>2): Ditto.
7134 (*truncdfsf_fast_mixed): Remove insn pattern.
7135 (*truncdfsf_fast_i387): Ditto.
7136 (*truncdfsf_mixed): Ditto.
7137 (*truncdfsf_i387): Ditto.
7138 (*truncdfsf2_i387_1): Ditto.
7139 (*truncxfsf2_mixed): Ditto.
7140 (*truncxfdf2_mixed): Ditto.
7141 (*truncxf<mode>2_i387_noop): Ditto. Update callers
7142 to call gen_truncxf<mode>2 instead.
7143 (*truncxf<mode>2_i387): Remove.
7144 (reg->reg splitters): Remove splitter pattern.
7145 (reg->mem splitters): Ditto.
7147 (truncdfsf2): New insn pattern.
7148 (truncxf<mode>2): Ditto.
7150 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7152 * tree-ssa-math-opts.c (is_mult_by): New function.
7153 (is_square_of): Use the above.
7154 (optimize_recip_sqrt): New function.
7155 (pass_cse_reciprocals::execute): Use the above.
7157 2018-09-05 Richard Biener <rguenther@suse.de>
7160 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7161 to zero-init the emplaced vec.
7163 2018-09-05 Martin Liska <mliska@suse.cz>
7165 PR tree-optimization/87205
7166 * tree-switch-conversion.c (pass_lower_switch::execute):
7167 Group cases for switch statements.
7169 2018-09-05 Richard Biener <rguenther@suse.de>
7171 PR tree-optimization/87217
7172 * tree-ssa-sccvn.c (vuse_valueize): New.
7173 (vn_reference_lookup_pieces): Use it.
7174 (vn_reference_lookup): Likewise.
7176 2018-09-05 Nathan Sidwell <nathan@acm.org>
7179 * stor-layout.c (place_field): Scan forwards to check last
7180 bitfield when ms_bitfield_placement is in effect.
7182 2018-09-05 Richard Biener <rguenther@suse.de>
7185 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7188 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
7189 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
7191 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7192 * config.gcc (extra_objs): Build it.
7193 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7195 * config/aarch64/aarch64-passes.def
7196 (pass_tag_collision_avoidance): New pass.
7197 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7198 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7199 (aarch64_classify_address): Remove static qualifier.
7200 (aarch64_address_info, aarch64_address_type): Move to...
7201 * config/aarch64/aarch64-protos.h: ... here.
7202 (make_pass_tag_collision_avoidance): New function.
7203 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7206 2018-09-05 Martin Liska <mliska@suse.cz>
7208 * doc/gcov.texi: Update documentation of humar
7210 * gcov.c (format_count): Print one decimal place, it provides
7211 more fine number of situations like '1G' vs. '1.4G'.
7213 2018-09-05 Martin Liska <mliska@suse.cz>
7216 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7217 * optc-gen.awk: Allow 'Var' for Deprecated options in order
7218 to generate a MASK value.
7220 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
7223 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7224 if frame pointer isn't used.
7225 (compute_frame_pointer_to_fb_displacement): Likewise.
7227 2018-09-04 Jakub Jelinek <jakub@redhat.com>
7230 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7231 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7232 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7233 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7234 and OPTION_MASK_ISA_XSAVEC_UNSET.
7236 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
7238 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7239 XOR operations in NAND case.
7241 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
7243 * wide-int-range.cc (wide_int_range_convert): New.
7244 * wide-int-range.h (wide_int_range_convert): New.
7245 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7246 code into wide_int_range_convert.
7247 (extract_range_into_wide_ints): Do not munge anti range constants
7248 into the entire domain. Just return the range back.
7250 2018-09-04 Martin Liska <mliska@suse.cz>
7252 * genmatch.c (output_line_directive): Add new argument
7254 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7256 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
7258 * doc/invoke.texi (Option Summary): Add whitespace.
7260 * doc/invoke.texi (Option Summary): Add -Waligned-new.
7262 2018-09-04 Richard Biener <rguenther@suse.de>
7264 PR tree-optimization/87211
7265 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7266 backedge value we're supposed to treat as VARYING also number
7267 the PHI to VARYING in case it got a different value-number already.
7269 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
7271 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7272 (extract_range_from_binary_expr_1): Do not call
7273 vrp_can_optimize_bit_op.
7274 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7276 (wide_int_range_get_mask_and_bounds): New.
7277 (wide_int_range_optimize_bit_op): New.
7278 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7279 (wide_int_range_bit_and): Same.
7280 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7281 (wide_int_range_optimize_bit_op): New.
7282 (wide_int_range_get_mask_and_bounds): New.
7284 2018-09-04 Richard Biener <rguenther@suse.de>
7286 PR tree-optimization/87176
7287 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7288 variable. When value-numbering a virtual PHI node make sure
7289 to not value-number to the backedge value.
7291 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
7293 * doc/extend.texi (Long Long, Hex Floats): Document support for
7294 long long and hex floats in more recent versions of ISO C++.
7296 2018-09-03 Richard Biener <rguenther@suse.de>
7298 PR tree-optimization/87177
7299 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7302 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7304 * bb-reorder.c (edge_order): Convert to C-qsort-style
7305 tri-state comparator.
7306 (reorder_basic_blocks_simple): Change std::stable_sort to
7309 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7311 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7312 tri-state comparator.
7313 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7315 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7317 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
7318 (mergesort): ... here as maximum count for using netsort.
7319 (gcc_qsort): Set nlim to 3 if stable sort is requested.
7320 (gcc_stablesort): New.
7321 * system.h (gcc_stablesort): Declare.
7323 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7325 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7326 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
7327 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
7329 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
7331 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7332 lxsdx and stxsdx alternatives.
7333 (*mov<mode>_hardfloat64): Ditto.
7334 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7336 2018-09-03 Richard Biener <rguenther@suse.de>
7338 PR tree-optimization/87200
7339 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7342 2018-09-03 Martin Liska <mliska@suse.cz>
7344 PR tree-optimization/87201
7345 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7346 Fix parenthesis in an expression.
7348 2018-09-03 Richard Biener <rguenther@suse.de>
7350 PR tree-optimization/87197
7351 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7352 visited. CSE the VN_INFO hashtable lookup.
7354 PR tree-optimization/87169
7355 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7356 iterating make sure there's no extra backedges from irreducible
7357 regions feeding the header. Mark the destination block
7360 2018-09-03 Martin Liska <mliska@suse.cz>
7363 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7364 * common/common-targhooks.c (default_get_valid_option_values):
7366 * common/common-targhooks.h (default_get_valid_option_values):
7368 * common/config/i386/i386-common.c: Move processor_target_table
7370 (ix86_get_valid_option_values): New function.
7371 (TARGET_GET_VALID_OPTION_VALUES): New macro.
7372 * config/i386/i386.c (struct ptt): Move to i386-common.c.
7373 (PTA_*): Move all defined masks into i386-common.c.
7374 (ix86_function_specific_restore): Use new processor_cost_table.
7375 * config/i386/i386.h (struct ptt): Moved from i386.c.
7376 (struct pta): Likewise.
7377 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7378 * doc/tm.texi.in: Likewise.
7379 * opt-suggestions.c (option_proposer::suggest_option):
7380 Pass prefix to build_option_suggestions.
7381 (option_proposer::get_completions): Likewise.
7382 (option_proposer::build_option_suggestions): Use the new target
7384 * opts.c (struct option_help_tuple): New struct.
7385 (print_filtered_help): Use the new target hook.
7387 2018-09-03 Martin Liska <mliska@suse.cz>
7390 * predict.c (set_even_probabilities): Add likely_edges
7391 argument and handle cases where we have precisely one
7393 (combine_predictions_for_bb): Catch also likely_edges.
7394 (tree_predict_by_opcode): Handle gswitch statements.
7395 * tree-cfg.h (find_case_label_for_value): New declaration.
7396 (find_taken_edge_switch_expr): Likewise.
7397 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7398 Find pivot in decision tree based on probabily, not by number of
7401 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
7403 * doc/standards.texi (Standards): Update Objective-C reference.
7405 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7407 * doc/install.texi (Prerequisites): Update link for MPC.
7409 2018-09-01 Michael Matz <matz@suse.de>
7411 PR tree-optimization/87074
7412 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7413 PHIs for outer-loop uses.
7415 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7417 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7418 * doc/invoke.texi (C Dialect Options): Ditto.
7420 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7422 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7424 2018-08-31 Richard Biener <rguenther@suse.de>
7426 PR tree-optimization/87168
7427 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7428 (rpo_elim::eliminate_avail): When OP was not visited it must
7431 2018-08-31 David Malcolm <dmalcolm@redhat.com>
7433 * tree-vrp.c (copy_value_range): Convert param "from" from
7434 "value_range *" to "const value_range *".
7435 (range_is_null): Likewise for param "vr".
7436 (range_int_cst_p): Likewise.
7437 (range_int_cst_singleton_p): Likewise.
7438 (symbolic_range_p): Likewise.
7439 (value_ranges_intersect_p): Likewise for both params.
7440 (value_range_nonnegative_p): Likewise for param "vr".
7441 (value_range_constant_singleton): Likewise.
7442 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7443 (extract_range_into_wide_ints): Likewise for param "vr".
7444 (extract_range_from_multiplicative_op): Likewise for params "vr0"
7446 (vrp_can_optimize_bit_op): Likewise.
7447 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7449 (extract_range_from_unary_expr): Likewise.
7450 (debug_value_range): Likewise for param "vr".
7451 (value_range::dump): Add "const" qualifier.
7452 (vrp_prop::check_array_ref): Convert local "vr" from
7453 "value_range *" to "const value_range *".
7454 (vrp_prop::check_mem_ref): Likewise.
7455 (vrp_prop::visit_stmt): Likewise for local "old_vr".
7456 (vrp_intersect_ranges_1): Likewise for param "vr_1".
7457 (vrp_intersect_ranges): Likewise.
7458 (simplify_stmt_for_jump_threading): Likewise for local "vr".
7459 (vrp_prop::vrp_finalize): Likewise.
7460 * tree-vrp.h (value_range::dump): Add "const" qualifier.
7461 (vrp_intersect_ranges): Add "const" qualifier to params as above.
7462 (extract_range_from_unary_expr): Likewise.
7463 (value_range_constant_singleton): Likewise.
7464 (symbolic_range_p): Likewise.
7465 (copy_value_range): Likewise.
7466 (extract_range_from_binary_expr_1): Likewise.
7467 (range_int_cst_p): Likewise.
7468 (vrp_set_zero_nonzero_bits): Likewise.
7469 (range_int_cst_singleton_p): Likewise.
7471 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
7473 * config/aarch64/arm_neon.h (vabsd_s64): New.
7474 (vnegd_s64): Likewise.
7476 2018-08-31 Martin Jambor <mjambor@suse.cz>
7478 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7480 2018-08-31 Martin Liska <mliska@suse.cz>
7482 * ipa-icf.c (sem_item::add_type): Use
7483 sem_item::m_type_hash_cache.
7484 * ipa-icf.h: Move the cache from sem_item_optimizer
7487 2018-08-31 Nathan Sidwell <nathan@acm.org>
7489 * doc/extend.texi (Backwards Compatibility): Remove implicit
7490 extern C leeway of () being (...).
7492 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7494 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7496 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
7500 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7502 2018-08-31 Jakub Jelinek <jakub@redhat.com>
7505 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7506 gen_int_mode. Formatting fixes.
7508 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
7510 * target.def (custom_function_descriptors): Improve documentation.
7511 * doc/tm.texi.in (Trampolines): Expand discussion of function
7512 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7513 beginning of the section.
7514 * doc/tm.texi: Regenerated.
7516 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
7518 * cfg.h (class auto_edge_flag): Spell out the template-id of the
7519 base class in the initializer list. This is a workaround for
7520 building with older GCC.
7521 (class auto_bb_flag): Likewise.
7523 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
7525 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7526 (altivec_vcmpequ<VI_char>_p): Remove star.
7527 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7529 (expand_strncmp_vec_sequence): New function.
7530 (emit_final_str_compare_vec): New function.
7531 (expand_strn_compare): Add support for vector strncmp.
7532 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7533 length specification to bytes.
7534 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7535 (vcmpnezb_p): New pattern.
7536 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7537 for option -mstring-compare-inline-limit.
7539 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
7541 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7542 (PTA_SKYLAKE): Add PTA_AES.
7543 (PTA_GOLDMONT): Likewise.
7545 2018-08-29 Jan Hubicka <jh@suse.cz>
7548 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7549 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7551 2018-08-29 Jan Hubicka <jh@suse.cz>
7553 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7555 (hash_tree): Do not visit TYPE_STUB_DECL.
7556 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7557 stream TYPE_STUB_DECL.
7558 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7559 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7560 after free_lang_data.
7561 (type_in_anonymous_namespace_p): Likewise.
7563 2018-08-29 Jan Hubicka <jh@suse.cz>
7565 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7566 comment that it has to be even number.
7567 (class sreal): Change m_sig type to int32_t.
7568 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7569 int64_t for temporary calculations.
7570 (sreal_verify_basics): Drop one bit from minimum and maximum.
7572 2018-08-30 Richard Biener <rguenther@suse.de>
7574 PR tree-optimization/87147
7575 * tree-ssa-sccvn.c (SSA_VISITED): New function.
7576 (visit_phi): When the degenerate result is from the backedge and
7577 we didn't visit its definition yet drop to VARYING.
7578 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7580 2018-08-29 Jan Hubicka <jh@suse.cz>
7582 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7584 (hash_tree): Likewise.
7586 2018-08-29 Jan Hubicka <jh@suse.cz>
7588 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
7589 and TYPE_NEXT_REF_TO.
7591 2018-08-29 Jan Hubicka <jh@suse.cz>
7593 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7594 comment that it has to be even number.
7595 (class sreal): Change m_sig type to int32_t.
7596 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7597 int64_t for temporary calculations.
7598 (sreal_verify_basics): Drop one bit from minimum and maximum.
7600 2018-08-30 Tamar Christina <tamar.christina@arm.com>
7602 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
7604 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
7607 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
7608 if to_add is negative.
7610 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
7613 * builtins.c (c_strlen): Improve range checks.
7615 2018-08-29 Martin Sebor <msebor@redhat.com>
7616 Jeff Law <law@redhat.com>
7618 PR tree-optimization/86714
7619 PR tree-optimization/86711
7620 * builtins.c (c_strlen): Add arguments to call to string_constant.
7621 * expr.c (string_constant): Add argument. Detect missing nul
7622 terminator and outermost declaration it's missing in.
7623 * expr.h (string_constant): Add argument.
7624 * fold-const.c (read_from_constant_string): Add arguments to call to
7626 (c_getstr): Likewise.
7627 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7629 * tree-ssa-strlen.c (get_stridx): Likewise.
7631 2018-08-29 Jan Hubicka <jh@suse.cz>
7633 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
7634 Do not stream DECL_VINDEX.
7635 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
7636 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
7637 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
7639 2018-08-29 Richard Biener <rguenther@suse.de>
7641 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
7642 virtual operands that are not default defs to honor region
7644 (rpo_vn_valueize): Remove ineffective code here.
7646 2018-08-29 Richard Biener <rguenther@suse.de>
7648 PR tree-optimization/87132
7649 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
7650 when skipping defs reachable over backedges.
7652 2018-08-29 Richard Biener <rguenther@suse.de>
7654 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
7655 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
7656 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
7657 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
7658 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
7659 (vn_reference_lookup_3): Remove use of const_parms.
7660 (free_rpo_vn): Do not free const_parms.
7661 (do_rpo_vn): Do not call init_const_parms.
7662 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
7663 SSA_NAME_POINTS_TO_READONLY_MEMORY.
7664 (call_may_clobber_ref_p_1): Likewise.
7666 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
7669 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
7671 (-ftree-scev-cprop): Document.
7673 2018-08-29 Jan Hubicka <jh@suse.cz>
7675 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
7677 (sreal constructor): Update.
7678 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
7679 sreal:operator/): Update.
7681 2018-08-29 Martin Liska <mliska@suse.cz>
7683 * tree-switch-conversion.c (switch_conversion::expand):
7684 Strenghten assumption about gswitch statements.
7686 2018-08-29 Richard Biener <rguenther@suse.de>
7688 PR tree-optimization/87117
7689 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
7690 re-value-number released SSA VDEFs.
7692 2018-08-29 Richard Biener <rguenther@suse.de>
7694 PR tree-optimization/87126
7695 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
7697 2018-08-28 Jim Wilson <jimw@sifive.com>
7699 * config/riscv/pic.md: Rewrite.
7700 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
7702 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
7703 (SOFTF, default_load, softload, softstore): New.
7705 2018-08-28 Jeff Law <law@redhat.com>
7707 * fold-const.c (fold_binary_loc): Remove recently added assert.
7709 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
7711 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
7712 to OP parmeter of generated function.
7714 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
7716 PR tree-optimization/87009
7717 * match.pd: Add boolean optimizations.
7719 2018-08-28 Martin Sebor <msebor@redhat.com>
7722 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
7723 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
7724 (pass_walloca::gate): Use it.
7725 (alloca_call_type): Same.
7726 (pass_walloca::execute): Same.
7727 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
7729 2018-08-28 David Malcolm <dmalcolm@redhat.com>
7731 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
7732 GCC_VERSION for usage of "__gcc_dump_printf__" format from
7735 2018-08-28 Richard Biener <rguenther@suse.de>
7737 PR tree-optimization/87124
7738 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
7739 constants before looking up avail.
7741 2018-08-28 Jakub Jelinek <jakub@redhat.com>
7744 * calls.c (maybe_warn_nonstring_arg): Punt early if
7745 warn_stringop_overflow is zero. Don't call get_range_strlen
7746 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
7747 Swap comparison operands to have constants on rhs. Only use
7748 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
7749 increment lenrng[0].
7751 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
7753 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
7754 use of tree_to_shwi. Remove duplicated test for the size being
7755 a whole number of bytes.
7757 2018-08-28 Richard Biener <rguenther@suse.de>
7759 PR tree-optimization/87117
7760 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
7761 Handle removed stmt without LHS (GIMPLE_NOP).
7763 2018-08-28 Richard Biener <rguenther@suse.de>
7765 PR tree-optimization/87117
7766 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
7767 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
7769 2018-08-28 Richard Biener <rguenther@suse.de>
7771 PR tree-optimization/87117
7772 * tree-ssa-pre.c (compute_avail): Do not make expressions
7773 with predicated values available.
7774 (get_expr_value_id): Assert we do not run into predicated value
7777 2018-08-28 Richard Biener <rguenther@suse.de>
7779 PR tree-optimization/87117
7780 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
7781 get virtual operands.
7782 (get_expr_operands): Handle STRING_CST like other decls.
7784 2018-08-28 Martin Liska <mliska@suse.cz>
7786 * tree.h: Update documentation of fndecl_built_in_p
7790 2018-08-27 Jeff Law <law@redhat.com>
7791 PR tree-optimization/87110
7792 * tree-ssa-dse.c (compute_trims): Handle non-constant
7795 2018-08-27 Martin Sebor <msebor@redhat.com>
7797 PR tree-optimization/86914
7798 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
7800 2018-08-27 Martin Sebor <msebor@redhat.com>
7802 PR tree-optimization/87112
7803 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
7804 the type of the bound argument.
7806 2018-08-27 Jeff Law <law@redhat.com>
7808 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
7809 type does not have a TYPE_SIZE_UNIT.
7811 2018-08-27 Steve Ellcey <sellcey@cavium.com>
7813 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
7814 with include of backend.h.
7816 2018-08-27 Richard Biener <rguenther@suse.de>
7818 PR tree-optimization/86927
7819 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
7820 use const cond reduction code.
7822 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
7824 PR tree-optimization/85758
7825 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
7827 2018-08-27 David Malcolm <dmalcolm@redhat.com>
7830 * diagnostic-show-locus.c (class layout_range): Update for
7831 conversion of show_caret_p to a tri-state.
7832 (layout_range::layout_range): Likewise.
7833 (make_range): Likewise.
7834 (layout::maybe_add_location_range): Likewise.
7835 (layout::should_print_annotation_line_p): Don't show annotation
7836 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
7837 (layout::get_state_at_point): Update for conversion of
7838 show_caret_p to a tri-state. Bail out early for
7839 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
7840 underlining or source colorization.
7841 (gcc_rich_location::add_location_if_nearby): Update for conversion
7842 of show_caret_p to a tri-state.
7843 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
7844 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
7846 (selftest::test_one_liner_labels): Likewise.
7847 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
7848 conversion of show_caret_p to a tri-state.
7849 * pretty-print.c (text_info::set_location): Likewise.
7850 * pretty-print.h (text_info::set_location): Likewise.
7851 * substring-locations.c (format_warning_n_va): Likewise.
7852 * tree-diagnostic.c (default_tree_printer): Likewise.
7853 * tree-pretty-print.c (newline_and_indent): Likewise.
7855 2018-08-27 David Malcolm <dmalcolm@redhat.com>
7858 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
7859 line above for line-insertion fix-it hints.
7860 (selftest::test_fixit_insert_containing_newline): Update the
7861 expected results, and add a test with line-numbering enabled.
7863 2018-08-27 Martin Liska <mliska@suse.cz>
7866 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
7867 params with DECL_HAS_VALUE_EXPR_P.
7869 2018-08-27 Martin Liska <mliska@suse.cz>
7871 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
7872 selected expansion strategy.
7874 2018-08-27 Martin Liska <mliska@suse.cz>
7876 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
7877 * builtins.c (is_builtin_fn): Likewise.
7878 * attribs.c (diag_attr_exclusions): Use new function
7879 fndecl_built_in_p and remove check for FUNCTION_DECL if
7881 (builtin_mathfn_code): Likewise.
7882 (fold_builtin_expect): Likewise.
7883 (fold_call_expr): Likewise.
7884 (fold_builtin_call_array): Likewise.
7885 (fold_call_stmt): Likewise.
7886 (set_builtin_user_assembler_name): Likewise.
7887 (is_simple_builtin): Likewise.
7888 * calls.c (gimple_alloca_call_p): Likewise.
7889 (maybe_warn_nonstring_arg): Likewise.
7890 * cfgexpand.c (expand_call_stmt): Likewise.
7891 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
7892 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
7893 (cgraph_node::verify_node): Likewise.
7894 * cgraphclones.c (build_function_decl_skip_args): Likewise.
7895 (cgraph_node::create_clone): Likewise.
7896 * config/arm/arm.c (arm_insert_attributes): Likewise.
7897 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
7898 * dse.c (scan_insn): Likewise.
7899 * expr.c (expand_expr_real_1): Likewise.
7900 * fold-const.c (operand_equal_p): Likewise.
7901 (fold_binary_loc): Likewise.
7902 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
7903 * gimple-low.c (lower_stmt): Likewise.
7904 * gimple-pretty-print.c (dump_gimple_call): Likewise.
7905 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7907 * gimple.c (gimple_build_call_from_tree): Likewise.
7908 (gimple_call_builtin_p): Likewise.
7909 (gimple_call_combined_fn): Likewise.
7910 * gimplify.c (gimplify_call_expr): Likewise.
7911 (gimple_boolify): Likewise.
7912 (gimplify_modify_expr): Likewise.
7913 (gimplify_addr_expr): Likewise.
7914 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
7915 * ipa-cp.c (determine_versionability): Likewise.
7916 * ipa-fnsummary.c (compute_fn_summary): Likewise.
7917 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
7918 * ipa-split.c (visit_bb): Likewise.
7919 (split_function): Likewise.
7920 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
7921 * lto-cgraph.c (input_node): Likewise.
7922 * lto-streamer-out.c (write_symbol): Likewise.
7923 * omp-low.c (setjmp_or_longjmp_p): Likewise.
7924 (lower_omp_1): Likewise.
7925 * predict.c (strip_predict_hints): Likewise.
7926 * print-tree.c (print_node): Likewise.
7927 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
7928 * trans-mem.c (is_tm_irrevocable): Likewise.
7929 (is_tm_load): Likewise.
7930 (is_tm_simple_load): Likewise.
7931 (is_tm_store): Likewise.
7932 (is_tm_simple_store): Likewise.
7933 (is_tm_abort): Likewise.
7934 (tm_region_init_1): Likewise.
7935 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
7936 * tree-cfg.c (verify_gimple_call): Likewise.
7937 (move_stmt_r): Likewise.
7938 (stmt_can_terminate_bb_p): Likewise.
7939 * tree-eh.c (lower_eh_constructs_2): Likewise.
7940 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
7941 * tree-inline.c (remap_gimple_stmt): Likewise.
7942 (copy_bb): Likewise.
7943 (estimate_num_insns): Likewise.
7944 (fold_marked_statements): Likewise.
7945 * tree-sra.c (scan_function): Likewise.
7946 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
7947 (optimize_stack_restore): Likewise.
7948 (pass_fold_builtins::execute): Likewise.
7949 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
7950 (mark_all_reaching_defs_necessary_1): Likewise.
7951 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
7952 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
7953 (pass_forwprop::execute): Likewise.
7954 * tree-ssa-loop-im.c (stmt_cost): Likewise.
7955 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
7956 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
7957 * tree-ssa-strlen.c (get_string_length): Likewise.
7958 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
7959 (find_func_aliases_for_call): Likewise.
7960 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
7961 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
7962 * tree-tailcall.c (find_tail_calls): Likewise.
7963 * tree.c (need_assembler_name_p): Likewise.
7964 (free_lang_data_in_decl): Likewise.
7965 (get_call_combined_fn): Likewise.
7966 * ubsan.c (is_ubsan_builtin_p): Likewise.
7967 * varasm.c (incorporeal_function_p): Likewise.
7968 * tree.h (DECL_BUILT_IN): Remove and replace with
7970 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
7971 (fndecl_built_in_p): New.
7973 2018-08-27 Martin Liska <mliska@suse.cz>
7975 PR tree-optimization/86847
7976 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
7977 Dump also subtree probability.
7978 (switch_decision_tree::do_jump_if_equal): New function.
7979 (switch_decision_tree::emit_case_nodes): Handle special
7980 situations in balanced tree that can be emitted much simpler.
7981 Fix calculation of probabilities that happen in tree expansion.
7982 * tree-switch-conversion.h (struct cluster): Add
7984 (struct simple_cluster): Likewise.
7985 (struct case_tree_node): Add new function has_child.
7986 (do_jump_if_equal): New.
7988 2018-08-27 Martin Liska <mliska@suse.cz>
7990 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
7991 Add new argument to bit_test_cluster constructor.
7992 (bit_test_cluster::emit): Set bits really number of values
7994 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
7995 probability argument.
7996 * tree-switch-conversion.h (struct bit_test_cluster):
7997 Add m_handles_entire_switch.
7999 2018-08-27 Martin Liska <mliska@suse.cz>
8001 PR tree-optimization/86702
8002 * tree-switch-conversion.c (jump_table_cluster::emit):
8003 Make probabilities even for values in jump table
8004 according to number of cases handled.
8005 (switch_decision_tree::compute_cases_per_edge): Pass
8006 argument to reset_out_edges_aux function.
8007 (switch_decision_tree::analyze_switch_statement): Likewise.
8008 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8011 2018-08-27 Martin Liska <mliska@suse.cz>
8013 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8014 cfun argument explicitly.
8015 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8016 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8017 function gimple_switch_default_bb.
8018 (convert_switch_statements):
8020 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8021 * stmt.c (label_to_block_fn): Use label_to_block and pass
8022 cfun argument explicitly and use gimple_switch_label_bb.
8023 (expand_case): Likewise.
8024 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8025 cfun argument explicitly. Likewise.
8026 (make_edges_bb): Likewise.
8027 (make_cond_expr_edges): Likewise.
8028 (get_cases_for_edge): Likewise.
8029 (make_gimple_switch_edges): Likewise.
8030 (label_to_block_fn): Likewise.
8031 (label_to_block): Likewise.
8032 (make_goto_expr_edges): Likewise.
8033 (make_gimple_asm_edges): Likewise.
8034 (main_block_label): Likewise.
8035 (group_case_labels_stmt): Likewise.
8036 (find_taken_edge_computed_goto): Likewise.
8037 (find_taken_edge_switch_expr): Likewise.
8038 (gimple_verify_flow_info): Likewise.
8039 (gimple_redirect_edge_and_branch): Likewise.
8040 (gimple_switch_label_bb): New function.
8041 (gimple_switch_default_bb): Likewise.
8042 (gimple_switch_edge): Likewise.
8043 (gimple_switch_default_edge): Likewise.
8044 * tree-cfg.h (label_to_block_fn): Remove and replace ...
8045 (label_to_block): ... with this.
8046 (gimple_switch_label_bb): New.
8047 (gimple_switch_default_bb): Likewise.
8048 (gimple_switch_edge): Likewise.
8049 (gimple_switch_default_edge): Likewise.
8050 * tree-cfgcleanup.c (convert_single_case_switch): Use
8051 new gimple functions and pass new argument to label_to_block.
8052 (cleanup_control_flow_bb):
8053 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8054 cfun argument explicitly.
8055 (make_eh_edges): Likewise.
8056 (redirect_eh_dispatch_edge): Likewise.
8057 (lower_resx): Likewise.
8058 (lower_eh_dispatch): Likewise.
8059 (maybe_remove_unreachable_handlers): Likewise.
8060 (unsplit_eh): Likewise.
8061 (cleanup_empty_eh): Likewise.
8062 (verify_eh_edges): Likewise.
8063 (verify_eh_dispatch_edge): Likewise.
8064 * tree-ssa-dom.c (record_edge_info): Likewise.
8065 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8066 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8067 (thread_through_normal_block): Likewise.
8068 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8069 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8070 * tree-switch-conversion.c (switch_conversion::collect): Use new
8072 (switch_conversion::check_final_bb): Likewise.
8073 (switch_conversion::gather_default_values): Pass new argument
8075 (switch_conversion::build_constructors): Likewise.
8076 (switch_decision_tree::compute_cases_per_edge): Use new
8077 gimple_switch_edge function.
8078 (switch_decision_tree::analyze_switch_statement): Pass new argument
8080 (switch_decision_tree::try_switch_expansion): Use
8081 gimple_switch_default_edge.
8082 * tree-vrp.c (find_switch_asserts): Pass new argument
8084 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8085 (vr_values::simplify_switch_using_ranges): Likewise.
8087 2018-08-27 Richard Biener <rguenther@suse.de>
8089 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8090 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8092 * tree-ssa-sccvn.h (struct vn_pval): New structure.
8093 (struct vn_nary_op_s): Add unwind_to member. Add
8094 predicated_values flag and put result into a union together
8095 with a linked list of vn_pval.
8096 (struct vn_ssa_aux): Add name member to make maintaining
8097 a map of SSA name to vn_ssa_aux possible. Remove no longer
8098 needed info, dfsnum, low, visited, on_sccstack, use_processed
8099 and range_info_anti_range_p members.
8100 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8101 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8103 (vn_valueize): New global.
8104 (vn_context_bb): Likewise.
8105 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8106 VN_INFO_PTR_INFO): Remove.
8107 * tree-ssa-sccvn.c: ... (rewrite)
8108 (pass_fre::execute): For -O2+ initialize loops and run
8109 RPO VN in optimistic mode (iterating). For -O1 and -Og
8110 run RPO VN in non-optimistic mode.
8111 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8112 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8113 * doc/invoke.texi (sccvn-max-scc-size): Remove.
8114 (rpo-vn-max-loop-depth): Document.
8115 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8116 when valuezing the VUSE signals we walked out of the region.
8117 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8118 (phi_translate): Set VN context block to use for availability
8120 (compute_avail): Likewise.
8121 (pre_valueize): New function.
8122 (pass_pre::execute): Adjust to the RPO VN API.
8124 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8125 (propagate_constants_for_unrolling): Remove.
8126 (tree_unroll_loops_completely): Perform value-numbering
8127 on the unrolled bodies loop parent.
8129 2018-08-27 Richard Biener <rguenther@suse.de>
8131 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8132 for partial antic compute.
8134 2018-08-27 Jakub Jelinek <jakub@redhat.com>
8136 PR rtl-optimization/87065
8137 * combine.c (simplify_if_then_else): Formatting fix.
8138 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8140 (known_cond): Don't return const_true_rtx for vector modes. Use
8141 CONST0_RTX instead of const0_rtx. Formatting fixes.
8143 2018-08-27 Martin Liska <mliska@suse.cz>
8145 PR gcov-profile/87069
8146 * gcov.c (process_file): Record files already processed
8147 and warn about a file being processed multiple times.
8149 2018-08-27 Martin Liska <mliska@suse.cz>
8152 * config/aarch64/aarch64.c (aarch64_override_options_internal):
8153 Set default values for x_aarch64_*_string strings.
8154 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8155 prefix. For -mabi do not print '=ABI' in help and use
8156 <option_value> format for -msve-vector-bits and -moverride
8159 2018-08-26 Jeff Law <law@redhat.com>
8161 * config/mips/frame-header-opt.c: Include "backend.h" rather than
8164 2018-08-26 Marek Polacek <polacek@redhat.com>
8166 PR c++/87029, Implement -Wredundant-move.
8167 * doc/invoke.texi: Document -Wredundant-move.
8169 2018-08-25 Martin Sebor <msebor@redhat.com>
8171 PR tree-optimization/87059
8172 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8173 to the same type as the other.
8174 * fold-const.c (fold_binary_loc): Assert expectation.
8176 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
8178 * config/darwin.c (machopic_legitimize_pic_address): Clean up
8179 extraneous parentheses, dead code section and formatting.
8181 2018-08-24 David Malcolm <dmalcolm@redhat.com>
8184 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8185 wide enough for jumps in the line-numbering to be visible.
8186 (layout::print_gap_in_line_numbering): New member function.
8187 (layout::calculate_line_spans): When using line numbering, merge
8188 line spans that are only 1 line apart.
8189 (diagnostic_show_locus): When printing line numbers, show gaps in
8190 line numbering directly, rather than printing headers.
8191 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8192 line-numbering with multiple line spans.
8193 (selftest::test_fixit_insert_containing_newline_2): Add test of
8194 line-numbering, in which the spans are close enough to be merged.
8196 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
8198 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8199 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
8200 * tree-vrp.c (range_is_nonnull): Remove.
8201 (range_includes_zero_p): Accept value_range instead of min/max.
8202 (extract_range_from_binary_expr_1): Do not early bail on
8204 Use range_includes_zero_p instead of range_is_nonnull.
8205 (extract_range_from_unary_expr): Use range_includes_zero_p instead
8206 of range_is_nonnull.
8207 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
8208 special case VR_ANTI_RANGE.
8209 (vrp_finalize): Same.
8210 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8212 (range_is_nonnull): Remove.
8213 * vr-values.c (vrp_stmt_computes_nonzero): Use
8214 range_includes_zero_p instead of range_is_nonnull.
8215 (extract_range_basic): Pass value_range to range_includes_zero_p
8216 instead of range_is_nonnull.
8218 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
8220 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8221 * emit-rtl.h (rtl_data): Remove return_bnd.
8222 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8223 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8224 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8225 (POINTER_BOUNDS_MODE): Remove definition.
8226 (make_pointer_bounds_mode): Remove.
8227 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8228 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8229 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8230 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8231 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8232 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8233 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8235 * config/i386/i386-modes.def (BND32, BND64): Remove.
8236 * config/i386/i386.c (dbx_register_map): Remove bound registers.
8237 (dbx64_register_map): Ditto.
8238 (svr4_dbx_register_map): Ditto.
8239 (indirect_thunk_bnd_needed): Remove.
8240 (indirect_thunks_bnd_used): Ditto.
8241 (indirect_return_bnd_needed): Ditto.
8242 (indirect_return_via_cx_bnd): Ditto.
8243 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8244 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8245 (output_indirect_thunk): Ditto. Remove need_prefix argument.
8246 (output_indirect_thunk_function): Remove handling of
8247 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8248 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8249 (ix86_save_reg): Remove handling of crtl->return_bnd.
8250 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8251 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8252 and UNSPEC_BNDLX_ADDR.
8253 (ix86_output_indirect_branch_via_reg): Remove handling of
8254 indirect_thunk_prefix_bnd.
8255 (ix86_output_indirect_branch_via_push): Ditto.
8256 (ix86_output_function_return): Ditto.
8257 (ix86_output_indirect_function_return): Ditto.
8258 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8259 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8260 (CALL_USED_REGISTERS): Ditto.
8261 (REG_ALLOC_ORDER): Update for removal of bound registers.
8262 (HI_REGISTER_NAMES): Ditto.
8263 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8264 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8265 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8266 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8267 (FIRST_PSEUDO_REG): Update.
8268 (BND): Remove mode iterator.
8269 * config/i386/predicates.md (bnd_mem_operator): Remove.
8271 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
8273 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8276 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
8278 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8279 the case in which the permute needs only a single element and
8280 repeats for every vector of the result. Extend that case to
8281 handle variable-length vectors.
8282 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8284 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
8287 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8288 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8290 2018-08-24 Richard Biener <rguenther@suse.de>
8292 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8293 bb_flags_allocated members.
8294 (auto_flag): New RAII class for allocating flags.
8295 (auto_edge_flag): New RAII class for allocating edge flags.
8296 (auto_bb_flag): New RAII class for allocating bb flags.
8297 * cfgloop.c (verify_loop_structure): Allocate temporary edge
8299 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8300 in favor of temporarily allocated BB flag.
8301 * hsa-brig.c: Re-order includes.
8302 * hsa-dump.c: Likewise.
8303 * hsa-regalloc.c: Likewise.
8304 * print-rtl.c: Likewise.
8305 * profile-count.c: Likewise.
8307 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
8310 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8313 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
8316 * wide-int-range.cc (wide_int_range_div): Do not ignore result
8317 from wide_int_range_multiplicative_op.
8319 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8321 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8324 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
8326 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8329 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
8331 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8332 full GENERIC expression used for replacement.
8334 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
8336 * tree-vrp.c (abs_extent_range): Remove.
8337 (extract_range_into_wide_ints): Pass wide ints by reference.
8338 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8339 Pass wide ints by reference in all calls to
8340 extract_range_into_wide_ints.
8341 * wide-int-range.cc (wide_int_range_div): New.
8342 * wide-int-range.h (wide_int_range_div): New.
8343 (wide_int_range_includes_zero_p): New.
8344 (wide_int_range_zero_p): New.
8346 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
8348 * config/aarch64/aarch64.md (arches): New enum.
8349 (arch): New enum attr.
8350 (arch_enabled): New attr.
8351 (enabled): Now uses arch_enabled only.
8352 (simd, sve, fp16): Removed attribute.
8353 (fp): Attr now defined in terms of 'arch'.
8354 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8355 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8356 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8357 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8358 attributes into 'arch'.
8359 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8360 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8361 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8362 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8363 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8364 'simd' attribute into 'arch'.
8365 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8366 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8367 Convert use of 'fp' attribute to 'arch'.
8368 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8369 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8371 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8372 (different modes) Merge 'fp' and 'simd' into 'arch'.
8373 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8376 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
8378 PR rtl-optimization/87026
8379 * expmed.c (canonicalize_comparison): If we can no longer create
8380 pseudoregisters, don't.
8382 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
8385 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8387 * config/arm/arm.c (speculation_barrier_libfunc): New static
8389 (arm_init_libfuncs): Initialize it.
8390 (arm_emit_speculation_barrier): New function.
8391 * config/arm/arm.md (speculation_barrier): Call
8392 arm_emit_speculation_barrier for architectures that do not have
8394 (speculation_barrier_insn): Only match on Armv7 or later.
8396 2018-08-23 Richard Biener <rguenther@suse.de>
8399 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8402 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
8404 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8405 of single-vector TBLs.
8406 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8409 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
8412 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8413 aarch64_evpc_tbl guard.
8415 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
8417 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8420 2018-08-22 Martin Sebor <msebor@redhat.com>
8423 * tree-pretty-print.c (pretty_print_string): Add argument.
8424 (dump_generic_node): Call to pretty_print_string with string size.
8426 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
8428 PR rtl-optimization/86771
8429 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8430 of two SETs into those two SETs, one to be placed at i2, if that SETs
8431 destination is modified between i2 and i3.
8433 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8435 PR tree-optimization/86725
8436 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8438 (vect_analyze_scalar_cycles_1): Check it.
8440 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8442 PR tree-optimization/86725
8443 * tree-vect-loop.c (vect_is_simple_reduction): When treating
8444 an outer loop phi as a double reduction, make sure that the
8445 single user of the phi result is an inner loop phi.
8447 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8449 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8450 grouped stores with gaps to a strided group.
8452 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8454 * tree-vect-stmts.c (get_group_load_store_type)
8455 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8456 first statement in a group.
8458 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8460 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8461 the sequence used in gcc/gcc.c.
8463 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8466 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8469 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8471 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8472 Darwin10-specific unwinder-shim.
8473 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8474 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8475 New to cater for Darwin10 Rosetta.
8477 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8479 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8480 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8482 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8487 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8488 (dwarf2out_switch_text_section): Generate a local label for the
8489 second function sub-section and apply it as the second FDE start
8491 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8492 second sub-section start.
8494 2018-08-22 Richard Biener <rguenther@suse.de>
8496 PR tree-optimization/86988
8497 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8499 2018-08-22 Richard Biener <rguenther@suse.de>
8501 PR tree-optimization/86945
8502 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8504 2018-08-22 Alexandre Oliva <oliva@adacore.com>
8506 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8507 a comment about how uses of r2 for .sdata2 come about.
8509 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
8511 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8513 2018-08-21 Marek Polacek <polacek@redhat.com>
8515 PR c++/86981, Implement -Wpessimizing-move.
8516 * doc/invoke.texi: Document -Wpessimizing-move.
8518 2018-08-21 Jan Hubicka <jh@suse.cz>
8520 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8521 * tree.h (is_redundant_typedef): Remove.
8522 * dwarf2out.c (is_redundant_typedef): Turn into static function.
8524 2018-08-21 Jan Hubicka <jh@suse.cz>
8526 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8529 2018-08-21 Tamar Christina <tamar.christina@arm.com>
8531 * expmed.c (extract_low_bits): Reject invalid subregs early.
8533 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
8536 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8539 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8541 * config/vxworks.h: Guard vxworks_asm_out_constructor and
8542 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8543 * config/vxworks.c: Likewise.
8545 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8547 * config/vxworks.c: Set targetm.have_ctors_dtors
8548 if HAVE_INITFINI_ARRAY_SUPPORT.
8549 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8550 if HAVE_INITFINI_ARRAY_SUPPORT.
8552 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
8554 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8555 default search path for VxWorks < 7.
8557 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
8559 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8560 (get_format_string): Refer to c_getstr.
8562 2018-08-21 Tom de Vries <tdevries@suse.de>
8564 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8565 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8566 (debuginfo_early_stop): Declare.
8567 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8568 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8569 (debuginfo_early_stop): New function.
8570 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8571 and debuginfo_early_stop.
8572 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8573 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8574 (general_init): Call debuginfo_early_init.
8575 (finalize): Call debuginfo_fini.
8576 (do_compile): Call debuginfo_init.
8577 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8579 (@item -fdump-debug, @item -fdump-earlydebug): Add.
8581 2018-08-21 Tom de Vries <tdevries@suse.de>
8583 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8584 flag_dump_noaddr and flag_dump_unnumbered.
8586 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
8588 * wide-int-range.cc (wide_int_range_abs): New.
8589 (wide_int_range_order_set): Rename from wide_int_range_min_max.
8590 * wide-int-range.h (wide_int_range_abs): New.
8591 (wide_int_range_min_max): New.
8592 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
8593 case to call wide_int_range_abs.
8594 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
8595 (extract_range_from_abs_expr): Delete.
8597 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
8600 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
8601 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
8604 2018-08-20 Nathan Sidwell <nathan@acm.org>
8605 Jeff Law <law@redhat.com>
8607 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
8608 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
8610 2018-08-20 David Malcolm <dmalcolm@redhat.com>
8613 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
8614 (decl_attributes): Likewise.
8615 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
8617 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
8618 * diagnostic-core.h (class auto_diagnostic_group): New class.
8619 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
8620 (diagnostic_report_diagnostic): Handle the first diagnostics within
8622 (emit_diagnostic): Add auto_diagnostic_group instance.
8624 (inform_n): Likewise.
8625 (warning): Likewise.
8626 (warning_at): Likewise.
8627 (warning_n): Likewise.
8628 (pedwarn): Likewise.
8629 (permerror): Likewise.
8631 (error_n): Likewise.
8632 (error_at): Likewise.
8634 (fatal_error): Likewise.
8635 (internal_error): Likewise.
8636 (internal_error_no_backtrace): Likewise.
8637 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
8638 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
8639 * diagnostic.h (struct diagnostic_context): Add fields
8640 "diagnostic_group_nesting_depth",
8641 "diagnostic_group_emission_count", "begin_group_cb",
8643 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
8644 Add auto_diagnostic_group instance(s).
8645 (find_explicit_erroneous_behavior): Likewise.
8646 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
8647 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
8648 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
8649 (gimplify_va_arg_expr): Likewise.
8650 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
8651 (HSA_SORRY_AT): Likewise.
8652 * ipa-devirt.c (compare_virtual_tables): Likewise.
8653 (warn_odr): Likewise.
8654 * multiple_target.c (expand_target_clones): Likewise.
8655 * opts-common.c (cmdline_handle_error): Likewise.
8656 * reginfo.c (globalize_reg): Likewise.
8657 * substring-locations.c (format_warning_n_va): Likewise.
8658 * tree-inline.c (expand_call_inline): Likewise.
8659 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
8660 * tree-ssa-loop-niter.c
8661 (do_warn_aggressive_loop_optimizations): Likewise.
8662 * tree-ssa-uninit.c (warn_uninit): Likewise.
8663 * tree.c (warn_deprecated_use): Likewise.
8665 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
8668 * config/i386/i386.md (eh_return): Always update EH return
8669 address in word_mode.
8671 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
8673 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
8674 TARGET_SPLIT_COMPLEX_ARG is defined.
8676 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
8678 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
8680 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
8683 * expr.c (expand_assignment): Assert that bitpos is positive.
8684 (store_field): Likewise
8685 (expand_expr_real_1): Make sure that bitpos is positive.
8686 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
8689 2018-08-20 Nathan Sidwell <nathan@acm.org>
8691 * Makefile.in (CPP_ID_DATA_H): Delete.
8692 (CPP_INTERNAL_H): Don't add it.
8693 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
8694 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
8696 2018-08-20 Richard Biener <rguenther@suse.de>
8698 PR tree-optimization/78655
8699 * tree-vrp.c (extract_range_from_binary_expr_1): Make
8700 pointer + offset nonnull if either operand is nonnull work.
8702 2018-08-20 Tom de Vries <tdevries@suse.de>
8704 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
8705 unless the referenced die describes the added property using
8706 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
8707 Otherwise, add a DW_AT_location to the referenced die.
8709 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
8712 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
8713 register_operand when calling ix86_set_reg_reg_cost.
8714 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
8715 Set *total to 0 for operands that satisfy x86_64_immediate_operand
8716 predicate and to 1 otherwise.
8718 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
8720 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
8721 emit a diagnostic that it is not supported and reset the option.
8722 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
8723 supported and consume the option. (ASM_FINAL_SPEC): New.
8725 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
8727 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
8730 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
8732 C-SKY port: Documentation
8734 * doc/extend.texi (C-SKY Function Attributes): New section.
8735 * doc/invoke.texi (Option Summary): Add C-SKY options.
8736 (C-SKY Options): New section.
8737 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
8739 2018-08-17 Jojo <jijie_rong@c-sky.com>
8740 Huibin Wang <huibin_wang@c-sky.com>
8741 Sandra Loosemore <sandra@codesourcery.com>
8742 Chung-Lin Tang <cltang@codesourcery.com>
8744 C-SKY port: Backend implementation
8746 * config/csky/*: New.
8747 * common/config/csky/*: New.
8749 2018-08-17 Jojo <jijie_rong@c-sky.com>
8750 Huibin Wang <huibin_wang@c-sky.com>
8751 Sandra Loosemore <sandra@codesourcery.com>
8752 Chung-Lin Tang <cltang@codesourcery.com>
8753 Andrew Jenner <andrew@codesourcery.com>
8755 C-SKY port: Configury
8757 * config.gcc (csky-*-*): New.
8758 * configure.ac: Add csky to targets for dwarf2 debug_line support.
8759 * configure: Regenerated.
8761 2018-08-17 David Malcolm <dmalcolm@redhat.com>
8763 * dump-context.h: Include "dumpfile.h".
8764 (dump_context::dump_printf_va): Convert final param from va_list
8765 to va_list *. Convert from ATTRIBUTE_PRINTF to
8766 ATTRIBUTE_GCC_DUMP_PRINTF.
8767 (dump_context::dump_printf_loc_va): Likewise.
8768 * dumpfile.c: Include "stringpool.h".
8769 (make_item_for_dump_printf_va): Delete.
8770 (make_item_for_dump_printf): Delete.
8771 (class dump_pretty_printer): New class.
8772 (dump_pretty_printer::dump_pretty_printer): New ctor.
8773 (dump_pretty_printer::emit_items): New member function.
8774 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
8776 (dump_pretty_printer::emit_item): New member function.
8777 (dump_pretty_printer::stash_item): New member function.
8778 (dump_pretty_printer::format_decoder_cb): New member function.
8779 (dump_pretty_printer::decode_format): New member function.
8780 (dump_context::dump_printf_va): Reimplement in terms of
8781 dump_pretty_printer.
8782 (dump_context::dump_printf_loc_va): Convert final param from va_list
8784 (dump_context::begin_scope): Reimplement call to
8785 make_item_for_dump_printf.
8786 (dump_printf): Update for change to dump_printf_va.
8787 (dump_printf_loc): Likewise.
8788 (selftest::test_capture_of_dump_calls): Convert "stmt" from
8789 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
8790 with %T, %E, and %G.
8791 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
8792 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
8793 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
8794 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
8795 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
8796 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
8797 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
8798 within a dump_printf_loc call to "%wu".
8799 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
8800 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
8801 missing space after "=".
8802 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
8803 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
8804 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
8805 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
8806 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
8807 duplicate "vectorized" from message.
8809 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
8811 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
8812 polyNxK_t element's TYPE_STRING_FLAG.
8814 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
8816 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
8817 (they were unnamed before). Fix comments.
8819 2018-08-17 Nathan Sidwell <nathan@acm.org>
8821 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
8823 2018-08-17 Richard Biener <rguenther@suse.de>
8825 PR tree-optimization/86841
8826 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
8828 2018-08-17 Martin Liska <mliska@suse.cz>
8830 * common.opt: Remove Warn, Init and Report for options with
8831 Ignore/Deprecated flag. Warning is done automatically for
8833 * config/i386/i386.opt: Likewise.
8834 * config/ia64/ia64.opt: Likewise.
8835 * config/rs6000/rs6000.opt: Likewise.
8836 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
8837 Remove usage of flag_check_pointer_bounds.
8838 * lto-wrapper.c (merge_and_complain): Do not handle
8839 OPT_fcheck_pointer_bounds.
8840 (append_compiler_options): Likewise.
8841 * opt-functions.awk: Do not handle Deprecated.
8842 * optc-gen.awk: Check that Var, Report and Init are not
8843 used for an option with Ignore/Deprecated flag.
8844 * opts-common.c (decode_cmdline_option): Do not report
8846 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
8848 * opts.h (struct cl_option): Remove cl_deprecated flag.
8849 (CL_ERR_DEPRECATED): Remove error enum value.
8851 2018-08-17 Richard Biener <rguenther@suse.de>
8854 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
8855 across a va-arg-pack using call adjust its return value accordingly.
8857 2018-08-16 Martin Sebor <msebor@redhat.com>
8859 PR tree-optimization/86853
8860 * gimple-ssa-sprintf.c (struct format_result): Rename member.
8861 (struct fmtresult): Add member and initialize it in ctors.
8862 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
8863 (format_string): Handle %S the same as %ls. Set MAYFAIL.
8864 (format_directive): Set POSUNDER4K when MAYFAIL is set.
8865 (parse_directive): Handle %C same as %c.
8866 (sprintf_dom_walker::compute_format_length): Adjust.
8867 (is_call_safe): Adjust.
8869 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8871 * builtins.c (c_strlen): Add new parameter eltsize. Use it
8872 for determining how to count the elements.
8873 * builtins.h (c_strlen): Adjust prototype.
8874 * expr.c (string_constant): Add new parameter mem_size.
8875 Set *mem_size appropriately.
8876 * expr.h (string_constant): Adjust protoype.
8877 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
8878 * gimple-fold.h (get_range_strlen): Adjust prototype.
8879 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
8880 (format_string): Call get_string_length with eltsize.
8882 2018-08-16 David Malcolm <dmalcolm@redhat.com>
8884 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
8885 to emit the span, rather than setting it as the prefix.
8887 2018-08-16 David Malcolm <dmalcolm@redhat.com>
8889 * diagnostic-show-locus.c (layout::start_annotation_line): Add
8890 "margin_char" parameter, defaulting to space. Use it in place
8891 of pp_space for the initial part of the margin.
8892 (layout::print_leading_fixits): Use '+' when filling the margin
8893 of line-insertion fix-it hints.
8895 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8897 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
8900 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8902 * config/rs6000/altivec.md: Don't set length attribute to the default
8904 * config/rs6000/darwin.md: Ditto.
8905 * config/rs6000/dfp.md: Ditto.
8906 * config/rs6000/htm.md: Ditto.
8907 * config/rs6000/rs6000.md: Ditto.
8908 * config/rs6000/sync.md: Ditto.
8909 * config/rs6000/vsx.md: Ditto.
8911 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8913 * config/rs6000/altivec.md: Don't set length attribute to the default
8914 value, for branch instructions.
8915 * config/rs6000/darwin.md: Ditto.
8916 * config/rs6000/rs6000.md: Ditto.
8918 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
8920 * config/rs6000/rs6000.md (length): Always define as const_int 4.
8921 (unnamed conditional branch define_insn): Set length to 4 or 8
8922 depending on offset.
8923 (<bd>_<mode>): Similar, for alternative 0.
8924 (<bd>tf_<mode>): Ditto.
8926 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8928 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
8930 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
8932 * doc/rtl.texi: Replace old RTX class names with new names.
8935 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
8937 * expmed.h (canonicalize_comparison): New declaration.
8938 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
8939 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
8940 * optabs.c (prepare_cmp_insn): Likewise.
8941 * rtl.h (unsigned_condition_p): New function which checks if a
8942 comparison operator is unsigned.
8944 2018-08-16 Nathan Sidwell <nathan@acm.org>
8946 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
8947 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
8949 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8952 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
8953 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
8954 (mov<mov>): ..this and enable unconditionally.
8956 2018-08-16 Tamar Christina <tamar.christina@arm.com>
8958 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
8960 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
8962 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
8963 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
8964 "Common" with "Target".
8966 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
8968 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
8969 * doc/invoke.texi (mmitigate-rop): Remove.
8970 * config/i386/i386.c: Do not include "regrename.h".
8971 (ix86_rop_should_change_byte_p, reg_encoded_number)
8972 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
8974 (ix86_reorg): Remove call to ix86_mitigate_rop.
8975 * config/i386/i386.md (attr "modrm_class"): Remove.
8976 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
8977 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
8978 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
8980 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
8982 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
8983 allow folding of mergeh() and mergel() for the float and double types.
8984 (fold_mergehl_helper): Rework to handle building a permute tree
8987 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
8989 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
8992 2018-08-15 David Malcolm <dmalcolm@redhat.com>
8994 * common.opt (fdiagnostics-show-labels): New option.
8995 * diagnostic-show-locus.c (class layout_range): Add field
8997 (class layout): Add field "m_show_labels_p".
8998 (layout_range::layout_range): Add param "label" and use it to
9000 (make_range): Pass in NULL for new "label" param of layout_range's
9002 (layout::layout): Initialize m_show_labels_p.
9003 (layout::maybe_add_location_range): Pass in loc_range->m_label
9004 when constructing layout_range instances.
9005 (struct line_label): New struct.
9006 (layout::print_any_labels): New member function.
9007 (layout::print_line): Call it if label-printing is enabled.
9008 (selftest::test_one_liner_labels): New test.
9009 (selftest::test_diagnostic_show_locus_one_liner): Call it.
9010 * diagnostic.c (diagnostic_initialize): Initialize
9011 context->show_labels_p.
9012 * diagnostic.h (struct diagnostic_context): Add field
9014 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9015 -fno-diagnostics-show-labels.
9016 * dwarf2out.c (gen_producer_string): Add
9017 OPT_fdiagnostics_show_labels to the ignored options.
9018 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9020 (gcc_rich_location::maybe_add_expr): Likewise.
9021 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9022 label" param, defaulting to NULL.
9023 (gcc_rich_location::add_expr): Add "label" param.
9024 (gcc_rich_location::maybe_add_expr): Likewise.
9025 (class text_range_label): New class.
9026 (class range_label_for_type_mismatch): New class.
9027 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9028 of format_warning_va.
9029 (fmtwarn_n): Likewise for new params of format_warning_n_va.
9030 * lto-wrapper.c (merge_and_complain): Add
9031 OPT_fdiagnostics_show_labels to the "pick one setting" options.
9032 (append_compiler_options): Likewise to the dropped options.
9033 (append_diag_options): Likewise to the passed-on options.
9034 * opts.c (common_handle_option): Handle the new option.
9035 * selftest-diagnostic.c
9036 (test_diagnostic_context::test_diagnostic_context): Enable
9038 * substring-locations.c: Include "gcc-rich-location.h".
9039 (format_warning_n_va): Add "fmt_label" and "param_label" params
9040 and use them as appropriate.
9041 (format_warning_va): Add "fmt_label" and "param_label" params,
9042 passing them on to format_warning_n_va.
9043 (format_warning_at_substring): Likewise.
9044 (format_warning_at_substring_n): Likewise.
9045 * substring-locations.h (format_warning_va): Add "fmt_label" and
9046 "param_label" params.
9047 (format_warning_n_va): Likewise.
9048 (format_warning_at_substring): Likewise.
9049 (format_warning_at_substring_n): Likewise.
9050 * toplev.c (general_init): Initialize global_dc->show_labels_p.
9052 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
9055 * builtins.c (expand_builtin_memcmp): Do not expand the call
9056 when overflow is detected.
9058 2018-08-15 Martin Sebor <msebor@redhat.com>
9060 PR tree-optimization/71625
9061 * config/aarch64/aarch64-builtins.c
9062 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9064 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
9066 * config/s390/s390.c (s390_reorg): Remove loop.
9068 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
9071 (darwin_function_switched_text_sections): Delete.
9072 * gcc/config/darwin.h
9073 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9075 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
9078 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9079 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
9080 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9082 2018-08-15 Martin Liska <mliska@suse.cz>
9084 PR tree-optimization/86925
9085 * predict.c (expr_expected_value_1): When taking
9086 later predictor, assign also probability.
9087 Use fold_build2_initializer_loc in order to fold
9088 the expression in -frounding-math.
9090 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
9092 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9094 (expand_vec_perm_1): Try the new method.
9096 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
9099 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9100 Check whether lra_live_max_point is 0 before dividing.
9102 2018-08-14 Martin Sebor <msebor@redhat.com>
9104 PR tree-optimization/86650
9105 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9106 (vrp_prop::check_mem_ref): Same.
9108 2018-08-13 Liu Hao <lh_mouse@126.com>
9110 * pretty-print.c (eat_esc_sequence): Swap the foreground and
9111 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9112 and clear it thereafter, as it only works for DBCS.
9114 2018-08-13 Liu Hao <lh_mouse@126.com>
9116 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9117 handle returned by _get_osf_handle().
9119 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9121 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9122 for folding vec_perm.
9124 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9126 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9127 Add support for gimple-folding of vec_pack() and vec_unpack()
9130 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9132 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9133 vec_xst variants to the list.
9134 (rs6000_gimple_fold_builtin): Add support for folding unaligned
9135 vector loads and stores.
9137 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
9139 * config.gcc (rs6000-ibm-aix4.x): Delete.
9140 (rs6000-ibm-aix5.1): Delete.
9141 (rs6000-ibm-aix5.2): Delete.
9142 (rs6000-ibm-aix5.3): Delete.
9143 * config/rs6000/aix43.h: Delete.
9144 * config/rs6000/aix51.h: Delete.
9145 * config/rs6000/aix52.h: Delete.
9146 * config/rs6000/t-aix43: Delete.
9148 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
9150 * config/s390/s390.c (s390_decompose_constant_pool_ref):
9152 (s390_decompose_address): Factor out constant pool ref
9155 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9157 * config/nds32/nds32-predicates.c
9158 (nds32_can_use_bclr_p): Change return type as bool.
9159 (nds32_can_use_bset_p): Ditto.
9160 (nds32_can_use_btgl_p): Ditto.
9161 (nds32_can_use_bitci_p): Ditto.
9162 * config/nds32/nds32-protos.h
9163 (nds32_can_use_bclr_p): Change declaration.
9164 (nds32_can_use_bset_p): Ditto.
9165 (nds32_can_use_btgl_p): Ditto.
9166 (nds32_can_use_bitci_p): Ditto.
9168 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9170 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9171 Support -msched-prolog-epilog option.
9172 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9174 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9176 * common/config/nds32/nds32-common.c
9177 (nds32_option_optimization_table): Enalbe -malways-align.
9179 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9181 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9183 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9184 OPT_misr_secure_ case.
9185 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9186 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9187 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9189 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9190 (nds32_isr_info): New field security_level.
9191 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9192 * config/nds32/nds32.md (return_internal): Consider critical attribute.
9193 * config/nds32/nds32.opt (misr-secure): New option.
9194 * config/nds32/nds32_init.inc: New file.
9195 * config/nds32/nds32_isr.h: New file.
9197 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
9199 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9200 Update comment for atomic instructions.
9201 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9202 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9204 (atomic_loaddi): Revise fence expansion to only emit fence prior to
9205 load for __ATOMIC_SEQ_CST model.
9206 (atomic_loaddi_1): Remove float register target.
9207 (atomic_storedi): Handle CONST_INT values.
9208 (atomic_storedi_1): Remove float register source. Add special case
9210 (memory_barrier): New expander and insn.
9212 2018-08-11 Jakub Jelinek <jakub@redhat.com>
9214 PR tree-optimization/86835
9215 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9216 new_stmt after def_gsi, make sure to insert new_square_stmt after
9217 that stmt, not 2 stmts before it.
9219 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
9222 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9225 2018-08-10 Martin Liska <mliska@suse.cz>
9228 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9230 * builtins.c (expand_builtin_expect_with_probability):
9232 (expand_builtin_expect_with_probability): New function.
9233 (build_builtin_expect_predicate): Add new argumnet probability
9234 for BUILT_IN_EXPECT_WITH_PROBABILITY.
9235 (fold_builtin_expect):
9238 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9239 * builtins.h (fold_builtin_expect): Set new argument.
9240 * doc/extend.texi: Document __builtin_expect_with_probability.
9241 * doc/invoke.texi: Likewise.
9242 * gimple-fold.c (gimple_fold_call): Pass new argument.
9243 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9244 also BUILT_IN_EXPECT_WITH_PROBABILITY.
9245 * predict.c (get_predictor_value): New function.
9246 (expr_expected_value): Add new argument probability. Assume
9247 that predictor and probability are always non-null.
9248 (expr_expected_value_1): Likewise. For __builtin_expect and
9249 __builtin_expect_with_probability set probability. Handle
9250 combination in binary expressions.
9251 (tree_predict_by_opcode): Simplify code by simply calling
9252 get_predictor_value.
9253 (pass_strip_predict_hints::execute): Add handling of
9254 BUILT_IN_EXPECT_WITH_PROBABILITY.
9255 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9257 * tree.h (DECL_BUILT_IN_P): New function.
9259 2018-08-10 Martin Liska <mliska@suse.cz>
9261 PR tree-optimization/85799
9262 * passes.def: Add argument for pass_strip_predict_hints.
9263 * predict.c (class pass_strip_predict_hints): Add new argument
9265 (strip_predictor_early): New function.
9266 (pass_strip_predict_hints::execute): Call the function to
9268 (strip_predict_hints): New function.
9269 * predict.def: Fix comment.
9271 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
9273 * Makefile.in: Clarify which tm.texi to copy over to assert the
9274 right to grant a GFDL license for all.
9276 2018-08-09 Jeff Law <law@redhat.com>
9278 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9281 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
9283 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9286 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9288 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9289 reductions for variable-length vectors.
9291 2018-08-09 David Malcolm <dmalcolm@redhat.com>
9294 * common.opt (fdiagnostics-show-line-numbers): New option.
9295 * diagnostic-show-locus.c (class layout): Add fields
9296 "m_show_line_numbers_p" and "m_linenum_width";
9297 (num_digits): New function.
9298 (test_num_digits): New function.
9299 (layout::layout): Initialize new fields. Update m_x_offset
9300 logic to handle any left margin.
9301 (layout::print_source_line): Print line number when requested.
9302 (layout::start_annotation_line): New member function.
9303 (layout::print_annotation_line): Call it.
9304 (layout::print_leading_fixits): Likewise.
9305 (layout::print_trailing_fixits): Likewise. Update calls to
9306 move_to_column for new parameter.
9307 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9308 it to potentially call start_annotation_line.
9309 (layout::show_ruler): Call start_annotation_line.
9310 (selftest::test_line_numbers_multiline_range): New selftest.
9311 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9312 and selftest::test_line_numbers_multiline_range.
9313 * diagnostic.c (diagnostic_initialize): Initialize
9314 show_line_numbers_p.
9315 * diagnostic.h (struct diagnostic_context): Add field
9316 "show_line_numbers_p".
9317 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9318 -fno-diagnostics-show-line-numbers.
9319 * dwarf2out.c (gen_producer_string): Add
9320 OPT_fdiagnostics_show_line_numbers to the ignored options.
9321 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9322 one setting" options.
9323 (append_compiler_options): Likewise to the dropped options.
9324 (append_diag_options): Likewise to the passed-on options.
9325 * opts.c (common_handle_option): Handle the new option.
9326 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9328 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
9330 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9331 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
9332 third argument of type "const signed char" to descriptions of
9333 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9334 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9335 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9336 __builtin_bcdsub_ov functions.
9338 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9340 PR tree-optimization/86858
9341 * tree-vect-loop.c (vect_is_simple_reduction): Restore
9342 flow_bb_inside_loop_p calls.
9344 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9346 PR tree-optimization/86871
9347 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9348 instead of gimple_assign_lhs.
9350 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
9353 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9354 register constraint to operand 0.
9355 (add<mode>3_carryinC): Likewise.
9356 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9358 2018-08-09 Martin Liska <mliska@suse.cz>
9361 * common.opt: Remove extra line.
9363 2018-08-09 Martin Liska <mliska@suse.cz>
9365 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9366 at the end of a line, make first letter capital and end up
9367 a sentence with a dot.
9368 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9369 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9370 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9371 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9372 (PARAM_MAX_ISL_OPERATIONS): Likewise.
9373 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9374 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9375 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9376 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9377 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9378 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9379 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9380 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9381 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9382 (PARAM_TREE_REASSOC_WIDTH): Likewise.
9383 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9384 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9385 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9387 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
9390 * config/s390/s390.c (s390_option_override_internal): Reduce the
9391 stack-clash-protection-probe-interval param if it would be too big
9394 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
9397 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9398 * config/m68k/m68k.c (handle_move_double): Don't call
9399 m68k_final_prescan_insn.
9400 (m68k_adjust_decorated_operand): Renamed from
9401 m68k_final_prescan_insn, remove first and third operand and
9403 (print_operand): Call it.
9404 (print_operand_address): Call it.
9406 2018-08-08 Nathan Sidwell <nathan@acm.org>
9408 * diagnostic.c (diagnostic_report_current_module): Use
9409 linemap_included_from & linemap_included_from_linemap.
9411 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
9413 * config/aarch64/aarch64-cores.def: Add phecda core.
9414 * config/aarch64/aarch64-tune.md: Regenerate.
9415 * doc/invoke.texi: Add phecda core.
9417 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
9420 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9422 * config/s390/s390.md ("movti"): Add more alternatives for
9423 constant to GPR copies.
9425 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
9427 * config/s390/s390.c: Fix whitespace damage throughout the file.
9428 * config/s390/s390.h: Likewise.
9429 * config/s390/tpf.h: Likewise.
9431 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
9433 * config/s390/s390.c (s390_loadrelative_operand_p):
9434 Remove TARGET_CPU_ZARCH usages.
9435 (s390_rtx_costs): Likewise.
9436 (s390_legitimate_constant_p): Likewise.
9437 (s390_cannot_force_const_mem): Likewise.
9438 (legitimate_reload_constant_p): Likewise.
9439 (s390_preferred_reload_class): Likewise.
9440 (legitimize_pic_address): Likewise.
9441 (legitimize_tls_address): Likewise.
9442 (s390_split_branches): Removed.
9443 (s390_add_execute): Removed.
9444 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9445 (s390_mainpool_start): Likewise.
9446 (s390_mainpool_finish): Likewise.
9447 (s390_mainpool_cancel): Removed.
9448 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9449 (s390_chunkify_cancel): Likewise.
9450 (s390_return_addr_rtx): Likewise.
9451 (s390_register_info): Remove split_branches_pending_p uages.
9452 (s390_optimize_register_info): Likewise.
9453 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9454 split_branches_pending_p usages.
9455 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9456 (s390_load_got): Likewise.
9457 (s390_expand_split_stack_prologue): Likewise.
9458 (output_asm_nops): Likewise.
9459 (s390_function_profiler): Likewise.
9460 (s390_emit_call): Likewise.
9461 (s390_conditional_register_usage): Likewise.
9462 (s390_optimize_prologue): Likewise.
9463 (s390_reorg): Remove TARGET_CPU_ZARCH and
9464 split_branches_pending_p usages.
9465 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9467 (s390_output_indirect_thunk_function): Likewise.
9468 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9469 (TARGET_CPU_ZARCH_P): Removed.
9470 (struct machine_function): Remove split_branches_pending_p.
9471 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9473 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
9475 * common/config/s390/s390-common.c (processor_flags_table):
9477 * config.gcc: Remove with_arch/with_tune support.
9478 * config/s390/2064.md: Remove cpu attribute comparisons.
9479 * config/s390/driver-native.c (s390_host_detect_local_cpu):
9481 * config/s390/linux.h (ASM_SPEC):
9482 Remove -march support.
9483 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9484 Use a table to get an arch level.
9485 * config/s390/s390-opts.h (enum processor_type):
9487 * config/s390/s390.c
9488 (processor_table): Remove entries, add arch_level values.
9489 (s390_issue_rate): Remove cases.
9490 (s390_option_override): Adjust
9491 s390_option_override_internal() call.
9492 (s390_option_override_internal): Remove deprecation warning.
9493 (s390_valid_target_attribute_tree): Adjust
9494 s390_option_override_internal() call.
9495 * config/s390/s390.h (struct s390_processor):
9496 Share with s390-c.c, add arch_level field.
9497 * config/s390/s390.md:
9498 Remove occurrences in cpu attribute.
9499 * config/s390/s390.opt: Remove -march/-mtune support.
9500 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9501 * doc/invoke.texi: Remove deprecation warning.
9503 2018-08-08 Luis Machado <luis.machado@linaro.org>
9505 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9507 (qdf24xx_tunings): Set vector cost structure to
9508 qdf24xx_vector_cost.
9510 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9511 <register_sextend>: Set to 3.
9513 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
9516 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9517 * config/aarch64/aarch64-simd.md
9518 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9519 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9520 (aarch64_frecpx<mode>): ...this new pattern.
9521 * config/aarch64/aarch64-simd-builtins.def: Remove comment
9522 about aarch64_frecp<FRECP:frecp_suffix><mode>.
9524 2018-08-07 Martin Liska <mliska@suse.cz>
9527 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9528 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9529 * predict.def (PRED_MALLOC_NONNULL): New predictor.
9530 * doc/extend.texi: Document that malloc attribute adds
9533 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
9536 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9537 Define to speculation_safe_value_not_needed.
9539 2018-08-06 Jeff Law <law@redhat.com>
9541 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9542 the vr_values instance to cprop_into_stmt.
9543 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9544 (cprop_operand): Also query EVRP to determine if OP is a constant.
9546 2018-08-06 Nathan Sidwell <nathan@acm.org>
9548 * diagnostic.c (diagnostic_report_current_module): Reroll
9549 included-at loop. Translate text.
9551 2018-08-06 David Malcolm <dmalcolm@redhat.com>
9553 * function-tests.c (selftest::test_expansion_to_rtl): Call
9554 free_after_compilation.
9556 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9558 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9560 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
9562 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9563 loops with memory block operations from getting unrolled.
9565 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
9568 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9569 Define to speculation_safe_value_not_needed.
9571 2018-08-06 Jeff Law <law@redhat.com>
9573 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9576 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
9579 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9580 with all enabled __intN types.
9582 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9584 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9586 * alias.c (record_set): Check for clobber high.
9587 * cfgexpand.c (expand_gimple_stmt): Likewise.
9588 * combine-stack-adj.c (single_set_for_csa): Likewise.
9589 * combine.c (find_single_use_1): Likewise.
9590 (set_nonzero_bits_and_sign_copies): Likewise.
9591 (get_combine_src_dest): Likewise.
9592 (is_parallel_of_n_reg_sets): Likewise.
9593 (try_combine): Likewise.
9594 (record_dead_and_set_regs_1): Likewise.
9595 (reg_dead_at_p_1): Likewise.
9596 (reg_dead_at_p): Likewise.
9597 * dce.c (deletable_insn_p): Likewise.
9598 (mark_nonreg_stores_1): Likewise.
9599 (mark_nonreg_stores_2): Likewise.
9600 * df-scan.c (df_find_hard_reg_defs): Likewise.
9601 (df_uses_record): Likewise.
9602 (df_get_call_refs): Likewise.
9603 * dwarf2out.c (mem_loc_descriptor): Likewise.
9604 * haifa-sched.c (haifa_classify_rtx): Likewise.
9605 * ira-build.c (create_insn_allocnos): Likewise.
9606 * ira-costs.c (scan_one_insn): Likewise.
9607 * ira.c (equiv_init_movable_p): Likewise.
9608 (rtx_moveable_p): Likewise.
9609 (interesting_dest_for_shprep): Likewise.
9610 * jump.c (mark_jump_label_1): Likewise.
9611 * postreload-gcse.c (record_opr_changes): Likewise.
9612 * postreload.c (reload_cse_simplify): Likewise.
9613 (struct reg_use): Add source expr.
9614 (reload_combine): Check for clobber high.
9615 (reload_combine_note_use): Likewise.
9616 (reload_cse_move2add): Likewise.
9617 (move2add_note_store): Likewise.
9618 * print-rtl.c (print_pattern): Likewise.
9619 * recog.c (decode_asm_operands): Likewise.
9620 (store_data_bypass_p): Likewise.
9621 (if_test_bypass_p): Likewise.
9622 * regcprop.c (kill_clobbered_value): Likewise.
9623 (kill_set_value): Likewise.
9624 * reginfo.c (reg_scan_mark_refs): Likewise.
9625 * reload1.c (maybe_fix_stack_asms): Likewise.
9626 (eliminate_regs_1): Likewise.
9627 (elimination_effects): Likewise.
9628 (mark_not_eliminable): Likewise.
9629 (scan_paradoxical_subregs): Likewise.
9630 (forget_old_reloads_1): Likewise.
9631 * reorg.c (find_end_label): Likewise.
9632 (try_merge_delay_insns): Likewise.
9633 (redundant_insn): Likewise.
9634 (own_thread_p): Likewise.
9635 (fill_simple_delay_slots): Likewise.
9636 (fill_slots_from_thread): Likewise.
9637 (dbr_schedule): Likewise.
9638 * resource.c (update_live_status): Likewise.
9639 (mark_referenced_resources): Likewise.
9640 (mark_set_resources): Likewise.
9641 * rtl.c (copy_rtx): Likewise.
9642 * rtlanal.c (reg_referenced_p): Likewise.
9643 (single_set_2): Likewise.
9644 (noop_move_p): Likewise.
9645 (note_stores): Likewise.
9646 * sched-deps.c (sched_analyze_reg): Likewise.
9647 (sched_analyze_insn): Likewise.
9649 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9651 * cse.c (invalidate_reg): New function extracted from...
9652 (invalidate): ...here.
9653 (canonicalize_insn): Check for clobber high.
9654 (invalidate_from_clobbers): invalidate clobber highs.
9655 (invalidate_from_sets_and_clobbers): Likewise.
9656 (count_reg_usage): Check for clobber high.
9657 (insn_live_p): Likewise.
9658 * cselib.c (cselib_expand_value_rtx_1):Likewise.
9659 (cselib_invalidate_regno): Check for clobber in setter.
9660 (cselib_invalidate_rtx): Pass through setter.
9661 (cselib_invalidate_rtx_note_stores):
9662 (cselib_process_insn): Check for clobber high.
9663 * cselib.h (cselib_invalidate_rtx): Add operand.
9665 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9667 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
9668 (mark_not_eliminable): Likewise.
9669 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
9670 * lra-lives.c (process_bb_lives): Check for clobber high.
9671 * lra.c (new_insn_reg): Remember clobber highs.
9672 (collect_non_operand_hard_regs): Check for clobber high.
9673 (lra_set_insn_recog_data): Likewise.
9674 (add_regs_to_insn_regno_info): Likewise.
9675 (lra_update_insn_regno_info): Likewise.
9677 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9679 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
9680 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
9682 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9684 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
9685 (copy_insn_1): Likewise.
9686 (gen_hard_reg_clobber_high): New gen function.
9687 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
9688 * genemit.c (gen_exp): Likewise.
9689 (gen_emit_seq): Pass through info.
9690 (gen_insn): Check for CLOBBER_HIGH.
9691 (gen_expand): Pass through info.
9692 (gen_split): Likewise.
9693 (output_add_clobbers): Likewise.
9694 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
9695 (remove_clobbers): Likewise.
9696 * rtl.h (gen_hard_reg_clobber_high): New declaration.
9698 2018-08-06 Alan Hayward <alan.hayward@arm.com>
9700 * doc/rtl.texi (clobber_high): Add.
9701 (parallel): Add in clobber high
9702 * rtl.c (rtl_check_failed_code3): Add function.
9703 * rtl.def (CLOBBER_HIGH): Add expression.
9704 * rtl.h (RTL_CHECKC3): Add macro.
9705 (rtl_check_failed_code3): Add declaration.
9706 (XC3EXP): Add macro.
9708 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
9711 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
9712 cfun->machine->max_used_stack_alignment if needed.
9714 2018-08-04 Martin Sebor <msebor@redhat.com>
9716 PR tree-optimization/86571
9717 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
9720 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
9723 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9726 2018-08-03 Jeff Law <law@redhat.com>
9729 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9730 Define to speculation_safe_value_not_needed.
9732 2018-08-03 David Malcolm <dmalcolm@redhat.com>
9734 * doc/gcov.texi (-x): Remove duplicate "to".
9735 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
9736 (-Wif-not-aligned): Remove duplicate "is".
9737 (-flto): Remove duplicate "the".
9738 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
9739 duplicate "v5.00.b".
9740 (MSP430 Options): Remove duplicate "and" from the description
9741 of "-mgprel-sec=regexp".
9742 (x86 Options): Remove duplicate copies of "vmldLog102" and
9743 vmlsLog104 from description of "-mveclibabi=type".
9745 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
9747 * internal-fn.h (first_commutative_argument): Declare.
9748 * internal-fn.c (first_commutative_argument): New function.
9749 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
9750 restrictions for pattern statements. Use first_commutative_argument
9751 to look for commutative operands in calls to internal functions.
9753 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
9755 * Makefile.in (wide-int-range.o): New.
9756 * tree-vrp.c: Move all the wide_int_* functions to...
9757 * wide-int-range.cc: ...here.
9758 * tree-vrp.h: Move all the wide_int_* prototypes to...
9759 * wide-int-range.h: ...here.
9761 2018-08-03 Tom de Vries <tdevries@suse.de>
9763 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
9765 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
9766 * except.c (output_function_exception_table): Do early exit if
9767 targetm_common.except_unwind_info (&global_options) == UI_NONE.
9769 2018-08-03 Martin Liska <mliska@suse.cz>
9771 * predict.c (dump_prediction): Change to 2 digits
9774 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
9776 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
9778 (falkor_am_1_gtov_gtov): ... a new insn reservation.
9780 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
9782 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
9783 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
9784 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
9785 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
9786 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
9788 2018-08-02 David Malcolm <dmalcolm@redhat.com>
9790 * diagnostic-show-locus.c (diagnostic_show_locus): Use
9791 pp_take_prefix when saving the existing prefix.
9792 * diagnostic.c (diagnostic_append_note): Likewise.
9793 * langhooks.c (lhd_print_error_function): Likewise.
9794 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
9795 param's type. Free the existing prefix.
9796 (pp_take_prefix): New function.
9797 (pretty_printer::pretty_printer): Drop the prefix parameter.
9798 Rename the length parameter to match the comment.
9799 (pretty_printer::~pretty_printer): Free the prefix.
9800 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
9802 (struct pretty_printer): Drop the "const" from "prefix" field's
9803 type and clarify memory management.
9804 (pp_set_prefix): Drop the "const" from the 2nd param.
9805 (pp_take_prefix): New decl.
9807 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
9809 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
9810 for word_mode_ok here instead of passing as argument.
9811 (expand_block_compare): Change select_block_compare_mode() call.
9812 (expand_strncmp_gpr_sequence): New function.
9813 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
9815 2018-08-02 Jeff Law <law@redhat.com>
9818 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9819 Define to speculation_safe_value_not_needed.
9822 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9823 Define to speculation_safe_value_not_needed.
9825 2018-08-02 Tom de Vries <tdevries@suse.de>
9828 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
9829 function. Return UI_TARGET unconditionally.
9830 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
9831 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
9833 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
9835 * genemit.c (print_overload_test): Fix typo.
9837 2018-08-02 Richard Biener <rguenther@suse.de>
9839 PR tree-optimization/86816
9840 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
9841 which checks for value availability before querying it.
9842 (gvn_uses_equal): Use it.
9843 (same_succ_hash): Likewise.
9844 (gimple_equal_p): Likewise.
9846 2018-08-02 Nick Clifton <nickc@redhat.com>
9849 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9850 Define to speculation_safe_value_not_needed.
9853 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9854 Define to speculation_safe_value_not_needed.
9857 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9858 Define to speculation_safe_value_not_needed.
9861 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9862 Define to speculation_safe_value_not_needed.
9865 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9866 Define to speculation_safe_value_not_needed.
9869 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9870 Define to speculation_safe_value_not_needed.
9873 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9874 Define to speculation_safe_value_not_needed.
9877 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9878 Define to speculation_safe_value_not_needed.
9881 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
9882 speculation_safe_value_not_needed.
9885 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
9886 to speculation_safe_value_not_needed.
9888 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
9890 * doc/md.texi: Expand the documentation of instruction names
9891 to mention port-local uses. Document '@' in pattern names.
9892 * read-md.h (overloaded_instance, overloaded_name): New structs.
9894 (md_reader::handle_overloaded_name): New member function.
9895 (md_reader::get_overloads): Likewise.
9896 (md_reader::m_first_overload): New member variable.
9897 (md_reader::m_next_overload_ptr): Likewise.
9898 (md_reader::m_overloads_htab): Likewise.
9899 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
9900 m_next_overload_ptr and m_overloads_htab.
9901 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
9902 (get_mode_token, get_code_token, get_int_token): New functions.
9903 (map_attr_string): Add an optional argument that passes back
9904 the associated iterator.
9905 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
9906 (md_reader::handle_overloaded_name, add_overload_instance): New
9908 (apply_iterators): Handle '@' names. Report an error if '@'
9909 is used without iterators.
9910 (initialize_iterators): Initialize the new iterator_group fields.
9911 * genopinit.c (handle_overloaded_code_for)
9912 (handle_overloaded_gen): New functions.
9913 (main): Use them to print declarations of maybe_code_for_* and
9914 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
9915 * genemit.c (print_overload_arguments, print_overload_test)
9916 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
9917 (main): Use it to print definitions of maybe_code_for_* and
9918 maybe_gen_* functions.
9919 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
9920 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
9921 instead of explicit mode checks.
9922 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
9923 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
9924 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
9925 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
9926 (aarch64_expand_compare_and_swap): Likewise
9927 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
9928 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
9929 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
9930 (aarch64_constant_pool_reload_icode): Delete.
9931 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
9932 instead of aarch64_constant_pool_reload_icode. Use
9933 code_for_aarch64_reload_mov instead of explicit mode checks.
9934 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
9935 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
9936 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
9937 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
9938 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
9939 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
9940 (aarch64_atomic_load_op_code): Delete.
9941 (aarch64_emit_atomic_load_op): Likewise.
9942 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
9943 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
9944 instead of aarch64_emit_atomic_load_op.
9945 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
9946 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
9947 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
9948 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
9949 character before the pattern name.
9950 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
9951 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
9952 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
9953 (aarch64_frecps<mode>): Likewise.
9954 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
9955 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
9956 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
9957 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
9958 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
9960 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
9962 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
9963 Allow HFmode constants if TARGET_FP_F16INST.
9965 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
9968 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
9969 No longer check last store for clobber of address register.
9971 2018-08-02 Martin Liska <mliska@suse.cz>
9973 PR gcov-profile/86817
9974 * gcov.c (process_all_functions): New function.
9976 (process_file): Move functions processing to
9977 process_all_functions.
9979 2018-08-02 David Malcolm <dmalcolm@redhat.com>
9981 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
9982 "const" to the "gimple *" and "rtx_insn *" parameters.
9983 * dumpfile.h (dump_user_location_t::dump_user_location_t):
9985 (dump_location_t::dump_location_t): Likewise.
9987 2018-08-01 Martin Sebor <msebor@redhat.com>
9989 PR tree-optimization/86650
9990 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
9991 rather than a "gcall *". Directly pass the data of interest
9992 to percent_K_format, rather than building a temporary CALL_EXPR
9994 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
9995 (gimple_fold_builtin_strncat): Adjust.
9996 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
9997 gcall* argument with gimple*.
9998 * gimple-ssa-warn-restrict.c (check_call): Same.
9999 (wrestrict_dom_walker::before_dom_children): Same.
10000 (builtin_access::builtin_access): Same.
10001 (check_bounds_or_overlap): Same
10002 (maybe_diag_overlap): Same.
10003 (maybe_diag_offset_bounds): Same.
10004 * tree-diagnostic.c (default_tree_printer): Move usage of
10005 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10007 * tree-pretty-print.c (percent_K_format): Add argument.
10008 * tree-pretty-print.h: Add argument.
10009 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10010 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10011 (maybe_diag_stxncpy_trunc): Same.
10012 (handle_builtin_stxncpy): Same.
10013 (handle_builtin_strcat): Same.
10015 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10017 * match.pd: Optimise pointer range checks.
10019 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10021 PR tree-optimization/86758
10022 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10023 to remove pattern statements.
10025 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10027 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10028 result of dfs_enumerate_from when constructing stmt_vec_infos,
10029 instead of additionally calling get_loop_body.
10031 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10033 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10035 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10036 When creating an iv, assert that the step is not known to be zero.
10037 (vect_setup_realignment): Update call accordingly.
10038 * tree-vect-stmts.c (vectorizable_store): Likewise.
10039 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
10041 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10043 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10044 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10045 (vectorizable_reduction): Likewise.
10046 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10047 (vect_detect_hybrid_slp_stmts): Likewise.
10048 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10050 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
10052 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10053 (wide_int_set_zero_nonzero_bits): ...this.
10054 (zero_nonzero_bits_from_vr): Rename to...
10055 (vrp_set_zero_nonzero_bits): ...this.
10056 (extract_range_from_multiplicative_op_1): Abstract wide int
10058 (wide_int_range_multiplicative_op): ...here.
10059 (extract_range_from_binary_expr_1): Extract wide int binary
10060 operations into their own functions.
10061 (wide_int_range_lshift): New.
10062 (wide_int_range_can_optimize_bit_op): New.
10063 (wide_int_range_shift_undefined_p): New.
10064 (wide_int_range_bit_xor): New.
10065 (wide_int_range_bit_ior): New.
10066 (wide_int_range_bit_and): New.
10067 (wide_int_range_trunc_mod): New.
10068 (extract_range_into_wide_ints): New.
10069 (vrp_shift_undefined_p): New.
10070 (extract_range_from_multiplicative_op): New.
10071 (vrp_can_optimize_bit_op): New.
10072 * tree-vrp.h (value_range::dump): New.
10073 (wide_int_range_multiplicative_op): New.
10074 (wide_int_range_lshift):New.
10075 (wide_int_range_shift_undefined_p): New.
10076 (wide_int_range_bit_xor): New.
10077 (wide_int_range_bit_ior): New.
10078 (wide_int_range_bit_and): New.
10079 (wide_int_range_trunc_mod): New.
10080 (zero_nonzero_bits_from_bounds): Rename to...
10081 (wide_int_set_zero_nonzero_bits): ...this.
10082 (zero_nonzero_bits_from_vr): Rename to...
10083 (vrp_set_zero_nonzero_bits): ...this.
10084 (range_easy_mask_min_max): Rename to...
10085 (wide_int_range_can_optimize_bit_op): this.
10086 * vr-values.c (simplify_bit_ops_using_ranges): Rename
10087 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10089 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10091 * tree-vectorizer.h (vect_orig_stmt): New function.
10092 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10093 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10094 (vect_create_epilog_for_reduction): Likewise.
10095 (vectorizable_live_operation): Likewise.
10096 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10097 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10098 * tree-vect-stmts.c (vectorizable_call): Likewise.
10099 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10101 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10103 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10105 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10106 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10107 (vect_transform_loop): Likewise.
10108 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10110 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10112 * tree-vectorizer.h (vect_schedule_slp): Return void.
10113 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10114 (vect_schedule_slp): Likewise.
10116 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10118 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10120 (vect_transform_loop): Update calls accordingly. Schedule SLP
10121 instances before the main loop, if any exist.
10123 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10125 PR tree-optimization/86749
10126 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10127 If the lhs is used in a COND_EXPR, check that it is being used
10128 as the "then" or "else" value.
10130 2018-08-01 Tom de Vries <tdevries@suse.de>
10133 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10134 speculation_safe_value_not_needed.
10136 2018-08-01 Richard Biener <rguenther@suse.de>
10138 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10139 as base and offset.
10141 2018-08-01 Martin Liska <mliska@suse.cz>
10143 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10144 format how successful transformation is dumped.
10145 (gimple_mod_pow2_value_transform): Likewise.
10146 (gimple_mod_subtract_transform): Likewise.
10147 (gimple_stringops_transform): Likewise.
10149 2018-08-01 Martin Liska <mliska@suse.cz>
10151 PR value-prof/35543
10152 * value-prof.c (interesting_stringop_to_profile_p):
10153 Simplify the code and add BUILT_IN_MEMMOVE.
10154 (gimple_stringops_transform): Likewise.
10156 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
10158 * config/aarch64/aarch64-simd.md
10159 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10160 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10161 use GPI iterator instead of SI mode.
10163 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
10165 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10166 rs6000_speculation_barrier.
10167 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10168 new barrier pattern name.
10170 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
10172 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10173 (speculation_barrier): New insn.
10175 2018-08-01 Richard Biener <rguenther@suse.de>
10178 * graphite.h: Include isl/id.h and isl/space.h to allow build
10181 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
10184 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10185 mode for COFF targets.
10186 * defaults.h (TARGET_COFF): Define.
10187 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10188 TARGET_COFF): Define.
10189 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10190 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10191 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10193 2018-07-31 Alexandre Oliva <oliva@adacore.com>
10195 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10196 * gimple-streamer-out.c (output_bb): Save it.
10197 * lto-streamer-in.c (input_struct_function_base): Restore
10198 instance discriminator if available. Create map on demand.
10199 * lto-streamer-out.c (output_struct_function_base): Save it if
10201 * final.c (decl_to_instance_map): Document LTO strategy.
10203 2018-07-31 Alexandre Oliva <oliva@adacore.com>
10204 Olivier Hainque <hainque@adacore.com>
10206 * debug.h (decl_to_instance_map_t): New type.
10207 (decl_to_instance_map): Declare.
10208 (maybe_create_decl_to_instance_map): New inline function.
10209 * final.c (bb_discriminator, last_bb_discriminator): New statics,
10210 to track basic block discriminators.
10211 (final_start_function_1): Initialize them.
10212 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10214 (decl_to_instance_map): New variable.
10215 (map_decl_to_instance, maybe_set_discriminator): New functions.
10216 (notice_source_line): Set discriminator.
10218 2018-07-31 Ian Lance Taylor <iant@golang.org>
10220 * targhooks.c (default_have_speculation_safe_value): Add
10223 2018-07-31 David Malcolm <dmalcolm@redhat.com>
10225 * dump-context.h: Include "pretty-print.h".
10226 (dump_context::refresh_dumps_are_enabled): New decl.
10227 (dump_context::emit_item): New decl.
10228 (class dump_context): Add fields "m_test_pp" and
10230 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10231 (temp_dump_context::get_dumped_text): New decl.
10232 (class temp_dump_context): Add field "m_pp".
10233 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10234 (dump_context::refresh_dumps_are_enabled): ...and add a test for
10236 (set_dump_file): Update for above change.
10237 (set_alt_dump_file): Likewise.
10238 (dump_loc): New overload, taking a pretty_printer *.
10239 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
10240 to any test pretty-printer.
10241 (make_item_for_dump_gimple_stmt): New function, adapted from
10242 optinfo::add_gimple_stmt.
10243 (dump_context::dump_gimple_stmt): Call it, and use the result,
10244 eliminating the direct usage of dump_file and alt_dump_file in
10245 favor of indirectly using them via emit_item.
10246 (make_item_for_dump_gimple_expr): New function, adapted from
10247 optinfo::add_gimple_expr.
10248 (dump_context::dump_gimple_expr): Call it, and use the result,
10249 eliminating the direct usage of dump_file and alt_dump_file in
10250 favor of indirectly using them via emit_item.
10251 (make_item_for_dump_generic_expr): New function, adapted from
10253 (dump_context::dump_generic_expr): Call it, and use the result,
10254 eliminating the direct usage of dump_file and alt_dump_file in
10255 favor of indirectly using them via emit_item.
10256 (make_item_for_dump_printf_va): New function, adapted from
10257 optinfo::add_printf_va.
10258 (make_item_for_dump_printf): New function.
10259 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10260 and use the result, eliminating the direct usage of dump_file and
10261 alt_dump_file in favor of indirectly using them via emit_item.
10262 (make_item_for_dump_dec): New function.
10263 (dump_context::dump_dec): Call it, and use the result,
10264 eliminating the direct usage of dump_file and alt_dump_file in
10265 favor of indirectly using them via emit_item.
10266 (make_item_for_dump_symtab_node): New function, adapted from
10267 optinfo::add_symtab_node.
10268 (dump_context::dump_symtab_node): Call it, and use the result,
10269 eliminating the direct usage of dump_file and alt_dump_file in
10270 favor of indirectly using them via emit_item.
10271 (dump_context::begin_scope): Reimplement, avoiding direct usage
10272 of dump_file and alt_dump_file in favor of indirectly using them
10274 (dump_context::emit_item): New member function.
10275 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10276 Set up test pretty-printer on the underlying context. Call
10277 refresh_dumps_are_enabled.
10278 (temp_dump_context::~temp_dump_context): Call
10279 refresh_dumps_are_enabled.
10280 (temp_dump_context::get_dumped_text): New member function.
10281 (selftest::verify_dumped_text): New function.
10282 (ASSERT_DUMPED_TEXT_EQ): New macro.
10283 (selftest::test_capture_of_dump_calls): Run all tests twice, with
10284 and then without optinfo enabled. Add uses of
10285 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
10286 * dumpfile.h: Update comment for the dump_* API.
10287 * optinfo-emit-json.cc
10288 (selftest::test_building_json_from_dump_calls): Update for new
10289 param for temp_dump_context ctor.
10290 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10291 and "m_owned" field.
10292 (optinfo_item::~optinfo_item): Likewise.
10293 (optinfo::add_item): New member function.
10294 (optinfo::emit): Update comment.
10295 (optinfo::add_string): Delete.
10296 (optinfo::add_printf): Delete.
10297 (optinfo::add_printf_va): Delete.
10298 (optinfo::add_gimple_stmt): Delete.
10299 (optinfo::add_gimple_expr): Delete.
10300 (optinfo::add_tree): Delete.
10301 (optinfo::add_symtab_node): Delete.
10302 (optinfo::add_dec): Delete.
10303 * optinfo.h (class dump_context): New forward decl.
10304 (optinfo::add_item): New decl.
10305 (optinfo::add_string): Delete.
10306 (optinfo::add_printf): Delete.
10307 (optinfo::add_printf_va): Delete.
10308 (optinfo::add_gimple_stmt): Delete.
10309 (optinfo::add_gimple_expr): Delete.
10310 (optinfo::add_tree): Delete.
10311 (optinfo::add_symtab_node): Delete.
10312 (optinfo::add_dec): Delete.
10313 (optinfo::add_poly_int): Delete.
10314 (optinfo_item::optinfo_item): Remove "owned" param.
10315 (class optinfo_item): Remove field "m_owned".
10317 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10319 PR middle-end/86705
10320 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10321 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10322 requested variable alignment.
10323 (expand_one_ssa_partition): Likewise.
10324 (expand_one_var): Likewise.
10326 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10328 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10329 to speculation_safe_value_not_needed.
10331 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10333 * targhooks.h (speculation_safe_value_not_needed): New prototype.
10334 * targhooks.c (speculation_safe_value_not_needed): New function.
10335 * target.def (have_speculation_safe_value): Update documentation.
10336 * doc/tm.texi: Regenerated.
10338 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10340 * config/aarch64/iterators.md (ALLI_TI): New iterator.
10341 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10343 (despeculate_copy<ALLI:mode>_insn): New insn.
10344 (despeculate_copyti_insn): New insn.
10345 (despeculate_simple<ALLI:mode>): New insn
10346 (despeculate_simpleti): New insn.
10347 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10349 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10350 aarch64_speculation_safe_value.
10351 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10353 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10355 * config/aarch64/aarch64-speculation.cc: New file.
10356 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10357 before pass_reorder_blocks.
10358 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10360 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10361 X14 and X15 when tracking speculation.
10362 * config/aarch64/aarch64.md (register name constants): Add
10363 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10364 (unspec): Add UNSPEC_SPECULATION_TRACKER.
10365 (speculation_barrier): New insn attribute.
10366 (cmp<mode>): Allow SP in comparisons.
10367 (speculation_tracker): New insn.
10368 (speculation_barrier): Add speculation_barrier attribute.
10369 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10370 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10371 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10373 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10375 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10376 aarch64_track_speculation is true.
10377 (tb<optab><mode>1): Likewise.
10378 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10379 generate CB[N]Z when tracking speculation.
10380 (aarch64_split_compare_and_swap): Likewise.
10381 (aarch64_split_atomic_op): Likewise.
10383 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10385 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10387 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10389 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10390 (speculation_barrier): New insn.
10392 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10394 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10395 * config/arm/arm.md (speculation_barrier): New expand.
10396 (speculation_barrier_insn): New pattern.
10398 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10400 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10401 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10402 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10403 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10405 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10406 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10407 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10408 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10409 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10410 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10411 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10412 * builtins.c (expand_speculation_safe_value): New function.
10413 (expand_builtin): Call it.
10414 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10415 * doc/extend.texi: Document __builtin_speculation_safe_value.
10416 * doc/md.texi: Document "speculation_barrier" pattern.
10417 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10418 TARGET_HAVE_SPECULATION_SAFE_VALUE.
10419 * doc/tm.texi: Regenerated.
10420 * target.def (have_speculation_safe_value, speculation_safe_value): New
10422 * targhooks.c (default_have_speculation_safe_value): New function.
10423 (default_speculation_safe_value): New function.
10424 * targhooks.h (default_have_speculation_safe_value): Add prototype.
10425 (default_speculation_safe_value): Add prototype.
10427 2018-07-31 David Malcolm <dmalcolm@redhat.com>
10429 * dump-context.h (dump_context::dump_loc): New decl.
10430 * dumpfile.c (dump_context::dump_loc): New member function.
10431 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10432 and dump_gimple_stmt.
10433 (dump_context::dump_gimple_expr_loc): Likewise, using
10435 (dump_context::dump_generic_expr_loc): Likewise, using
10437 (dump_context::dump_printf_loc_va): Likewise, using
10439 (dump_context::begin_scope): Explicitly using the global function
10440 "dump_loc", rather than the member function.
10442 2018-07-31 Martin Sebor <msebor@redhat.com>
10444 PR tree-optimization/86741
10445 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10447 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
10449 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10450 depend on whether prefetch instructions will be emitted or not.
10451 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10452 will be emitted or not.
10453 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10454 (TARGET_SETMEM_PFD): New macros.
10456 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10458 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10459 (NULL_STMT_VEC_INFO): Delete.
10460 (stmt_vec_info::operator*): Likewise.
10461 (stmt_vec_info::operator gimple *): Likewise.
10462 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10463 of NULL_STMT_VEC_INFO.
10464 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10465 (vect_reassociating_reduction_p): Likewise.
10466 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10467 (vectorizable_store): Likewise.
10468 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10469 (vec_info::free_stmt_vec_infos): Likewise.
10471 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10473 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10474 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10476 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10478 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10479 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10480 (vec_info::free_stmt_vec_info): New private member functions.
10481 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10482 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10483 * tree-parloops.c (gather_scalar_reductions): Remove calls to
10484 set_stmt_vec_info_vec and free_stmt_vec_infos.
10485 * tree-vect-loop.c (_loop_vec_info): Remove call to
10486 set_stmt_vec_info_vec.
10487 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10488 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10489 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10490 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10491 (vec_info::free_stmt_vec_info): ...these new functions. Remove
10492 assignments in {vec_info::,}new_stmt_vec_info that are redundant
10493 with the clearing in the xcalloc.
10494 (stmt_vec_info_vec): Delete.
10495 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10496 (vectorize_loops): Likewise.
10497 (vec_info::~vec_info): Remove argument from call to
10498 free_stmt_vec_infos.
10499 (vec_info::add_stmt): Remove vinfo argument from call to
10502 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10504 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10505 rather than a gimple stmt.
10506 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
10507 information for pattern statements when passed the original
10508 statement; instead wait to be passed the pattern statement itself.
10509 Don't call set_vinfo_for_stmt here.
10510 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10511 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10512 stmt_vec_infos here.
10513 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10514 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10515 stmt_vec_infos entry.
10517 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10519 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10520 * tree-vectorizer.c (vec_info::replace_stmt): New function.
10521 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10522 * tree-vect-stmts.c (vectorizable_call): Likewise.
10523 (vectorizable_simd_clone_call): Likewise.
10525 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10527 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10528 * tree-vectorizer.c (vec_info::remove_stmt): New function.
10529 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10530 * tree-vect-loop.c (vect_transform_loop): Likewise.
10531 * tree-vect-slp.c (vect_schedule_slp): Likewise.
10532 * tree-vect-stmts.c (vect_remove_stores): Likewise.
10534 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10536 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10537 (vect_dr_stmt): Delete.
10538 * tree-vectorizer.c (vec_info::lookup_dr): New function.
10539 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10541 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10542 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10543 (vect_verify_datarefs_alignment, vect_peeling_supportable)
10544 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10545 (vect_analyze_data_refs): Likewise.
10546 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
10548 (vect_find_same_alignment_drs): Likewise.
10549 (vect_slp_analyze_node_dependences): Update calls accordingly.
10550 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
10551 instead of DR_VECT_AUX.
10552 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10553 of a vector data references. Use vec_info::lookup_dr instead of
10555 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10556 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
10557 instead of DR_VECT_AUX.
10559 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10561 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10563 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10565 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10566 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10567 (vect_gen_prolog_loop_niters): Likewise.
10569 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10571 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10572 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10573 (vect_known_alignment_in_bytes, vect_dr_behavior)
10574 (vect_get_scalar_dr_size): Take references as dr_vec_infos
10575 instead of data_references. Update calls to other routines for
10576 which the same change has been made.
10577 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10578 dr_vec_infos instead of stmt_vec_infos.
10579 (vect_analyze_data_ref_dependence): Update call accordingly.
10580 (vect_slp_analyze_data_ref_dependence)
10581 (vect_record_base_alignments): Use DR_VECT_AUX.
10582 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10583 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10584 (vector_alignment_reachable_p, vect_get_data_access_cost)
10585 (vect_peeling_supportable, vect_analyze_group_access_1)
10586 (vect_analyze_group_access, vect_analyze_data_ref_access)
10587 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10588 (vect_compile_time_alias, vect_small_gap_p)
10589 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
10590 (vect_supportable_dr_alignment): Take references as dr_vec_infos
10591 instead of data_references. Update calls to other routines for
10592 which the same change has been made.
10593 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
10594 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10595 (vect_slp_analyze_and_verify_node_alignment)
10596 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10597 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10598 (vect_setup_realignment): Use dr_vec_infos. Update calls after
10600 (_vect_peel_info::dr): Replace with...
10601 (_vect_peel_info::dr_info): ...this new field.
10602 (vect_peeling_hash_get_most_frequent)
10603 (vect_peeling_hash_choose_best_peeling): Update accordingly.
10604 (vect_peeling_hash_get_lowest_cost):
10605 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
10606 routines for which the same change has been made.
10607 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
10609 * tree-vect-loop-manip.c (get_misalign_in_elems)
10610 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
10612 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10613 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
10614 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
10615 (get_group_load_store_type, get_negative_load_store_type)
10616 (vect_get_data_ptr_increment, vectorizable_store)
10617 (vectorizable_load): Likewise.
10618 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
10619 Update calls to other routines for which the same change has been made.
10621 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10623 * tree-vectorizer.h (vec_info::move_dr): New member function.
10624 (dataref_aux): Rename to...
10625 (dr_vec_info): ...this and add "dr" and "stmt" fields.
10626 (_stmt_vec_info::dr_aux): Update accordingly.
10627 (_stmt_vec_info::data_ref_info): Delete.
10628 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
10629 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
10630 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
10631 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
10633 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
10634 (STMT_VINFO_DR_INFO): New macro.
10635 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
10636 (set_dr_misalignment): Update after rename of dataref_aux.
10637 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
10638 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
10639 initialization of STMT_VINFO_DATA_REF.
10640 * tree-vectorizer.c (vec_info::move_dr): New function.
10641 * tree-vect-patterns.c (vect_recog_bool_pattern)
10642 (vect_recog_mask_conversion_pattern)
10643 (vect_recog_gather_scatter_pattern): Use it.
10644 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
10645 the "dr" and "stmt" fields of dr_vec_info instead of
10646 STMT_VINFO_DATA_REF.
10648 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10650 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
10651 (is_pattern_stmt_p): Use it.
10652 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
10653 on pattern statements.
10655 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10657 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
10658 original stmt as a stmt_vec_info rather than a gimple stmt.
10659 (vect_pattern_recog_1): Take the statement directly as a
10660 stmt_vec_info, rather than via a gimple_stmt_iterator.
10661 Update call to vect_mark_pattern_stmts.
10662 (vect_pattern_recog): Update calls accordingly.
10664 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10666 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
10667 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
10668 a vect_def_type for the first argument.
10669 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
10670 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
10671 operand if it isn't defined by a vectorized statement.
10672 (vect_build_gather_load_calls): Remove the mask_dt argument and
10673 update calls to vect_get_vec_def_for_stmt_copy.
10674 (vectorizable_bswap): Likewise the dt argument.
10675 (vectorizable_call): Update calls to vectorizable_bswap and
10676 vect_get_vec_def_for_stmt_copy.
10677 (vectorizable_simd_clone_call, vectorizable_assignment)
10678 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
10679 (vectorizable_comparison): Update calls to
10680 vect_get_vec_def_for_stmt_copy.
10681 (vectorizable_store): Likewise. Remove now-unnecessary calls to
10682 vect_is_simple_use.
10683 (vect_get_loop_based_defs): Remove dt argument and update call
10684 to vect_get_vec_def_for_stmt_copy.
10685 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
10686 and vect_get_vec_def_for_stmt_copy.
10687 (vectorizable_load): Update calls to vect_build_gather_load_calls
10688 and vect_get_vec_def_for_stmt_copy.
10689 * tree-vect-loop.c (vect_create_epilog_for_reduction)
10690 (vectorizable_reduction, vectorizable_live_operation): Update calls
10691 to vect_get_vec_def_for_stmt_copy.
10693 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10695 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
10696 and gimple stmt arguments with a stmt_vec_info.
10697 (vect_record_base_alignments): Update calls accordingly.
10698 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
10699 and gimple stmt arguments with a stmt_vec_info.
10700 (vect_build_slp_tree_1): Remove vinfo argument and update call
10701 to vect_record_max_nunits.
10702 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
10703 and vect_record_max_nunits.
10705 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10707 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
10708 file and take a stmt_vec_info instead of a gimple stmt.
10709 (supportable_widening_operation, vect_finish_replace_stmt)
10710 (vect_finish_stmt_generation, vect_get_store_rhs)
10711 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
10712 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
10713 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
10714 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
10715 (vect_create_data_ref_ptr, bump_vector_ptr)
10716 (vect_permute_store_chain, vect_setup_realignment)
10717 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
10718 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
10719 (vectorizable_reduction, vectorizable_induction)
10720 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
10721 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
10722 than gimple stmts as arguments.
10723 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
10724 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
10725 (can_group_stmts_p, vect_check_gather_scatter)
10726 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
10727 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
10728 (vect_permute_load_chain, vect_shift_permute_load_chain)
10729 (vect_transform_grouped_load)
10730 (vect_record_grouped_load_vectors): Likewise.
10731 * tree-vect-loop.c (vect_fixup_reduc_chain)
10732 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
10733 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
10734 (vectorizable_reduction, vectorizable_induction)
10735 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
10736 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
10737 (vect_get_load_store_mask): Likewise.
10738 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
10739 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
10740 * tree-vect-stmts.c (vect_mark_relevant)
10741 (is_simple_and_all_uses_invariant)
10742 (exist_non_indexing_operands_for_use_p, process_use)
10743 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
10744 (vect_get_vec_def_for_operand, vect_get_vec_defs)
10745 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
10746 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
10747 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
10748 (get_negative_load_store_type, get_load_store_type)
10749 (vect_check_load_store_mask, vect_check_store_rhs)
10750 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
10751 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10752 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10753 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10754 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
10755 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
10756 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
10757 (supportable_widening_operation): Likewise.
10759 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10761 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
10762 a stmt_vec_info instead of a gcall.
10763 (vect_check_gather_scatter): Update call accordingly.
10764 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
10766 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
10767 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
10768 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
10769 instead of a gimple stmt.
10770 (vect_transform_loop): Update calls accordingly.
10771 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
10772 stmt_vec_infos instead of gimple stmts.
10773 (vect_analyze_slp_instance): Update use accordingly.
10774 * tree-vect-stmts.c (read_vector_array, write_vector_array)
10775 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
10776 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
10777 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
10778 (vect_gen_widened_results_half, vect_get_loop_based_defs)
10779 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
10780 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
10781 down to subroutines.
10783 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10785 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
10786 of the worklist from a vector of gimple stmts to a vector of
10788 * tree-vect-stmts.c (vect_mark_relevant, process_use)
10789 (vect_mark_stmts_to_be_vectorized): Likewise
10791 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10793 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
10794 statement before passing it to vect_analyze_stmt.
10795 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
10796 the chain of phi vector definitions. Track the exit phi via its
10798 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
10799 STMT_VINFO_REDUC_DEF.
10800 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
10801 stmt_vec_infos to handle the statement chains.
10802 (vect_get_slp_defs): Record the first statement in the node
10803 using a stmt_vec_info.
10804 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
10805 statements here and pass their stmt_vec_info down to subroutines.
10806 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
10807 down to vect_finish_stmt_generation.
10808 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
10809 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
10810 stmt_vec_infos to subroutines.
10811 (vect_remove_stores): Use stmt_vec_infos to handle the statement
10814 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10816 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
10817 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
10818 (vect_permute_store_chain, vect_setup_realignment)
10819 (vect_permute_load_chain, vect_shift_permute_load_chain)
10820 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
10821 stmts internally, and when passing values to other vectorizer routines.
10822 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10823 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
10824 (vect_analyze_loop_operations, get_initial_def_for_reduction)
10825 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
10826 (vectorizable_reduction, vectorizable_induction)
10827 (vectorizable_live_operation, vect_transform_loop_stmt)
10828 (vect_transform_loop): Likewise.
10829 * tree-vect-patterns.c (vect_reassociating_reduction_p)
10830 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
10831 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
10832 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10833 (vect_slp_analyze_node_operations_1): Likewise.
10834 * tree-vect-stmts.c (vect_mark_relevant, process_use)
10835 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
10836 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
10837 (vect_finish_stmt_generation_1, get_group_load_store_type)
10838 (get_load_store_type, vect_build_gather_load_calls)
10839 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
10840 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
10841 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
10842 (vectorizable_store, vectorizable_load, vectorizable_condition)
10843 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
10844 (supportable_widening_operation): Likewise.
10845 (vect_get_vector_types_for_stmt): Likewise.
10846 * tree-vectorizer.h (vect_dr_behavior): Likewise.
10848 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10850 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10851 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
10852 (vect_permute_store_chain, vect_permute_load_chain)
10853 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
10854 repeated stmt_vec_info lookups.
10855 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
10856 (vect_update_ivs_after_vectorizer): Likewise.
10857 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
10858 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
10859 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
10860 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10861 (vect_bb_slp_scalar_cost): Likewise.
10862 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
10864 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10866 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
10867 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
10868 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
10869 to get gassigns and gcalls, rather than operating on generc gimple
10871 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
10872 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
10873 (vectorizable_load, vect_analyze_stmt): Likewise.
10874 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
10876 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10878 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
10879 return stmt_vec_infos rather than gimple stmts. Do not accept
10881 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
10883 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10884 Update use of get_later_stmt.
10885 (vect_get_constant_vectors): Update call accordingly.
10886 (vect_schedule_slp_instance): Likewise
10887 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
10888 (vect_slp_analyze_instance_dependence): Likewise.
10889 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
10891 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10893 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
10894 (stmt_info_for_cost::stmt_info): ...this new field.
10895 (add_stmt_costs): Update accordingly.
10896 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
10897 (vect_get_known_peeling_cost): Likewise.
10898 (vect_estimate_min_profitable_iters): Likewise.
10899 * tree-vect-stmts.c (record_stmt_cost): Likewise.
10901 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10903 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
10904 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10905 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10907 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
10909 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10911 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
10912 a gimple stmt to a stmt_vec_info.
10913 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
10915 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10917 * tree-vectorizer.h (vec_info::grouped_stores): Change from
10918 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
10919 (_loop_vec_info::reduction_chains): Likewise.
10920 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
10922 * tree-vect-slp.c (vect_analyze_slp): Likewise.
10924 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10926 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
10927 a gimple stmt to a stmt_vec_info.
10928 (_stmt_vec_info::next_element): Likewise.
10929 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
10930 (vect_slp_analyze_and_verify_node_alignment)
10931 (vect_analyze_group_access_1, vect_analyze_group_access)
10932 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
10933 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
10934 (vect_supportable_dr_alignment): Update accordingly.
10935 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
10936 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
10937 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
10938 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
10939 * tree-vect-slp.c (vect_build_slp_tree_1)
10940 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
10941 (vect_split_slp_store_group, vect_analyze_slp_instance)
10942 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
10943 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
10944 (get_group_load_store_type, get_load_store_type)
10945 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
10946 (vect_transform_stmt, vect_remove_stores): Likewise.
10948 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10950 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
10951 than a gimple stmt.
10952 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
10953 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
10954 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
10955 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
10956 (vector_alignment_reachable_p, vect_get_data_access_cost)
10957 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
10958 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
10959 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
10960 (vect_analyze_group_access_1, vect_analyze_group_access)
10961 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
10962 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
10963 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
10964 result of vect_dr_stmt and use the stmt_vec_info instead of
10965 the associated gimple stmt.
10966 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10967 (vect_gen_prolog_loop_niters): Likewise.
10968 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10970 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10972 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
10973 to a vec<stmt_vec_info>.
10974 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
10975 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
10976 vec<stmt_vec_info>.
10977 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
10978 to a vec<stmt_vec_info>.
10979 (bst_traits::value_type, bst_traits::value_type): Likewise.
10980 (bst_traits::hash): Update accordingly.
10981 (vect_get_and_check_slp_defs): Change the stmts parameter from
10982 a vec<gimple *> to a vec<stmt_vec_info>.
10983 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
10984 (vect_build_slp_tree): Likewise.
10985 (vect_build_slp_tree_2): Likewise. Update uses of
10986 SLP_TREE_SCALAR_STMTS.
10987 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
10988 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
10989 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
10990 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
10991 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
10992 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
10993 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
10994 (vect_get_constant_vectors, vect_get_slp_defs)
10995 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
10996 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
10997 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
10998 instead of gimple stmts.
10999 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11000 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11001 (vect_slp_analyze_instance_dependence): Update uses of
11002 SLP_TREE_SCALAR_STMTS.
11003 (vect_slp_analyze_and_verify_node_alignment): Likewise.
11004 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11005 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11006 (get_initial_defs_for_reduction): Likewise.
11007 (vect_create_epilog_for_reduction): Likewise.
11008 (vectorize_fold_left_reduction): Likewise.
11009 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11010 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11011 (can_vectorize_live_stmts): Likewise.
11013 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11015 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11016 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11017 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11019 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11021 (gather_scalar_reductions): Update after above interface changes.
11022 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11023 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11025 (vect_force_simple_reduction): Likewise.
11026 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11027 LOOP_VINFO_REDUCTIONS.
11028 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11030 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11032 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11033 a gimple stmt to a stmt_vec_info.
11034 * tree-vect-loop.c (vect_active_double_reduction_p)
11035 (vect_force_simple_reduction, vectorizable_reduction): Update
11038 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11040 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11041 vec<gimple *> to a vec<stmt_vec_info>.
11042 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11043 the reduction_phis argument from a vec<gimple *> to a
11044 vec<stmt_vec_info>.
11045 (vectorizable_reduction): Likewise the phis local variable that
11046 is passed to vect_create_epilog_for_reduction. Update for new type
11047 of SLP_TREE_VEC_STMTS.
11048 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11049 (vectorizable_live_operation): Likewise.
11050 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11051 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11053 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11055 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11056 a gimple stmt to a stmt_vec_info.
11057 (vectorizable_condition, vectorizable_live_operation)
11058 (vectorizable_reduction, vectorizable_induction): Pass back the
11059 vectorized statement as a stmt_vec_info.
11060 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11061 use of STMT_VINFO_VEC_STMT.
11062 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11063 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11064 as stmt_vec_infos rather than gimple stmts.
11065 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11066 to a stmt_vec_info.
11067 (vectorizable_live_operation): Likewise.
11068 (vectorizable_reduction, vectorizable_induction): Likewise,
11069 updating use of STMT_VINFO_VEC_STMT.
11070 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11071 of STMT_VINFO_VEC_STMT.
11072 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11073 (vectorizable_simd_clone_call, vectorizable_conversion)
11074 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11075 (vectorizable_store, vectorizable_load, vectorizable_condition)
11076 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11077 from a gimple stmt to a stmt_vec_info.
11078 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
11079 pointer to a stmt_vec_info to the vectorizable_* routines.
11081 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11083 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11084 a gimple stmt to a stmt_vec_info.
11085 (is_pattern_stmt_p): Update accordingly.
11086 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11087 (vect_record_grouped_load_vectors): Likewise.
11088 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11089 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11090 (vect_model_reduction_cost): Likewise.
11091 (vect_create_epilog_for_reduction): Likewise.
11092 (vectorizable_reduction, vectorizable_induction): Likewise.
11093 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11094 Return the stmt_vec_info for the pattern statement.
11095 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11096 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11097 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11098 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11099 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11100 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11101 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11102 (free_stmt_vec_info, vect_is_simple_use): Likewise.
11104 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11106 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11107 (vect_finish_stmt_generation): Likewise.
11108 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11109 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11110 (vect_build_gather_load_calls): Use the return value of the above
11111 functions instead of a separate call to vinfo_for_stmt. Use narrow
11112 scopes for the input gimple stmt and wider scopes for the associated
11113 stmt_vec_info. Use vec_info::lookup_def when setting these
11114 stmt_vec_infos from an SSA_NAME definition.
11115 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11116 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11117 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11118 (vectorizable_store, vectorizable_load, vectorizable_condition)
11119 (vectorizable_comparison): Likewise.
11120 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11121 (vectorizable_reduction): Likewise.
11123 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11125 * tree-vectorizer.h (vect_is_simple_use): Add an optional
11126 stmt_vec_info * parameter before the optional gimple **.
11127 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11128 (process_use, vect_get_vec_def_for_operand_1): Update callers.
11129 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11130 * tree-vect-loop.c (vectorizable_reduction): Likewise.
11131 (vectorizable_live_operation): Likewise.
11132 * tree-vect-patterns.c (type_conversion_p): Likewise.
11133 (vect_look_through_possible_promotion): Likewise.
11134 (vect_recog_rotate_pattern): Likewise.
11135 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11137 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11139 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11140 a typedef to a wrapper class.
11141 (NULL_STMT_VEC_INFO): New macro.
11142 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11143 (stmt_vec_info::operator*): New function.
11144 (stmt_vec_info::operator gimple *): Likewise.
11145 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11146 (add_stmt_costs): Likewise.
11147 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11148 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11149 (vect_get_known_peeling_cost): Likewise.
11150 (vect_estimate_min_profitable_iters): Likewise.
11151 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11152 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11153 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11154 (vectorizable_store, free_stmt_vec_infos): Likewise.
11155 (new_stmt_vec_info): Change return type of xcalloc to
11158 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11160 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11161 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11162 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11163 a single_imm_use-based sequence.
11164 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11166 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11168 * tree-vectorizer.h (vec_info::lookup_def): Declare.
11169 * tree-vectorizer.c (vec_info::lookup_def): New function.
11170 * tree-vect-patterns.c (vect_get_internal_def): Use it.
11171 (vect_widened_op_tree): Likewise.
11172 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11173 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11174 (vectorizable_reduction): Likewise.
11175 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11177 (vect_is_slp_reduction): Update calls accordingly. Use
11178 vec_info::lookup_def.
11179 (vect_is_simple_reduction): Likewise
11180 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11182 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11184 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11185 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11186 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11188 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11189 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11190 (vect_update_vf_for_slp, vect_analyze_loop_operations)
11191 (vect_is_slp_reduction, vectorizable_induction)
11192 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11193 * tree-vect-patterns.c (vect_init_pattern_stmt):
11194 (vect_determine_min_output_precision_1, vect_determine_precisions)
11195 (vect_pattern_recog): Likewise.
11196 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11197 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11198 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11199 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11200 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11201 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
11202 info field from a loop to a loop_vec_info.
11204 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11206 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11207 (vec_info::add_stmt): Declare.
11208 * tree-vectorizer.c (vec_info::add_stmt): New function.
11209 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11210 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11211 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11212 (vectorizable_induction): Likewise.
11213 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11214 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11215 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11216 (vectorizable_load): Likewise.
11217 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11218 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11219 (vect_recog_gather_scatter_pattern): Likewise.
11220 (append_pattern_def_seq): Likewise. Remove a check that is
11221 performed by add_stmt itself.
11223 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11225 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11226 which make_ssa_name was called with new_stmt before new_stmt
11229 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11231 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11233 (vect_is_slp_reduction): ...here...
11234 (vect_is_simple_reduction): ...and here. Remove repetition of tests
11235 that are already known to be false.
11237 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11239 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11240 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
11241 STMT_VINFO_NUM_SLP_USES when it's true.
11242 (vect_free_slp_instance): Add a final_p parameter and pass it to
11243 vect_free_slp_tree.
11244 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11245 (vect_analyze_slp_instance): Likewise.
11246 (vect_slp_analyze_operations): Likewise.
11247 (vect_slp_analyze_bb_1): Likewise.
11248 * tree-vectorizer.c (vec_info): Likewise.
11249 * tree-vect-loop.c (vect_transform_loop): Likewise.
11251 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11253 * tree-vect-loop.c (vectorizable_reduction): Assert that the
11254 function is not called for second and subsequent members of
11257 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11259 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11260 cases for nested loops from here to ...
11261 (vect_create_epilog_for_reduction): ...here. Only call
11262 vect_is_simple_use for inner-loop reductions.
11264 2018-07-31 Martin Liska <mliska@suse.cz>
11266 PR gcov-profile/85338
11267 PR gcov-profile/85350
11268 PR gcov-profile/85372
11269 * profile.c (struct location_triplet): New.
11270 (struct location_triplet_hash): Likewise.
11271 (output_location): Do not output a BB that
11272 is already recorded for a line.
11273 (branch_prob): Use streamed_locations.
11275 2018-07-31 Martin Liska <mliska@suse.cz>
11277 PR gcov-profile/85370
11278 * coverage.c (coverage_begin_function): Do not mark target
11279 clones as artificial functions.
11281 2018-07-31 Martin Liska <mliska@suse.cz>
11283 PR gcov-profile/83813
11284 PR gcov-profile/84758
11285 PR gcov-profile/85217
11286 PR gcov-profile/85332
11287 * profile.c (branch_prob): Do not record GOTO expressions
11288 for GIMPLE statements which locations are already streamed.
11290 2018-07-31 Olivier Hainque <hainque@adacore.com>
11292 * gcc.c (handle_spec_function): Accept a soft_matched_part
11293 argument, as do_spec_1. Pass it down to ...
11294 (eval_spec_function): Accept a soft_matched_part argument,
11295 and pass it down to ...
11296 (do_spec_2): Accept a soft_matched_part argument, and pass
11297 it down to do_spec_1.
11298 (do_spec_1): Pass soft_matched_part to handle_spec_function.
11299 (handle_braces): Update call to handle_spec_function.
11300 (driver::set_up_specs): Update calls to do_spec_2.
11301 (compare_debug_dump_opt_spec_function): Likewise.
11302 (compare_debug_self_opt_spec_function): Likewise.
11304 2018-07-31 Olivier Hainque <hainque@adacore.com>
11306 * common.opt (nolibc): New option.
11307 * doc/invoke.texi (Link Options): Document it.
11308 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11309 * config/alpha/linux.h: Likewise.
11310 * config/arc/elf.h: Likewise.
11311 * config/arm/uclinux-elf.h: Likewise.
11312 * config/arm/unknown-elf.h: Likewise.
11313 * config/avr/avrlibc.h: Likewise.
11314 * config/bfin/bfin.h: Likewise.
11315 * config/bfin/linux.h: Likewise.
11316 * config/bfin/uclinux.h: Likewise.
11317 * config/darwin.h: Likewise.
11318 * config/darwin10.h: Likewise.
11319 * config/darwin12.h: Likewise.
11320 * config/gnu-user.h: Likewise.
11321 * config/lm32/uclinux-elf.h: Likewise.
11322 * config/pa/pa-hpux11.h: Likewise.
11323 * config/pa/pa64-hpux.h: Likewise.
11324 * config/sparc/sparc.h: Likewise.
11326 2018-07-31 Olivier Hainque <hainque@adacore.com>
11328 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11329 undefined variables.
11331 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
11334 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11335 instead of GEN_INT.
11337 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
11339 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11340 terminated string literal.
11342 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
11344 PR rtl-optimization/85160
11345 * combine.c (is_just_move): New function.
11346 (try_combine): Allow combining two instructions into two if neither of
11347 the original instructions was a move.
11349 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
11352 * doc/extend.texi (Global Register Variables): Discourage use of type
11354 (Local Register Variables): Likewise.
11356 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
11358 PR tree-optimization/86506
11359 * hwint.h (ceil_log2): Resync with hwint.c implementation.
11361 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
11364 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11365 hard_regno, make sure no insn between `from` and `to` clobbers it.
11367 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
11368 Tom de Vries <tdevries@suse.de>
11370 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11371 (PTX_DEFAULT_RUNTIME_DIM): ... this.
11372 (nvptx_goacc_validate_dims): Set default worker and gang dims to
11373 PTX_DEFAULT_RUNTIME_DIM.
11374 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11376 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
11378 * config/pa/pa.c (pa_output_addr_vec): Align address table.
11379 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11380 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11382 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
11384 * config/rs6000/constraints.md (wG constraint): Delete, no longer
11386 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11387 predicate to reflect toc fusion has been deleted.
11388 (toc_fusion_mem_raw): Delete, no longer used.
11389 (toc_fusion_mem_wrapped): Likewise.
11390 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11392 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11393 Delete, no longer used.
11394 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11395 meant to be used for toc fusion.
11396 (rs6000_debug_print_mode): Delete toc fusion debugging.
11397 (rs6000_debug_reg_global): Likewise.
11398 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11399 fusion and secondary reload support that were never used.
11400 (rs6000_option_override_internal): Delete TOC fusion, that was only
11401 partially defined, and it did not work unless you also used the
11403 (rs6000_legitimate_address_p): Delete TOC fusion support.
11404 (rs6000_opt_masks): Likewise.
11405 (fusion_wrap_memory_address): Delete function, no longer used.
11406 (fusion_split_address); Delete TOC fusion support.
11407 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11408 longer used with toc fusion being deleted.
11409 (TARGET_TOC_FUSION_FP): Likewise.
11410 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11412 (toc fusion spliter): Delete TOC fusion support.
11413 (toc_fusionload_<mode>): Likewise.
11414 (toc_fusionload_di): Likewise.
11415 (fusion_gpr_load_<mode>): Delete generator function, this insn no
11416 longer needs to be named. Rename predicate to delete TOC fusion.
11417 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11418 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11419 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11420 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11421 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11423 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
11425 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11426 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
11427 __int128 in built-in function prototypes.
11428 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11429 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11431 2018-07-27 Martin Sebor <msebor@redhat.com>
11433 PR tree-optimization/86696
11434 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11435 types, including enums.
11436 (handle_char_store): Be prepared for the above function to fail.
11438 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
11440 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11441 when optimization level is lower than 2 or optimize for size.
11443 2018-07-26 Martin Sebor <msebor@redhat.com>
11445 PR tree-optimization/86043
11446 PR tree-optimization/86042
11447 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11448 (get_string_cst_length): Rename...
11449 (get_min_string_length): ...to this. Add argument.
11450 (handle_char_store): Extend to handle multi-character stores by
11452 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
11453 * tree.h (initializer_zerop): Add argument.
11455 2018-07-26 Jakub Jelinek <jakub@redhat.com>
11457 PR middle-end/86660
11458 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11459 declare target to variables if they have always,{to,from,tofrom} map
11462 2018-07-26 Martin Liska <mliska@suse.cz>
11465 * lto-wrapper.c: Add linker_output as prefix
11466 for ltrans_output_file.
11468 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
11470 PR rtl-optimization/85805
11471 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11472 value for hard registers if that was written in the same mode.
11474 2018-07-26 Martin Liska <mliska@suse.cz>
11476 PR gcov-profile/86536
11477 * gcov.c (format_gcov): Use printf format %.*f directly
11478 and do not handle special values.
11480 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11482 * common/config/arc/arc-common.c (arc_option_optimization_table):
11483 Update default optimizations for size.
11485 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11487 * config/arc/arc.md (movsf_insn): Add short instruction selection.
11488 * config/arc/constraints.md (CfZ): New constraint.
11489 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11490 (subsf3_fpu): Likewise.
11491 (cmpsf_fpu): Likewise.
11492 (cmpsf_fpu_uneq): Likewise.
11494 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11496 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11498 (arc_is_uncached_mem_p): uncached applies to both the variable and
11501 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11503 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11506 2018-07-25 David Malcolm <dmalcolm@redhat.com>
11508 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11509 field "m_scopes" from vec to auto_vec.
11511 2018-07-25 Martin Liska <mliska@suse.cz>
11513 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11516 2018-07-25 Richard Biener <rguenther@suse.de>
11519 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11520 special wrt context_die late.
11521 (gen_subprogram_die): Re-use DIEs in local scope.
11523 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
11525 PR tree-optimization/86644
11526 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
11528 2018-07-25 Martin Liska <mliska@suse.cz>
11530 PR middle-end/86645
11531 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11532 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11534 2018-07-25 Martin Liska <mliska@suse.cz>
11537 * params.def: Explain ASan abbreviation and provide
11538 a documentation link.
11540 2018-07-24 Martin Sebor <msebor@redhat.com>
11542 PR tree-optimization/86622
11543 PR tree-optimization/86532
11544 * builtins.h (string_length): Declare.
11545 * builtins.c (c_strlen): Correct handling of non-constant offsets.
11546 (check_access): Be prepared for non-constant length ranges.
11547 (string_length): Make extern.
11548 * expr.c (string_constant): Only handle the minor non-constant
11549 array index. Use string_constant to compute the length of
11550 a generic string constant.
11552 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
11554 PR tree-optimization/86618
11555 * tree-vect-stmts.c (vectorizable_call): Don't take the address
11556 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11558 2018-07-24 David Malcolm <dmalcolm@redhat.com>
11560 PR tree-optimization/86636
11561 * json.cc (json::object::set): Fix comment. Add assertions.
11562 (json::array::append): Move here from json.h. Add comment and an
11564 (json::string::string): Likewise.
11565 * json.h (json::array::append): Move to json.cc.
11566 (json::string::string): Likewise.
11567 * optinfo-emit-json.cc
11568 (optrecord_json_writer::impl_location_to_json): Assert that we
11569 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11570 wrapper around it. Expand the location once, rather than three
11572 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11573 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11575 (optrecord_json_writer::optinfo_to_json): Likewise, in four
11576 places. Fix some overlong lines.
11578 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
11580 * config/aarch64/aarch64-simd.md
11581 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11582 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11583 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11584 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11585 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11586 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11587 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11588 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
11589 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
11591 2018-07-24 Jakub Jelinek <jakub@redhat.com>
11593 PR middle-end/86627
11594 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
11595 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
11596 and abs_d == d, do the power of two handling if profitable.
11598 2018-07-24 Richard Biener <rguenther@suse.de>
11600 * match.pd: Add BIT_FIELD_REF canonicalizations.
11602 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
11605 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
11607 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
11609 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
11610 terminated STRING_CST object.
11612 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
11614 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
11616 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
11618 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
11619 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
11620 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
11621 the elements into a register.
11622 (rs6000_split_v4si_init_di_reg): Delete.
11623 (rs6000_split_v4si_init): Delete.
11624 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
11625 (vsx_init_v4si): Rewrite as a define_expand.
11627 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
11629 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
11630 zero_extend argument from memory): New.
11632 2018-07-22 Martin Sebor <msebor@redhat.com>
11635 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
11636 diagnosing calls with unknown arguments unless -Walloca-larger-than
11637 is restricted to less than PTRDIFF_MAX bytes.
11639 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
11641 * doc/gcov.texi (Invoking Gcov): Editorial changes.
11643 2018-07-20 David Malcolm <dmalcolm@redhat.com>
11645 * pretty-print.c (text_info::set_location): Remove redundant
11646 "line_table" parameter from call to rich_location::set_range.
11648 2018-07-20 Martin Sebor <msebor@redhat.com>
11650 PR middle-end/82063
11651 * builtins.c (expand_builtin_alloca): Adjust.
11652 * calls.c (alloc_max_size): Simplify.
11653 * cgraphunit.c (cgraph_node::expand): Adjust.
11654 * common.opt (larger_than_size, warn_frame_larger_than): Remove
11656 (frame_larger_than_size): Same.
11657 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
11658 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
11660 * doc/invoke.texi (GCC Command Options): Document option arguments.
11661 Explain byte-size arguments and suffixes.
11662 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
11663 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
11664 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
11665 * doc/options.texi (UInteger): Expand.
11666 (Host_Wide_Int, ByteSize): Document new properties.
11667 * final.c (final_start_function_1): Include sizes in an error message.
11668 * function.c (frame_offset_overflow): Same.
11669 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
11670 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
11671 Diagnose unbounded alloca calls only for limits of less than
11673 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
11674 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
11676 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
11677 only for limits of less than PTRDIFF_MAX.
11678 * langhooks-def.h (lhd_handle_option): Change function argument
11680 * langhooks.c (lhd_handle_option): Same.
11681 * langhooks.h (handle_option): Same.
11682 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
11684 (var_type, var_type_struct): Same.
11685 (var_set): Handle ByteSize flag.
11686 * optc-gen.awk: Add comments to output to ease debugging. Make
11687 use of HOST_WIDE_INT where appropriate.
11688 * opts-gen-save.awk: Use %lx to format unsigned long.
11689 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
11690 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
11691 arguments. Parse bytes-size suffixes.
11692 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
11693 (enum_value_to_arg): Same.
11694 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
11695 (handle_option): Adjust.
11696 (generate_option): Change function argument to HOST_WIDE_INT.
11697 (cmdline_handle_error): Adjust.
11698 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
11699 (set_option): Change function argument to HOST_WIDE_INT.
11700 (option_enabled): Handle cl_host_wide_int.
11701 (get_option_state): Handle CLVC_SIZE.
11702 (control_warning_option): Same.
11703 * opts.c (common_handle_option): Change function argument to
11704 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
11705 OPT_Wvla_larger_than_.
11706 * opts.h (enum cl_var_type): Add an enumerator.
11707 * stor-layout.c (layout_decl): Print a more meaningful warning.
11708 * toplev.c (output_stack_usage): Adjust.
11710 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
11712 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
11713 call to inline_expand_builtin_string_cmp.
11714 (expand_builtin_strcmp): Likewise.
11715 (expand_builtin_strncmp): Likewise.
11716 (inline_string_cmp): Delete the last parameter, change char_type_node
11717 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
11719 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
11720 the inlining expansion on target where the type of the call has same or
11721 narrower precision than unsigned char.
11723 2018-07-20 David Malcolm <dmalcolm@redhat.com>
11725 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
11726 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
11727 * common.opt (fsave-optimization-record): New option.
11728 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
11729 * doc/invoke.texi (-fsave-optimization-record): New option.
11730 * dumpfile.c: Include "optinfo-emit-json.h".
11731 (struct kv_pair): Move to coretypes.h.
11732 (optgroup_options): Make non-static.
11733 (dump_context::end_scope): Call
11734 optimization_records_maybe_pop_dump_scope.
11735 * dumpfile.h (optgroup_options): New decl.
11736 * json.cc: New file.
11737 * json.h: New file.
11738 * optinfo-emit-json.cc: New file.
11739 * optinfo-emit-json.h: New file.
11740 * optinfo.cc: Include "optinfo-emit-json.h".
11741 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
11742 (optinfo_enabled_p): Check optimization_records_enabled_p.
11743 (optinfo_wants_inlining_info_p): Likewise.
11744 * optinfo.h: Update comment.
11745 * profile-count.c (profile_quality_as_string): New function.
11746 * profile-count.h (profile_quality_as_string): New decl.
11747 (profile_count::quality): New accessor.
11748 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
11749 and optinfo_emit_json_cc_tests.
11750 * selftest.h (selftest::json_cc_tests): New decl.
11751 (selftest::optinfo_emit_json_cc_tests): New decl.
11752 * toplev.c: Include "optinfo-emit-json.h".
11753 (compile_file): Call optimization_records_finish.
11754 (do_compile): Call optimization_records_start.
11755 * tree-ssa-live.c: Include optinfo.h.
11756 (remove_unused_scope_block_p): Retain inlining information if
11757 optinfo_wants_inlining_info_p returns true.
11759 2018-07-20 Richard Biener <rguenther@suse.de>
11762 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
11763 to cover -flto-partition=none.
11765 2018-07-20 Martin Liska <mliska@suse.cz>
11767 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
11768 (get_decl_source_range): Remove unused function.
11770 2018-07-20 Richard Biener <rguenther@suse.de>
11772 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
11773 (struct vn_phi_s): Likewise.
11774 (struct vn_reference_s): Likewise.
11775 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
11776 for searching the slot of an entry known to be in the hash itself.
11777 (vn_phi_hasher::equal): Likewise.
11778 (vn_reference_hasher::equal): Likewise.
11779 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
11781 (optimistic_info, current_info): Remove, keeping only valid_info.
11782 (vn_reference_lookup_1): Remove fallback lookup.
11783 (vn_reference_lookup_2): Likewise.
11784 (vn_nary_op_lookup_1): Likewise.
11785 (vn_phi_lookup): Likewise.
11786 (vn_nary_build_or_lookup_1): Make sure to not chain the built
11788 (vn_reference_insert): Adjust, chain the inserted hash element
11789 at last_inserted_ref.
11790 (vn_reference_insert_pieces): Likewise.
11791 (visit_reference_op_call): Likewise.
11792 (vn_nary_op_insert_into): Chain the inserted hash element at
11793 last_inserted_nary.
11794 (vn_nary_op_insert_pieces): Adjust.
11795 (vn_nary_op_insert): Likewise.
11796 (vn_nary_op_insert_stmt): Likewise.
11797 (vn_phi_insert): Adjust, chain the inserted hash element at
11799 (process_scc): Remove clearing and copying the optimistic
11800 table. Instead remove elements inserted during an optimistic
11801 iteration from the single table we maintain.
11802 (init_scc_vn): Adjust.
11803 (free_scc_vn): Likewise.
11804 (sccvn_dom_walker::record_cond): Likewise.
11805 (sccvn_dom_walker::after_dom_children): Likewise.
11807 2018-07-19 Martin Sebor <msebor@redhat.com>
11809 PR tree-optimization/84047
11810 PR tree-optimization/83776
11811 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
11812 (check_array_bounds): Call it.
11814 2018-07-19 Martin Sebor <msebor@redhat.com>
11816 * align.h (align_flags): Use member initialization.
11818 2018-07-19 David Malcolm <dmalcolm@redhat.com>
11820 * Makefile.in (OBJS): Add optinfo.o.
11821 * coretypes.h (class symtab_node): New forward decl.
11822 (struct cgraph_node): New forward decl.
11823 (class varpool_node): New forward decl.
11824 * dump-context.h: New file.
11825 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
11827 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
11828 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
11829 (set_alt_dump_file): Likewise.
11830 (dump_context::~dump_context): New dtor.
11831 (dump_gimple_stmt): Move implementation to...
11832 (dump_context::dump_gimple_stmt): ...this new member function.
11833 Add the stmt to any pending optinfo, creating one if need be.
11834 (dump_gimple_stmt_loc): Move implementation to...
11835 (dump_context::dump_gimple_stmt_loc): ...this new member function.
11836 Start a new optinfo and add the stmt to it.
11837 (dump_gimple_expr): Move implementation to...
11838 (dump_context::dump_gimple_expr): ...this new member function.
11839 Add the stmt to any pending optinfo, creating one if need be.
11840 (dump_gimple_expr_loc): Move implementation to...
11841 (dump_context::dump_gimple_expr_loc): ...this new member function.
11842 Start a new optinfo and add the stmt to it.
11843 (dump_generic_expr): Move implementation to...
11844 (dump_context::dump_generic_expr): ...this new member function.
11845 Add the tree to any pending optinfo, creating one if need be.
11846 (dump_generic_expr_loc): Move implementation to...
11847 (dump_context::dump_generic_expr_loc): ...this new member
11848 function. Add the tree to any pending optinfo, creating one if
11850 (dump_printf): Move implementation to...
11851 (dump_context::dump_printf_va): ...this new member function. Add
11852 the text to any pending optinfo, creating one if need be.
11853 (dump_printf_loc): Move implementation to...
11854 (dump_context::dump_printf_loc_va): ...this new member function.
11855 Start a new optinfo and add the stmt to it.
11856 (dump_dec): Move implementation to...
11857 (dump_context::dump_dec): ...this new member function. Add the
11858 value to any pending optinfo, creating one if need be.
11859 (dump_context::dump_symtab_node): New member function.
11860 (dump_context::get_scope_depth): New member function.
11861 (dump_context::begin_scope): New member function.
11862 (dump_context::end_scope): New member function.
11863 (dump_context::ensure_pending_optinfo): New member function.
11864 (dump_context::begin_next_optinfo): New member function.
11865 (dump_context::end_any_optinfo): New member function.
11866 (dump_context::s_current): New global.
11867 (dump_context::s_default): New global.
11868 (dump_scope_depth): Delete global.
11869 (dumpfile_ensure_any_optinfo_are_flushed): New function.
11870 (dump_symtab_node): New function.
11871 (get_dump_scope_depth): Reimplement in terms of dump_context.
11872 (dump_begin_scope): Likewise.
11873 (dump_end_scope): Likewise.
11874 (selftest::temp_dump_context::temp_dump_context): New ctor.
11875 (selftest::temp_dump_context::~temp_dump_context): New dtor.
11876 (selftest::verify_item): New function.
11877 (ASSERT_IS_TEXT): New macro.
11878 (ASSERT_IS_TREE): New macro.
11879 (ASSERT_IS_GIMPLE): New macro.
11880 (selftest::test_capture_of_dump_calls): New test.
11881 (selftest::dumpfile_c_tests): Call it.
11882 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
11883 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
11884 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
11885 descriptive comment.
11886 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
11887 (dump_node, dump_bb): Move these unrelated decls.
11888 (class dump_manager): Add leading comment.
11889 * optinfo.cc: New file.
11890 * optinfo.h: New file.
11892 2018-07-19 Michael Collison <michael.collison@arm.com>
11893 Richard Henderson <rth@redhat.com>
11895 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
11896 (subti): Handle op1 zero.
11897 (subvti4, usub4ti4): New.
11898 (*sub<GPI>3_compare1_imm): New.
11899 (sub<GPI>3_carryinCV): New.
11900 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
11901 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
11903 2018-07-19 Michael Collison <michael.collison@arm.com>
11904 Richard Henderson <rth@redhat.com>
11906 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
11907 (addti3): Create simpler code if low part is already known to be 0.
11908 (addvti4, uaddvti4): New.
11909 (*add<GPI>3_compareC_cconly_imm): New.
11910 (*add<GPI>3_compareC_cconly): New.
11911 (*add<GPI>3_compareC_imm): New.
11912 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
11913 handle constants within this pattern..
11914 (*add<GPI>3_compareV_cconly_imm): New.
11915 (*add<GPI>3_compareV_cconly): New.
11916 (*add<GPI>3_compareV_imm): New.
11917 (add<GPI>3_compareV): New.
11918 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
11919 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
11920 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
11921 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
11922 with 'comparison' operator.
11923 (*add<GPI>3_compareV_cconly_imm): Ditto.
11924 (*add<GPI>3_compareV_cconly): Ditto.
11925 (*add<GPI>3_compareV_imm): Ditto.
11926 (add<GPI>3_compareV): Ditto.
11927 (add<mode>3_carryinC): Ditto.
11928 (*add<mode>3_carryinC_zero): Ditto.
11929 (*add<mode>3_carryinC): Ditto.
11930 (add<mode>3_carryinV): Ditto.
11931 (*add<mode>3_carryinV_zero): Ditto.
11932 (*add<mode>3_carryinV): Ditto.
11934 2018-07-19 Michael Collison <michael.collison@arm.com>
11935 Richard Henderson <rth@redhat.com>
11937 * config/aarch64/aarch64-modes.def (CC_V): New.
11938 * config/aarch64/aarch64-protos.h
11939 (aarch64_addti_scratch_regs): Declare
11940 (aarch64_subvti_scratch_regs): Declare.
11941 (aarch64_expand_subvti): Declare.
11942 (aarch64_gen_unlikely_cbranch): Declare
11943 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
11944 for signed overflow using CC_Vmode.
11945 (aarch64_get_condition_code_1): Handle CC_Vmode.
11946 (aarch64_gen_unlikely_cbranch): New function.
11947 (aarch64_addti_scratch_regs): New function.
11948 (aarch64_subvti_scratch_regs): New function.
11949 (aarch64_expand_subvti): New function.
11951 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11953 * config/aarch64/aarch64-option-extensions.def: New entry for profile
11955 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
11956 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
11959 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11962 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11963 address check not strict.
11965 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
11967 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
11969 (store_pair_lanes<mode>): Likewise.
11970 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
11971 enum value 'ADDR_QUERY_LDP_STP_N'.
11972 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
11973 (aarch64_print_address_internal): Add declaration.
11974 (aarch64_print_ldpstp_address): Remove.
11975 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
11976 (aarch64_print_operand): Change printing of 'y'.
11977 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
11978 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
11979 'true' rather than '1'.
11980 * config/aarch64/constraints.md (Uml): Likewise.
11981 (Uml): Rename to Umn.
11984 2018-07-19 Richard Biener <rguenther@suse.de>
11986 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
11988 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
11989 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
11990 (vn_reference_hasher): Likewise.
11991 (struct vn_tables_s): Remove obstack and alloc-pool members.
11992 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
11993 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
11994 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
11995 (vn_reference_insert_pieces): Likewise.
11996 (alloc_vn_nary_op_noinit): Adjust.
11997 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
11998 (vn_phi_eq): Adjust.
11999 (shared_lookup_phiargs): Remove.
12000 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12001 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12002 (visit_reference_op_call): Likewise.
12003 (copy_nary, copy_phi, copy_reference): Remove.
12004 (process_scc): Rewind the obstack when iterating. Do not
12005 copy the elements to valid_info but just move them from one
12006 hashtable to the other.
12007 (allocate_vn_table): Adjust.
12008 (free_vn_table): Likewise.
12009 (init_scc_vn): Likewise.
12010 (free_scc_vn): Likewise.
12012 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
12015 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12016 indirect_return as function type attribute.
12017 (ix86_attribute_table): Change indirect_return to function
12019 * doc/extend.texi: Update indirect_return attribute.
12021 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
12023 * wide-int.h (widest2_int): New.
12024 * gimple-fold.c (arith_overflowed_p): Use it.
12025 * tree.h (widest2_int_cst): New.
12026 * tree-vrp.c (wide_int_binop_overflow): Rename from
12027 vrp_int_const_binop.
12028 Rewrite to work on trees.
12029 (extract_range_from_multiplicative_op_1): Abstract code to...
12030 (wide_int_range_min_max): ...here.
12031 (wide_int_range_cross_product): ...and here.
12032 (extract_range_from_binary_expr_1): Abstract overflow code to...
12033 (wide_int_range_mult_wrapping): ...here.
12034 * tree-vrp.h (wide_int_range_cross_product): New.
12035 (wide_int_range_mult_wrapping): New.
12037 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
12038 Julia Koval <julia.koval@intel.com>
12040 * config/i386/x86-tune-costs.h (skylake_memcpy,
12041 skylake_memset): Replace rep_prefix with unrolling for size 512.
12043 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
12045 PR middle-end/86544
12046 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12047 comparision with EQ_EXPR in last stmt.
12049 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
12051 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12052 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12053 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12054 previously known as "PowerPC AltiVec Built-in Functions". Move
12055 some material to new subsubsections "PowerPC AltiVec Built-in
12056 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12058 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12059 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12060 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12061 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12063 2018-07-18 Richard Biener <rguenther@suse.de>
12065 PR tree-optimization/86557
12066 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12069 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
12071 * config/s390/s390.c (s390_function_profiler): Generate CFI.
12073 2018-07-17 Jeff Law <law@redhat.com>
12075 * config/arm/arm.c (get_label_padding): Update for recent
12076 changes to label_to_alignment.
12078 PR tree-optimization/86010
12079 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12081 * config/mips/mips.c (vr4130_align_insns): Update for recent
12082 changes to label_to_alignment.
12084 * config/frv/frv.c (frv_label_align): Update for recent changes
12085 to label_to_alignment.
12087 * config/nios2/nios2.c (nios2_label_align): Update for recent
12088 changes which dropped ALIGN_LABELS_LOG.
12090 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
12092 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12093 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12095 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
12097 * config/arc/arc.c (arc_label_align): Use align_labels instead of
12098 deprecated align_labels_log.
12100 2018-07-17 Richard Biener <rguenther@suse.de>
12103 * dwarf2out.c (init_sections_and_labels): Always generate
12104 a debug_line_str_section for early LTO debug.
12105 (dwarf2out_finish): Reset debug_line_str_hash output early.
12106 Bump counter for extra dwarf5 .debug_loc labels to not conflict
12108 (dwarf2out_early_finish): Output debug_line_str.
12110 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
12112 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12113 index register on z196 or later.
12115 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
12117 * config/s390/s390.c (s390_default_align): Set default function
12119 (s390_override_options_after_change): Call s390_default align.
12120 (s390_option_override_internal): Call s390_default_align.
12121 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12123 2018-07-17 Jakub Jelinek <jakub@redhat.com>
12125 PR middle-end/86542
12126 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12127 to _looptemp_ clauses, other than the first two.
12129 2018-07-17 Martin Liska <mliska@suse.cz>
12131 * opts.c: Do not enable OPT_falign_* for -Os.
12133 2018-07-17 Martin Liska <mliska@suse.cz>
12135 * align.h (MAX_CODE_ALIGN): New.
12136 (MAX_CODE_ALIGN_VALUE): New.
12137 * common/config/i386/i386-common.c (ix86_handle_option):
12138 (MAX_CODE_ALIGN): Moved to align.h.
12139 * final.c (MAX_CODE_ALIGN): Likewise.
12140 * opts.c (parse_and_check_align_values):
12141 (MAX_CODE_ALIGN): Likewise.
12142 (MAX_CODE_ALIGN_VALUE): Likewise.
12144 2018-07-17 Martin Liska <mliska@suse.cz>
12146 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12147 in order to fulfil coding style.
12148 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12149 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12150 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12151 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12152 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12153 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12154 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12155 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12156 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12158 2018-07-17 Martin Liska <mliska@suse.cz>
12160 * align.h: New file.
12161 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12163 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12164 align_flags of label_to_alignment.
12165 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12167 * config/m68k/m68k.c: Do not use removed align_labels_value and
12169 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12170 (LOOP_ALIGN): Likewise.
12171 (LABEL_ALIGN): Likewise.
12172 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12173 Remove not used macro.
12174 (rs6000_loop_align): Change return type to align_flags.
12175 (rs6000_loop_align_max_skip): Remove.
12176 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12177 Change return type to align_flags.
12178 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12179 Remove not used macro.
12180 (rs6000_loop_align): Change return type to align_flags.
12181 (rs6000_loop_align_max_skip): Remove.
12182 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12183 * config/rx/rx-protos.h (rx_align_for_label): Make it
12185 * config/rx/rx.c (rx_align_for_label): Change return type
12187 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12189 into align_flags class.
12190 (LABEL_ALIGN): Likewise.
12191 (LOOP_ALIGN): Likewise.
12192 * config/s390/s390.c (s390_label_align): Use align_flags
12194 (s390_asm_output_function_label): Likewise.
12195 * config/sh/sh.c (sh_override_options_after_change):
12196 Use align_flags class directly without macros.
12197 (find_barrier): Likewise.
12198 (barrier_align): Likewise.
12199 (sh_loop_align): Likewise.
12200 * config/spu/spu.c (spu_option_override):
12201 Use align_flags_tuple::get_value instead of removed macros.
12202 (spu_sched_init): Likewise.
12203 * config/spu/spu.h (GTY): Likewise.
12204 * config/visium/visium.c (visium_option_override):
12205 Set "8" as default secondary alignment.
12206 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12207 in order to guarantee secondary alignment of 8.
12208 * coretypes.h: Include align.h header file.
12209 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12210 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12211 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12212 * doc/tm.texi.in: Likewise.
12213 * final.c (struct label_alignment): Remove not used structure.
12214 (LABEL_ALIGN): Change type to align_flags.
12215 (LOOP_ALIGN): Likewise.
12216 (JUMP_ALIGN): Likewise.
12217 (default_loop_align_max_skip): Remove.
12218 (default_label_align_max_skip): Likewise.
12219 (default_jump_align_max_skip): Likewise.
12220 (default_label_align_after_barrier_max_skip):
12221 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12222 (LABEL_TO_MAX_SKIP): Remove.
12223 (label_to_alignment): Return align_flags type instead of integer.
12224 (label_to_max_skip): Remove.
12225 (align_fuzz): Use align_flags type.
12226 (compute_alignments): Use align_flags type and use align_flags::max
12227 to combine multiple alignments.
12228 (grow_label_align): Grow vec instead of C array.
12229 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12230 (shorten_branches): Use align_flags type and use align_flags::max
12231 to combine multiple alignments.
12232 (final_scan_insn_1): Remove usage of secondary alignment that comes
12233 from label alignment, but instead use proper secondary alignment
12234 which is computed in grow_label_align.
12235 * flags.h (struct align_flags_tuple): Move to align.h.
12236 (struct align_flags): Likewise.
12237 (state_align_loops): Rename to align_loops.
12238 (state_align_jumps): Rename to align_jumps.
12239 (state_align_labels): Rename to align_labels.
12240 (state_align_functions): Rename to align_functions.
12241 (align_loops_log): Remove.
12242 (align_jumps_log): Remove.
12243 (align_labels_log): Remove.
12244 (align_functions_log): Remove.
12245 (align_loops_max_skip): Remove.
12246 (align_jumps_max_skip): Remove.
12247 (align_labels_max_skip): Remove.
12248 (align_functions_max_skip): Remove.
12249 (align_loops_value): Remove.
12250 (align_jumps_value): Remove.
12251 (align_labels_value): Remove.
12252 (align_functions_value): Remove.
12253 * output.h (label_to_alignment): Change return type to align_flags.
12254 (label_to_max_skip): Remove.
12255 * target.def: Remove loop_align_max_skip, label_align_max_skip,
12256 jump_align_max_skip macros.
12257 * targhooks.h (default_loop_align_max_skip): Remove.
12258 (default_label_align_max_skip): Likewise.
12259 (default_jump_align_max_skip): Likewise.
12260 (default_label_align_after_barrier_max_skip): Remove.
12261 * toplev.c (read_log_maxskip): Use ::normalize function.
12262 (parse_N_M): Remove not used argument and also call ::normalize.
12263 (parse_alignment_opts): Do not pass unused arguments.
12264 * varasm.c (assemble_start_function): Use directly align_functions
12265 instead of removed macros.
12266 * system.h: Do not poison removed macros.
12268 2018-07-17 Jakub Jelinek <jakub@redhat.com>
12270 PR middle-end/86539
12271 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12272 and cond temporaries don't have reference type if iterator has
12273 pointer type. For init use &for_pre_body instead of pre_p if
12274 for_pre_body is non-empty.
12276 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12278 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12279 double-double modes to SFmode directly directly.
12280 (trunc<mode>sf2_fprs): Delete.
12282 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12284 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12285 for conversions between IFmode and the decimal floating point modes.
12286 (init_float128_ieee): Use the correct names for conversions between
12287 KFmode and the decimal floating point modes.
12289 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12291 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12292 for the conversions between TDmode and IFmode.
12293 (init_float128_ieee): Use more correct names for the conversions
12294 between TDmode and KFmode.
12296 2018-07-16 Jakub Jelinek <jakub@redhat.com>
12298 PR tree-optimization/86526
12299 * builtins.c (expand_builtin_memcmp): Formatting fixes.
12300 (inline_expand_builtin_string_cmp): Likewise.
12301 (inline_string_cmp): Likewise. Use c_readstr instead of
12302 builtin_memcpy_read_str. Add unit_mode temporary.
12304 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
12306 PR middle-end/86528
12307 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12308 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12310 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12312 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12313 Alphabetize prototypes of built-in functions, separating out
12314 built-in functions that are listed in this section but should be
12315 described elsewhere.
12317 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
12320 * expmed.c (emit_store_flag): Do not emit setcc followed by a
12321 conditional move when trapping comparison was split to a
12322 non-trapping one (and vice versa).
12324 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12326 * config/s390/s390.c (s390_function_profiler): Generate nops
12327 instead of profiler call sequences.
12328 * config/s390/s390.opt: Add the new option.
12330 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12332 * config/s390/s390.c (s390_function_profiler): Generate
12333 __mcount_loc section.
12334 * config/s390/s390.opt: Add the new option.
12336 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12338 * common.opt: Add the new warning.
12339 * config/s390/s390.c (s390_function_profiler): Emit "brasl
12340 %r0,__fentry__" when -mfentry is specified.
12341 (s390_option_override_internal): Disallow -mfentry for 31-bit
12343 * config/s390/s390.opt: Add the new option.
12345 2018-07-16 Richard Biener <rguenther@suse.de>
12348 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12349 for function-local FUNCTION_DECL and RESULT_DECL immediately.
12351 2018-07-16 Martin Liska <mliska@suse.cz>
12354 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12357 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
12359 * config/arc/arcHS.md: Update ARCHS scheduling rules.
12361 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
12363 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12365 * config/arc/arc-cpus.def (hs4x): New cpu.
12367 * config/arc/arc-tables.opt: Regenerate.
12368 * config/arc/arc.c (arc_sched_issue_rate): New function.
12369 (TARGET_SCHED_ISSUE_RATE): Define.
12370 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12371 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12373 (attr tune): Add ARCHS4x tune values.
12374 (attr tune_dspmpy): Define.
12375 (*tst): Correct instruction type.
12376 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12377 * config/arc/arcHS4x.md: New file.
12378 * config/arc/fpu.md: Update instruction type attributes.
12379 * config/arc/t-multilib: Regenerate.
12381 2018-07-16 Tom de Vries <tdevries@suse.de>
12384 * var-tracking.c (vt_initialize): Fix pre_dec handling.
12386 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12388 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12391 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
12393 PR tree-optimization/86514
12394 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12395 conversion to a boolean type from a type with greater precision.
12397 2018-07-16 Tom de Vries <tdevries@suse.de>
12399 * var-tracking.c (vt_initialize): Print adjusted insn slim if
12400 dump_flags request TDF_SLIM.
12402 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
12404 * fold-const.c (int_const_binop_1): Abstract...
12405 (wide_int_binop): ...wide int code here.
12406 (poly_int_binop): ...poly int code here.
12407 Abstract the rest of int_const_binop_1 into int_const_binop.
12408 * fold-const.h (wide_int_binop): New.
12409 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12410 Remove useless PLUS/MINUS_EXPR case.
12411 (zero_nonzero_bits_from_vr): Move wide int code...
12412 (zero_nonzero_bits_from_bounds): ...here.
12413 (extract_range_from_binary_expr_1): Move mask optimization code...
12414 (range_easy_mask_min_max): ...here.
12415 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12416 (range_easy_mask_min_max): New.
12418 2018-07-15 Jeff Law <law@redhat.com>
12421 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12424 2018-07-14 Jim Wilson <jimw@sifive.com>
12426 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12428 2018-07-14 Paul Koning <ni1d@arrl.net>
12430 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12432 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
12434 * lto-streamer-out.c (copy_function_or_variable): Dump info about
12437 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
12438 Steve Munroe <munroesj52@gmail.com>
12440 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12441 (_mm_andnot_si128): Likewise.
12442 (_mm_or_si128): Likewise.
12443 (_mm_xor_si128): Likewise.
12445 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
12447 PR middle-end/78809
12448 * builtins.c (expand_builtin_memcmp): Inline the calls first
12449 when result_eq is false.
12450 (expand_builtin_strcmp): Inline the calls first.
12451 (expand_builtin_strncmp): Likewise.
12452 (inline_string_cmp): New routine. Expand a string compare
12453 call by using a sequence of char comparison.
12454 (inline_expand_builtin_string_cmp): New routine. Inline expansion
12455 a call to str(n)cmp/memcmp.
12456 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12458 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12460 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
12462 * config/arm/driver-arm.c: Include arm-native.h.
12463 (host_detect_local_cpu): Use auto-generated data tables.
12464 (vendors, arm_cpu_table): Delete. Move part information to ...
12465 * config/arm/arm-cpus.in: ... here.
12466 * config/arm/parsecpu.awk (gen_native): New function.
12467 (vendor, part): New CPU fields.
12468 (END): Add support for building the native CPU detection tables.
12469 * config/arm/t-arm (arm-native.h): Add build rule.
12470 (driver-arm.o): Add dependency on arm-native.h.
12472 2018-07-13 Richard Biener <rguenther@suse.de>
12474 PR middle-end/85974
12475 * match.pd (addr1 - addr2): Allow either of the operand to
12478 2018-07-13 Tom de Vries <tdevries@suse.de>
12480 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12483 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
12485 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12486 arrays instead of numbered variables.
12488 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
12490 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12491 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12493 2018-07-13 Richard Biener <rguenther@suse.de>
12496 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12497 instead of get_context_die.
12499 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
12500 Richard Biener <rguenther@suse.de>
12502 PR middle-end/86489
12503 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12504 that the loop latch destination where phi is defined.
12506 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
12508 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12509 (riscv_expand_epilogue): Add assertion to check interrupt mode.
12510 (riscv_set_current_function): Extract getting interrupt type to new
12512 (riscv_get_interrupt_type): New function.
12513 (riscv_merge_decl_attributes): New function, checking interrupt type is
12515 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12517 2018-07-12 Paul Koning <ni1d@arrl.net>
12519 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12522 2018-07-12 Paul Koning <ni1d@arrl.net>
12524 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12525 zero reference, add doloop_end instead.
12526 * doc/md.texi (decrement_and_branch_until_zero): Remove.
12527 (Looping patterns): Remove decrement_and_branch_until_zero. Add
12528 detail for doloop_end.
12530 2018-07-12 Martin Sebor <msebor@redhat.com>
12533 * attribs.c (decl_attributes): Reject conflicting attributes before
12534 calling attribute handlers.
12536 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
12538 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12540 (gcc::dump_manager::get_dump_file_name): likewise.
12541 (dump_begin): Likewise.
12542 * dumpfile.h (dump_begin): Update prototype.
12543 (gcc::dump_manager::get_dump_file_name,
12544 gcc::dump_manager::get_dump_file_name): Update prototype.
12546 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
12548 * internal-fn.h (vectorizable_internal_fn_p): New function.
12549 * tree-vect-slp.c (compatible_calls_p): Likewise.
12550 (vect_build_slp_tree_1): Remove nops argument. Handle calls
12551 to internal functions.
12552 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12554 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
12556 * fold-const.h (inverse_conditions_p): Declare.
12557 * fold-const.c (inverse_conditions_p): New function.
12558 * match.pd: Use inverse_conditions_p. Add folds of view_converts
12559 that test the inverse condition of a conditional internal function.
12560 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12561 * internal-fn.c (internal_fn_mask_index): Handle conditional
12562 internal functions.
12563 (vectorized_internal_fn_supported_p): New function.
12564 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12565 (any_pred_load_store): Replace with...
12566 (need_to_predicate): ...this new variable.
12567 (redundant_ssa_names): New variable.
12568 (ifcvt_can_use_mask_load_store): Move initial checks to...
12569 (ifcvt_can_predicate): ...this new function. Handle tree codes
12570 for which a conditional internal function exists.
12571 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12572 instead of ifcvt_can_use_mask_load_store. Update after variable
12574 (predicate_load_or_store): New function, split out from
12575 predicate_mem_writes.
12576 (check_redundant_cond_expr): New function.
12577 (value_available_p): Likewise.
12578 (predicate_rhs_code): Likewise.
12579 (predicate_mem_writes): Rename to...
12580 (predicate_statements): ...this. Use predicate_load_or_store
12581 and predicate_rhs_code.
12582 (combine_blocks, tree_if_conversion): Update after above name changes.
12583 (ifcvt_local_dce): Handle redundant_ssa_names.
12584 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12585 general conditional functions.
12586 * tree-vect-stmts.c (vectorizable_call): Likewise.
12588 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
12589 Alan Hayward <alan.hayward@arm.com>
12590 David Sherwood <david.sherwood@arm.com>
12592 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
12593 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
12594 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
12595 plus and minus and convert them into IFN_COND_FMA-based sequences.
12596 (convert_mult_to_fma): Handle conditional plus and minus.
12598 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
12600 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
12601 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
12602 (cond_fnms_optab): New optabs.
12603 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
12604 internal functions.
12605 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
12606 * internal-fn.h (get_conditional_internal_fn): Declare.
12607 (get_unconditional_internal_fn): Likewise.
12608 * internal-fn.c (cond_ternary_direct): New macro.
12609 (expand_cond_ternary_optab_fn): Likewise.
12610 (direct_cond_ternary_optab_supported_p): Likewise.
12611 (FOR_EACH_COND_FN_PAIR): Likewise.
12612 (get_conditional_internal_fn): New function.
12613 (get_unconditional_internal_fn): Likewise.
12614 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
12615 (gimple_match_op::gimple_match_op): Add a new overload for 5
12617 (gimple_match_op::set_op): Likewise.
12618 (gimple_resimplify5): Declare.
12619 * genmatch.c (decision_tree::gen): Generate simplifications for
12621 * gimple-match-head.c (gimple_simplify): Define an overload for
12622 5 operands. Handle calls with 5 arguments in the top-level overload.
12623 (convert_conditional_op): Handle conversions from unconditional
12624 internal functions to conditional ones.
12625 (gimple_resimplify5): New function.
12626 (build_call_internal): Pass a fifth operand.
12627 (maybe_push_res_to_seq): Likewise.
12628 (try_conditional_simplification): Try converting conditional
12629 internal functions to unconditional internal functions.
12630 Handle 3-operand unconditional forms.
12631 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
12632 Define ternary equivalents of the current rules for binary conditional
12633 internal functions.
12634 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
12635 ternary operations.
12636 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
12637 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
12638 (optab): Handle them.
12639 (SVE_COND_FP_TERNARY): New int iterator.
12640 (sve_fmla_op, sve_fmad_op): New int attributes.
12641 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
12642 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
12643 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
12645 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
12647 * target.def (preferred_else_value): New target hook.
12648 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
12649 * doc/tm.texi: Regenerate.
12650 * targhooks.h (default_preferred_else_value): Declare.
12651 * targhooks.c (default_preferred_else_value): New function.
12652 * internal-fn.h (conditional_internal_fn_code): Declare.
12653 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
12654 (get_conditional_internal_fn): Use it.
12655 (conditional_internal_fn_code): New function.
12656 * gimple-match.h (gimple_match_cond): New struct.
12657 (gimple_match_op): Add a cond member function.
12658 (gimple_match_op::gimple_match_op): Update all forms to take a
12660 * genmatch.c (expr::gen_transform): Use the same condition as res_op
12661 for the suboperation, but don't specify a particular else_value.
12662 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
12663 (visit_nary_op, visit_reference_op_load): Pass
12664 gimple_match_cond::UNCOND to the gimple_match_op constructor.
12665 * gimple-match-head.c: Include tree-eh.h
12666 (convert_conditional_op): New function.
12667 (maybe_resimplify_conditional_op): Likewise.
12668 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
12669 (gimple_resimplify2): Likewise.
12670 (gimple_resimplify3): Likewise.
12671 (gimple_resimplify4): Likewise.
12672 (maybe_push_res_to_seq): Return null for conditional operations.
12673 (try_conditional_simplification): New function.
12674 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
12676 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
12678 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
12680 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
12682 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
12684 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
12686 (hash_tree): Do not hash DECL_FCONTEXT
12687 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
12688 Do not stream DECL_FCONTEXT.
12689 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
12690 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
12692 2018-07-12 Richard Biener <rguenther@suse.de>
12695 * dwarf2out.c (gen_block_die): Only output blocks when they have
12696 at least one !DECL_IGNORED_P variable.
12698 2018-07-12 Richard Biener <rguenther@suse.de>
12701 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
12702 Remove -mieee-fp handling.
12704 2018-07-12 Richard Biener <rguenther@suse.de>
12706 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
12707 left-over from last patch.
12709 2018-07-12 Jakub Jelinek <jakub@redhat.com>
12711 PR tree-optimization/86492
12712 * gimple-ssa-store-merging.c
12713 (imm_store_chain_info::coalesce_immediate_stores): Call
12714 check_no_overlap even for the merge_overlapping case. Formatting fix.
12716 2018-07-12 Richard Biener <rguenther@suse.de>
12718 PR middle-end/86479
12719 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
12720 move possibly trapping operations into the conditional.
12722 2018-07-12 Richard Biener <rguenther@suse.de>
12724 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
12725 (vn_lookup_simplify_result): Remove recursion limit applied
12727 (vn_nary_build_or_lookup_1): Adjust.
12728 (try_to_simplify): Likewise.
12729 * gimple-match-head.c (gimple_resimplify1): Instead apply one
12731 (gimple_resimplify2): Likewise.
12732 (gimple_resimplify3): Likewise.
12733 (gimple_resimplify4): Likewise.
12735 2018-07-11 Jakub Jelinek <jakub@redhat.com>
12737 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
12738 Use __mmask64 type instead of __mmask8 for __M argument.
12739 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
12740 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
12742 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
12743 __mmask16 for __M argument.
12744 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
12745 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
12746 to __mmask16 instead of __mmask8.
12747 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
12748 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
12749 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
12750 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
12751 instead of __mmask16 for __U argument.
12752 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
12753 __mmask16 instead of __mmask8 for __U argument.
12754 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
12756 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
12758 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
12760 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
12762 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
12764 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
12766 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
12768 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
12770 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
12772 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
12774 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
12775 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
12776 return type as well as __M argument type and all casts from __mmask8
12778 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
12779 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
12780 return type as well as __M argument type and all casts from __mmask8
12782 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
12783 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
12784 return type as well as __M argument type and all casts from __mmask8
12786 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
12787 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
12788 return type as well as __M argument type and all casts from __mmask8
12790 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
12791 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
12794 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
12796 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
12797 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
12800 2018-07-11 Paul Koning <ni1d@arrl.net>
12802 * doc/md.texi (define_subst): Document how multiple occurrences of
12803 the same argument in the replacement pattern are handled.
12805 2018-07-11 Paul Koning <ni1d@arrl.net>
12807 * doc/extend.texi (Common Variable Attributes): Move "mode" into
12808 alphabetical order.
12809 (Common Type Attributes): Add "mode" attribute.
12811 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
12813 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
12814 stream DECL_ORIGINAL_TYPE.
12815 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
12816 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
12817 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
12818 Do not walk original type.
12819 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
12821 (write_ts_decl_non_common_tree_pointers): Do not stream
12823 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
12824 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
12826 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
12828 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
12829 thread twice from the same starting edge.
12831 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
12833 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
12834 * gimple.c (gimple_call_nonnull_result_p): ...here...
12835 (gimple_call_nonnull_arg): ...and here.
12836 * gimple.h (gimple_call_nonnull_result_p): New.
12837 (gimple_call_nonnull_arg): New.
12839 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
12841 * config/arm/arm-cpus.in: Move information from fpu field of each
12842 cpu definition to the isa field.
12843 * config/arm/parsecpu.awk (fpu): Delete match rule.
12844 (gen_comm_data): Don't add bits from the CPU's FPU entry.
12846 2018-07-11 Richard Biener <rguenther@suse.de>
12849 * dwarf2out.c (init_sections_and_labels): Use
12850 output_asm_line_debug_info consistently.
12851 (dwarf2out_early_finish): Likewise.
12852 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
12855 2018-07-11 Richard Biener <rguenther@suse.de>
12857 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
12858 Rework father_bb setting in a way to avoid propagating constants
12859 multiple times on a loop body.
12861 2018-07-10 Mark Wielaard <mark@klomp.org>
12864 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
12866 2018-07-10 Richard Biener <rguenther@suse.de>
12868 * hash-map.h (hash_map::iterator::operator*): Return
12869 references to key and value.
12871 2018-07-10 Jakub Jelinek <jakub@redhat.com>
12874 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
12875 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
12876 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
12877 (gimplify_omp_for): For composite loops, move outer
12878 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
12879 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
12880 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
12881 TREE_LIST for both the original class iterator and the "last" helper
12882 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
12883 loop, remember has_decl_expr from outer composite loops for the
12884 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
12886 2018-07-09 Martin Sebor <msebor@redhat.com>
12888 PR middle-end/77357
12889 PR middle-end/86428
12890 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
12891 accessing implicitly initialized array elements.
12892 * expr.c (string_constant): Handle string initializers of
12893 character arrays within aggregates.
12894 * gimple-fold.c (fold_array_ctor_reference): Add argument.
12895 Store element offset. As a special case, handle zero size.
12896 (fold_nonarray_ctor_reference): Same.
12897 (fold_ctor_reference): Add argument. Store subobject offset.
12898 * gimple-fold.h (fold_ctor_reference): Add argument.
12900 2018-07-09 Paul Koning <ni1d@arrl.net>
12902 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
12903 (pdp11_insn_cost): New function.
12904 (pdp11_md_asm_adjust): New function.
12905 (TARGET_INVALID_WITHIN_DOLOOP): Define.
12906 (pdp11_rtx_costs): Update to match machine better.
12907 (output_addr_const_pdp11): Correct format mismatch warnings.
12908 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
12909 * config/pdp11/pdp11.md: General change to add base_cost and/or
12910 length attributes for use by new pdp11_insn_cost function.
12911 (MIN_BRANCH): Correct definition.
12913 (doloop_end): Use standard pattern name for looping pattern.
12914 (doloop_end_nocc): New.
12915 (movsf): Add another constraint alternative.
12916 (zero_extendqihi2): Add constraint alternatives for not in place
12918 (zero_extendhisi2): Remove.
12919 (shift patterns): Add CC handling variants.
12923 (define_peephole2): New peephole to recognize mov that sets CC for
12926 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12928 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
12929 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
12932 2018-07-09 Jakub Jelinek <jakub@redhat.com>
12935 * real.c (real_nextafter): Return true if result is denormal.
12937 2018-07-09 Martin Liska <mliska@suse.cz>
12939 * common.opt: Add back wrongly removed attribute.
12941 2018-07-09 Richard Biener <rguenther@suse.de>
12944 * dwarf2out.c (gen_block_die): For an early generated DIE
12945 always output high/low PC attributes.
12947 2018-07-09 Tom de Vries <tdevries@suse.de>
12949 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
12950 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
12951 onto VAR_DECL with abstract origin.
12953 2018-07-07 Jim Wilson <jimw@sifive.com>
12955 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
12957 2018-07-07 Tom de Vries <tdevries@suse.de>
12959 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
12961 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
12963 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
12965 (combine_bound): Use wide-int overflow calculation instead of
12967 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
12969 * fold-const.c (int_const_binop_2): Same.
12970 (extract_muldiv_1): Same.
12971 (fold_div_compare): Same.
12972 (fold_abs_const): Same.
12974 * poly-int.h (add): Same.
12978 * predict.c (predict_iv_comparison): Same.
12979 * profile-count.c (slow_safe_scale_64bit): Same.
12980 * simplify-rtx.c (simplify_const_binary_operation): Same.
12981 * tree-chrec.c (tree_fold_binomial): Same.
12982 * tree-data-ref.c (split_constant_offset_1): Same.
12983 * tree-if-conv.c (idx_within_array_bound): Same.
12984 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
12985 * tree-ssa-phiopt.c (minmax_replacement): Same.
12986 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
12987 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
12988 * vr-values.c (vr_values::adjust_range_with_scev): Same.
12989 * wide-int.cc (wi::add_large): Same.
12990 (wi::mul_internal): Same.
12991 (wi::sub_large): Same.
12992 (wi::divmod_internal): Same.
12993 * wide-int.h: Change overflow type to overflow_type for neg, add,
12994 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
12995 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
12996 mul_internal, divmod_internal.
12997 (overflow_type): New enum.
12998 (accumulate_overflow): New.
13000 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
13002 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13003 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13005 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
13007 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13008 argument is checked for zero before entering loop, avoid checking again.
13010 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
13012 * gimplify.h (generic_expr_could_trap_p): Set as global function.
13013 * gimplify.h (generic_expr_could_trap_p): Likwise.
13014 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13016 2018-07-06 Jakub Jelinek <jakub@redhat.com>
13018 PR tree-optimization/86401
13019 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13020 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13021 (fold_bit_and_mask): ... here. New helper function for match.pd.
13022 * fold-const.h (fold_bit_and_mask): Declare.
13023 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13025 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
13028 * target.def (translate_mode_attribute): New hook.
13029 * targhooks.h (default_translate_mode_attribute): Declare.
13030 * targhooks.c (default_translate_mode_attribute): New function.
13031 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13032 * doc/tm.texi: Regenerate.
13033 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13034 (rs6000_translate_mode_attribute): New function.
13036 2018-07-06 Paul Koning <ni1d@arrl.net>
13038 * doc/md.texi (define_split): Document DONE and FAIL.
13039 (define_peephole2): Ditto.
13041 2018-07-05 Jeff Law <law@redhat.com>
13043 PR tree-optimization/86010
13044 * tree-ssa-dse.c (compute_trims): More aggressively trim at
13045 both the head and tail of mem* and str* calls.
13047 2018-07-05 Jim Wilson <jimw@sifive.com>
13049 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13051 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
13053 * config/aarch64/aarch64-simd.md: correct flags text for
13054 MIN_EXPR replacement.
13056 2018-07-05 James Clarke <jrtc27@jrtc27.com>
13058 * configure: Regenerated.
13060 2018-07-05 Carl Love <cel@us.ibm.com>
13062 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13063 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13064 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13065 VSX_BUILTIN_DOUBLEL_V4SF.
13067 2018-07-05 Martin Sebor <msebor@redhat.com>
13070 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13071 than its domain to compute its the upper bound of a char array.
13073 2018-07-05 Nathan Sidwell <nathan@acm.org>
13075 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13076 * doc/cpp.texi: Update comment.
13077 * doc/tm.texi: Rebuilt.
13078 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13079 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13080 * doc/extend.texi (Backwards Compatibility): Clarify it is system
13081 headers affected by extern "C".
13082 * system.h: Poison NO_IMPLICIT_EXTERN_C.
13083 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13084 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13085 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13086 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13087 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13088 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13089 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13090 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13091 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13092 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13093 config/sparc/sp64-elf.h, config/spu/spu.h,
13094 config/stormy16/stormy16.h, config/v850/v850.h,
13095 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13096 define NO_IMPLICIT_EXTERN_C.
13097 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13099 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13102 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13103 instead of GET_MODE_SIZE when comparing Units.
13105 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13108 * rtlanal.c (set_noop_p): Constrain on mode change,
13109 include hard-reg-set.h
13111 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13113 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13115 2018-07-05 Jakub Jelinek <jakub@redhat.com>
13118 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
13121 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13123 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13125 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
13128 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13130 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13132 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
13134 PR middle-end/86380
13135 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13137 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
13139 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13140 neg_*_op* variables.
13142 2018-07-04 Martin Liska <mliska@suse.cz>
13144 * tree-switch-conversion.c: Define
13145 max_ratio_for_speed and max_ratio_for_size constants.
13147 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
13148 Martin Liska <mliska@suse.cz>
13150 PR middle-end/66240
13153 * common.opt: Rename align options with 'str_' prefix.
13154 * common/config/i386/i386-common.c (set_malign_value): New
13156 (ix86_handle_option): Use it to set -falign-* options/
13157 * config/aarch64/aarch64-protos.h (struct tune_params): Change
13158 type from int to string.
13159 * config/aarch64/aarch64.c: Update default values from int
13161 * config/alpha/alpha.c (alpha_override_options_after_change):
13163 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13164 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13165 max skip conditionally.
13166 * config/i386/freebsd.h (SUBALIGN_LOG): New.
13167 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13168 max skip conditionally.
13169 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13170 max skip conditionally.
13171 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13172 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13173 max skip conditionally.
13174 * config/i386/i386.c (struct ptt): Change type from int to
13176 (ix86_default_align): Set default values.
13177 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13178 max skip conditionally.
13179 * config/i386/iamcu.h (SUBALIGN_LOG): New.
13180 (ASM_OUTPUT_MAX_SKIP_ALIGN):
13181 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13182 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13183 max skip conditionally.
13184 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13185 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13186 * config/i386/x86-64.h (SUBALIGN_LOG): New.
13187 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13188 max skip conditionally.
13189 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13190 * config/ia64/ia64.c (ia64_option_override): Set default values
13191 for alignment options.
13192 * config/m68k/m68k.c: Handle new str_align_* options.
13193 * config/mips/mips.c (mips_set_compression_mode): Change
13195 (mips_option_override): Set default values for options.
13196 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13198 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13200 * config/rx/rx.c (rx_option_override): Likewise.
13201 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13202 (LABEL_ALIGN): Use align_labels_log.
13203 (LOOP_ALIGN): Use align_loops_align.
13204 * config/s390/s390.c (s390_asm_output_function_label): Use new
13206 * config/sh/sh.c (sh_override_options_after_change):
13207 Change type of constants.
13208 * config/spu/spu.c (spu_sched_init): Likewise.
13209 * config/sparc/sparc.c (sparc_option_override): Set default
13210 values for options.
13211 * config/visium/visium.c (visium_option_override): Likewise.
13212 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13213 emit p2align format with last argument if it's not needed.
13214 * doc/invoke.texi: Document extended format of -falign-*.
13215 * final.c: Use align_labels alignment.
13216 * flags.h (struct target_flag_state): Change type to use
13218 (struct align_flags_tuple): New.
13219 (struct align_flags): Likewise.
13220 (align_loops_log): Redefine macro to use new types.
13221 (align_loops_max_skip): Redefine macro to use new types.
13222 (align_jumps_log): Redefine macro to use new types.
13223 (align_jumps_max_skip): Redefine macro to use new types.
13224 (align_labels_log): Redefine macro to use new types.
13225 (align_labels_max_skip): Redefine macro to use new types.
13226 (align_functions_log): Redefine macro to use new types.
13227 (align_loops): Redefine macro to use new types.
13228 (align_jumps): Redefine macro to use new types.
13229 (align_labels): Redefine macro to use new types.
13230 (align_functions): Redefine macro to use new types.
13231 (align_functions_max_skip): Redefine macro to use new types.
13232 (align_loops_value): New macro.
13233 (align_jumps_value): New macro.
13234 (align_labels_value): New macro.
13235 (align_functions_value): New macro.
13236 * function.c (invoke_set_current_function_hook): Propagate
13237 alignment values from flags to global variables default in
13239 * ipa-icf.c (sem_function::equals_wpa): Use
13240 cl_optimization_option_eq instead of memcmp.
13241 * lto-streamer.h (cl_optimization_stream_out): Support streaming
13243 (cl_optimization_stream_in): Likewise.
13244 * optc-save-gen.awk: Support strings in cl_optimization.
13245 * opth-gen.awk: Likewise.
13246 * opts.c (finish_options): Remove error checking of invalid
13248 (MAX_CODE_ALIGN): Remove.
13249 (MAX_CODE_ALIGN_VALUE): Likewise.
13250 (parse_and_check_align_values): New function.
13251 (check_alignment_argument): Likewise.
13252 (common_handle_option): Use check_alignment_argument.
13253 * opts.h (parse_and_check_align_values): Declare.
13254 * toplev.c (init_alignments): Remove.
13255 (read_log_maxskip): New.
13256 (parse_N_M): Likewise.
13257 (parse_alignment_opts): Likewise.
13258 (backend_init_target): Remove usage of init_alignments.
13259 * toplev.h (parse_alignment_opts): Declare.
13260 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13262 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13263 * tree.c (cl_option_hasher::equal): New.
13264 * varasm.c: Use new global macros.
13266 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
13268 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13269 Use a simpler align directive also if MAXSKIP = ALIGN-1.
13270 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13271 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13272 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13273 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13274 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13275 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13276 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13277 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13278 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13279 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13280 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13282 2018-07-04 Martin Liska <mliska@suse.cz>
13283 Jonathan Wakely <jwakely@redhat.com>
13285 * coverage.c: Use correct type.
13286 * doc/invoke.texi: Language correction.
13288 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
13291 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13292 ENDBRANCH for non-tail call which may return via indirect branch.
13293 * doc/extend.texi: Document indirect_return attribute.
13295 2018-07-03 Martin Sebor <msebor@redhat.com>
13297 PR tree-optimization/86274
13298 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13300 (format_floating): Correct handling of infinities and NaNs.
13302 2018-07-03 Martin Sebor <msebor@redhat.com>
13304 * print-tree.c (print_real_cst): New function.
13305 (print_node_brief): Call it.
13306 (print_node): Ditto.
13308 2018-07-03 Jeff Law <law@redhat.com>
13310 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13311 into a single pattern.
13313 * config/h8300/h8300.md (ors code_iterator): New.
13314 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13315 a single pattern and single splitter.
13316 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13317 (iorqi3_1, xorqi3_1): Likewise.
13318 (iorqi3, xorqi3 expanders): Similarly.
13320 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13321 (movmd_internal) into a single pattern using the P mode iterator.
13322 (movmd splitters): Similarly.
13323 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13324 (movsd splitters): Similarly.
13326 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13327 ADDB, ADDW and ADDL into a single ADD attribute which selects the
13328 right table based on the size of the operand.
13329 * config/h8300/h8300.md (length_table): Corresponding changes. All
13330 references to "addb", "addw" and "addl" changed to "add".
13331 (btst patterns): Merge two variants into a single pattern.
13332 (tstqi, tsthi): Likewise.
13333 (addhi3_incdec, addsi3_incdec): Likewise.
13334 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13335 (mulhi3, mulsi3): Likewise.
13336 (udivhi3, udivsi3): Likewise.
13337 (divhi3, divsi3): Likewise.
13338 (andorqi3, andorhi3, andorsi3): Likewise.
13340 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
13343 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13344 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13346 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13348 PR tree-optimization/85694
13349 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13350 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13352 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13353 (<u>avg<mode>3_ceil): New patterns.
13355 2018-07-03 David Malcolm <dmalcolm@redhat.com>
13357 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13358 scan-tree-dump directive.
13359 * gcc.dg/vect/slp-perm-2.c: Likewise.
13360 * gcc.dg/vect/slp-perm-3.c: Likewise.
13361 * gcc.dg/vect/slp-perm-5.c: Likewise.
13362 * gcc.dg/vect/slp-perm-6.c: Likewise.
13363 * gcc.dg/vect/slp-perm-7.c: Likewise.
13364 * gcc.dg/vect/slp-perm-8.c: Likewise.
13366 2018-07-03 Marek Polacek <polacek@redhat.com>
13368 PR middle-end/86202
13369 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13371 2018-07-03 Richard Biener <rguenther@suse.de>
13374 * tree-ssa-structalias.c (find_func_clobbers): Properly
13375 handle indirect calls.
13377 2018-07-03 Jeff Law <law@redhat.com>
13379 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13380 (shifts): New code iterator.
13381 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13382 expander. Fix HImode handling on H8/SX.
13383 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13384 (subqi3, subhi3, subsi3 expanders): Likewise.
13385 (andqi3, andhi3, andsi3 expanders): Likewise.
13386 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13387 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13388 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13389 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13390 (zero_extendqihi2, zero_extendqisi2): Likewise.
13391 (extendqihi2, extendqisi2): Likewise.
13392 (rotlqi3, rotlhi3, rotlsi3): Likewise.
13393 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13394 (rotlqi3_1, rotlhi3_1): Likewise.
13395 (logicalhi3_sn, logicalsi3_sn): Likewise.
13396 (logicalhi3, logicalsi3): Likewise.
13398 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13400 * tree-vect-patterns.c (vect_recog_rotate_pattern)
13401 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13402 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13403 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13404 type to append_pattern_def_seq instead of creating a stmt_vec_info
13406 (build_mask_conversion): Likewise. Remove vinfo argument.
13407 (vect_add_conversion_to_patterm): Likewise, renaming to...
13408 (vect_add_conversion_to_pattern): ...this.
13409 (vect_recog_mask_conversion_pattern): Update call to
13410 build_mask_conversion. Pass the vector type to
13411 append_pattern_def_seq here too.
13412 (vect_recog_gather_scatter_pattern): Update call to
13413 vect_add_conversion_to_pattern.
13415 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13417 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13418 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13419 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13420 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13421 STMT_VINFO_PATTERN_DEF_SEQ to null here.
13422 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13423 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13424 append_pattern_def_seq instead of new_pattern_def_seq.
13425 (vect_recog_divmod_pattern): Do both of the above.
13426 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13429 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13431 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13432 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13433 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13434 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13435 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13436 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13437 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13438 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13439 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13440 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13441 parameter with a single stmt_vec_info.
13442 (vect_recog_func_ptr): Likewise.
13443 (vect_recog_gather_scatter_pattern): Likewise, folding in...
13444 (vect_try_gather_scatter_pattern): ...this.
13445 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13446 the stmt_vec_info of the statement to be matched. Don't clear
13447 STMT_VINFO_RELATED_STMT.
13448 (vect_pattern_recog): Update call accordingly.
13450 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13452 PR tree-optimization/85694
13453 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13454 (uavgM3_ceil): Document new optabs.
13455 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13456 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13458 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13459 (savg_ceil_optab): New optabs.
13460 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13461 (vect_vect_recog_func_ptrs): Add it.
13462 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13463 constant directly from the associated lhs.
13465 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13467 * tree-vect-patterns.c (vect_split_statement): New function.
13468 (vect_convert_input): Use it to try to split an existing cast.
13470 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13472 * poly-int.h (print_hex): New function.
13473 * dumpfile.h (dump_dec, dump_hex): Declare.
13474 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13475 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13476 min_input_precision, operation_precision and operation_sign.
13477 * tree-vect-patterns.c (vect_get_range_info): New function.
13478 (vect_same_loop_or_bb_p, vect_single_imm_use)
13479 (vect_operation_fits_smaller_type): Delete.
13480 (vect_look_through_possible_promotion): Add an optional
13481 single_use_p parameter.
13482 (vect_recog_over_widening_pattern): Rewrite to use new
13483 stmt_vec_info infomration. Handle one operation at a time.
13484 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13485 (vect_truncatable_operation_p, vect_set_operation_type)
13486 (vect_set_min_input_precision): New functions.
13487 (vect_determine_min_output_precision_1): Likewise.
13488 (vect_determine_min_output_precision): Likewise.
13489 (vect_determine_precisions_from_range): Likewise.
13490 (vect_determine_precisions_from_users): Likewise.
13491 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13492 (vect_vect_recog_func_ptrs): Put over_widening first.
13494 (vect_pattern_recog): Call vect_determine_precisions.
13496 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13498 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13499 statements that have been replaced by further pattern statements.
13500 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13502 2018-07-03 Richard Biener <rguenther@suse.de>
13504 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13505 always set *dt. Dump vectype in vectype overload.
13506 * dumpfile.h (dump_gimple_expr): New function.
13507 (dump_gimple_expr_loc): Likewise.
13508 * dumpfile.c (dump_gimple_expr): New function.
13509 (dump_gimple_expr_loc): Likewise.
13511 2018-07-02 Jeff Law <law@redhat.com>
13513 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13514 the H8/300, H8/300H and H8/S variants into a single pattern.
13515 (movhi_h8300, movqi_h8300hs): Similarly.
13516 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13517 (QHI mode iterator): New.
13519 * config/h8300/h8300.md: Remove trailing whitespace.
13521 2018-07-02 Jim Wilson <jimw@sifive.com>
13523 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13524 instead of emit_insn for interrupt returns.
13525 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13526 (riscv_sret, riscv_uret): Likewise.
13528 2018-07-02 David Malcolm <dmalcolm@redhat.com>
13530 * pretty-print.c (selftest::test_pp_format): Move save and restore
13531 of quotes to class auto_fix_quotes, and add an instance.
13532 * selftest.c: Include "intl.h".
13533 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13534 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13535 * selftest.h (selftest::auto_fix_quotes): New class.
13537 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
13539 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13540 (aarch64_sve_prepare_conditional_op): Remove.
13541 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13542 Allow aarch64_simd_reg_or_zero as select operand; remove
13543 the aarch64_sve_prepare_conditional_op call.
13544 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13545 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13546 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13547 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13548 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13549 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13550 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13551 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13552 and a splitters to match all of the *_any patterns.
13553 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13555 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13556 (SVE_COND_FP_BINARY_REV): Remove.
13557 (sve_int_op_rev, sve_fp_op_rev): New.
13558 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13559 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13560 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13561 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13562 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13563 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13564 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13565 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13566 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13568 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13569 Remove match_dup 1 from the inner unspec.
13570 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13572 * config/aarch64/aarch64.md (movprfx): New attr.
13573 (length): Default movprfx to 8.
13574 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13575 (*madd<SVE_I>, *msub<SVE_I): Likewise.
13576 (*<su>mul<SVE_I>3_highpart): Likewise.
13577 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13578 (*v<ASHIFT><SVE_I>3): Likewise.
13579 (*<su><MAXMIN><SVE_I>3): Likewise.
13580 (*<su><MAXMIN><SVE_F>3): Likewise.
13581 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13582 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13583 (*div<SVE_F>4): Likewise.
13585 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
13587 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13590 2018-07-02 Richard Biener <rguenther@suse.de>
13592 PR tree-optimization/86363
13593 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
13594 memset argument refers to a non-variable address.
13596 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
13598 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
13599 {PLUS,MINUS}_EXPR code to...
13600 (adjust_symbolic_bound): ...here,
13601 (combine_bound): ...here,
13602 (set_value_range_with_overflow): ...and here.
13604 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
13606 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
13608 (extract_range_from_abs_expr): ...here.
13610 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
13612 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
13613 -fno-omit-frame-pointer when not optimizing.
13615 2018-07-02 Martin Liska <mliska@suse.cz>
13618 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
13619 (propagate_nothrow): Likewise.
13621 2018-07-02 Martin Liska <mliska@suse.cz>
13624 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
13626 2018-07-02 David Malcolm <dmalcolm@redhat.com>
13628 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
13629 function in r262149, changing "loc" param from source_location to
13630 const dump_location_t &.
13631 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
13632 declaration, as above.
13634 2018-07-01 Paul Koning <ni1d@arrl.net>
13636 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
13637 -munit-asm, -mgnu-asm, -mdec-asm.
13638 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
13639 (pdp11_output_labelref): New.
13640 (pdp11_output_def): New.
13641 (pdp11_output_addr_vec_elt): New.
13642 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
13643 %# and %@ format codes.
13644 (pdp11_option_override): New.
13645 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
13646 (pdp11_output_ident): New.
13647 (pdp11_asm_named_section): New.
13648 (pdp11_asm_init_sections): New.
13649 (pdp11_file_start): New.
13650 (pdp11_file_end): New.
13651 (output_ascii): Use .ascii/.asciz for -mdec-asm.
13652 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
13653 %o, like %c but octal.
13654 (pdp11_option_override): New.
13655 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
13657 (DATA_SECTION_ASM_OP): Ditto.
13658 (READONLY_DATA_SECTION_ASM_OP): New.
13659 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
13660 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
13661 (ASM_OUTPUT_LABELREF): Ditto.
13662 (ASM_OUTPUT_DEF): Ditto.
13663 (ASM_OUTPUT_EXTERNAL): New.
13664 (ASM_OUTPUT_SOURCE_FILENAME): New.
13665 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
13666 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
13667 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
13668 %# and %@ format codes.
13669 * config/pdp11/pdp11.opt (mgnu-asm): New.
13670 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
13671 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
13672 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
13674 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
13676 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
13677 dereferencing path[] beyond its length.
13679 (debug_all_paths): New.
13680 (rewire_first_differing_edge): New.
13681 (adjust_paths_after_duplication): New.
13682 (duplicate_thread_path): Call adjust_paths_after_duplication.
13684 (thread_through_all_blocks): Add new argument to
13685 duplicate_thread_path.
13687 2018-06-30 Jim Wilson <jimw@sifive.com>
13689 * config/riscv/predicates.md (p2m1_shift_operand): New.
13690 (high_mask_shift_operand): New.
13691 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
13692 pattern using p2m1_shift_operand.
13693 (lshsi3_zero_extend_3+2): New combiner pattern using
13694 high_mask_shift_operand.
13696 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13698 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
13700 (vect_recog_rotate_pattern): ...here.
13701 (vect_convert_input): Try to insert casts of invariants in the
13703 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
13704 preheader to be empty.
13706 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13708 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
13709 vector type. If given, install it in the new statement's
13710 STMT_VINFO_VECTYPE.
13711 (vect_element_precision): New function.
13712 (vect_unpromoted_value): New struct.
13713 (vect_unpromoted_value::vect_unpromoted_value): New function.
13714 (vect_unpromoted_value::set_op): Likewise.
13715 (vect_look_through_possible_promotion): Likewise.
13716 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
13717 (vect_widened_op_tree, vect_convert_input): Likewise.
13718 (vect_convert_inputs, vect_convert_output): Likewise.
13719 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
13720 to handle the optional cast of the multiplication result and
13721 vect_widened_op_tree to detect the widened multiplication itself.
13722 Do not require the input and output of promotion casts to have
13723 the same sign, but base the signedness of the operation on the
13724 input rather than the result. If the pattern includes two
13725 promotions, check that those promotions have the same sign.
13726 Do not restrict the MULT_EXPR handling to a double-width result;
13727 handle quadruple-width results and wider. Use vect_convert_inputs
13728 to convert the inputs to the common type.
13729 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
13730 to handle the optional cast of the ABS result. Also allow a sign
13731 change or a sign extension between the ABS and MINUS.
13732 Use vect_widened_op_tree to detect the widened subtraction and use
13733 vect_convert_inputs to convert the inputs to the common type.
13734 (vect_handle_widen_op_by_const): Delete.
13735 (vect_recog_widen_op_pattern): New function.
13736 (vect_recog_widen_mult_pattern): Use it.
13737 (vect_recog_widen_shift_pattern): Likewise.
13738 (vect_recog_widen_sum_pattern): Use
13739 vect_look_through_possible_promotion to handle the promoted
13742 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13744 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
13745 the containing gimple_seq *.
13746 * gimple-iterator.h (gsi_for_stmt): Declare it.
13747 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
13748 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
13749 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13750 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13751 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
13753 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
13755 (vect_mark_pattern_stmts): ...here. Handle cases in which the
13756 statement being replaced is part of an existing pattern
13757 definition sequence, inserting the new pattern statements before
13759 (vect_pattern_recog_1): Don't return a bool. If the statement
13760 is already part of a pattern, instead apply pattern matching
13761 to the pattern definition statements. Don't clear the
13762 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
13763 (vect_pattern_recog): Don't break after the first match;
13764 continue processing the pattern definition statements instead.
13765 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
13767 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13769 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
13770 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13771 (vect_recog_widen_sum_pattern): Use it.
13773 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13775 * tree-vect-loop.c (vectorizable_reduction): Assert that the
13776 phi is not a pattern statement and has not been replaced by
13777 a pattern statement.
13778 * tree-vect-patterns.c (type_conversion_p): Don't check
13779 STMT_VINFO_IN_PATTERN_P.
13780 (vect_recog_vector_vector_shift_pattern): Likewise.
13781 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
13782 the pattern statement rather than the original statement; check
13783 directly for a WIDEN_MULT_EXPR here.
13784 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
13785 vect_is_simple_use to return the pattern statement rather
13786 than the original statement; use is_pattern_stmt_p to check
13787 for such a pattern statement.
13788 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
13789 to return the pattern statement rather than the original statement;
13790 don't do the same transformation here.
13791 (vect_is_simple_use): If the defining statement has been replaced
13792 by a pattern statement, return the pattern statement instead.
13793 Remove the corresponding (local) transformation from the vectype
13796 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
13798 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
13799 end and default to null.
13800 * tree-vect-loop.c (vect_create_epilog_for_reduction)
13801 (vectorizable_reduction): Update calls accordingly, dropping the
13802 gimple ** argument if the passed-back statement isn't needed.
13803 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
13804 (vect_recog_rotate_pattern): Likewise.
13805 (vect_recog_mask_conversion_pattern): Likewise.
13806 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
13807 (vect_mask_constant_operand_p): Likewise.
13808 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
13809 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
13810 (get_group_load_store_type, get_load_store_type): Likewise.
13811 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
13812 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
13813 (vectorizable_conversion, vectorizable_assignment): Likewise.
13814 (vectorizable_shift, vectorizable_operation): Likewise.
13815 (vectorizable_store, vect_is_simple_cond): Likewise.
13816 (vectorizable_condition, vectorizable_comparison): Likewise.
13817 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
13818 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
13819 and move it to the end. Cope with null def_stmt_outs.
13821 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
13823 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
13825 2018-06-29 Jeff Law <law@redhat.com>
13827 * config/v850/v850.c (v850_legitimate_address_p): Handle large
13828 displacements for TARGET_V850E2V3 and newer.
13829 (TARGET_LRA_P): Remove. Defaults to LRA now.
13830 * config/v850/v850.md (sign23byte_load): Remove.
13831 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
13832 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
13834 2018-06-29 Martin Liska <mliska@suse.cz>
13837 * coverage.c (coverage_init): Mangle full path name.
13838 * doc/invoke.texi: Document the change.
13839 * gcov-io.c (mangle_path): New.
13840 * gcov-io.h (mangle_path): Likewise.
13841 * gcov.c (mangle_name): Use mangle_path for path mangling.
13843 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13845 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
13846 if starting source register is not even.
13848 2018-06-29 Martin Liska <mliska@suse.cz>
13850 PR tree-optimization/86263
13851 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
13852 Make edge redirection.
13854 2018-06-29 David Malcolm <dmalcolm@redhat.com>
13856 * dumpfile.c (dump_loc): Add indentation based on scope depth.
13857 (dump_scope_depth): New variable.
13858 (get_dump_scope_depth): New function.
13859 (dump_begin_scope): New function.
13860 (dump_end_scope): New function.
13861 * dumpfile.h (get_dump_scope_depth): New declaration.
13862 (dump_begin_scope): New declaration.
13863 (dump_end_scope): New declaration.
13864 (class auto_dump_scope): New class.
13865 (AUTO_DUMP_SCOPE): New macro.
13866 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
13869 2018-06-29 Richard Biener <rguenther@suse.de>
13871 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
13872 compute_all_dependences succeeds.
13873 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
13874 exceed --param loop-max-datarefs-for-datadeps.
13876 2018-06-29 Jakub Jelinek <jakub@redhat.com>
13878 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
13880 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
13883 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
13884 alternative 0 in preferred_for_speed attribute.
13886 2018-06-28 Paul Koning <ni1d@arrl.net>
13888 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
13889 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
13890 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
13891 * config/pdp11/pdp11.md: Correct "length" attribute calculation
13892 for shift insn patterns.
13894 2018-06-28 David Malcolm <dmalcolm@redhat.com>
13896 * cgraph.c (cgraph_node::get_body): Replace assignments to
13897 "dump_file" with calls to set_dump_file.
13898 * dumpfile.c (alt_dump_file): Make static, and group with...
13899 (alt_flags): ...this definition.
13900 (dumps_are_enabled): New variable.
13901 (refresh_dumps_are_enabled): New function.
13902 (set_dump_file): New function.
13903 (set_alt_dump_file): New function.
13904 (gcc::dump_manager::dump_start): Replace assignments to
13905 "dump_file" and "alt_dump_file" with calls to set_dump_file and
13907 (gcc::dump_manager::dump_finish): Likewise.
13908 * dumpfile.h (alt_dump_file): Delete decl.
13909 (dumps_are_enabled): New variable decl.
13910 (set_dump_file): New function decl.
13911 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
13913 * tree-nested.c (lower_nested_functions): Replace assignments to
13914 "dump_file" with calls to set_dump_file.
13916 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
13918 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
13919 goto_locus of each outgoing edge of each basic block.
13921 2018-06-28 Richard Biener <rguenther@suse.de>
13923 * dwarf2out.c (decl_scope_table): Remove.
13924 (push_decl_scope): Likewise.
13925 (pop_decl_scope): Likewise.
13926 (gen_type_die_for_member): Do not call push/pop_decl_scope.
13927 (gen_struct_or_union_type_die): Likewise.
13928 (gen_tagged_type_die): Likewise.
13929 (dwarf2out_init): Do not initialize decl_scope_table.
13930 (dwarf2out_c_finalize): Do not free it.
13932 2018-06-28 Richard Biener <rguenther@suse.de>
13934 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
13935 deciding whether to not re-use a DIE.
13937 2018-06-28 Richard Biener <rguenther@suse.de>
13939 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
13940 DW_AT_abstract_origin attribute.
13942 2018-06-28 Martin Liska <mliska@suse.cz>
13944 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
13945 Use newly introduced constants.
13946 * tree-switch-conversion.h (struct jump_table_cluster):
13947 Define max_ratio_for_size and max_ratio_for_speed.
13949 2018-06-28 Martin Liska <mliska@suse.cz>
13951 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
13952 Add new checking assert to catch invalid state.
13953 (jump_table_cluster::can_be_handled): Handle single case
13955 (jump_table_cluster::is_beneficial): Bail out for such case.
13956 (bit_test_cluster::find_bit_tests):
13957 Add new checking assert to catch invalid state.
13958 (bit_test_cluster::can_be_handled): Handle single case
13960 (bit_test_cluster::is_beneficial): Bail out for such case.
13961 (switch_decision_tree::analyze_switch_statement):
13964 2018-06-28 Martin Liska <mliska@suse.cz>
13966 * common.opt: Introduce -completion option.
13967 * gcc.c (driver_handle_option): Handle it.
13968 (driver::main): Print completions if completion
13970 * opt-suggestions.c (option_proposer::get_completions):
13972 (option_proposer::suggest_completion): Likewise.
13973 (option_proposer::find_param_completions): Likewise.
13974 (verify_autocompletions): Likewise.
13975 (test_completion_valid_options): Likewise.
13976 (test_completion_valid_params): Likewise.
13977 (in_completion_p): Likewise.
13978 (empty_completion_p): Likewise.
13979 (test_completion_partial_match): Likewise.
13980 (test_completion_garbage): Likewise.
13981 (opt_proposer_c_tests): Likewise.
13982 * opt-suggestions.h: Declare new functions.
13983 * opts.c (common_handle_option): Handle OPT__completion_.
13984 * selftest-run-tests.c (selftest::run_tests): Add
13985 opt_proposer_c_tests.
13986 * selftest.c (assert_str_startswith): New.
13987 * selftest.h (assert_str_startswith): Likewise.
13988 (opt_proposer_c_tests): New.
13989 (ASSERT_STR_STARTSWITH): Likewise.
13991 2018-06-28 Martin Liska <mliska@suse.cz>
13993 * Makefile.in: Add opt-suggestions.o.
13994 * gcc-main.c: Include opt-suggestions.h.
13995 * gcc.c (driver::driver): Likewise.
13996 (driver::~driver): Remove m_option_suggestions.
13997 (driver::build_option_suggestions): Moved to option_proposer.
13998 (driver::suggest_option): Likewise.
13999 (driver::handle_unrecognized_options): Use option_proposer.
14000 * gcc.h (class driver): Add new memver m_option_proposer.
14001 * opt-suggestions.c: New file.
14002 * opt-suggestions.h: New file.
14004 2018-06-28 Martin Liska <mliska@suse.cz>
14006 * vec.h (class auto_string_vec): New (moved from auto_argvec).
14007 (auto_string_vec::~auto_string_vec): Likewise.
14009 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
14011 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14012 prevent_decl_creation_for_types fields up and add reset_location field.
14013 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14014 statement if id->reset_location is true.
14015 (copy_edges_for_bb): Do not set goto_locus on the new edges if
14016 id->reset_location is true.
14017 (copy_phis_for_bb): Force input_location on the arguments if
14018 id->reset_location is true.
14019 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14020 is set on the function to be inlined.
14022 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
14024 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14026 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
14028 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14029 registers for Pmode.
14030 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14031 hard registers for the clobbered pseudo.
14033 2018-06-27 Paul Koning <ni1d@arrl.net>
14035 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14036 mutually exclusive options.
14037 * config/pdp11/constraints.md (h): New constraint.
14038 (O): Update definition to match shift code generation.
14039 (D): New constraint.
14040 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14042 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14044 (output_jump): Change arguments.
14045 (pdp11_fixed_cc_regs): New function.
14046 (pdp11_cc_mode): Ditto.
14047 (pdp11_expand_shift): Ditto.
14048 (pdp11_assemble_shift): Ditto.
14049 (pdp11_small_shift): Ditto.
14050 (pdp11_branch_cost): Remove.
14051 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14053 (pdp11_register_move_cost): Update for CC registers.
14054 (pdp11_rtx_costs): Add case for LSHIFTRT.
14055 (pdp11_output_jump): Add CCNZ mode conditional branches.
14056 (notice_update_cc_on_set): Remove.
14057 (pdp11_cc_mode): New function.
14058 (simple_memory_operand): Correct pre/post decrement case.
14059 (no_side_effect_operand): New function.
14060 (pdp11_regno_reg_class): Add CC_REGS class.
14061 (pdp11_fixed_cc_regs): New function.
14062 (pdp11_small_shift): New function.
14063 (pdp11_expand_shift): New function to expand shift insns.
14064 (pdp11_assemble_shift): New function to output shifts.
14065 (pdp11_branch_cost): Remove.
14066 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14067 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14068 (WCHAR_TYPE): Ditto.
14069 (PTRDIFF_TYPE): Ditto.
14070 (ADJUST_INSN_LENGTH): New macro.
14071 (FIXED_REGISTERS): Add CC registers.
14072 (CALL_USED_REGISTERS): Ditto.
14073 (reg_class): Ditto.
14074 (REG_CLASS_NAMES): Ditto.
14075 (REG_CLASS_CONTENTS): Ditto.
14076 (SELECT_CC_MODE): Use new function.
14077 (TARGET_FLAGS_REGNUM): New macro.
14078 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14079 (cc0_reg_rtx): Remove.
14080 (CC_STATUS_MDEP): Remove.
14081 (CC_STATUS_MDEFP_INIT): Remove.
14082 (CC_IN_FPU): Remove.
14083 (NOTICE_UPDATE_CC): Remove.
14084 (REGISTER_NAMES): Add CC registers.
14085 (BRANCH_COST): Change to constant 1.
14086 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14088 * config/pdp11/pdp11.opt (mbcopy): Remove.
14089 (mbcopy-builtin): Remove.
14090 (mbranch-cheap): Remove.
14091 (mbranch-expensive): Remove.
14092 * config/pdp11/predicates.md (expand_shift_operand): Update to
14093 match shift code generation.
14094 (ccnz_operator): New predicate.
14095 * doc/invoke.texi (PDP-11 Options): Remove deleted options
14096 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14097 Remove non-existent option -mabshi, -mno-abshi. Document mutually
14099 * doc/md.texi (PDP-11): Document new D and h constraints. Update
14100 description of O constraint.
14102 2018-06-27 Jeff Law <law@redhat.com>
14103 Austin Law <austinklaw@gmail.com>
14105 * config/v850/v850.md (addsi3_set_flags): New pattern.
14106 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14107 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14108 (zero_extendhisi2_v850_set_flags): Likewise.
14109 (zero_extendqisi2_v850_set_flags): Likewise.
14110 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14111 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14112 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14114 * config/v850/v850-protos.h (notice_update_cc): Remove.
14115 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14116 (v850_print_operand): Handle 'D' and "d".
14117 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14118 Add handling of arithmetic/logical operations compared against zero.
14119 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14120 Do not look at v850_compare_op, instead get mode from last argument.
14121 (v850_gen_compare): Remove
14122 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14123 after reload for prologue insns.
14124 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14126 (construct_save_jarl): Likewise.
14127 (TARGET_FLAGS_REGNUM): Define.
14128 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14129 (NOTICE_UPDATE_CC): Remove.
14130 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14131 than cc0. Conditionalize on reload_completed.
14132 (cmpsi_insn, setfcc_insn): Likewise.
14133 (tst1 splitter): Turn into define_and_split which sets the flags
14135 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14136 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14137 (cstoresf4, cstoredf4): Clobber the flags.
14138 (cmpsi, cmpsf, cmpdf): Remove expanders.
14139 (setf_insn): Remove pattern.
14140 (addsi3): Turn into define_and_split which clobbers the flags after
14141 reload and a suitable pattern (addsi3_clobber_flags) for use after
14143 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14144 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14145 (ashrsi3, ashrsi3_v850e2): Likewise.
14146 (bins): Clobber the flags.
14147 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14148 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14149 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14150 (call_value_internal_short, call_value_internal_long): Likewise.
14151 (callt_save_interrupt, callt_return_interrupt): Likewise.
14152 (save_interrupt, return_interrupt): Likewise.
14153 (callt_save_all_interrupt, save_all_interrupt): Likewise.
14154 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14155 (restore_all_interrupt, _restore_all_interrupt): Likewise.
14156 (All FP comparisons): Only allow after reload has completed.
14158 (divh, divhu): Tweak output template.
14159 (branch_z_normal, branch_z_invert): Remove
14160 (branch_nz_normal, branch_nz_invert): Likewise.
14161 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14163 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14164 * config/v850/v850.c (notice_update_cc): Remove.
14165 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
14166 (CC_NO_CARRY): Likewise.
14167 (NOTICE_UPDATE_CC): Define to nothing.
14168 * config/v850/v850.md: Remove block comment on cc0 handling
14169 Remove "cc" attribute from all patterns. Remove cc_status handling
14170 from all patterns. Minor formatting fixes.
14172 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14174 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14175 (cortex-a76.cortex-a55): Likewise.
14176 * config/aarch64/aarch64-tune.md: Regenerate.
14177 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14178 cortex-a76.cortex-a55.
14180 2018-06-27 Jeff Law <law@redhat.com>
14182 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14183 (MULTILIB_DIRNAMES): Similarly.
14185 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
14187 * gimple.h (gimple_return_retbnd): Delete.
14188 (gimple_return_set_retbnd): Likewise.
14189 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14190 gimple_return_set_retbnd.
14191 * gimple-pretty-print.c (dump_gimple_return): Remove call to
14192 gimple_return_retbnd and adjust.
14193 * tree-inline.h (struct copy_body_data): Remove retbnd field.
14194 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14195 Explicitly return NULL in a couple more cases. Move assertion
14196 on debug statements and remove unreachable code.
14197 (reset_debug_binding): Do not test id->retbnd.
14198 (expand_call_inline): Do not set it.
14200 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
14202 * configure.ac: Add --disable-gcov option.
14203 * configure: Regenerate.
14204 * Makefile.in: Honour @enable_gcov@.
14205 * doc/install.texi: Document --disable-gcov.
14207 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14209 * config/arm/arm-cpus.in (cortex-a76): New entry.
14210 (cortex-a76.cortex-a55): Likewise.
14211 * config/arm/arm-tables.opt: Regenerate.
14212 * config/arm/arm-tune.md: Likewise.
14213 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14214 * doc/invoke.texi (ARM Options): Document cortex-a76 and
14215 cortex-a76.cortex-a55.
14217 2018-06-27 Tamar Christina <tamar.christina@arm.com>
14220 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14222 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
14224 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14226 (EPILOGUE_USES): Likewise.
14228 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
14230 * tree-inline.c (remap_location): New function extracted from...
14231 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
14232 (copy_phis_for_bb): ...here. Call remap_location.
14233 (copy_cfg_body): Adjust call to copy_edges_for_bb.
14235 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
14237 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14238 unaligned vsx for 16B memset.
14240 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
14243 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14244 ieee128_float_type_node to long_double_type_node unless
14245 TARGET_LONG_DOUBLE_128 is set.
14247 2018-06-26 David Malcolm <dmalcolm@redhat.com>
14249 * cfgloop.c (get_loop_location): Convert return type from
14250 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14251 by implicit construction from rtx_insn *, and using
14252 dump_user_location_t::from_function_decl for the fallback case.
14253 * cfgloop.h (get_loop_location): Convert return type from
14254 location_t to dump_user_location_t.
14255 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14256 dump_printf_loc to pass in a dump_location_t rather than a
14257 location_t, via the gimple stmt.
14258 * coverage.c (get_coverage_counts): Update calls to
14259 dump_printf_loc to pass in dump_location_t rather than a
14261 * doc/optinfo.texi (Dump types): Convert example of
14262 dump_printf_loc from taking "locus" to taking "insn". Update
14263 description of the "_loc" calls to cover dump_location_t.
14264 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14266 (dump_user_location_t::dump_user_location_t): New constructors,
14267 from gimple *stmt and rtx_insn *.
14268 (dump_user_location_t::from_function_decl): New function.
14269 (dump_loc): Make static.
14270 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14271 const dump_location_t &.
14272 (dump_generic_expr_loc): Delete.
14273 (dump_printf_loc): Convert param "loc" from location_t to
14274 const dump_location_t &.
14275 (selftest::test_impl_location): New function.
14276 (selftest::dumpfile_c_tests): New function.
14277 * dumpfile.h: Include "profile-count.h".
14278 (class dump_user_location_t): New class.
14279 (struct dump_impl_location_t): New struct.
14280 (class dump_location_t): New class.
14281 (dump_printf_loc): Convert 2nd param from source_location to
14282 const dump_location_t &.
14283 (dump_generic_expr_loc): Delete.
14284 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14285 const dump_location_t &.
14286 * gimple-fold.c (fold_gimple_assign): Update call to
14287 dump_printf_loc to pass in a dump_location_t rather than a
14288 location_t, via the gimple stmt.
14289 (gimple_fold_call): Likewise.
14290 * gimple-loop-interchange.cc
14291 (loop_cand::analyze_iloop_reduction_var): Update for change to
14292 check_reduction_path.
14293 (tree_loop_interchange::interchange): Update for change to
14294 find_loop_location.
14295 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14296 change in return-type of find_loop_location.
14297 (graphite_regenerate_ast_isl): Likewise.
14298 * graphite-optimize-isl.c (optimize_isl): Likewise.
14299 * graphite.c (graphite_transform_loops): Likewise.
14300 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14301 pass in a dump_location_t rather than a location_t, via the
14303 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14304 * ipa.c (walk_polymorphic_call_targets): Likewise.
14305 * loop-unroll.c (report_unroll): Convert "locus" param from
14306 location_t to dump_location_t.
14307 (decide_unrolling): Update for change to get_loop_location's
14309 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14310 location_t to dump_user_location_t.
14311 (grid_find_single_omp_among_assignments_1): Updates calls to
14312 dump_printf_loc to pass in a dump_location_t rather than a
14313 location_t, via the gimple stmt.
14314 (grid_parallel_clauses_gridifiable): Convert "tloc" from
14315 location_t to dump_location_t. Updates calls to dump_printf_loc
14316 to pass in a dump_location_t rather than a location_t, via the
14318 (grid_inner_loop_gridifiable_p): Likewise.
14319 (grid_dist_follows_simple_pattern): Likewise.
14320 (grid_gfor_follows_tiling_pattern): Likewise.
14321 (grid_target_follows_gridifiable_pattern): Likewise.
14322 (grid_attempt_target_gridification): Convert initialization
14323 of local "grid" from memset to zero-initialization; FIXME: does
14324 this require C++11? Update call to dump_printf_loc to pass in a
14325 optinfo_location rather than a location_t, via the gimple stmt.
14326 * profile.c (read_profile_edge_counts): Updates call to
14327 dump_printf_loc to pass in a dump_location_t rather than a
14329 (compute_branch_probabilities): Likewise.
14330 * selftest-run-tests.c (selftest::run_tests): Call
14332 * selftest.h (dumpfile_c_tests): New decl.
14333 * tree-loop-distribution.c (pass_loop_distribution::execute):
14334 Update for change in return type of find_loop_location.
14335 * tree-parloops.c (parallelize_loops): Likewise.
14336 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14337 "locus" from location_t to dump_user_location_t.
14338 (canonicalize_loop_induction_variables): Likewise.
14339 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14340 for change in return type of find_loop_location.
14341 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14342 to dump_printf_loc to pass in a dump_location_t rather than a
14343 location_t, via the stmt.
14344 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14346 * tree-vect-loop-manip.c (find_loop_location): Convert return
14347 type from source_location to dump_user_location_t.
14348 (vect_do_peeling): Update for above change.
14349 (vect_loop_versioning): Update for change in type of
14351 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14352 from location_t to dump_user_location_t.
14353 (vect_estimate_min_profitable_iters): Update for change in type
14355 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14356 location_t to dump_location_t.
14357 (vect_slp_bb): Update for change in type of vect_location.
14358 * tree-vectorizer.c (vect_location): Convert from source_location
14359 to dump_user_location_t.
14360 (try_vectorize_loop_1): Update for change in vect_location's type.
14361 (vectorize_loops): Likewise.
14362 (increase_alignment): Likewise.
14363 * tree-vectorizer.h (vect_location): Convert from source_location
14364 to dump_user_location_t.
14365 (find_loop_location): Convert return type from source_location to
14366 dump_user_location_t.
14367 (check_reduction_path): Convert 1st param from location_t to
14368 dump_user_location_t.
14369 * value-prof.c (check_counter): Update call to dump_printf_loc to
14370 pass in a dump_user_location_t rather than a location_t; update
14371 call to error_at for change in type of "locus".
14372 (check_ic_target): Update call to dump_printf_loc to
14373 pass in a dump_user_location_t rather than a location_t, via the
14376 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
14378 * config/s390/s390.h (enum processor_flags): Do not use
14379 default tune parameter when -march was specified.
14381 2018-06-26 Jakub Jelinek <jakub@redhat.com>
14384 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14385 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14388 2018-06-26 Richard Biener <rguenther@suse.de>
14390 PR tree-optimization/86287
14392 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14393 (vect_analyze_loop): Initialize n_stmts.
14395 2018-06-26 Richard Biener <rguenther@suse.de>
14397 PR middle-end/86271
14398 * fold-const.c (fold_convertible_p): Pointer extension
14401 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
14404 * dwarf2out.c (loc_list_has_views): Adjust comments.
14405 (dw_loc_list): Split single cross-partition range with
14408 2018-06-25 Jeff Law <law@redhat.com>
14410 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14411 on -mbig-switch by default.
14413 * config/v850/predicates.md (const_float_1_operand): Fix match_code
14415 (const_float_0_operand): Remove unused predicate.
14416 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14417 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14418 (recipsf2): New expander. Original pattern now called
14420 (recipdf2, recipdf2_insn): Similarly.
14421 (rsqrtsf2, rsqrtsf2_insn): Similarly
14422 (rsqrtdf2, rsqrtdf2_insn): Similarly
14424 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
14426 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14427 Simplify logic for FreeBSD (twice).
14429 2018-06-25 Martin Sebor <msebor@redhat.com>
14431 PR tree-optimization/86204
14432 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14433 a strnlen result if it's less than the length of the string.
14435 2018-06-25 Martin Sebor <msebor@redhat.com>
14437 PR tree-optimization/85700
14438 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14439 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14440 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14442 2018-06-25 Martin Sebor <msebor@redhat.com>
14444 * doc/extend.texi (Zero-length arrays): Update and clarify.
14446 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
14448 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14449 added IEEE/IBM long double multilib support on PowerPC little
14450 endian Linux systems.
14451 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14452 (MULTILIB_DEFAULTS): Likewise.
14453 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14455 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14456 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14457 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14459 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
14461 PR middle-end/86311
14462 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14463 (REORDER_45): Likewise.
14465 2018-06-25 Jeff Law <law@redhat.com>
14467 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14468 dividend to 32 bits. Adjust length.
14469 (udivmodhi4): Cleanup output template. Fix length.
14471 2018-06-25 Carl Love <cel@us.ibm.com>
14473 * config/rs6000/vsx.md: Change word selector to prefered location.
14475 2018-06-25 Richard Biener <rguenther@suse.de>
14477 PR tree-optimization/86304
14478 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14479 epilogue-if-converted loops as well.
14481 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
14483 * lto-section-out.c (lto_begin_section): Do not print section
14484 name for noaddr and unnumbered dumps.
14486 2018-06-25 Richard Biener <rguenther@suse.de>
14488 * tree-vectorizer.h (struct vec_info_shared): New structure
14489 with parts split out from struct vec_info and loop_nest from
14490 struct _loop_vec_info.
14491 (struct vec_info): Adjust accordingly.
14492 (struct _loop_vec_info): Likewise.
14493 (LOOP_VINFO_LOOP_NEST): Adjust.
14494 (LOOP_VINFO_DATAREFS): Likewise.
14495 (LOOP_VINFO_DDRS): Likewise.
14496 (struct _bb_vec_info): Likewise.
14497 (BB_VINFO_DATAREFS): Likewise.
14498 (BB_VINFO_DDRS): Likewise.
14499 (struct _stmt_vec_info): Add dr_aux member.
14500 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14501 (DR_MISALIGNMENT_UNINITIALIZED): New.
14502 (set_dr_misalignment): Adjust.
14503 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14504 (vect_analyze_loop): Adjust prototype.
14505 (vect_analyze_loop_form): Likewise.
14506 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14507 Compute dependences lazily.
14508 (vect_record_base_alignments): Use shared datarefs/ddrs.
14509 (vect_verify_datarefs_alignment): Likewise.
14510 (vect_analyze_data_refs_alignment): Likewise.
14511 (vect_analyze_data_ref_accesses): Likewise.
14512 (vect_analyze_data_refs): Likewise.
14513 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14514 constructor parameter for shared part.
14515 (vect_analyze_loop_form): Pass in shared part and adjust.
14516 (vect_analyze_loop_2): Pass in storage for the number of
14517 stmts. Move loop nest finding to the caller. Compute
14519 (vect_analyze_loop): Pass in shared part.
14520 (vect_transform_loop): Verify shared datarefs are unchanged.
14521 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14522 constructor parameter for shared part.
14523 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14524 (vect_slp_bb): Verify shared datarefs are unchanged before
14526 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14528 (new_stmt_vec_info): Initialize DR_AUX misalignment to
14529 DR_MISALIGNMENT_UNINITIALIZED.
14530 * tree-vectorizer.c (vec_info::vec_info): Add constructor
14531 parameter for shared part.
14532 (vec_info::~vec_info): Adjust.
14533 (vec_info_shared::vec_info_shared): New.
14534 (vec_info_shared::~vec_info_shared): Likewise.
14535 (vec_info_shared::save_datarefs): Likewise.
14536 (vec_info_shared::check_datarefs): Likewise.
14537 (try_vectorize_loop_1): Construct shared part live for analyses
14538 of a single loop for multiple vector sizes.
14539 * tree-parloops.c (gather_scalar_reductions): Adjust.
14541 2018-06-25 Richard Biener <rguenther@suse.de>
14543 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14544 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14545 (vect_analyze_data_refs): Remove similar code from here and
14546 simplify accordingly.
14548 2018-06-25 Richard Biener <rguenther@suse.de>
14550 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14551 for reverse storage order accesses rather than asserting
14552 they cannot happen here.
14554 2018-06-25 Tom de Vries <tdevries@suse.de>
14557 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14558 Use data16 instead of .byte for insn prefix.
14560 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
14563 * parser.c (make_char_string_pack): Pass this literal chars
14564 through cpp_interpret_string.
14565 (cp_parser_userdef_numeric_literal): Check the result of
14566 make_char_string_pack.
14568 2018-06-24 Maya Rashish <coypu@sdf.org>
14570 * ginclude/stddef.h: Simplify conditions around avoiding
14571 re-definition of __size_t.
14573 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
14575 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14576 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14578 2018-06-22 Maya Rashish <coypu@sdf.org>
14580 * doc/invoke.texi (mno-fancy-math-387): Update for changes
14581 made to OpenBSD and NetBSD through the years.
14583 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
14585 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14586 behavior of vec_pack (vector double, vector double) to match
14587 behavior of vec_float2 (vector double, vector double).
14589 2018-06-22 Olivier Hainque <hainque@adacore.com>
14591 * gimplify.c (gimplify_function_tree): Prevent creation
14592 of a trampoline for the address of the current function
14593 passed to entry/exit instrumentation hooks.
14595 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
14598 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
14601 2018-06-22 Martin Liska <mliska@suse.cz>
14603 PR tree-optimization/86263
14604 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14605 Bail out if is_enabled is false.
14606 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
14608 (jump_table_cluster::is_enabled): New function.
14610 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
14612 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
14613 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
14614 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
14615 (lto_input_ts_binfo_tree_pointers): Likewise.
14616 * tree-streamer-out.c (streamer_write_tree_bitfields,
14617 write_ts_binfo_tree_pointers): Likewise.
14618 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
14620 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
14622 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
14624 2018-06-22 Martin Liska <mliska@suse.cz>
14626 * symbol-summary.h (get): Make it pure and inline move
14627 functionality from ::get function.
14628 (get): Remove and inline into ::get and ::get_create.
14629 (get_create): Move code from ::get function.
14631 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14634 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
14635 -x assembler-with-cpp.
14637 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14639 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
14640 _FILE_OFFSET_BITS=64 for C++.
14642 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
14644 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
14645 conversion insn that shows up when pr85657-3.c is compiled using
14646 IEEE 128-bit long double.
14647 (neg<mode>2_internal): Use the correct mode to check whether the
14648 mode is IBM extended.
14649 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
14650 multiply and divide external functions from being created more
14653 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
14655 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
14657 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
14658 the edge can be forwarded.
14659 (cfg_layout_merge_blocks): Likewise.
14661 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
14663 * except.c (finish_eh_generation): Commit edge insertions only after
14664 the EH edges have been redirected from post-landing to landing pads.
14666 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
14668 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
14669 create_tmp_var_for to create the FRAME decl.
14670 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
14672 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
14674 * tree-inline.c (copy_edges_for_bb): Minor tweak.
14675 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
14676 debug statement when resetting its value.
14677 (expand_call_inline): Copy the locus of the call onto the assignment
14678 of the return value, if any. Use local variable in more cases.
14680 2018-06-21 Martin Liska <mliska@suse.cz>
14682 * ipa-pure-const.c (propagate_nothrow): Use
14683 funct_state_summaries->get.
14684 (dump_malloc_lattice): Likewise.
14685 (propagate_malloc): Likewise.
14687 2018-06-21 Richard Biener <rguenther@suse.de>
14689 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
14690 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
14691 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
14693 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
14694 BLOCK_ABSTRACT_ORIGIN unconditionally.
14696 2018-06-21 David Malcolm <dmalcolm@redhat.com>
14698 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
14700 * ipa-reference.c (ipa_reference_c_finalize): Delete
14701 ipa_ref_opt_sum_summaries and set it to NULL.
14703 2018-06-21 Tom de Vries <tdevries@suse.de>
14705 PR tree-optimization/85859
14706 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
14707 test with comment from bb_no_side_effects_p.
14709 2018-06-21 Richard Biener <rguenther@suse.de>
14711 PR tree-optimization/86232
14712 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
14713 max for constant niter.
14715 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
14717 * config/aarch64/aarch64-simd.md
14718 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
14720 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
14722 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
14723 Make opernads of the unspec commutative.
14725 2018-06-21 Richard Biener <rguenther@suse.de>
14727 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
14728 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
14729 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
14730 (vect_analyze_data_ref_dependence): Re-order checks to deal with
14732 (vect_record_base_alignments): Do not record base alignment
14733 for gathers or scatters.
14734 (vect_compute_data_ref_alignment): Drop return value that is always
14735 true. Bail out early for gathers or scatters.
14736 (vect_enhance_data_refs_alignment): Bail out early for gathers
14738 (vect_find_same_alignment_drs): Likewise.
14739 (vect_analyze_data_refs_alignment): Remove dead code.
14740 (vect_slp_analyze_and_verify_node_alignment): Likewise.
14741 (vect_analyze_data_refs): For possible gathers or scatters do
14742 not create an alternate DR, just check their possible validity
14743 and mark them. Adjust DECL_NONALIASED handling to not rely
14744 on DR_BASE_ADDRESS.
14745 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
14746 update inits of gathers or scatters.
14747 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
14748 Also copy gather/scatter flag to pattern vinfo.
14750 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
14752 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14753 behavior of vec_packsu (vector unsigned long long, vector unsigned
14754 long long) to match behavior of vec_packs with same signature.
14756 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
14757 Thomas Schwinge <thomas@codesourcery.com>
14758 Cesar Philippidis <cesar@codesourcery.com>
14760 * gimplify.c (gimplify_scan_omp_clauses): Add support for
14761 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14762 (gimplify_adjust_omp_clauses): Likewise.
14763 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
14764 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
14765 (gimplify_omp_target_update): Update handling of acc update and
14767 * omp-low.c (install_var_field): Remove unused parameter
14768 base_pointers_restrict.
14769 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
14770 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
14772 (omp_target_base_pointers_restrict_p): Delete.
14773 (scan_omp_target): Update call to scan_sharing_clauses.
14774 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
14776 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
14777 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
14778 (convert_local_omp_clauses): Likewise.
14779 * tree-pretty-print.c (dump_omp_clause): Likewise.
14780 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
14782 (omp_clause_code_name): Likewise.
14784 2018-06-20 Jakub Jelinek <jakub@redhat.com>
14787 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
14790 PR tree-optimization/86231
14791 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
14792 anti-range don't overwrite *vr0min before using it to compute *vr0max.
14794 2018-06-20 Tom de Vries <tdevries@suse.de>
14796 PR tree-optimization/86097
14797 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
14798 iv type if signedness of iv type is not the same as that of *nit.
14800 2018-06-20 Jakub Jelinek <jakub@redhat.com>
14802 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
14803 EDGE_EH edges, verify they are all EDGE_EH.
14805 2018-06-20 Maya Rashish <coypu@sdf.org>
14807 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
14809 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14811 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
14812 * config/aarch64/aarch64.c (xgene1_tunings): Add
14813 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
14814 (aarch64_mode_valid_for_sched_fusion_p):
14815 Allow 16-byte modes.
14816 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
14817 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
14819 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
14821 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
14822 * config/aarch64/iterators.md (VQ2): New mode iterator.
14824 2018-06-20 Martin Liska <mliska@suse.cz>
14826 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14827 Change default ratio from 10 to 8.
14829 2018-06-20 Martin Liska <mliska@suse.cz>
14831 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14833 (bit_test_cluster::find_bit_tests): Likewise.
14834 (switch_decision_tree::analyze_switch_statement): Find clusters.
14835 * tree-switch-conversion.h (struct jump_table_cluster): Document
14838 2018-06-20 Martin Liska <mliska@suse.cz>
14840 * tree-switch-conversion.c (switch_conversion::collect):
14841 Record m_uniq property.
14842 (switch_conversion::expand): Bail out for special conditions.
14843 (group_cluster::~group_cluster): New.
14844 (group_cluster::group_cluster): Likewise.
14845 (group_cluster::dump): Likewise.
14846 (jump_table_cluster::emit): New.
14847 (switch_decision_tree::fix_phi_operands_for_edges): New.
14848 (struct case_node): Remove struct.
14849 (jump_table_cluster::can_be_handled): New.
14850 (case_values_threshold): Moved to header.
14851 (reset_out_edges_aux): Likewise.
14852 (jump_table_cluster::is_beneficial): New.
14853 (bit_test_cluster::can_be_handled): Likewise.
14854 (add_case_node): Remove.
14855 (bit_test_cluster::is_beneficial): New.
14856 (case_bit_test::cmp): New.
14857 (bit_test_cluster::emit): New.
14858 (expand_switch_as_decision_tree_p): Remove.
14859 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
14860 (fix_phi_operands_for_edge): Likewise.
14861 (switch_decision_tree::analyze_switch_statement): New.
14862 (compute_cases_per_edge): Move ...
14863 (switch_decision_tree::compute_cases_per_edge): ... here.
14864 (try_switch_expansion): Likewise.
14865 (switch_decision_tree::try_switch_expansion): Likewise.
14866 (record_phi_operand_mapping): Likewise.
14867 (switch_decision_tree::record_phi_operand_mapping): Likewise.
14868 (emit_case_decision_tree): Likewise.
14869 (switch_decision_tree::emit): Likewise.
14870 (balance_case_nodes): Likewise.
14871 (switch_decision_tree::balance_case_nodes): Likewise.
14872 (dump_case_nodes): Likewise.
14873 (switch_decision_tree::dump_case_nodes): Likewise.
14874 (emit_jump): Likewise.
14875 (switch_decision_tree::emit_jump): Likewise.
14876 (emit_cmp_and_jump_insns): Likewise.
14877 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
14878 (emit_case_nodes): Likewise.
14879 (switch_decision_tree::emit_case_nodes): Likewise.
14880 (conditional_probability): Remove.
14881 * tree-switch-conversion.h (enum cluster_type): New.
14883 (struct cluster): Likewise.
14884 (cluster::cluster): Likewise.
14885 (struct simple_cluster): Likewise.
14886 (simple_cluster::simple_cluster): Likewise.
14887 (struct group_cluster): Likewise.
14888 (struct jump_table_cluster): Likewise.
14889 (struct bit_test_cluster): Likewise.
14890 (struct min_cluster_item): Likewise.
14891 (struct case_tree_node): Likewise.
14892 (case_tree_node::case_tree_node): Likewise.
14893 (jump_table_cluster::case_values_threshold): Likewise.
14894 (struct case_bit_test): Likewise.
14895 (struct switch_decision_tree): Likewise.
14896 (struct switch_conversion): Likewise.
14897 (switch_decision_tree::reset_out_edges_aux): Likewise.
14899 2018-06-20 Martin Liska <mliska@suse.cz>
14901 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
14902 (hoist_edge_and_branch_if_true): Likewise.
14903 (expand_switch_using_bit_tests_p): Likewise.
14904 (struct case_bit_test): Likewise.
14905 (case_bit_test_cmp): Likewise.
14906 (emit_case_bit_tests): Likewise.
14907 (switch_conversion::switch_conversion): New class.
14908 (struct switch_conv_info): Remove old struct.
14909 (collect_switch_conv_info): More to ...
14910 (switch_conversion::collect): ... this.
14911 (check_range): Likewise.
14912 (switch_conversion::check_range): Likewise.
14913 (check_all_empty_except_final): Likewise.
14914 (switch_conversion::check_all_empty_except_final): Likewise.
14915 (check_final_bb): Likewise.
14916 (switch_conversion::check_final_bb): Likewise.
14917 (create_temp_arrays): Likewise.
14918 (switch_conversion::create_temp_arrays): Likewise.
14919 (free_temp_arrays): Likewise.
14920 (gather_default_values): Likewise.
14921 (switch_conversion::gather_default_values): Likewise.
14922 (build_constructors): Likewise.
14923 (switch_conversion::build_constructors): Likewise.
14924 (constructor_contains_same_values_p): Likewise.
14925 (switch_conversion::contains_same_values_p): Likewise.
14926 (array_value_type): Likewise.
14927 (switch_conversion::array_value_type): Likewise.
14928 (build_one_array): Likewise.
14929 (switch_conversion::build_one_array): Likewise.
14930 (build_arrays): Likewise.
14931 (switch_conversion::build_arrays): Likewise.
14932 (gen_def_assigns): Likewise.
14933 (switch_conversion::gen_def_assigns): Likewise.
14934 (prune_bbs): Likewise.
14935 (switch_conversion::prune_bbs): Likewise.
14936 (fix_phi_nodes): Likewise.
14937 (switch_conversion::fix_phi_nodes): Likewise.
14938 (gen_inbound_check): Likewise.
14939 (switch_conversion::gen_inbound_check): Likewise.
14940 (process_switch): Use the newly created class.
14941 (switch_conversion::expand): New.
14942 (switch_conversion::~switch_conversion): New.
14943 * tree-switch-conversion.h: New file.
14945 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14947 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
14948 tree-vect-patterns.c.
14949 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
14950 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
14951 (vect_recog_sad_pattern): Likewise.
14952 (vect_recog_widen_sum_pattern): Likewise.
14953 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
14954 (vect_recog_widen_shift_pattern): Remove the type_in argument.
14955 (vect_recog_rotate_pattern): Likewise.
14956 (vect_recog_mult_pattern): Likewise.
14957 (vect_recog_vector_vector_shift_pattern): Likewise.
14958 (vect_recog_divmod_pattern): Likewise.
14959 (vect_recog_mixed_size_cond_pattern): Likewise.
14960 (vect_recog_bool_pattern): Likewise.
14961 (vect_recog_mask_conversion_pattern): Likewise.
14962 (vect_try_gather_scatter_pattern): Likewise.
14963 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
14964 (vect_recog_over_widening_pattern): Likewise.
14965 (vect_recog_gather_scatter_pattern): Likewise.
14966 (vect_recog_func_ptr): Move from tree-vectorizer.h
14967 (vect_vect_recog_func_ptrs): Move further down the file.
14968 (vect_recog_func): Likewise. Remove the third argument.
14969 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
14970 (vect_pattern_recog_1): Expect the pattern function to do any
14971 necessary target tests. Also expect it to provide a vector type.
14972 Remove the type_in handling.
14974 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14976 * tree-vect-patterns.c (vect_pattern_detected): New function.
14977 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
14978 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
14979 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
14980 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
14981 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
14982 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
14983 (vect_recog_mask_conversion_pattern)
14984 (vect_try_gather_scatter_pattern): Likewise.
14986 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14988 * tree-vect-patterns.c (vect_get_internal_def): New function.
14989 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14990 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
14991 (search_type_for_mask_1): Use it.
14993 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
14995 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
14996 redundant WIDEN_SUM_EXPR handling.
14997 (vect_recog_sad_pattern): Likewise.
14999 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15001 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15002 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15003 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15004 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15005 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15007 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15009 * tree-vect-stmts.c (vectorizable_call): Make sure that we
15010 use the stmt_vec_info of the original bb statement for the
15011 new zero assignment, even if the call is part of a pattern.
15013 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15015 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15016 that the sequence is attached to the original statement rather
15017 than the pattern statement.
15018 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15019 PATTERN_DEF_SEQ from the original statement rather than
15020 the main pattern statement.
15021 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15022 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15023 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15025 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15027 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15028 definition statements before the early exit for statements that aren't
15030 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15032 (vect_transform_loop): ...here. Process pattern definition
15033 statements without first checking whether the main pattern
15034 statement is live or relevant.
15036 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
15038 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15039 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15041 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
15043 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15044 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15045 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15046 (expand_block_compare): Change select_block_compare_mode call.
15047 (expand_strncmp_align_check): Use new functions, fix comment.
15048 (emit_final_str_compare_gpr): New function.
15049 (expand_strn_compare): Refactor and clean up code.
15050 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15052 2018-06-19 Tony Reix <tony.reix@atos.com>
15053 Damien Bergamini <damien.bergamini@atos.com>
15054 David Edelsohn <dje.gcc@gmail.com>
15056 * collect2.c (static_obj): New variable.
15057 (static_libs): New variable.
15058 (is_in_list): Uncomment declaration.
15059 (main): Track AIX libraries linked statically.
15060 (is_in_list): Uncomment definition.
15061 (scan_prog_file): Don't add AIX shared libraries initializer
15062 to constructor list if linking statically.
15064 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
15066 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15068 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15070 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
15072 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15075 2018-06-19 Martin Liska <mliska@suse.cz>
15077 * config/i386/i386.c (ix86_can_inline_p): Do not use
15078 ipa_fn_summaries::get_create.
15079 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15081 (devirtualization_time_bonus): Likewise.
15082 (ipcp_propagate_stage): Likewise.
15083 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15084 (edge_set_predicate): Likewise.
15085 (evaluate_conditions_for_known_args): Likewise.
15086 (evaluate_properties_for_edge): Likewise.
15087 (ipa_call_summary::reset): Tranform to ...
15088 (ipa_call_summary::~ipa_call_summary): ... this.
15089 (ipa_fn_summary::reset): Transform to ...
15090 (ipa_fn_summary::~ipa_fn_summary): ... this.
15091 (ipa_fn_summary_t::remove): Rename to ...
15092 (ipa_fn_summary_t::remove_callees): ... this.
15093 (ipa_fn_summary_t::duplicate): Use placement new
15094 instead of memory copy.
15095 (ipa_call_summary_t::duplicate): Likewise.
15096 (ipa_call_summary_t::remove): Remove.
15097 (dump_ipa_call_summary): Change get_create to get.
15098 (ipa_dump_fn_summary): Dump only when summary exists.
15099 (analyze_function_body): Use symbol_summary::get instead
15101 (compute_fn_summary): Likewise.
15102 (estimate_edge_devirt_benefit): Likewise.
15103 (estimate_edge_size_and_time): Likewise.
15104 (inline_update_callee_summaries): Likewise.
15105 (remap_edge_change_prob): Likewise.
15106 (remap_edge_summaries): Likewise.
15107 (ipa_merge_fn_summary_after_inlining): Likewise.
15108 (write_ipa_call_summary): Likewise.
15109 (ipa_fn_summary_write): Likewise.
15110 (ipa_free_fn_summary): Likewise.
15111 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15112 (struct ipa_call_summary): Likewise.
15113 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15115 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15116 (estimate_size_after_inlining): Likewise.
15117 (estimate_growth): Likewise.
15118 (growth_likely_positive): Likewise.
15119 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15120 (inline_call): Likewise.
15121 * ipa-inline.c (caller_growth_limits): Likewise.
15122 (can_inline_edge_p): Likewise.
15123 (can_inline_edge_by_limits_p): Likewise.
15124 (compute_uninlined_call_time): Likewise.
15125 (compute_inlined_call_time): Likewise.
15126 (want_inline_small_function_p): Likewise.
15127 (edge_badness): Likewise.
15128 (update_caller_keys): Likewise.
15129 (update_callee_keys): Likewise.
15130 (inline_small_functions): Likewise.
15131 (inline_to_all_callers_1): Likewise.
15132 (dump_overall_stats): Likewise.
15133 (early_inline_small_functions): Likewise.
15134 (early_inliner): Likewise.
15135 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15136 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15137 * ipa-pure-const.c (malloc_candidate_p): Likewise.
15138 * ipa-split.c (execute_split_functions): Likewise.
15139 * symbol-summary.h: Likewise.
15140 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15142 2018-06-19 Richard Biener <rguenther@suse.de>
15144 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15145 (vectorize_loops): ... here. Fix dbgcnt handling.
15146 (try_vectorize_loop): Wrap try_vectorize_loop_1.
15148 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
15151 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15152 ieee128 argument takes up only one (vector) register, not two (floating
15155 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
15157 * gimplify.c (gimplify_init_constructor): Really never clear for an
15158 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15160 2018-06-19 Richard Biener <rguenther@suse.de>
15162 PR tree-optimization/86179
15163 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15164 after failed recognition.
15166 2018-06-18 Martin Sebor <msebor@redhat.com>
15168 PR middle-end/85602
15169 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15170 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15171 Handle integer subtraction.
15172 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15173 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15175 2018-06-18 David Malcolm <dmalcolm@redhat.com>
15177 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15178 param from rtx to rtx_insn *.
15179 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15181 (frv_ifcvt_modify_insn): Likwise.
15182 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15183 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
15184 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15185 as_a <rtx_insn *> cast to local "unprotected_region" once
15186 it's been established that it's not NULL or pc_rtx.
15187 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15188 param "sethi" from rtx to rtx_insn *.
15189 (nds32_group_float_insns): Likewise for param "insn".
15190 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15192 (vax_output_int_subtract): Likewise.
15193 * config/vax/vax.c (vax_output_int_add): Likewise for param
15195 (vax_output_int_subtract): Likewise.
15196 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15197 (emit_pattern_after): Likewise for param "after".
15198 (emit_insn_after): Likewise.
15199 (emit_jump_insn_after): Likewise.
15200 (emit_call_insn_after): Likewise.
15201 (emit_debug_insn_after): Likewise.
15202 (emit_pattern_before): Likewise for param "before".
15203 (emit_insn_before): Likewise.
15204 (emit_jump_insn_before): Likewise.
15205 * final.c (get_insn_template): Likewise for param "insn", removing
15207 * output.h (get_insn_template): Likewise for 2nd param.
15208 * rtl.h (emit_insn_before): Likewise.
15209 (emit_jump_insn_before): Likewise.
15210 (emit_debug_insn_before_noloc): Likewise.
15211 (emit_insn_after): Likewise.
15212 (emit_jump_insn_after): Likewise.
15213 (emit_call_insn_after): Likewise.
15214 (emit_debug_insn_after): Likewise.
15215 (set_insn_deleted): Likewise for param.
15217 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
15220 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15221 floating point modes, so that IFmode is numerically greater than
15222 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15223 to declare the ordering. This prevents IFmode from being
15224 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15225 machine. Include rs6000-modes.h to share the fractional values
15226 between genmodes* and the rest of the compiler.
15227 (IFmode): Likewise.
15228 (KFmode): Likewise.
15229 (TFmode): Likewise.
15230 * config/rs6000/rs6000-modes.h: New file.
15231 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15232 meaning of rs6000_long_double_size so that 126..128 selects an
15233 appropriate 128-bit floating point type.
15234 (rs6000_option_override_internal): Likewise.
15235 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15236 (TARGET_LONG_DOUBLE_128): Change the meaning of
15237 rs6000_long_double_size so that 126..128 selects an appropriate
15238 128-bit floating point type.
15239 (LONG_DOUBLE_TYPE_SIZE): Update comment.
15240 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15241 source and destination to match the standard usage.
15242 (truncifkf2): Likewise.
15243 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15244 ISA 2.07 to use an explicit clobber, instead of passing in a
15246 (copysign<mode>3_soft): Likewise.
15248 2018-06-18 David Malcolm <dmalcolm@redhat.com>
15250 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15251 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15252 (vect_slp_analyze_instance_dependence): Likewise.
15253 (vect_enhance_data_refs_alignment): Likewise.
15254 (vect_analyze_data_refs_alignment): Likewise.
15255 (vect_slp_analyze_and_verify_instance_alignment
15256 (vect_analyze_data_ref_accesses): Likewise.
15257 (vect_prune_runtime_alias_test_list): Likewise.
15258 (vect_analyze_data_refs): Likewise.
15259 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15260 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15261 (vect_analyze_scalar_cycles_1): Likewise.
15262 (vect_get_loop_niters): Likewise.
15263 (vect_analyze_loop_form_1): Likewise.
15264 (vect_update_vf_for_slp): Likewise.
15265 (vect_analyze_loop_operations): Likewise.
15266 (vect_analyze_loop): Likewise.
15267 (vectorizable_induction): Likewise.
15268 (vect_transform_loop): Likewise.
15269 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15270 * tree-vect-slp.c (vect_analyze_slp): Likewise.
15271 (vect_make_slp_decision): Likewise.
15272 (vect_detect_hybrid_slp): Likewise.
15273 (vect_slp_analyze_operations): Likewise.
15274 (vect_slp_bb): Likewise.
15275 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15276 (vectorizable_bswap): Likewise.
15277 (vectorizable_call): Likewise.
15278 (vectorizable_simd_clone_call): Likewise.
15279 (vectorizable_conversion): Likewise.
15280 (vectorizable_assignment): Likewise.
15281 (vectorizable_shift): Likewise.
15282 (vectorizable_operation): Likewise.
15283 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15285 2018-06-18 Martin Sebor <msebor@redhat.com>
15287 PR tree-optimization/81384
15288 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15289 * builtins.c (expand_builtin_strnlen): New function.
15290 (expand_builtin): Call it.
15291 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15292 * builtins.def (BUILT_IN_STRNLEN): New.
15293 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15294 Warn for bounds in excess of maximum object size.
15295 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15296 single-value ranges. Handle strnlen.
15297 (handle_builtin_strlen): Handle strnlen.
15298 (strlen_check_and_optimize_stmt): Same.
15299 * doc/extend.texi (Other Builtins): Document strnlen.
15301 2018-06-18 Maya Rashish <coypu@sdf.org>
15303 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15304 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15305 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15307 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15309 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15311 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15313 * tree.c (escaped_string::escape): Replace cast to char * by
15314 const_cast<char *> (unescaped).
15316 2018-06-18 Nick Clifton <nickc@redhat.com>
15319 * tree.c (escaped_string): New class. Converts an unescaped
15320 string into its escaped equivalent.
15321 (warn_deprecated_use): Use the new class to convert the
15322 deprecation message, if present.
15323 (test_escaped_strings): New self test.
15324 (test_c_tests): Add test_escaped_strings.
15325 * doc/extend.texi (deprecated): Add a note that the
15326 deprecation message is affected by the -fmessage-length
15327 option, and that control characters will be escaped.
15328 (#pragma GCC error): Document this pragma.
15329 (#pragma GCC warning): Likewise.
15330 * doc/invoke.texi (-fmessage-length): Document this option's
15331 effect on the #warning and #error preprocessor directives and
15332 the deprecated attribute.
15334 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
15336 * tree.c (decl_value_expr_lookup): Revert latest change.
15337 (decl_value_expr_insert): Likewise.
15339 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
15341 * gimplify.c (nonlocal_vlas): Delete.
15342 (nonlocal_vla_vars): Likewise.
15343 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15345 (gimplify_body): Do not create and destroy nonlocal_vlas.
15346 * tree-nested.c: Include diagnostic.h.
15347 (use_pointer_in_frame): Tweak.
15348 (lookup_field_for_decl): Add assertion and declare the transformation.
15349 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15350 internal error when the reference is in a wrong context. Do not
15351 create a debug decl by default.
15352 (note_nonlocal_block_vlas): Delete.
15353 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15354 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
15355 create a debug decl by default.
15356 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15357 call is in a wrong context.
15358 (fixup_vla_decls): New function.
15359 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
15360 debug variables were created.
15361 * tree.c (decl_value_expr_lookup): Add checking assertion.
15362 (decl_value_expr_insert): Likewise.
15364 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
15366 PR middle-end/82479
15367 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15368 * tree-scalar-evolution.c (interpret_expr): Likewise.
15369 (expression_expensive_p): Likewise.
15370 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15371 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15372 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15373 (ssa_defined_by_minus_one_stmt_p): New.
15375 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
15377 PR middle-end/64946
15378 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15379 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15380 * dojump.c (do_jump): Likewise.
15381 * expr.c (expand_expr_real_2): Check operand type's sign.
15382 * fold-const.c (const_unop): Handle ABSU_EXPR.
15383 (fold_abs_const): Likewise.
15384 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15385 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15386 (strip_sign_op_1): Likesise.
15387 * match.pd: Add new pattern to generate ABSU_EXPR.
15388 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15389 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15390 * tree-eh.c (operation_could_trap_helper_p): Likewise.
15391 * tree-inline.c (estimate_operator_cost): Likewise.
15392 * tree-pretty-print.c (dump_generic_node): Likewise.
15393 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15394 * tree.def (ABSU_EXPR): New.
15396 2018-06-16 Jakub Jelinek <jakub@redhat.com>
15398 PR middle-end/86095
15399 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15400 documented as preserved for backward compatibility only.
15401 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15403 PR rtl-optimization/86108
15404 * bb-reorder.c (create_forwarder_block): Renamed to ...
15405 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
15406 jump from new landing pad to the second part.
15407 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15410 2018-06-15 Jakub Jelinek <jakub@redhat.com>
15412 PR middle-end/85878
15413 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15414 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15415 Only call store_expr for halves if the mode is the same.
15417 PR middle-end/86123
15418 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15419 Fix up comment formatting.
15421 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
15423 * typed-splay-tree.h (typed_splay_tree::remove): New function.
15424 (typed_splay_tree::closure,
15425 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15426 (typed_splay_tree::typed_splay_tree,
15427 typed_splay_tree::operator =): Declared private.
15428 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15429 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15430 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15431 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15432 typed_splay_tree::splay_tree_splay,
15433 typed_splay_tree::splay_tree_foreach_helper,
15434 typed_splay_tree::splay_tree_insert,
15435 typed_splay_tree::splay_tree_remove,
15436 typed_splay_tree::splay_tree_lookup,
15437 typed_splay_tree::splay_tree_predecessor,
15438 typed_splay_tree::splay_tree_successor,
15439 typed_splay_tree::splay_tree_min,
15440 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15441 (typed_splay_tree::root, typed_splay_tree::comp,
15442 typed_splay_tree::delete_key,
15443 typed_splay_tree::delete_value): New data members.
15444 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15445 typed_splay_tree::remove.
15447 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
15449 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15450 -mginv and -mno-ginv to the assembler.
15451 * config/mips/mips.opt (-mcrc): New option.
15452 (-mginv): Likewise.
15453 * doc/invoke.text (-mcrc): Document.
15454 (-mginv): Likewise.
15456 2018-06-15 Nick Clifton <nickc@redhat.com>
15459 * tree.c (escaped_string): New class. Converts an unescaped
15460 string into its escaped equivalent.
15461 (warn_deprecated_use): Use the new class to convert the
15462 deprecation message, if present.
15463 (test_escaped_strings): New self test.
15464 (test_c_tests): Add test_escaped_strings.
15465 * doc/extend.texi (deprecated): Add a note that the
15466 deprecation message is affected by the -fmessage-length
15467 option, and that control characters will be escaped.
15468 (#pragma GCC error): Document this pragma.
15469 (#pragma GCC warning): Likewise.
15470 * doc/invoke.texi (-fmessage-length): Document this option's
15471 effect on the #warning and #error preprocessor directives and
15472 the deprecated attribute.
15474 2018-06-15 Richard Biener <rguenther@suse.de>
15476 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15477 here, also noting vector size used.
15478 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
15479 size used in MSG_OPTIMIZED_LOCATIONS dump.
15480 (pass_slp_vectorize::execute): Adjust.
15482 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
15485 * config/arc/arc.c (arc_return_address_register): Fix
15488 2018-06-15 Richard Biener <rguenther@suse.de>
15490 PR middle-end/86159
15491 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15492 leave useless conversion stripping to force_gimple_operand_gsi.
15493 (gimplify_build2): Likewise.
15494 (gimplify_build1): Likewise.
15496 2018-06-15 Richard Biener <rguenther@suse.de>
15498 PR middle-end/86076
15499 * tree-cfg.c (move_stmt_op): unshare invariant addresses
15500 before adjusting their block.
15502 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
15504 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15505 multilibs for *-*-rtems*.
15506 * config/riscv/t-rtems: New file.
15508 2018-06-14 Jakub Jelinek <jakub@redhat.com>
15510 PR middle-end/86122
15511 * match.pd ((A +- CST1) +- CST2): Punt if last resort
15512 unsigned_type_for returns NULL.
15515 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15516 subregs of multi-word pseudos unless the float mode has word size.
15518 2018-06-14 Richard Biener <rguenther@suse.de>
15520 PR middle-end/86139
15521 * tree-vect-generic.c (build_word_mode_vector_type): Remove
15522 duplicate and harmful type_hash_canon.
15523 * tree.c (type_hash_canon): Assert we didn't find ourselves.
15525 2018-06-14 Richard Biener <rguenther@suse.de>
15528 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15531 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
15533 * config/rtems.h (STDINT_LONG32): Define.
15535 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
15536 Prachi Godbole <prachi.godbole@imgtec.com>
15538 * config/mips/mips-cpus.def: Define P6600.
15539 * config/mips/mips-tables.opt: Regenerate.
15540 * config/mips/mips.c (mips_ucbranch_type): New enum.
15541 (mips_rtx_cost_data): Add support for P6600.
15542 (mips_issue_rate): Likewise.
15543 (mips_multipass_dfa_lookahead): Likewise.
15544 (mips_avoid_hazard): Likewise.
15545 (mips_reorg_process_insns): Likewise.
15546 (mips_classify_branch_p6600): New function.
15547 * config/mips/mips.h (TUNE_P6600): New define.
15548 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15549 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15550 * config/mips/mips.md: Include p6600.md.
15551 (processor): Add p6600.
15552 * config/mips/p6600.md: New file.
15553 * doc/invoke.texi: Add p6600 to supported architectures.
15555 2018-06-13 Martin Sebor <msebor@redhat.com>
15557 PR tree-optimization/86114
15558 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15560 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15562 2018-06-13 Richard Biener <rguenther@suse.de>
15564 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15565 Properly set vector type of the intermediate stmt.
15566 * tree-vect-stmts.c (vectorizable_operation): The destination
15567 var always has vectype_out type.
15569 2018-06-13 Jeff Law <law@redhat.com>
15571 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15572 integer 0 for argument to print_rtl_with_bb.
15573 (rl78_reorg): Likewise.
15575 2018-06-13 David Malcolm <dmalcolm@redhat.com>
15577 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15578 from rtx to rtx_insn *.
15579 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15581 (add_sched_insns_for_speculation): Likewise for local "target",
15582 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15583 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15584 from rtx_insn ** to rtx_code_label **.
15585 (reorg_emit_nops): Likewise.
15586 (c6x_reorg): Likewise for local "call_labels".
15587 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15589 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
15590 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
15591 the loops over LABEL_REFS.
15592 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
15594 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15595 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
15596 (split_branches): Strengthen local "olabel" from rtx to
15597 rtx_insn *, adding a safe_as_a cast.
15598 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
15600 (add_insn_after): Likewise for first two params.
15601 (add_insn_before): Likewise.
15602 (remove_insn): Likewise for param.
15603 (emit_pattern_before_noloc): Likewise for second and third params.
15604 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
15605 (emit_call_insn_before_noloc): Likewise.
15606 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
15608 (emit_barrier_before): Likewise.
15609 (emit_label_before): Strengthen "label" param from "rtx" to
15610 "rtx_code_label *". Strengthen "before" param from "rtx" to
15612 (emit_insn_after_1): Strengthen "after" param from "rtx" to
15614 (emit_pattern_after_noloc): Likewise.
15615 (emit_insn_after_noloc): Likewise.
15616 (emit_jump_insn_after_noloc): Likewise.
15617 (emit_call_insn_after_noloc): Likewise.
15618 (emit_debug_insn_after_noloc): Likewise.
15619 (emit_barrier_after): Likewise.
15620 (emit_label_after): Likewise for both params.
15621 (emit_pattern_after_setloc): Likewise for "after" param. Convert
15622 "loc" param from "int" to "location_t".
15623 (emit_insn_after_setloc): Likewise.
15624 (emit_jump_insn_after_setloc): Likewise.
15625 (emit_call_insn_after_setloc): Likewise.
15626 (emit_debug_insn_after_setloc): Likewise.
15627 (emit_pattern_before_setloc): Likewise for "before" param. Convert
15628 "loc" param from "int" to "location_t".
15629 (emit_pattern_before): Convert NULL_RTX to NULL.
15630 (emit_insn_before_setloc): Convert "loc" param from "int" to
15632 (emit_jump_insn_before_setloc): Likewise.
15633 (emit_call_insn_before_setloc): Likewise.
15634 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
15635 rtx_insn *. Convert "loc" param from "int" to "location_t".
15636 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
15637 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
15638 Convert 3rd param from "int" to "location_t".
15639 (emit_barrier_before, emit_barrier_after, next_real_insn):
15640 Strengthen param from rtx to rtx_insn *.
15641 (emit_label_before): Strengthen 1st param from "rtx" to
15642 "rtx_code_label *". Strengthen 2nd param from "rtx" to
15644 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
15645 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
15646 Strengthen 2nd param from "rtx" to "rtx_insn *".
15647 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
15648 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
15649 Likewise. Convert 3rd param from "int" to "location_t".
15650 (emit_label_after): Strengthen 1st param from "rtx" to
15651 "rtx_code_label *".
15652 (next_real_insn, remove_insn): Strengthen param from "rtx" to
15654 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
15655 from "rtx" to "rtx_insn *".
15657 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
15659 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
15660 bodies streamed in with -Q.
15661 * dumpfile.c (dump_files): Add lto-stream-out dump file.
15662 * dumpfile.h (tree_dump_index): Add lto_stream_out.
15663 * gimple-streamer-out.c: Include gimple-pretty-print.h
15664 (output_bb): Dump stmts streamed.
15665 * lto-section-out.c: Include print-tree.h
15666 (lto_begin_section): Dump sections created.
15667 (lto_output_decl_index): Dump decl encoded.
15668 * lto-streamer-out.c: Include print-tree.h
15669 (create_output_block): Dump output block created.
15670 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
15671 (output_function): Dump function output.
15672 (output_constructor): Dump constructor streamed.
15673 (write_global_stream): Output indexes encoded.
15674 (produce_asm_for_decls): Dump streams encoded.
15675 * lto-streamer.c (streamer_dump_file): New global var.
15676 * lto-streamer.h (streamer_dump_file): Declare.
15677 * passes.c (ipa_write_summaries): Initialize streamer dump.
15678 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
15681 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
15684 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
15685 offsets for register save directives. Emit a second batch of save
15686 directives, if need be, when the function accesses prior frames.
15688 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15690 * config/arc/fpu.md (fmasf4): Force operand to register.
15691 (fnmasf4): Likewise.
15693 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15695 * config/arc/arc-protos.h (arc_pad_return): Remove.
15696 * config/arc/arc.c (machine_function): Remove force_short_suffix
15698 (arc_print_operand): Adjust printing of '&'.
15699 (arc_verify_short): Remove conditional printing of short suffix.
15700 (arc_final_prescan_insn): Remove reference to size_reason.
15701 (pad_return): New function.
15702 (arc_reorg): Call pad_return.
15703 (arc_pad_return): Remove.
15704 (arc_init_machine_status): Remove reference to force_short_suffix.
15705 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
15706 (attr length): When attribute iscompact is true force to 2
15707 regardless; in the case of maybe check if we want to force the
15708 instruction to have 4 bytes length.
15709 (nopv): Change it to generate 4 byte long nop as well.
15710 (blockage): New pattern.
15711 (simple_return): Remove call to arc_pad_return.
15712 (p_return_i): Likewise.
15714 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15716 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
15718 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15720 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
15723 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15725 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
15726 for ARC700 and ARCv2.
15728 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
15731 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
15732 operands[2] instead of operands[1].
15735 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
15737 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
15738 case, check whether the outer register overlaps an unallocatable
15739 register, not just whether it fits the required class.
15741 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
15743 * poly-int.h (can_div_trunc_p): Add new overload in which all values
15745 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
15746 (memrefs_conflict_p): Likewise.
15747 (init_alias_analysis): Likewise.
15748 * cfgexpand.c (expand_debug_expr): Likewise.
15749 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
15750 * cse.c (fold_rtx): Likewise.
15751 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
15752 * expr.c (emit_block_move_hints): Likewise.
15753 (clear_storage_hints, push_block, emit_push_insn): Likewise.
15754 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
15755 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
15756 (emit_group_store): Likewise.
15757 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
15758 to read the PRE/POST_MODIFY increment.
15759 * calls.c (store_one_arg): Use strip_offset.
15760 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
15762 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
15764 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
15765 (simplify_binary_operation_1): Extend CONST_INT handling to
15767 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
15768 than a HOST_WIDE_INT.
15769 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
15771 (adjust_mems, add_stores): Update accodingly.
15772 (vt_canonicalize_addr): Track polynomial offsets.
15773 (emit_note_insn_var_location): Likewise.
15774 (vt_add_function_parameter): Likewise.
15775 (vt_initialize): Likewise.
15777 2018-06-12 Jeff Law <law@redhat.com>
15779 * config.gcc (alpha*-*-freebsd*): Remove.
15780 * config/alpha/freebsd.h: Remove.
15782 2018-06-12 David Malcolm <dmalcolm@redhat.com>
15785 * spellcheck-tree.c (levenshtein_distance): Rename to...
15786 (get_edit_distance): ...this, and update for underlying renaming.
15787 * spellcheck-tree.h (levenshtein_distance): Rename to...
15788 (get_edit_distance): ...this.
15789 * spellcheck.c (levenshtein_distance): Rename to...
15790 (get_edit_distance): ...this. Convert from Levenshtein distance
15791 to Damerau-Levenshtein distance by supporting transpositions of
15792 adjacent characters. Rename "v1" to "v_next" and "v0" to
15794 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
15795 (selftest::test_edit_distance_unit_test_oneway): ...this, and
15796 update for underlying renaming.
15797 (selftest::levenshtein_distance_unit_test): Rename to...
15798 (selftest::test_get_edit_distance_unit): ...this, and update for
15799 underlying renaming.
15800 (selftest::test_find_closest_string): Add example from PR 69968
15801 where transposition helps
15802 (selftest::test_metric_conditions): Update for renaming.
15803 (selftest::test_metric_conditions): Likewise.
15804 (selftest::spellcheck_c_tests): Likewise.
15805 * spellcheck.h (levenshtein_distance): Rename both overloads to...
15806 (get_edit_distance): ...this.
15807 (best_match::consider): Update for renaming.
15809 2018-06-12 Martin Sebor <msebor@redhat.com>
15811 PR tree-optimization/85259
15812 * builtins.c (compute_objsize): Handle constant offsets.
15813 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
15814 true iff a warning has been issued.
15815 * gimple.h (gimple_nonartificial_location): New function.
15816 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
15817 gimple_nonartificial_location and handle -Wno-system-headers.
15818 (handle_builtin_stxncpy): Same.
15820 2018-06-12 Martin Sebor <msebor@redhat.com>
15823 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
15825 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
15827 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
15828 BUILTIN_VEC_XST entries for pointer to double and long long.
15830 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
15833 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15835 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
15838 2018-06-12 Martin Liska <mliska@suse.cz>
15840 * doc/options.texi: Document IntegerRange.
15842 2018-06-12 Martin Liska <mliska@suse.cz>
15844 * config/i386/i386.opt: Make MPX-related options as Deprecated.
15845 * opt-functions.awk: Handle Deprecated flag.
15846 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
15848 (read_cmdline_option): Report warning for a deprecated option.
15849 * opts.h (struct cl_option): Add new field cl_deprecated.
15850 (CL_ERR_DEPRECATED): New.
15852 2018-06-12 Martin Liska <mliska@suse.cz>
15854 * doc/options.texi: Document Deprecated option flag.
15856 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15858 * config/arc/arc-arch.h (arc_extras): New enum.
15859 (arc_cpu_t):Add field extra.
15860 (arc_cpu_types): Consider the extras.
15861 * config/arc/arc-cpus.def: Add extras info.
15862 * config/arc/arc-opts.h (processor_type): Consider extra field.
15863 * config/arc/arc.c (arc_override_options): Handle extra field.
15865 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
15867 * config/arc/arc-arch.h: Update ARC_OPTX macro.
15868 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
15870 * config/arc/arc.c (arc_init): Update pic warning.
15871 (irq_range): Update irq range parsing warnings.
15872 (arc_override_options): Update various warning messages.
15873 (arc_handle_aux_attribute): Likewise.
15875 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
15877 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
15879 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
15881 * doc/sourcebuild.texi: Document usage of line number 0 in verify
15882 compiler messages directives.
15884 2018-06-12 Matthew Fortune <mfortune@gmail.com>
15886 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
15887 * config/mips/mips-tables.opt: Regenerate.
15888 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
15890 * doc/invoke.texi: Document -march=i6500.
15892 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
15894 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
15895 (i6400_gpmul): Add cpu_unit.
15896 (i6400_gpdiv): Likewise.
15897 (i6400_msa_add_d): Update reservations.
15898 (i6400_msa_int_add) Likewise.
15899 (i6400_msa_short_logic3) Likewise.
15900 (i6400_msa_short_logic2) Likewise.
15901 (i6400_msa_short_logic) Likewise.
15902 (i6400_msa_move) Likewise.
15903 (i6400_msa_cmp) Likewise.
15904 (i6400_msa_short_float2) Likewise.
15905 (i6400_msa_div_d) Likewise.
15906 (i6400_msa_long_logic1) Likewise.
15907 (i6400_msa_long_logic2) Likewise.
15908 (i6400_msa_mult) Likewise.
15909 (i6400_msa_long_float2) Likewise.
15910 (i6400_msa_long_float4) Likewise.
15911 (i6400_msa_long_float5) Likewise.
15912 (i6400_msa_long_float8) Likewise.
15913 (i6400_fpu_fadd): Include frint type.
15914 (i6400_fpu_store): New define_insn_reservation.
15915 (i6400_fpu_load): Likewise.
15916 (i6400_fpu_move): Likewise.
15917 (i6400_fpu_fcmp): Likewise.
15918 (i6400_fpu_fmadd): Likewise.
15919 (i6400_int_mult): Include imul3nc type and update reservation.
15920 (i6400_int_div): Include idiv3 type and update reservation.
15921 (i6400_int_load): Update to check type not move_type.
15922 (i6400_int_store): Likewise.
15923 (i6400_int_prefetch): Set zero latency.
15925 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
15927 * gcc.c: Document new %@{...} sequence.
15928 (LINK_COMMAND_SPEC): Use it for the -L switches.
15929 (cpp_unique_options): Use it for the -I switches.
15930 (at_file_argbuf): New global variable.
15931 (in_at_file): Likewise.
15932 (alloc_args): Create at_file_argbuf.
15933 (clear_args): Truncate at_file_argbuf.
15934 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
15935 (open_at_file): New function.
15936 (close_at_file): Likewise.
15937 (create_at_file): Delete.
15938 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
15941 (validate_switches_from_spec): Deal with %@{...} sequence.
15942 (validate_switches): Likewise.
15943 (driver::finalize): Call clear_args.
15945 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
15947 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
15949 2018-06-11 Martin Sebor <msebor@redhat.com>
15951 * doc/invoke.texi (-Wall): List -Wc++17-compat.
15952 (Wno-class-memaccess): Add @opindex.
15953 (Wno-templates, Wno-multiple-inheritance): Same.
15954 (Wno-virtual-inheritance, Wno-namespaces): Same.
15955 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
15956 (Wno-format-overflow, Wno-format-truncation): Same.
15957 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
15958 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
15959 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
15960 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
15961 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
15962 (Wno-misspelled-isr): Same.
15964 2018-06-11 Martin Sebor <msebor@redhat.com>
15966 * PR tree-optimization/86083
15967 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
15969 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
15971 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
15973 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
15976 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
15977 on the correct operand.
15978 (*movdi_internal64): Ditto.
15980 2018-06-11 Martin Liska <mliska@suse.cz>
15982 PR tree-optimization/86089
15983 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
15985 2018-06-11 Julia Koval <julia.koval@intel.com>
15987 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
15988 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
15989 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
15991 2018-06-11 Olivier Hainque <hainque@adacore.com>
15993 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
15994 for Ada with strict dwarf2.
15996 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
15999 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16002 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
16004 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16006 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
16008 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16011 2018-06-08 Martin Liska <mliska@suse.cz>
16013 * tree-cfg.h (debug_function): Fix argument type to match
16016 2018-06-08 Martin Liska <mliska@suse.cz>
16018 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16019 Remove usage of MPX-related (and removed) fields.
16020 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16022 2018-06-08 David Malcolm <dmalcolm@redhat.com>
16024 * cfg.c (debug): Use TDF_NONE rather than 0.
16025 * cfghooks.c (debug): Likewise.
16026 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16027 (struct dump_option_value_info): Convert to...
16028 (struct kv_pair): ...this template type.
16029 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16031 (optinfo_verbosity_options): Likewise.
16032 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16034 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16035 than int for "optgroup_flags" param.
16036 (dump_generic_expr_loc): Use dump_flags_t rather than int for
16038 (dump_dec): Likewise.
16039 (dump_finish): Use TDF_NONE rather than 0.
16040 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16041 rather than int for "optgroup_flags" param. Use TDF_NONE rather
16042 than 0. Update for change to option_ptr.
16043 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16044 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
16045 0. Update for changes to optinfo_verbosity_options and
16047 (opt_info_switch_p): Convert optgroup_flags from int to
16049 (dump_basic_block): Use dump_flags_t rather than int
16050 for "dump_kind" param.
16051 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16052 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16053 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16054 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16055 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16056 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16057 TDF_NONE): Convert from macros to...
16058 (enum dump_flag): ...this new enum.
16059 (dump_flags_t): Update to use enum.
16060 (operator|, operator&, operator~, operator|=, operator&=):
16061 Implement for dump_flags_t.
16062 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16063 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16064 Convert from macros to...
16065 (enum optgroup_flag): ...this new enum.
16066 (optgroup_flags_t): New typedef.
16067 (operator|, operator|=): Implement for optgroup_flags_t.
16068 (struct dump_file_info): Convert field "alt_flags" to
16069 dump_flags_t. Convert field "optgroup_flags" to
16071 (dump_basic_block): Use dump_flags_t rather than int for param.
16072 (dump_generic_expr_loc): Likewise.
16073 (dump_dec): Likewise.
16074 (dump_register): Convert param "optgroup_flags" to
16076 (opt_info_enable_passes): Likewise.
16077 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16079 * gimple-pretty-print.c (debug): Likewise.
16080 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16081 (merged_store_group::apply_stores): Likewise.
16082 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16083 * gimple.c (verify_gimple_pp): Likewise.
16084 * graphite-poly.c (print_pbb_body): Likewise.
16085 * passes.c (pass_manager::register_one_dump_file): Convert
16086 local "optgroup_flags" to optgroup_flags_t.
16087 * print-tree.c (print_node): Use TDF_NONE rather than 0.
16089 (debug_body): Likewise.
16090 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16091 to optgroup_flags_t.
16092 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16094 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16095 (convert_mult_to_fma): Likewise.
16096 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16097 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16098 * tree-vect-data-refs.c (dump_lower_bound): Convert param
16099 "dump_kind" to dump_flags_t.
16101 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
16103 * config/rs6000/rs6000.c (min, max): Delete.
16105 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
16107 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16108 -mabi=spe and -mabi=no-spe.
16110 2018-06-08 Martin Liska <mliska@suse.cz>
16112 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16113 where we expect an existing summary.
16115 2018-06-08 Martin Liska <mliska@suse.cz>
16117 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16118 * ipa-inline.h (estimate_edge_growth): Likewise.
16120 2018-06-08 Martin Liska <mliska@suse.cz>
16122 * cgraph.c (function_version_hasher::hash): Use
16123 cgraph_node::get_uid ().
16124 (function_version_hasher::equal):
16125 * cgraph.h (cgraph_node::get_uid): New method.
16126 * ipa-inline.c (update_caller_keys): Use
16127 cgraph_node::get_uid ().
16128 (update_callee_keys): Likewise.
16129 * ipa-utils.c (searchc): Likewise.
16130 (ipa_reduced_postorder): Likewise.
16131 * lto-cgraph.c (input_node): Likewise.
16132 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16133 * symbol-summary.h (symtab_insertion): Likewise.
16134 (symtab_removal): Likewise.
16135 (symtab_duplication): Likewise.
16136 * tree-pretty-print.c (dump_function_header): Likewise.
16137 * tree-sra.c (convert_callers_for_node): Likewise.
16139 2018-06-08 Martin Liska <mliska@suse.cz>
16141 * cgraph.c (symbol_table::create_edge): Always assign a new
16143 (symbol_table::free_edge): Do not recycle numbers.
16144 * cgraph.h (cgraph_edge::get): New method.
16145 * symbol-summary.h (symtab_removal): Use it.
16146 (symtab_duplication): Likewise.
16147 (call_summary::hashable_uid): Remove.
16149 2018-06-08 Martin Liska <mliska@suse.cz>
16151 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16152 (initialize_growth_caches): Remove.
16153 (free_growth_caches): Likewise.
16154 (do_estimate_edge_time): Use edge_growth_cache.
16155 (do_estimate_edge_size): Likewise.
16156 (do_estimate_edge_hints): Likewise.
16157 * ipa-inline.c (reset_edge_caches): Likewise.
16158 (recursive_inlining): Likewise.
16159 (inline_small_functions): Likewise.
16160 * ipa-inline.h (initialize_growth_caches): Remove.
16161 (estimate_edge_size): Likewise.
16162 (estimate_edge_time): Likewise.
16163 (estimate_edge_hints): Likewise.
16164 (reset_edge_growth_cache): Likewise.
16165 * symbol-summary.h (call_summary::remove): New method.
16167 2018-06-08 Martin Liska <mliska@suse.cz>
16169 * ipa-cp.c (class edge_clone_summary): New summary.
16170 (grow_edge_clone_vectors): Remove.
16171 (ipcp_edge_duplication_hook): Remove.
16172 (class edge_clone_summary_t): New call_summary class.
16173 (ipcp_edge_removal_hook): Remove.
16174 (edge_clone_summary_t::duplicate): New function.
16175 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16176 (create_specialized_node): Likewise.
16177 (ipcp_driver): Initialize edge_clone_summaries and do not
16180 2018-06-08 Martin Liska <mliska@suse.cz>
16182 * symbol-summary.h (get): New function.
16183 (call_summary::m_initialize_when_cloning): New class member.
16185 2018-06-08 Martin Liska <mliska@suse.cz>
16187 * cgraph.c (cgraph_node::remove): Do not recycle uid.
16188 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16189 (symbol_table::allocate_cgraph_symbol): Do not set uid.
16190 * passes.c (uid_hash_t): Record removed_nodes by their uids.
16191 (remove_cgraph_node_from_order): Use the removed_nodes set.
16192 (do_per_function_toporder): Likwise.
16193 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16194 instead of summary_uid.
16195 (symtab_removal): Likewise.
16196 (symtab_duplication): Likewise.
16198 2018-06-08 Martin Liska <mliska@suse.cz>
16200 * ipa-cp.c (ipcp_store_bits_results): Use
16201 ipcp_transformation_sum.
16202 (ipcp_store_vr_results): Likewise.
16203 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16205 (ipcp_transformation_initialize): ... this.
16206 (ipa_set_node_agg_value_chain):
16207 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16208 (write_ipcp_transformation_info): Likewise.
16209 (read_ipcp_transformation_info): Likewise.
16210 (ipcp_update_bits): Likewise.
16211 (ipcp_update_vr): Likewise.
16212 (ipcp_transform_function): Likewise.
16213 * ipa-prop.h: Rename ipcp_transformation_summary to
16214 ipcp_transformation.
16215 (class ipcp_transformation_t): New function summary.
16216 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16217 (ipa_get_agg_replacements_for_node): Likewise.
16219 2018-06-08 Martin Liska <mliska@suse.cz>
16221 * ipa-pure-const.c (struct funct_state_d): Do it class instead
16223 (class funct_state_summary_t): New function_summary class.
16224 (has_function_state): Remove.
16225 (get_function_state): Likewise.
16226 (set_function_state): Likewise.
16227 (add_new_function): Likewise.
16228 (funct_state_summary_t::insert): New function.
16229 (duplicate_node_data): Remove.
16230 (remove_node_data): Remove.
16231 (funct_state_summary_t::duplicate): New function.
16232 (register_hooks): Create new funct_state_summaries.
16233 (pure_const_generate_summary): Use it.
16234 (pure_const_write_summary): Likewise.
16235 (pure_const_read_summary): Likewise.
16236 (propagate_pure_const): Likewise.
16237 (propagate_nothrow): Likewise.
16238 (dump_malloc_lattice): Likewise.
16239 (propagate_malloc): Likewise.
16240 (execute): Do not register hooks, just remove summary
16242 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16245 2018-06-08 Martin Liska <mliska@suse.cz>
16247 * ipa-reference.c (remove_node_data): Remove.
16248 (duplicate_node_data): Likewise.
16249 (class ipa_ref_var_info_summary_t): New class.
16250 (class ipa_ref_opt_summary_t): Likewise.
16251 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16252 (get_reference_optimization_summary): Use
16253 ipa_ref_opt_sum_summaries.
16254 (set_reference_vars_info): Remove.
16255 (set_reference_optimization_summary): Likewise.
16256 (ipa_init): Create summaries.
16257 (init_function_info): Use function summary.
16258 (ipa_ref_opt_summary_t::duplicate): New function.
16259 (ipa_ref_opt_summary_t::remove): New function.
16260 (get_read_write_all_from_node): Fix GNU coding style.
16261 (propagate): Use function summary.
16262 (write_node_summary_p): Fix GNU coding style.
16263 (stream_out_bitmap): Likewise.
16264 (ipa_reference_read_optimization_summary): Use function summary.
16265 (ipa_reference_c_finalize): Do not release hooks.
16267 2018-06-08 Martin Liska <mliska@suse.cz>
16269 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16270 (analyze_function_body): Extract multiple calls of get_create.
16271 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16272 * ipa-inline.c (recursive_inlining): Use ::get method.
16273 * ipa-inline.h (estimate_edge_growth): Likewise.
16275 2018-06-08 Martin Liska <mliska@suse.cz>
16277 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16279 (hsa_function_summary::hsa_function_summary): Use the new enum
16281 (hsa_gpu_implementation_p): Use hsa_summaries::get.
16282 * hsa-gen.c (hsa_get_host_function): Likewise.
16283 (get_brig_function_name): Likewise.
16284 * ipa-hsa.c (process_hsa_functions): Likewise.
16285 (ipa_hsa_write_summary): Likewise.
16286 * symbol-summary.h (symtab_duplication): Use ::get function/
16287 (get): New function.
16289 2018-06-08 Martin Liska <mliska@suse.cz>
16291 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16293 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16294 (hsa_register_kernel): Likewise.
16295 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16296 * hsa-gen.c (hsa_get_host_function): Likewise.
16297 (get_brig_function_name): Likewise.
16298 (generate_hsa): Likewise.
16299 (pass_gen_hsail::execute): Likewise.
16300 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16301 (devirtualization_time_bonus): Likewise.
16302 (ipcp_propagate_stage): Likewise.
16303 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16304 (edge_set_predicate): Likewise.
16305 (evaluate_conditions_for_known_args): Likewise.
16306 (evaluate_properties_for_edge): Likewise.
16307 (ipa_fn_summary::reset): Likewise.
16308 (ipa_fn_summary_t::duplicate): Likewise.
16309 (dump_ipa_call_summary): Likewise.
16310 (ipa_dump_fn_summary): Likewise.
16311 (analyze_function_body): Likewise.
16312 (compute_fn_summary): Likewise.
16313 (estimate_edge_devirt_benefit): Likewise.
16314 (estimate_edge_size_and_time): Likewise.
16315 (estimate_calls_size_and_time): Likewise.
16316 (estimate_node_size_and_time): Likewise.
16317 (inline_update_callee_summaries): Likewise.
16318 (remap_edge_change_prob): Likewise.
16319 (remap_edge_summaries): Likewise.
16320 (ipa_merge_fn_summary_after_inlining): Likewise.
16321 (ipa_update_overall_fn_summary): Likewise.
16322 (read_ipa_call_summary): Likewise.
16323 (inline_read_section): Likewise.
16324 (write_ipa_call_summary): Likewise.
16325 (ipa_fn_summary_write): Likewise.
16326 (ipa_free_fn_summary): Likewise.
16327 * ipa-hsa.c (process_hsa_functions): Likewise.
16328 (ipa_hsa_write_summary): Likewise.
16329 (ipa_hsa_read_section): Likewise.
16330 * ipa-icf.c (sem_function::merge): Likewise.
16331 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16332 (do_estimate_edge_time): Likewise.
16333 (estimate_size_after_inlining): Likewise.
16334 (estimate_growth): Likewise.
16335 (growth_likely_positive): Likewise.
16336 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16337 (inline_call): Likewise.
16338 * ipa-inline.c (caller_growth_limits): Likewise.
16339 (can_inline_edge_p): Likewise.
16340 (can_inline_edge_by_limits_p): Likewise.
16341 (compute_uninlined_call_time): Likewise.
16342 (compute_inlined_call_time): Likewise.
16343 (want_inline_small_function_p): Likewise.
16344 (edge_badness): Likewise.
16345 (update_caller_keys): Likewise.
16346 (update_callee_keys): Likewise.
16347 (recursive_inlining): Likewise.
16348 (inline_small_functions): Likewise.
16349 (inline_to_all_callers_1): Likewise.
16350 (dump_overall_stats): Likewise.
16351 (early_inline_small_functions): Likewise.
16352 (early_inliner): Likewise.
16353 * ipa-inline.h (estimate_edge_growth): Likewise.
16354 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16355 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16356 * ipa-prop.h (IPA_NODE_REF): Likewise.
16357 (IPA_EDGE_REF): Likewise.
16358 * ipa-pure-const.c (malloc_candidate_p): Likewise.
16359 (propagate_malloc): Likewise.
16360 * ipa-split.c (execute_split_functions): Likewise.
16361 * symbol-summary.h: Rename get to get_create.
16363 (get_create): Likewise.
16364 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16366 2018-06-08 Martin Liska <mliska@suse.cz>
16368 * symbol-summary.h (release): Move definition out of class
16370 (symtab_removal): Likewise.
16371 (symtab_duplication): Likewise.
16373 2018-06-08 Martin Liska <mliska@suse.cz>
16375 * symbol-summary.h (function_summary): Move constructor
16376 implementation out of class declaration.
16377 (release): Likewise.
16378 (symtab_insertion): Likewise.
16379 (symtab_removal): Likewise.
16380 (symtab_duplication): Likewise.
16383 2018-06-08 Martin Liska <mliska@suse.cz>
16385 * Makefile.in: Remove support for MPX (macros, related functions,
16386 fields in cgraph_node, ...).
16387 * builtin-types.def (BT_BND): Likewise.
16388 (BT_FN_BND_CONST_PTR): Likewise.
16389 (BT_FN_CONST_PTR_BND): Likewise.
16390 (BT_FN_VOID_PTR_BND): Likewise.
16391 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16392 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16393 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16394 (expand_builtin_mempcpy_with_bounds): Likewise.
16395 (expand_builtin_memset_with_bounds): Likewise.
16396 (expand_builtin_memset_args): Likewise.
16397 (std_expand_builtin_va_start): Likewise.
16398 (expand_builtin): Likewise.
16399 (expand_builtin_with_bounds): Likewise.
16400 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16401 (DEF_LIB_BUILTIN_CHKP): Likewise.
16402 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16403 (DEF_CHKP_BUILTIN): Likewise.
16404 (BUILT_IN_MEMCPY): Likewise.
16405 (BUILT_IN_MEMMOVE): Likewise.
16406 (BUILT_IN_MEMPCPY): Likewise.
16407 (BUILT_IN_MEMSET): Likewise.
16408 (BUILT_IN_STPCPY): Likewise.
16409 (BUILT_IN_STRCAT): Likewise.
16410 (BUILT_IN_STRCHR): Likewise.
16411 (BUILT_IN_STRCPY): Likewise.
16412 (BUILT_IN_STRLEN): Likewise.
16413 (BUILT_IN_MEMCPY_CHK): Likewise.
16414 (BUILT_IN_MEMMOVE_CHK): Likewise.
16415 (BUILT_IN_MEMPCPY_CHK): Likewise.
16416 (BUILT_IN_MEMSET_CHK): Likewise.
16417 (BUILT_IN_STPCPY_CHK): Likewise.
16418 (BUILT_IN_STRCAT_CHK): Likewise.
16419 (BUILT_IN_STRCPY_CHK): Likewise.
16420 * calls.c (store_bounds): Likewise.
16421 (emit_call_1): Likewise.
16422 (special_function_p): Likewise.
16423 (maybe_warn_nonstring_arg): Likewise.
16424 (initialize_argument_information): Likewise.
16425 (finalize_must_preallocate): Likewise.
16426 (compute_argument_addresses): Likewise.
16427 (expand_call): Likewise.
16428 * cfgexpand.c (expand_call_stmt): Likewise.
16429 (expand_return): Likewise.
16430 (expand_gimple_stmt_1): Likewise.
16431 (pass_expand::execute): Likewise.
16432 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16433 (cgraph_node::remove): Likewise.
16434 (cgraph_node::dump): Likewise.
16435 (cgraph_node::verify_node): Likewise.
16436 * cgraph.h (chkp_function_instrumented_p): Likewise.
16437 (symtab_node::get_alias_target): Likewise.
16438 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16439 (cgraph_local_p): Likewise.
16440 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16441 (cgraph_edge::rebuild_references): Likewise.
16442 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16443 (walk_polymorphic_call_targets): Likewise.
16444 (cgraph_node::expand_thunk): Likewise.
16445 (symbol_table::output_weakrefs): Likewise.
16446 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16447 (ix86_handle_option): Likewise.
16448 * config/i386/constraints.md: Likewise.
16449 * config/i386/i386-builtin-types.def (BND): Likewise.
16453 * config/i386/i386-builtin.def (BDESC_END): Likewise.
16454 (BDESC_FIRST): Likewise.
16456 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16457 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16458 * config/i386/i386.c (enum reg_class): Likewise.
16459 (ix86_target_string): Likewise.
16460 (ix86_option_override_internal): Likewise.
16461 (ix86_conditional_register_usage): Likewise.
16462 (ix86_valid_target_attribute_inner_p): Likewise.
16463 (ix86_set_indirect_branch_type): Likewise.
16464 (ix86_set_current_function): Likewise.
16465 (ix86_function_arg_regno_p): Likewise.
16466 (init_cumulative_args): Likewise.
16467 (ix86_function_arg_advance): Likewise.
16468 (ix86_function_arg): Likewise.
16469 (ix86_pass_by_reference): Likewise.
16470 (ix86_function_value_regno_p): Likewise.
16471 (ix86_function_value_1): Likewise.
16472 (ix86_function_value_bounds): Likewise.
16473 (ix86_return_in_memory): Likewise.
16474 (ix86_setup_incoming_vararg_bounds): Likewise.
16475 (ix86_va_start): Likewise.
16476 (indirect_thunk_need_prefix): Likewise.
16477 (print_reg): Likewise.
16478 (ix86_print_operand): Likewise.
16479 (ix86_expand_call): Likewise.
16480 (ix86_output_function_return): Likewise.
16481 (reg_encoded_number): Likewise.
16482 (BDESC_VERIFYS): Likewise.
16483 (ix86_init_mpx_builtins): Likewise.
16484 (ix86_init_builtins): Likewise.
16485 (ix86_emit_cmove): Likewise.
16486 (ix86_emit_move_max): Likewise.
16487 (ix86_expand_builtin): Likewise.
16488 (ix86_builtin_mpx_function): Likewise.
16489 (ix86_get_arg_address_for_bt): Likewise.
16490 (ix86_load_bounds): Likewise.
16491 (ix86_store_bounds): Likewise.
16492 (ix86_load_returned_bounds): Likewise.
16493 (ix86_store_returned_bounds): Likewise.
16494 (ix86_class_likely_spilled_p): Likewise.
16495 (ix86_hard_regno_mode_ok): Likewise.
16496 (x86_order_regs_for_local_alloc): Likewise.
16497 (ix86_mitigate_rop): Likewise.
16498 (ix86_bnd_prefixed_insn_p): Likewise.
16499 (ix86_mpx_bound_mode): Likewise.
16500 (ix86_make_bounds_constant): Likewise.
16501 (ix86_initialize_bounds): Likewise.
16502 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16503 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16504 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16505 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16506 (TARGET_CHKP_BOUND_MODE): Likewise.
16507 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16508 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16509 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16510 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16511 * config/i386/i386.h (TARGET_MPX): Likewise.
16512 (TARGET_MPX_P): Likewise.
16513 (VALID_BND_REG_MODE): Likewise.
16514 (FIRST_BND_REG): Likewise.
16515 (LAST_BND_REG): Likewise.
16516 (enum reg_class): Likewise.
16517 (BND_REG_P): Likewise.
16518 (BND_REGNO_P): Likewise.
16519 (BNDmode): Likewise.
16520 (ADJUST_INSN_LENGTH): Likewise.
16521 * config/i386/i386.md: Likewise.
16522 * config/i386/i386.opt: Likewise.
16523 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16524 (defined): Likewise.
16525 (LINK_MPX): Likewise.
16526 (MPX_SPEC): Likewise.
16527 (LIBMPX_SPEC): Likewise.
16528 (LIBMPXWRAPPERS_SPEC): Likewise.
16529 (CHKP_SPEC): Likewise.
16530 * config/i386/predicates.md: Likewise.
16531 * dbxout.c (dbxout_type): Likewise.
16532 * doc/extend.texi: Likewise.
16533 * doc/invoke.texi: Likewise.
16534 * doc/md.texi: Likewise.
16535 * doc/tm.texi: Likewise.
16536 * doc/tm.texi.in: Likewise.
16537 * dwarf2out.c (is_base_type): Likewise.
16538 (gen_formal_types_die): Likewise.
16539 (gen_subprogram_die): Likewise.
16540 (gen_type_die_with_usage): Likewise.
16541 (gen_decl_die): Likewise.
16542 (dwarf2out_late_global_decl): Likewise.
16543 * expr.c (expand_assignment): Likewise.
16544 (emit_storent_insn): Likewise.
16545 (store_expr_with_bounds): Likewise.
16546 (store_expr): Likewise.
16547 (expand_expr_real_1): Likewise.
16548 * expr.h (store_expr_with_bounds): Likewise.
16549 * function.c (use_register_for_decl): Likewise.
16550 (struct bounds_parm_data): Likewise.
16551 (assign_parms_augmented_arg_list): Likewise.
16552 (assign_parm_find_entry_rtl): Likewise.
16553 (assign_parm_is_stack_parm): Likewise.
16554 (assign_parm_load_bounds): Likewise.
16555 (assign_bounds): Likewise.
16556 (assign_parms): Likewise.
16557 (expand_function_start): Likewise.
16558 * gcc.c (CHKP_SPEC): Likewise.
16559 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16560 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16561 (wrestrict_dom_walker::check_call): Likewise.
16562 * gimple.c (gimple_build_call_from_tree): Likewise.
16563 * gimple.h (enum gf_mask): Likewise.
16564 (gimple_call_with_bounds_p): Likewise.
16565 (gimple_call_set_with_bounds): Likewise.
16566 * gimplify.c (gimplify_init_constructor): Likewise.
16567 * ipa-cp.c (initialize_node_lattices): Likewise.
16568 (propagate_constants_across_call): Likewise.
16569 (find_more_scalar_values_for_callers_subset): Likewise.
16570 * ipa-hsa.c (process_hsa_functions): Likewise.
16571 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16572 * ipa-icf.c (sem_function::merge): Likewise.
16573 * ipa-inline.c (early_inliner): Likewise.
16574 * ipa-pure-const.c (warn_function_noreturn): Likewise.
16575 (warn_function_cold): Likewise.
16576 (propagate_pure_const): Likewise.
16577 * ipa-ref.h (enum GTY): Likewise.
16578 * ipa-split.c (find_retbnd): Likewise.
16579 (consider_split): Likewise.
16580 (split_function): Likewise.
16581 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16582 * ipa.c (walk_polymorphic_call_targets): Likewise.
16583 (symbol_table::remove_unreachable_nodes): Likewise.
16584 (process_references): Likewise.
16585 (cgraph_build_static_cdtor_1): Likewise.
16586 * lto-cgraph.c (lto_output_node): Likewise.
16587 (output_refs): Likewise.
16588 (compute_ltrans_boundary): Likewise.
16589 (input_overwrite_node): Likewise.
16590 (input_node): Likewise.
16591 (input_cgraph_1): Likewise.
16592 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
16593 * passes.c (pass_manager::execute_early_local_passes): Likewise.
16594 (class pass_chkp_instrumentation_passes): Likewise.
16595 (make_pass_chkp_instrumentation_passes): Likewise.
16596 * passes.def: Likewise.
16597 * rtl.h (struct GTY): Likewise.
16598 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
16599 * stor-layout.c (layout_type): Likewise.
16600 * symtab.c: Likewise.
16601 * target.def: Likewise.
16602 * targhooks.c (default_chkp_bound_type): Likewise.
16603 (default_chkp_bound_mode): Likewise.
16604 (default_builtin_chkp_function): Likewise.
16605 (default_chkp_function_value_bounds): Likewise.
16606 (default_chkp_make_bounds_constant): Likewise.
16607 (default_chkp_initialize_bounds): Likewise.
16608 * targhooks.h (default_chkp_bound_type): Likewise.
16609 (default_chkp_bound_mode): Likewise.
16610 (default_builtin_chkp_function): Likewise.
16611 (default_chkp_function_value_bounds): Likewise.
16612 (default_chkp_make_bounds_constant): Likewise.
16613 (default_chkp_initialize_bounds): Likewise.
16614 * toplev.c (compile_file): Likewise.
16615 (process_options): Likewise.
16616 * tree-core.h (DEF_BUILTIN): Likewise.
16617 (DEF_BUILTIN_CHKP): Likewise.
16618 * tree-inline.c (declare_return_variable): Likewise.
16619 (remap_gimple_stmt): Likewise.
16620 (copy_bb): Likewise.
16621 (initialize_inlined_parameters): Likewise.
16622 (expand_call_inline): Likewise.
16623 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
16624 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
16625 (make_pass_ipa_chkp_produce_thunks): Likewise.
16626 (make_pass_chkp): Likewise.
16627 (make_pass_chkp_opt): Likewise.
16628 (make_pass_chkp_instrumentation_passes): Likewise.
16629 * tree-pretty-print.c (dump_generic_node): Likewise.
16630 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
16631 * tree-ssa-dce.c (propagate_necessity): Likewise.
16632 (eliminate_unnecessary_stmts): Likewise.
16633 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
16634 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
16635 * tree-ssa-sccvn.h: Likewise.
16636 * tree-ssa-strlen.c (get_string_length): Likewise.
16637 (valid_builtin_call): Likewise.
16638 (adjust_last_stmt): Likewise.
16639 (handle_builtin_strchr): Likewise.
16640 (handle_builtin_strcpy): Likewise.
16641 (handle_builtin_stxncpy): Likewise.
16642 (handle_builtin_memcpy): Likewise.
16643 (handle_builtin_strcat): Likewise.
16644 (strlen_check_and_optimize_stmt): Likewise.
16645 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
16646 * tree-streamer-in.c: Likewise.
16647 * tree-streamer.c (record_common_node): Likewise.
16648 * tree.c (tree_code_size): Likewise.
16649 (wide_int_to_tree_1): Likewise.
16650 (type_contains_placeholder_1): Likewise.
16651 (build_common_tree_nodes): Likewise.
16652 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
16653 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
16654 (POINTER_BOUNDS_P): Likewise.
16655 (BOUNDED_TYPE_P): Likewise.
16656 (BOUNDED_P): Likewise.
16657 (CALL_WITH_BOUNDS_P): Likewise.
16658 (pointer_bounds_type_node): Likewise.
16659 * value-prof.c (gimple_ic): Likewise.
16660 * var-tracking.c (vt_add_function_parameters): Likewise.
16661 * varasm.c (make_decl_rtl): Likewise.
16662 (assemble_start_function): Likewise.
16663 (output_constant): Likewise.
16664 (maybe_assemble_visibility): Likewise.
16665 * varpool.c (ctor_for_folding): Likewise.
16666 * chkp-builtins.def: Remove.
16667 * ipa-chkp.c: Remove.
16668 * ipa-chkp.h: Remove.
16669 * rtl-chkp.c: Remove.
16670 * rtl-chkp.h: Remove.
16671 * tree-chkp-opt.c: Remove.
16672 * tree-chkp.c: Remove.
16673 * tree-chkp.h: Remove.
16675 2018-06-07 Carl Love <cel@us.ibm.com>
16677 * config/rs6000/vsx.md (vextract_fp_from_shorth,
16678 vextract_fp_from_shortl): Add BE support.
16680 2018-06-07 Paul Koning <ni1d@arrl.net>
16682 * compare-elim.c (try_merge_compare): Don't merge compare if
16683 address contains a side effect.
16684 (try_eliminate_compare): Likewise.
16686 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
16688 * config.gcc: Support "tremont".
16689 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
16690 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16692 * config/i386/i386.c (m_TREMONT): Define.
16693 (processor_target_table): Add "tremont".
16694 (PTA_TREMONT): Define.
16695 (ix86_lea_outperforms): Add TARGET_TREMONT.
16696 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
16697 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
16698 and M_INTEL_GOLDMONT_PLUS.
16699 (fold_builtin_cpu): Add "tremont".
16700 (ix86_add_stmt_cost): Add TARGET_TREMONT.
16701 (ix86_option_override_internal): Add "tremont".
16702 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
16703 (processor_type): Add PROCESSOR_TREMONT.
16704 * config/i386/x86-tune.def: Add m_TREMONT.
16705 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
16707 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16709 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
16710 symbol defined for msp430i* devices to be lower case.
16712 2018-06-07 Richard Biener <rguenther@suse.de>
16714 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
16715 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
16716 Properly wrap signed arithmetic if overflow wraps.
16718 2018-06-07 Jakub Jelinek <jakub@redhat.com>
16720 PR tree-optimization/69615
16721 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
16722 of a cast from a same precision integral SSA_NAME in a bb dominated
16723 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
16724 cast to utype if rhs2 has already a compatible type.
16726 2018-06-07 Richard Biener <rguenther@suse.de>
16728 PR tree-optimization/85935
16729 * graphite-scop-detection.c (find_params_in_bb): Analyze
16730 condition operands with respect to the correct loop. Assert
16731 the analysis doesn't fail.
16733 2018-06-04 Carl Love <cel@us.ibm.com>
16735 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
16736 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
16737 as it is slightly cheaper.
16738 (first_match_or_eos_index_<mode>):
16739 Calculate index using natural element order.
16740 (first_match_index_<mode>):
16741 Calculate index using natural element order.
16742 (first_match_or_eos_index_<mode>):
16743 Calculate index using natural order.
16744 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
16745 for BE and LE modes.
16746 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
16747 P9V_BUILTIN_VCLZLSBB_V16QI.
16748 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
16751 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
16753 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
16754 indentation and line wrap for many prototypes. Add missing
16755 @smallexample directives around block of prototypes for vec_xl and
16758 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
16760 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
16761 track if we pass or return IEEE 128-bit floating point.
16762 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
16763 C++ mangling that is compatible with GCC 8.1.
16764 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
16765 (init_cumulative_args): Note if we pass or return IEEE 128-bit
16766 floating point types.
16767 (rs6000_function_arg_advance_1): Likewise.
16768 (rs6000_mangle_type): Optionally generate mangled names that match
16769 what GCC 8.1 generated for IEEE 128-bit floating point types.
16770 (rs6000_globalize_decl_name): If we have an external function that
16771 passes or returns IEEE 128-bit types, generate a weak reference
16772 from the mangled name used in GCC 8.1 to the current mangled
16774 (rs6000_init_builtins): Make __ibm128 use the long double type if
16775 long double is IBM extended double. Make __float128 use the long
16776 double type if long double is IEEE 128-bit.
16779 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
16780 macro for __ibm128 built-in functions.
16781 (PACK_IF): Add __ibm128 pack/unpack functions.
16782 (UNPACK_IF): Likewise.
16783 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
16784 enable long double built-in functions if long double is IEEE
16785 128-bit floating point.
16786 (rs6000_invalid_builtin): Update long double built-in function
16788 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
16789 functions, adjust the built-in function to use the long double
16790 built-in function if __ibm128 and long double are the same type.
16791 * doc/extend.texi (PowerPC builtins): Update documention for
16792 __builtin_{,un}pack_longdouble. Add documentation for
16793 __builtin_{,un}pack_ibm128.
16795 2018-06-06 Jim Wilson <jimw@sifive.com>
16797 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
16798 (struct machine_function): New field interrupt_mode.
16799 (riscv_handle_type_attribute): New function. Add forward declaration.
16800 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
16801 (riscv_expand_epilogue): Check interrupt_mode field.
16802 (riscv_set_current_function): Check interrupt attribute args and
16803 set interrupt_mode field.
16804 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
16805 (riscv_sret, riscv_uret): New.
16806 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
16807 new arguments to interrupt attribute.
16809 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
16812 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
16813 Don't handle -mcpu=power8 if -mpower9-vector is also used.
16815 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
16817 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
16818 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
16819 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16820 several redundant entries.
16822 2018-06-06 David Malcolm <dmalcolm@redhat.com>
16824 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
16825 type from "rtx" to "rtx_insn *".
16826 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
16827 for local "call_insn", removing cast.
16828 (ix86_expand_call): Likewise, introducing a "call_insn" local.
16830 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
16832 PR tree-optimization/86066
16833 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
16834 for BIT_INSERT_EXPR stores.
16836 2018-06-06 Richard Biener <rguenther@suse.de>
16838 PR tree-optimization/86062
16839 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
16840 component refs ontop
16841 of to be offsetted base.
16843 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16845 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
16846 to be static and remove check on interrupt attribute name.
16848 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
16850 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
16851 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
16853 2018-06-05 Steve Ellcey <sellcey@cavium.com>
16856 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
16858 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
16859 Remove second argument, change how error is called.
16860 (aarch64_layout_arg): Remove second argument from
16861 aarch64_err_no_fpadvsimd call.
16862 (aarch64_init_cumulative_args): Ditto.
16863 (aarch64_gimplify_va_arg_expr): Ditto.
16864 * config/aarch64/aarch64.md (mov<mode>): Ditto.
16866 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
16868 * config/i386/i386.md (simple_return_indirect_internal): New expander.
16869 (*simple_return_indirect_internal<mode>): Rename from
16870 simple_return_indirect_internal. Use W mode iterator.
16871 (rstorssp): New expander.
16872 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
16873 (clrssbsy): New expander.
16874 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
16876 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
16878 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
16880 (cmse_check_pointed_object): Likewise.
16882 2018-06-05 Martin Liska <mliska@suse.cz>
16884 PR gcov-profile/47618
16885 * doc/invoke.texi: Document how -fprofile-dir format
16888 2018-06-05 Richard Biener <rguenther@suse.de>
16890 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
16891 removal pretend DOM info isn't available so we do not update
16892 it and only remove edges, not dominated blocks. Actually free
16893 DOM info in case we removed something. Remove unreachable blocks.
16894 (mfb_keep_latches): Work with either DOM info or marked backedges.
16895 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
16896 first. Mark backedges if DOM info isn't available.
16897 (Re-)compute DOM info after cleanup_control_flow_pre.
16899 2018-06-05 Richard Biener <rguenther@suse.de>
16901 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
16902 (locus_discrim_hasher::hash): Adjust.
16903 (locus_discrim_hasher::equal): Likewise.
16904 (next_discriminator_for_locus): Work on line directly.
16905 (same_line_p): Pass in expanded locus1 as well.
16906 (assign_discriminators): Avoid redundant location expansions.
16908 2018-06-05 Richard Biener <rguenther@suse.de>
16910 PR tree-optimization/86046
16911 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
16912 if required after clearing TREE_ADDRESSABLE.
16914 2018-06-05 Richard Biener <rguenther@suse.de>
16916 PR tree-optimization/86047
16917 * tree-ssa-loop.c (for_each_index): Glob handling of all
16918 decls and constants and really handle all of them.
16920 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16923 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
16924 qualifier_void_pointer and qualifier_const_void_pointer.
16925 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
16926 (arm_init_builtins): Handle the above.
16927 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
16928 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
16929 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
16932 2018-06-05 Martin Liska <mliska@suse.cz>
16934 * auto-profile.c (read_autofdo_file): Do not use
16935 gcov_ctr_summary struct.
16936 (afdo_callsite_hot_enough_for_early_inline): Likewise.
16937 * coverage.c (struct counts_entry): Likewise.
16938 (read_counts_file): Read just single summary entry.
16939 (get_coverage_counts): Use gcov_summary struct.
16940 * coverage.h (get_coverage_counts): Likewise.
16941 * gcov-dump.c (dump_working_sets): Likewise.
16942 (tag_summary): Dump just single summary.
16943 * gcov-io.c (gcov_write_summary): Write just histogram
16945 (gcov_read_summary): Read just single summary.
16946 (compute_working_sets): Use gcov_summary struct.
16947 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
16948 of GCOV_COUNTERS_SUMMABLE.
16949 (GCOV_COUNTERS_SUMMABLE): Remove.
16950 (GCOV_FIRST_VALUE_COUNTER): Replace with
16951 GCOV_COUNTER_V_INTERVAL.
16952 (struct gcov_ctr_summary): Remove.
16953 (struct gcov_summary): Directly use fields of former
16955 (compute_working_sets): Use gcov_summary struct.
16956 * gcov.c (read_count_file): Do not use ctrs fields.
16957 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
16959 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
16961 * profile.c: Likewise.
16962 * profile.h: Likewise.
16964 2018-06-05 Martin Liska <mliska@suse.cz>
16966 PR gcov-profile/84846
16967 * gcov.c (output_lines): Print working directory only
16968 in intermediate format.
16970 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
16972 * config/s390/s390-builtin-types.def: Add void function type.
16973 * config/s390/s390-builtins.def: Use the function type for the
16976 2018-06-04 Jim Wilson <jimw@sifive.com>
16978 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
16980 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
16981 and maybe_eh_return. Change regno to unsigned int. Use new args to
16982 handle EH_RETURN_DATA_REGNO registers properly.
16983 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
16984 (riscv_expand_epilogue): Update comment. Change argument name and
16985 type. Update code to use new name and type. Pass new args to
16986 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
16988 * config/riscv/riscv.md (NORMAL_RETURN): New.
16989 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
16990 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16991 (eh_return): Call gen_eh_return_internal and emit barrier.
16992 (eh_return_internal): Call riscv_expand_epilogue.
16994 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
16996 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
16997 bit_insertion field and declare can_be_merged_into method.
16998 (merged_store_group::can_be_merged_into): New method.
16999 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17000 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
17001 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17003 2018-06-04 Richard Biener <rguenther@suse.de>
17005 PR tree-optimization/85955
17006 * builtins.c (fold_builtin_sincos): Convert pointers to
17007 destination to appropriate type before dereferencing.
17009 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
17011 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17013 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
17015 * expr.c (expand_expr_real_1): Force the operand into memory if
17016 its TYPE_MODE is BLKmode and if there is no integer mode for
17017 the number of bits being extracted.
17019 2018-06-04 Jakub Jelinek <jakub@redhat.com>
17023 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17024 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17026 2018-06-04 Richard Biener <rguenther@suse.de>
17028 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17029 (cleanup_tree_cfg_noloop): ... single caller. Do
17030 start_recording_case_labels later.
17032 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
17034 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17035 to _IMMINTRIN_H_INCLUDED.
17036 * config/i386/pconfigintrin.h: Ditto.
17037 * config/i386/waitpkgintrin.h: Ditto.
17038 * config/i386/immintrin.h: Add includes for sgxintrin.h,
17039 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17040 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17041 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17042 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17043 waitpkgintrin.h and cldemoteintrin.h.
17045 2018-06-04 Richard Biener <rguenther@suse.de>
17047 PR tree-optimization/86038
17048 * tracer.c (find_best_successor): Check probability for
17049 being initialized, bail out if not.
17051 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
17054 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17055 of bits to ignore when comparing architectures.
17057 2018-06-04 Jakub Jelinek <jakub@redhat.com>
17059 PR tree-optimization/69615
17060 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17061 maximum or minimum of the type, try to merge it also as if
17062 range1 is + [-, x - 1] or + [x + 1, -].
17065 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17067 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
17069 PR tree-optimization/86034
17070 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17071 the unsigned bitfield type in a bit insertion sequence if it does not
17072 have a larger precision than the bitfield size.
17073 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17075 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
17077 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17079 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
17081 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17082 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17083 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17084 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17086 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
17088 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17089 Disable -fdelete-null-pointer-checks for ELF toolchain.
17091 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
17092 Kito Cheng <kito.cheng@gmail.com>
17094 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17095 (nds32le-*-*, nds32be-*-*): Integrate checking process.
17096 (nds32*-*-*): Add glibc and uclibc conditions.
17097 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17098 (TARGET_EXCEPT_UNWIND_INFO): Define.
17099 * config/nds32/elf.h: New file.
17100 * config/nds32/linux.h: New file.
17101 * config/nds32/nds32-elf.opt: New file.
17102 * config/nds32/nds32-linux.opt: New file.
17103 * config/nds32/nds32-fp-as-gp.c
17104 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17105 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17107 (nds32_asm_file_end): Ditto.
17108 (nds32_print_operand): Ditto.
17109 (nds32_insert_attributes): Ditto.
17110 (nds32_init_libfuncs): New function.
17111 (TARGET_HAVE_TLS): Define.
17112 (TARGET_INIT_LIBFUNCS): Define.
17113 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17115 (TARGET_ELF): Apply different mcmodel setting.
17116 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17117 been migrated into elf.h and linux.h files.
17118 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17119 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17120 (mcmodel): The content has been migrated into nds32-elf.opt and
17121 nds32-linux.opt files.
17122 * config/nds32/t-elf: New file.
17123 * config/nds32/t-linux: New file.
17125 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
17126 Shiva Chen <shiva0217@gmail.com>
17128 * config/nds32/constants.md (unspec_volatile_element): Add
17129 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17130 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17132 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17133 (make_pass_nds32_fp_as_gp): Declare.
17134 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17136 (nds32_asm_function_end_prologue): Remove unused asm output.
17137 (nds32_asm_function_begin_epilogue): Remove unused asm output.
17138 (nds32_asm_file_start): Output necessary fp_as_gp information.
17139 (nds32_option_override): Adjust register usage.
17140 (nds32_expand_prologue): Consider fp_as_gp situation.
17141 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17142 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17145 (simple_return): Ditto.
17146 (omit_fp_begin): Output special directive for fp_as_gp.
17147 (omit_fp_end): Output special directive for fp_as_gp.
17148 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17149 mforbid-fp-as-gp): New options.
17151 2018-06-01 Mark Wielaard <mark@klomp.org>
17153 * dwarf2out.c (dwarf2out_finish): Remove generation of
17154 DW_AT_loclists_base.
17156 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
17158 * gimple-ssa-store-merging.c: Include gimple-fold.h.
17159 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17160 (struct merged_store_group): Add bit_insertion field.
17161 (dump_char_array): Use standard hexadecimal format.
17162 (merged_store_group::merged_store_group): Set bit_insertion to false.
17163 (merged_store_group::apply_stores): Use optimal buffer size. Deal
17164 with BIT_INSERT_EXPR stores. Move up code updating the mask and
17165 also print the mask in the dump file.
17166 (pass_store_merging::gate): Minor tweak.
17167 (imm_store_chain_info::coalesce_immediate): Fix wrong association
17168 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
17169 stores with INTEGER_CST stores.
17170 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17171 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17172 and use it throughout. Generate bit insertion sequences if need be.
17173 (pass_store_merging::process_store): Remove redundant condition.
17174 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17176 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
17178 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17179 the 128-bit floating point types. Fix function comment.
17181 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17183 * config/aarch64/aarch64-simd.md
17184 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17186 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17189 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
17191 PR tree-optimization/85989
17192 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17194 (backprop::intersect_uses): Check it when deciding whether this
17195 is a backedge reference.
17196 (backprop::process_block): Add each phi to m_visited_phis
17197 after visiting it, then clear it at the end.
17199 2018-06-01 Richard Biener <rguenther@suse.de>
17201 * tree-vectorizer.h (vect_dr_stmt): New function.
17202 (vect_get_load_cost): Adjust.
17203 (vect_get_store_cost): Likewise.
17204 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17205 Use vect_dr_stmt instead of DR_SMTT.
17206 (vect_record_base_alignments): Likewise.
17207 (vect_calculate_target_alignment): Likewise.
17208 (vect_compute_data_ref_alignment): Likewise and make static.
17209 (vect_update_misalignment_for_peel): Likewise.
17210 (vect_verify_datarefs_alignment): Likewise.
17211 (vector_alignment_reachable_p): Likewise.
17212 (vect_get_data_access_cost): Likewise. Pass down
17213 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17214 (vect_get_peeling_costs_all_drs): Likewise.
17215 (vect_peeling_hash_get_lowest_cost): Likewise.
17216 (vect_enhance_data_refs_alignment): Likewise.
17217 (vect_find_same_alignment_drs): Likewise.
17218 (vect_analyze_data_refs_alignment): Likewise.
17219 (vect_analyze_group_access_1): Likewise.
17220 (vect_analyze_group_access): Likewise.
17221 (vect_analyze_data_ref_access): Likewise.
17222 (vect_analyze_data_ref_accesses): Likewise.
17223 (vect_vfa_segment_size): Likewise.
17224 (vect_small_gap_p): Likewise.
17225 (vectorizable_with_step_bound_p): Likewise.
17226 (vect_prune_runtime_alias_test_list): Likewise.
17227 (vect_analyze_data_refs): Likewise.
17228 (vect_supportable_dr_alignment): Likewise.
17229 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17230 (vect_gen_prolog_loop_niters): Likewise.
17231 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17232 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17234 (vect_recog_mask_conversion_pattern): Likewise.
17235 (vect_try_gather_scatter_pattern): Likewise.
17236 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17237 to vect_get_store_cost.
17238 (vect_get_store_cost): Get stmt_info instead of DR.
17239 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17240 (vect_get_load_cost): Get stmt_info instead of DR.
17242 2018-06-01 Richard Biener <rguenther@suse.de>
17244 PR middle-end/86017
17245 * gimple-fold.c (var_decl_component_p): Also allow offsetted
17246 vars wrapped in MEM_REFs.
17248 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
17250 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17251 Fix subreg tests so that we only return a choice between
17252 GENERAL_REGS and FP_REGS if the original classes included both.
17254 2018-06-01 Richard Biener <rguenther@suse.de>
17257 * tree-ssa-structalias.c (get_function_part_constraint):
17258 Handle NULL fi->decl.
17259 (find_func_aliases_for_call): Properly handle indirect
17260 fi from direct call.
17261 (find_func_clobbers): Likewise.
17262 (ipa_pta_execute): Likewise.
17263 (create_variable_info_for): For functions that are ifunc_resolver
17264 resolve to a varinfo that contains the result of the resolver call.
17265 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17268 2018-05-31 Michael Collison <michael.collison@arm.com>
17270 * config/aarch64/aarch64.md:
17271 (*fix_to_zero_extenddfdi2): New pattern.
17272 * gcc.target/aarch64/fix_extend1.c: New testcase.
17274 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
17276 PR middle-end/78809
17277 PR middle-end/83026
17278 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17279 and BUILT_IN_STRNCMP_EQ.
17280 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17281 BUILT_IN_STRNCMP_EQ.
17282 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17283 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17284 (gimple_fold_builtin): Likewise.
17285 * tree-ssa-strlen.c (compute_string_length): New function.
17286 (determine_min_obsize): New function.
17287 (handle_builtin_string_cmp): New function to handle calls to
17288 string compare functions.
17289 (strlen_optimize_stmt): Add handling to builtin string compare
17291 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17292 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17293 * tree.c (build_common_builtin_nodes): Add new defines of
17294 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17296 2018-05-31 Jakub Jelinek <jakub@redhat.com>
17299 * bb-reorder.c (pass_partition_blocks::gate): Return false for
17300 functions with naked attribute.
17302 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
17304 * config/i386/sse.md (avx_vec_concat<mode>):
17305 Substitute concat_tg_mode mode attribute with xtg_mode.
17306 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17307 (concat_tg_mode): Remove mode attribute.
17309 2018-05-31 Martin Sebor <msebor@redhat.com>
17312 * calls.c (alloc_max_size): Correct a logic error/typo.
17313 Treat excessive arguments as infinite. Warn for invalid arguments.
17314 * doc/invoke.texi (-Walloc-size-larger-than): Update.
17316 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
17319 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17320 and movx for Haswell.
17322 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
17323 Cesar Philippidis <cesar@codesourcery.com>
17325 PR middle-end/85879
17326 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17327 when emitting error on private/firstprivate reductions.
17328 * omp-low.c (lower_omp_target): Avoid reference-type processing
17329 on pointers for firstprivate clause.
17331 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
17333 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17336 * config/aarch64/aarch64-simd.md
17337 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17338 (aarch64_ld1_x3_<mode>): Likewise
17339 (aarch64_st1x2<VALLDIF:mode>): Likewise
17340 (aarch64_st1_x2_<mode>): Likewise
17341 (aarch64_st1x3<VALLDIF:mode>): Likewise
17342 (aarch64_st1_x3_<mode>): Likewise
17343 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17344 (vld1_s8_x3): Likewise.
17345 (vld1_u16_x3): Likewise.
17346 (vld1_s16_x3): Likewise.
17347 (vld1_u32_x3): Likewise.
17348 (vld1_s32_x3): Likewise.
17349 (vld1_u64_x3): Likewise.
17350 (vld1_s64_x3): Likewise.
17351 (vld1_f16_x3): Likewise.
17352 (vld1_f32_x3): Likewise.
17353 (vld1_f64_x3): Likewise.
17354 (vld1_p8_x3): Likewise.
17355 (vld1_p16_x3): Likewise.
17356 (vld1_p64_x3): Likewise.
17357 (vld1q_u8_x3): Likewise.
17358 (vld1q_s8_x3): Likewise.
17359 (vld1q_u16_x3): Likewise.
17360 (vld1q_s16_x3): Likewise.
17361 (vld1q_u32_x3): Likewise.
17362 (vld1q_s32_x3): Likewise.
17363 (vld1q_u64_x3): Likewise.
17364 (vld1q_s64_x3): Likewise.
17365 (vld1q_f16_x3): Likewise.
17366 (vld1q_f32_x3): Likewise.
17367 (vld1q_f64_x3): Likewise.
17368 (vld1q_p8_x3): Likewise.
17369 (vld1q_p16_x3): Likewise.
17370 (vld1q_p64_x3): Likewise.
17371 (vst1_s64_x2): Likewise.
17372 (vst1_u64_x2): Likewise.
17373 (vst1_f64_x2): Likewise.
17374 (vst1_s8_x2): Likewise.
17375 (vst1_p8_x2): Likewise.
17376 (vst1_s16_x2): Likewise.
17377 (vst1_p16_x2): Likewise.
17378 (vst1_s32_x2): Likewise.
17379 (vst1_u8_x2): Likewise.
17380 (vst1_u16_x2): Likewise.
17381 (vst1_u32_x2): Likewise.
17382 (vst1_f16_x2): Likewise.
17383 (vst1_f32_x2): Likewise.
17384 (vst1_p64_x2): Likewise.
17385 (vst1q_s8_x2): Likewise.
17386 (vst1q_p8_x2): Likewise.
17387 (vst1q_s16_x2): Likewise.
17388 (vst1q_p16_x2): Likewise.
17389 (vst1q_s32_x2): Likewise.
17390 (vst1q_s64_x2): Likewise.
17391 (vst1q_u8_x2): Likewise.
17392 (vst1q_u16_x2): Likewise.
17393 (vst1q_u32_x2): Likewise.
17394 (vst1q_u64_x2): Likewise.
17395 (vst1q_f16_x2): Likewise.
17396 (vst1q_f32_x2): Likewise.
17397 (vst1q_f64_x2): Likewise.
17398 (vst1q_p64_x2): Likewise.
17399 (vst1_s64_x3): Likewise.
17400 (vst1_u64_x3): Likewise.
17401 (vst1_f64_x3): Likewise.
17402 (vst1_s8_x3): Likewise.
17403 (vst1_p8_x3): Likewise.
17404 (vst1_s16_x3): Likewise.
17405 (vst1_p16_x3): Likewise.
17406 (vst1_s32_x3): Likewise.
17407 (vst1_u8_x3): Likewise.
17408 (vst1_u16_x3): Likewise.
17409 (vst1_u32_x3): Likewise.
17410 (vst1_f16_x3): Likewise.
17411 (vst1_f32_x3): Likewise.
17412 (vst1_p64_x3): Likewise.
17413 (vst1q_s8_x3): Likewise.
17414 (vst1q_p8_x3): Likewise.
17415 (vst1q_s16_x3): Likewise.
17416 (vst1q_p16_x3): Likewise.
17417 (vst1q_s32_x3): Likewise.
17418 (vst1q_s64_x3): Likewise.
17419 (vst1q_u8_x3): Likewise.
17420 (vst1q_u16_x3): Likewise.
17421 (vst1q_u32_x3): Likewise.
17422 (vst1q_u64_x3): Likewise.
17423 (vst1q_f16_x3): Likewise.
17424 (vst1q_f32_x3): Likewise.
17425 (vst1q_f64_x3): Likewise.
17426 (vst1q_p64_x3): Likewise.
17428 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17430 * config/msp430/msp430.c (msp430_output_labelref): Prepend
17431 user_label_prefix to name.
17433 * tree-core.h: Update comment about the format of NAME string
17434 passed to handler in attribute_spec.
17436 * config/msp430/msp430.md: Remove erroneous subreg expression from
17437 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17438 zero_extend{q,h}isi2.
17440 2018-05-30 Borislav Petkov <bp@suse.de>
17442 * doc/extend.texi: Document some architecture specific
17443 constraints and sort entries.
17445 2018-05-30 Martin Sebor <msebor@redhat.com>
17447 PR middle-end/85369
17448 * builtins.c (expand_builtin_stpcpy_1): New function.
17449 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17450 only if the former succeeds.
17452 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
17454 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17457 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17459 * doc/invoke.texi (-flinker-output): Document
17461 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17463 * passes.c (ipa_write_summaries): Only modify statements if body
17465 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17466 incrementally linking.
17467 (ipa_passes): Likewise.
17468 * lto-cgraph.c (lto_output_node): When incrementally linking do not
17469 pass down resolution info.
17470 * common.opt (flag_incremental_link): Update info.
17471 * gcc.c (plugin specs): Turn flinker-output=* to
17472 -plugin-opt=-linker-output-known
17473 * toplev.c (compile_file): Also cut compilation when doing incremental
17475 * flag-types. (enum lto_partition_model): Add
17476 LTO_LINKER_OUTPUT_NOLTOREL.
17477 (invoke.texi): Add -flinker-output docs.
17478 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17479 link same way as WPA; do not stream in dead initializers.
17481 * dwarf2out.c (dwarf2out_die_ref_for_decl,
17482 darf2out_register_external_decl): Support incremental link.
17484 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17486 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17488 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17490 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17491 it down to simple_object_copy_lto_debug_sections.
17492 (run_gcc): Determine incremental LTO link time and configure
17493 lto1 into non-wpa mode, disable renaming of debug sections.
17495 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
17497 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17498 descriptions of various incorrectly documented functions.
17500 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
17503 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17504 address check not strict.
17506 2018-05-30 Richard Biener <rguenther@suse.de>
17508 PR tree-optimization/85964
17509 * tracer.c (better_p): Drop initialized count check, we only
17510 call the function with initialized counts now.
17511 (find_best_successor): Do find a best edge if one
17512 has uninitialized count.
17513 (find_best_predecessor): Likewise. Do BB frequency check only
17514 if count is initialized.
17516 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
17518 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17519 (aarch64_ldrstr_offset_compare): New.
17520 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17521 load/store orderings.
17522 (aarch64_gen_adjusted_ldpstp): Likewise.
17524 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
17526 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17527 Check for subset of GENERAL_REGS and FP_REGS.
17528 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17531 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
17533 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17534 and wi::to_poly_offset. Add the current offset and then check
17535 whether the sum fits, rather than using an unchecked addition of
17536 a checked term. Check for a shwi rather than a uhwi.
17537 * expr.c (get_bit_range): Use tree_to_poly_uint64.
17538 (store_constructor): Use poly_int_tree_p.
17539 (expand_expr_real_1): Likewise.
17540 * function.c (assign_temp): Likewise.
17541 * fold-const.c (const_binop): Use poly_int_tree_p and
17542 wi::to_poly_offset.
17543 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
17545 * ipa-icf-gimple.c (func_checker::compare_operand): Use
17546 to_poly_offset for MEM offsets.
17547 * ipa-icf.c (sem_variable::equals): Likewise.
17548 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17549 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17550 wi::to_poly_offset for BIT_FIELD_REF offsets.
17551 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17552 wi::to_poly_offset.
17553 * var-tracking.c (emit_note_insn_var_location): Use
17554 tree_to_poly_uint64.
17556 2018-05-29 Jim Wilson <jimw@sifive.com>
17558 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17560 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
17563 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17564 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17566 (sse4_1_round<mode>2): Use nonimmediate_operand
17567 for operand 1 predicate.
17569 2018-05-29 Martin Sebor <msebor@redhat.com>
17570 Richard Biener <rguenther@suse.de>
17573 * calls.c (get_size_range): Call determine_value_range instead
17574 of get_value_range..
17575 * tree-vrp.h (determine_value_range): Declared new function.
17576 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17578 2018-05-29 Richard Biener <rguenther@suse.de>
17580 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17581 sure to use non-pattern stmts for get_earlier_stmt arguments.
17582 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17583 called on pattern stmts.
17584 (get_later_stmt): Likewise.
17586 2018-05-29 Martin Liska <mliska@suse.cz>
17588 PR gcov-profile/85759
17589 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
17592 2018-05-29 Jakub Jelinek <jakub@redhat.com>
17594 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
17596 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
17600 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17601 VEC_PACK_FLOAT_EXPR): New tree codes.
17602 * tree-pretty-print.c (op_code_prio): Handle
17603 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17604 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17605 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17606 * tree-inline.c (estimate_operator_cost): Likewise.
17607 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
17608 * fold-const.c (const_binop): Likewise.
17609 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
17610 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
17611 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
17612 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
17613 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
17614 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
17615 * expr.c (expand_expr_real_2): Likewise.
17616 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
17617 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
17618 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
17620 * optabs.c (expand_widen_pattern_expr): For
17621 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
17622 sign from result type rather than operand's type.
17623 (expand_binop_directly): For vec_packu_float_optab and
17624 vec_packs_float_optab allow result type to be different from operand's
17626 * optabs-tree.c (optab_for_tree_code): Handle
17627 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17628 VEC_PACK_FLOAT_EXPR. Formatting fixes.
17629 * tree-vect-generic.c (expand_vector_operations_1): Handle
17630 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
17631 VEC_PACK_FLOAT_EXPR.
17632 * tree-vect-stmts.c (supportable_widening_operation): Handle
17634 (supportable_narrowing_operation): Handle FLOAT_EXPR.
17635 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
17636 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
17637 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
17638 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
17640 (vec_pack<floatprefix>_float_<mode>): New expander.
17641 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
17643 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
17644 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
17645 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
17646 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
17647 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
17649 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
17650 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
17651 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
17652 VEC_PACK_FLOAT_EXPR): Document.
17654 2018-05-29 Richard Biener <rguenther@suse.de>
17656 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
17658 (stmt_vec_info_vec): Make pointer.
17659 (init_stmt_vec_info_vec): Remove.
17660 (free_stmt_vec_info_vec): Likewise.
17661 (set_stmt_vec_info_vec): New function.
17662 (free_stmt_vec_infos): Likewise.
17663 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
17664 (set_vinfo_for_stmt): Likewise.
17665 (get_earlier_stmt): Likewise.
17666 (get_later_stmt): Likewise.
17667 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
17668 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
17669 (vec_info::~vec_info): Free stmt_vec_infos.
17670 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
17671 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
17672 (pass_slp_vectorize::execute): Likewise.
17673 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
17674 (free_stmt_vec_info_vec): Likewise.
17675 (set_stmt_vec_info_vec): New function.
17676 (free_stmt_vec_infos): Likewise.
17677 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
17678 the global stmt_vec_info_vec.
17679 * tree-parloops.c (gather_scalar_reductions): Use
17680 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
17683 2018-05-29 Richard Biener <rguenther@suse.de>
17685 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
17687 2018-05-29 Martin Liska <mliska@suse.cz>
17688 David Malcolm <dmalcolm@redhat.com>
17690 * vec.c (test_reverse): New.
17691 (vec_c_tests): Add new test.
17692 * vec.h (vl_ptr>::reverse): New function.
17694 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
17696 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
17698 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
17701 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
17703 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
17705 2018-05-28 Richard Biener <rguenther@suse.de>
17707 PR tree-optimization/85933
17708 * tree-vect-data-refs.c (vect_record_base_alignments): Only
17709 look at stmts marked as vectorizable.
17711 2018-05-28 Richard Biener <rguenther@suse.de>
17713 PR tree-optimization/85934
17714 * tree-vect-generic.c (expand_vector_operations_1): Hoist
17715 vector boolean check before scalar optimization.
17717 2018-05-28 Jakub Jelinek <jakub@redhat.com>
17719 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
17722 2018-05-28 Mark Wielaard <mark@klomp.org>
17724 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
17725 if it is an expression containing a minus sign.
17727 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
17729 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
17731 2018-05-27 Paul Koning <ni1d@arrl.net>
17733 * config/pdp11/pdp11.md (truncsihi2): Remove.
17735 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
17736 Chung-Ju Wu <jasonwucj@gmail.com>
17738 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
17740 (unaligned_store_dw): Ditto.
17741 * config/nds32/nds32-memory-manipulation.c
17742 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
17743 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
17744 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
17745 (emit_setmem_word_loop): Rename to ...
17746 (emit_setmem_doubleword_loop): ... this.
17747 (nds32_gen_dup_4_byte_to_word_value): New function.
17748 (nds32_gen_dup_8_byte_to_double_word_value): New function.
17749 (nds32_expand_setmem_loop): Refine implementation.
17750 (nds32_expand_setmem_loop_v3m): Ditto.
17751 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
17754 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
17756 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
17758 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
17760 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
17761 (nds32_init_machine_status): Initialize machine->attr_naked_p and
17762 machine->attr_no_prologue_p.
17763 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
17764 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
17765 (nds32_expand_epilogue): Consider attr_naked_p.
17766 (nds32_expand_epilogue_v3pop): Likewise.
17767 (nds32_can_use_return_insn): Likewise.
17768 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
17769 attr_no_prologue_p fields.
17770 * config/nds32/nds32.opt (mret-in-naked-func): New option.
17772 2018-05-27 Jakub Jelinek <jakub@redhat.com>
17775 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
17777 * config/i386/sse.md
17778 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
17780 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
17782 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
17784 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
17786 (*<floatsuffix>floatv2div2sf2): Rename to ...
17787 (*float<floatunssuffix>v2div2sf2): ... this.
17788 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
17789 (float<floatunssuffix>v2div2sf2_mask): ... this.
17790 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
17791 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
17792 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
17794 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
17796 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17798 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
17800 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17802 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
17804 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
17805 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
17806 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
17807 gen_ufix_truncv8dfv8si2.
17808 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
17809 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
17810 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
17811 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
17812 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
17813 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
17814 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
17815 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
17817 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
17821 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
17823 2018-05-25 Jim Wilson <jimw@sifive.com>
17825 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
17826 * config/riscv/riscv.c (struct machine_function): Add
17827 interrupt_handler_p and attribute_checked_p fields.
17828 (riscv_attribute_table): Add interrupt.
17829 (riscv_interrupt_type_p): New.
17830 (riscv_save_reg_p): Save extra regs for interrupt handler.
17831 (riscv_use_save_libcall): Return false for interrupt handler.
17832 (riscv_first_stack_step): Add forward declaration.
17833 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
17834 for interrupt handler with large frame. Use it for saved reg list.
17835 (riscv_expand_prologue): Move flag_stack_usage_info support to
17836 eliminate duplication.
17837 (riscv_expand_epilogue): Generate mret for interrupt handler.
17838 (riscv_epilogue_uses): New.
17839 (riscv_can_use_return_insn): Return false for interrupt handler.
17840 (riscv_function_ok_for_sibcall): Likewise.
17841 (riscv_set_current_function): Add interrupt handler support.
17842 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
17843 * config/riscv/riscv.md (UNSPECV_MRET): New.
17845 (riscv_frflags, riscv_fsflags): Use tab after opcode.
17847 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
17849 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
17851 PR tree-optimization/85712
17852 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
17853 this candidate has already been replaced in-situ by a copy.
17855 2018-05-25 Jason Merrill <jason@redhat.com>
17857 PR c++/80485 - inline function non-zero address.
17858 * symtab.c (nonzero_address): Check DECL_COMDAT.
17860 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
17863 * config/alpha/alpha.md (ashlsi3): New insn pattern.
17864 (*ashlsi_se): Rename from *ashldi_se. Define as sign
17865 extension of SImode operation. Use const123_operand predicate.
17866 (*saddsi_1): Remove.
17867 (*saddl_se_1): Ditto.
17868 (*ssubsi_1): Ditto.
17869 (*ssubl_se_1): Ditto.
17870 * config/alpha/predicates.md (const123_operand): New predicate.
17871 * config/alpha/constraints.md (P): Use IN_RANGE.
17873 2018-05-25 Richard Biener <rguenther@suse.de>
17875 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
17877 (ref_maybe_used_by_stmt_p): Likewise.
17878 (stmt_may_clobber_ref_p): Likewise.
17879 (stmt_may_clobber_ref_p_1): Likewise.
17880 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
17882 (ref_maybe_used_by_stmt_p): Likewise.
17883 (stmt_may_clobber_ref_p): Likewise.
17884 (stmt_may_clobber_ref_p_1): Likewise.
17885 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
17886 the alias oracle to disambiguate DRs with stmts DR analysis
17888 (vect_analyze_data_refs): Do not give up on not analyzable
17889 DRs for BB vectorization. Remove code truncating the dataref
17892 2018-05-25 Jakub Jelinek <jakub@redhat.com>
17895 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17896 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
17897 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
17899 2018-05-25 Richard Biener <rguenther@suse.de>
17901 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
17902 function, combining stmt data ref gathering and fatal analysis
17904 (vect_analyze_data_refs): Remove now redudnant code and simplify.
17905 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
17906 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
17907 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
17908 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
17910 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17912 PR tree-optimization/85720
17913 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
17914 SCC if all partitions are builtins.
17915 (version_loop_by_alias_check): New parameter. Generate cancelable
17916 runtime alias check if all partitions are builtins.
17917 (distribute_loop): Update call to above function.
17919 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17921 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
17922 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
17923 (parm_default_def_partition_arg): Ditto.
17924 (set_parm_default_def_partition): Ditto.
17925 (get_parm_default_def_partitions): Ditto and make it static.
17926 (get_undefined_value_partitions): Ditto and make it static.
17927 (remove_ssa_form): Refactor call to init_var_map here.
17928 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
17929 computation for loop region.
17930 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
17931 (register_default_def): Delete.
17932 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
17933 (parm_default_def_partition_arg): Ditto.
17934 (set_parm_default_def_partition): Ditto.
17935 (get_parm_default_def_partitions): Ditto and make it static.
17936 (get_undefined_value_partitions): Ditto and make it static.
17937 (coalesce_with_default, coalesce_with_default): Update comment.
17938 (create_coalesce_list_for_region): New func factored out from
17939 create_outofssa_var_map.
17940 (populate_coalesce_list_for_outofssa): New func factored out from
17941 create_outofssa_var_map and coalesce_ssa_name.
17942 (create_outofssa_var_map): Delete.
17943 (coalesce_ssa_name): Refactor to support live range computation.
17944 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
17945 (get_parm_default_def_partitions): Delete.
17946 (get_undefined_value_partitions): Ditto.
17947 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
17948 computation for loop region.
17949 (new_tree_live_info, loe_visit_block): Ditto.
17950 (live_worklist, set_var_live_on_entry): Ditto.
17951 (calculate_live_on_exit, verify_live_on_entry): Ditto.
17952 * tree-ssa-live.h (struct _var_map): New fields.
17953 (init_var_map): Change decl.
17954 (region_contains_p): New.
17956 2018-05-25 Bin Cheng <bin.cheng@arm.com>
17958 * tree-ssa-live.h (live_merge_and_clear): Delete.
17960 2018-05-25 Richard Biener <rguenther@suse.de>
17963 * tree-dump.c (dequeue_and_dump): Remove access to removed
17964 operand 2 of a SWITCH_EXPR.
17966 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17968 * doc/sourcebuild.texi (vect_double_cond_arith): Include
17969 multiplication and division.
17970 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
17971 (cond_udiv@var{m}, cond_umod@var{m}): Document.
17972 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
17973 (cond_udiv_optab, cond_umod_optab): New optabs.
17974 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
17975 (IFN_COND_RDIV): New internal functions.
17976 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
17977 TRUNC_MOD_EXPR and RDIV_EXPR.
17978 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
17979 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
17981 (SVE_INT_BINARY): Include mult.
17982 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
17983 (optab, sve_int_op): Handle mult.
17984 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
17986 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
17987 for SVE_INT_BINARY_SD.
17989 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17991 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
17992 (optab, sve_int_op): Handle div and udiv.
17993 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
17994 for SVE_INT_BINARY_SD.
17995 (*<optab><mode>3): New insn for the same.
17997 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
17999 * tree-vect-patterns.c: Include predict.h.
18000 (vect_recog_divmod_pattern): Restrict check for division support
18001 to when optimizing for size.
18003 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18005 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18006 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18007 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18008 (gimple_match_op::set_op): Likewise.
18009 (gimple_resimplify4): Declare.
18010 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18011 (expr::gen_transform): Likewise.
18012 (decision_tree::gen): Generate a simplification routine for 4 operands.
18013 * gimple-match-head.c (gimple_simplify): Add an overload for
18014 4 operands. In the top-level function, handle up to 4 call
18015 arguments and call gimple_resimplify4.
18016 (gimple_resimplify4): New function.
18017 (build_call_internal): Pass a fourth operand.
18018 (maybe_push_to_seq): Likewise.
18019 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18020 Fold VEC_COND_EXPRs of an operation and a default value into
18021 an IFN_COND_* function if possible.
18022 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18024 (SVE_COND_FP_BINARY): Include them.
18025 (optab, sve_fp_op): Handle them.
18026 (SVE_INT_BINARY_REV): New code iterator.
18027 (SVE_COND_FP_BINARY_REV): New int iterator.
18028 (commutative): New int attribute.
18029 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18031 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18033 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18034 (*cond_<optab><mode>): New patterns for reversed operands.
18036 2018-05-25 Richard Biener <rguenther@suse.de>
18038 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18039 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18040 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18041 (STMT_VINFO_GROUPED_ACCESS): Adjust.
18042 * tree-vect-data-refs.c (everywhere): Adjust users.
18043 * tree-vect-loop.c (everywhere): Likewise.
18044 * tree-vect-slp.c (everywhere): Likewise.
18045 * tree-vect-stmts.c (everywhere): Likewise.
18046 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18048 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18050 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18052 (gcc_cv_as_section_exclude): ... this.
18053 Try Solaris as #exclude syntax.
18054 * configure: Regenerate.
18055 * config.in: Regenerate.
18056 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18058 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18059 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18061 * varasm.c (default_elf_asm_named_section): Don't check if
18062 HAVE_GAS_SECTION_EXCLUDE is defined.
18064 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18066 * doc/md.texi: Update the documentation of the cond_* optabs
18067 to mention the new final operand. Fix GET_MODE_NUNITS call.
18068 Describe the scalar case too.
18069 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18070 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18072 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18073 (get_conditional_internal_fn): Update comment.
18074 * tree-vect-loop.c (vectorizable_reduction): Pass the original
18075 accumulator value as a final argument to conditional functions.
18076 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18077 a define_expand and add an "else" operand. Assert for now that
18078 the else operand is equal to operand 2. Use SVE_INT_BINARY and
18079 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18080 (*cond_<optab><mode>): New patterns.
18081 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18082 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18083 (UNSPEC_COND_EOR): Delete.
18084 (optab): Remove associated mappings.
18085 (SVE_INT_BINARY): New code iterator.
18086 (sve_int_op): Remove int attribute and add "minus" to the code
18088 (SVE_COND_INT_OP): Delete.
18089 (SVE_COND_FP_OP): Rename to...
18090 (SVE_COND_FP_BINARY): ...this.
18092 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18094 * optabs.c (can_reuse_operands_p): New function.
18095 (maybe_legitimize_operands): Try to reuse the results for
18098 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
18100 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18101 Add {q} suffix to insn mnemonic.
18103 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
18105 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18106 (msp430_warn_func_return): New.
18108 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
18110 * fold-const.c (tree_nonzero_bits): New function.
18111 * fold-const.h (tree_nonzero_bits): Likewise.
18112 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18113 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18115 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
18119 * varasm.c (assemble_alias): Check ifunc_resolver only on
18122 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
18125 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18126 when memory input operand is handled.
18128 2018-05-24 Luis Machado <luis.machado@linaro.org>
18130 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18132 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18134 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
18136 * match.pd: Delay FMA folds until after vectorization.
18138 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
18141 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18142 address check not strict.
18144 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
18146 * gimple-match.h (gimple_match_op): New class.
18147 (mprts_hook): Replace parameters with a gimple_match_op *.
18148 (maybe_build_generic_op): Likewise.
18149 (gimple_simplified_result_is_gimple_val): Replace parameters with
18150 a const gimple_match_op *.
18151 (gimple_simplify): Replace code_helper * and tree * parameters with
18152 a gimple_match_op * parameter.
18153 (gimple_resimplify1): Replace code_helper *, tree and tree *
18154 parameters with a gimple_match_op * parameter.
18155 (gimple_resimplify2): Likewise.
18156 (gimple_resimplify3): Likewise.
18157 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18158 parameters with a gimple_match_op * parameter.
18159 * gimple-match-head.c (gimple_simplify): Change prototypes of
18160 auto-generated functions to take a gimple_match_op * instead of
18161 separate code_helper * and tree * parameters. Make the same
18162 change in the top-level overload and update calls to the
18163 gimple_resimplify routines. Update calls to the auto-generated
18164 functions and to maybe_push_res_to_seq in the publicly-facing
18165 operation-specific gimple_simplify overloads.
18166 (gimple_match_op::MAX_NUM_OPS): Define.
18167 (gimple_resimplify1): Replace rcode and ops with a single res_op
18168 parameter. Update call to gimple_simplify.
18169 (gimple_resimplify2): Likewise.
18170 (gimple_resimplify3): Likewise.
18171 (mprts_hook): Replace parameters with a gimple_match_op *.
18172 (maybe_build_generic_op): Likewise.
18173 (build_call_internal): Replace type, nargs and ops with
18174 a gimple_match_op *.
18175 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18176 with a single gimple_match_op *. Update calls to mprts_hook,
18177 build_call_internal and gimple_simplified_result_is_gimple_val.
18178 Factor out code that is common to the tree_code and combined_fn cases.
18179 * genmatch.c (expr::gen_transform): Replace tem_code and
18180 tem_ops with a gimple_match_op called tem_op. Update calls
18181 to the gimple_resimplify functions and maybe_push_res_to_seq.
18182 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18183 res_ops. Update call to the gimple_resimplify functions.
18184 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18185 (decision_tree::gen): Make the functions take a gimple_match_op *
18186 called res_op instead of separate res_code and res_ops parameters.
18187 Update call accordingly.
18188 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18189 and ops with a single res_op parameter. Update calls to
18190 maybe_build_generic_op and maybe_push_res_to_seq.
18191 (fold_stmt_1): Update calls to gimple_simplify and
18192 replace_stmt_with_simplification.
18193 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18194 and gimple_simplified_result_is_gimple_val.
18195 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18197 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18198 with a gimple_match_op *.
18199 (vn_nary_build_or_lookup): Likewise. Update call to
18200 vn_nary_build_or_lookup_1.
18201 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18202 gimple_match_op *. Update calls to the gimple_resimplify routines
18203 and to gimple_simplified_result_is_gimple_val.
18204 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18205 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18206 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18207 (visit_nary_op): Likewise.
18208 (visit_reference_op_load): Likewise.
18210 2018-05-23 Luis Machado <luis.machado@linaro.org>
18212 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18213 modifier for printing the step amount.
18215 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
18218 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18221 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
18223 * doc/sourcebuild.texi (Endianness): New subsubsection.
18225 2018-05-23 Luis Machado <luis.machado@linaro.org>
18227 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18228 <prefetch_dynamic_strides>: New const bool field.
18229 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18230 prefetch_dynamic_strides.
18231 (exynosm1_prefetch_tune): Likewise.
18232 (thunderxt88_prefetch_tune): Likewise.
18233 (thunderx_prefetch_tune): Likewise.
18234 (thunderx2t99_prefetch_tune): Likewise.
18235 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18237 (aarch64_override_options_internal): Update to set
18238 PARAM_PREFETCH_DYNAMIC_STRIDES.
18239 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18240 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18241 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18242 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18243 prefetch-dynamic-strides setting.
18245 2018-05-23 Luis Machado <luis.machado@linaro.org>
18247 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18248 <minimum_stride>: New const int field.
18249 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18250 minimum_stride field defaulting to -1.
18251 (exynosm1_prefetch_tune): Likewise.
18252 (thunderxt88_prefetch_tune): Likewise.
18253 (thunderx_prefetch_tune): Likewise.
18254 (thunderx2t99_prefetch_tune): Likewise.
18255 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18256 <default_opt_level>: Set to 3.
18257 (aarch64_override_options_internal): Update to set
18258 PARAM_PREFETCH_MINIMUM_STRIDE.
18259 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18260 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18261 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18262 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18263 stride is constant and is below the minimum stride threshold.
18265 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18267 * config/arm/arm-cpus.in (mode26): Delete.
18268 (armv4): Delete mode26 reference.
18269 * config/arm/arm.c (arm_configure_build_target): Delete use of
18272 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
18274 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18276 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18277 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
18279 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
18280 (floatunsdidf2): Ditto.
18282 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
18284 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18285 (fixuns_trunc<mode>si2_avx512f): Ditto.
18286 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18287 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18288 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18290 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
18292 PR rtl-optimization/79985
18293 * df-scan.c (df_insn_refs_collect): Remove special case for
18294 global registers and asm statements.
18296 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
18298 * extend.texi (Global Register Variables): Rewrite the bullet list.
18299 Note that the register is available for allocation. Note that access
18300 via inline asm must use constraints. Add note about async-signal
18301 handlers. Remove paragraph about automagic register selection.
18303 2018-05-23 Richard Biener <rguenther@suse.de>
18305 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18306 of fixed offset from memset VN.
18308 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
18310 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18311 first_interp field.
18312 (alloc_cand_and_find_basis): Initialize first_interp field.
18313 (slsr_process_mul): Modify first_interp field.
18314 (slsr_process_add): Likewise.
18315 (slsr_process_cast): Modify first_interp field for each new
18317 (slsr_process_copy): Likewise.
18318 (dump_candidate): Dump first_interp field.
18319 (replace_mult_candidate): Process all interpretations, not just
18321 (replace_rhs_if_not_dup): Likewise.
18322 (replace_one_candidate): Likewise.
18324 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
18326 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18328 (aarch64_needs_frame_chain): New function.
18329 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18331 2018-05-23 Sudakshina Das <sudi.das@arm.com>
18334 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18335 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18336 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18337 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18338 as true for strict-align.
18339 (aarch64_can_inline_p): Perform checks even when callee has no
18340 attributes to check for strict alignment.
18341 * doc/extend.texi (AArch64 Function Attributes): Document
18343 * doc/invoke.texi: (AArch64 Options): Likewise.
18345 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
18347 PR tree-optimization/85853
18348 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18349 the handling of the root of the node to...
18350 (vect_slp_analyze_node_operations_1): ...this new function,
18351 and run the whole thing with the child nodes' def types
18352 set according to their SLP node's def type.
18354 2018-05-23 Richard Biener <rguenther@suse.de>
18356 PR middle-end/85874
18357 * tree-data-ref.c (create_runtime_alias_checks): Defer
18358 and ignore overflow warnings.
18360 2018-05-23 Yury Gribov <tetra2005@gmail.com>
18362 PR tree-optimization/85822
18363 * tree-vrp.c (is_masked_range_test): Fix handling of negative
18366 2018-05-23 Richard Biener <rguenther@suse.de>
18368 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18369 memset constants via native_interpret_expr.
18371 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
18374 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18376 (cgraph_node::create_alias): Likewise.
18377 (cgraph_node::get_availability): Check ifunc_resolver instead
18378 of looking up ifunc attribute.
18379 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18380 * varasm.c (do_assemble_alias): Likewise.
18381 (assemble_alias): Likewise.
18382 (default_binds_local_p_3): Likewise.
18383 * cgraph.h (cgraph_node): Add ifunc_resolver.
18384 (cgraph_node::only_called_directly_or_aliased_p): Return false
18385 for IFUNC resolver.
18386 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18388 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18389 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18390 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18391 instead of looking up ifunc attribute.
18393 2018-05-22 Luis Machado <luis.machado@linaro.org>
18395 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18397 2018-05-22 Martin Sebor <msebor@redhat.com>
18399 PR middle-end/85359
18400 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18401 only when expasion succeeds.
18402 (expand_builtin_strcmp): Same.
18403 (expand_builtin_strncmp): Same.
18405 2018-05-22 Martin Sebor <msebor@redhat.com>
18407 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18409 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
18410 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18412 * config/aarch64/aarch64-ldpstp.md: Replace uses of
18413 aarch64_mem_pair_operand with memory_operand and delete operand swapping
18415 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18416 Add check for legitimate_address.
18417 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18418 (aarch64_swap_ldrstr_operands): New.
18419 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18422 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
18423 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18425 * config/aarch64/aarch64.md: New patterns to generate stp
18427 (store_pair_sw, store_pair_dw): New patterns to generate stp for
18428 single words and double words.
18429 (load_pair_sw, load_pair_dw): Likewise.
18430 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18432 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18434 * config/aarch64/aarch64-ldpstp.md: Modify peephole
18435 for different mode ldpstp and add peephole for merged zero stores.
18436 Likewise for loads.
18437 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18439 (aarch64_gen_store_pair): Rename calls to match new patterns.
18440 (aarch64_gen_load_pair): Rename calls to match new patterns.
18441 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18442 (load_pair<DREG:mode><DREG2:mode>): ... This.
18443 (store_pair<mode>): Rename to...
18444 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18445 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18446 New mode iterators.
18447 (V_INT_EQUIV): Handle SImode.
18448 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18451 2018-05-22 Martin Sebor <msebor@redhat.com>
18454 * calls.c (maybe_warn_nonstring_arg): Use string length to set
18455 or ajust the presumed bound on an operation to avoid unnecessary
18458 2018-05-22 Martin Sebor <msebor@redhat.com>
18460 PR tree-optimization/85826
18461 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18462 assuming that a DECL necesarily has a constant size.
18464 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
18466 PR middle-end/85862
18467 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18469 2018-05-22 Richard Biener <rguenther@suse.de>
18471 PR tree-optimization/85834
18472 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18473 non-constant and non-zero memset arguments.
18475 2018-05-22 Martin Liska <mliska@suse.cz>
18478 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18480 2018-05-22 Richard Biener <rguenther@suse.de>
18482 PR tree-optimization/85863
18483 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18484 comparisons when vectype is specified.
18485 (vectorizable_condition): Do not specify vectype for
18486 vect_is_simple_cond when SLP vectorizing.
18488 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
18491 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18492 define __ibm128 as long double.
18493 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18494 as a distinct type when IEEE 128-bit support is enabled.
18495 (init_float128_ieee): Fix up conversions between IFmode and IEEE
18496 128-bit types to use the correct functions.
18497 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18498 convert between 128-bit floating point types that have different
18499 modes but the same representation, instead of using gen_lowpart to
18501 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18503 (IFKF_reg): New attributes to give the register constraints for
18505 (extend<mode>tf2_internal): New insns to mark an explicit
18506 conversion between 128-bit floating point types that have a
18507 different mode but share the same representation.
18509 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
18511 PR tree-optimization/85814
18512 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18513 a null return from get_strinfo when unsharing the next
18514 strinfo in the chain.
18516 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18519 * varasm.c (weak_finish): Clean up weak_decls.
18521 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18523 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18524 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18525 UNSPEC_UADALP values.
18526 * config/aarch64/iterators.md (ABAL): New int iterator.
18529 (sur): Add mappings for the above.
18530 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18532 (aarch64_<sur>abal<mode>_4): Likewise.
18533 (aarch64_<sur>adalp<mode>_3): Likewise.
18534 (<sur>sadv16qi): New define_expand.
18536 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
18538 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18539 (*movdf_internal): Ditto.
18540 (*rcpsf2_sse): Ditto.
18541 (*rsqrtsf2_sse): Ditto.
18542 (*sqrt<mode>2_sse): Ditto.
18544 2018-05-21 Tamar Christina <tamar.christina@arm.com>
18546 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18548 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18549 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18550 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18551 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18553 * config/aarch64/arm_neon.h: Likewise.
18554 * config/aarch64/iterators.md (VQ_I): New.
18556 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
18558 * config.gcc: Add arc/t-multilib-linux to tmake_file for
18560 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18563 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
18565 * config/nds32/constraints.md (S): New constraint.
18566 * config/nds32/nds32.md (call_internal): Use constraint S.
18567 (call_value_internal): Likewise.
18568 (sibcall_internal): Likewise.
18569 (sibcall_value_internal): Likewise.
18571 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
18572 Chung-Ju Wu <jasonwucj@gmail.com>
18574 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18575 into consideration.
18577 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
18578 Chung-Ju Wu <jasonwucj@gmail.com>
18580 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18581 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18582 (nds32_rtx_costs_impl): Simplify.
18583 (nds32_address_cost_impl): Simplify.
18584 (nds32_init_rtx_costs): New function.
18585 (nds32_rtx_costs_speed_prefer): Likewise.
18586 (nds32_rtx_costs_size_prefer): Likewise.
18587 (nds32_address_cost_speed_prefer): Likewise.
18588 (nds32_address_cost_speed_fwprop): Likewise.
18589 (nds32_address_cost_size_prefer): Likewise.
18590 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
18591 * config/nds32/nds32.c (nds32_option_override): Use
18592 nds32_init_rtx_costs function.
18594 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
18596 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
18597 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
18598 (TARGET_PIPELINE_N8): Likewise.
18599 (TARGET_PIPELINE_N10): Likewise.
18600 (TARGET_PIPELINE_N13): Likewise.
18601 (TARGET_PIPELINE_GRAYWOLF): Likewise.
18603 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
18605 * config/nds32/nds32-fpu.md: Update copyright year.
18607 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18609 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
18611 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18613 * config/nds32/nds32.c
18614 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
18615 * config/nds32/nds32.opt (minline-asm-r15): New option.
18617 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18619 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
18621 * config/nds32/nds32.md (abssi2): New pattern.
18623 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
18625 * config/i386/i386.md (rex64namesuffix): New mode attribute.
18626 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
18627 Merge insn pattern from sse_cvtsi2ss<round_name> and
18628 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
18629 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
18630 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
18631 using SWI48 mode iterator.
18632 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
18633 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
18634 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
18635 pattern from sse_cvttss2si<round_saeonly_name>
18636 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
18637 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
18638 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
18639 using SWI48 mode iterator.
18640 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
18641 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
18642 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
18643 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
18644 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
18645 using SWI48 mode iterator.
18646 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
18647 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
18648 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
18649 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
18650 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
18651 SWI48 mode iterator.
18652 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
18653 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
18654 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
18655 pattern from sse_cvttsd2si<round_saeonly_name>
18656 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
18658 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18660 * config/nds32/nds32-md-auxiliary.c
18661 (nds32_valid_smw_lwm_base_p): Refine.
18662 (nds32_output_smw_single_word): Refine.
18663 (nds32_output_smw_double_word): New.
18664 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
18666 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18668 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
18669 (nds32_output_stack_pop): Refine.
18670 (nds32_expand_unaligned_load): Refine.
18671 (nds32_expand_unaligned_store): Refine.
18673 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18674 Chung-Ju Wu <jasonwucj@gmail.com>
18676 * config/nds32/constants.md: Add TP_REGNUM constant.
18677 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
18678 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
18680 * config/nds32/nds32-doubleword.md: Consider flag_pic.
18681 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
18682 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
18683 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
18684 and PIC code generation.
18685 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
18687 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
18689 * config/nds32/nds32.md: Support TLS and PIC.
18690 * config/nds32/nds32.c: Support TLS and PIC.
18691 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
18692 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
18695 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
18697 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
18698 mode with E_ prefix.
18700 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18701 Chung-Ju Wu <jasonwucj@gmail.com>
18703 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
18704 * config/nds32/nds32-md-auxiliary.c
18705 (symbolic_reference_mentioned_p): New.
18706 (nds32_legitimize_ict_address): New.
18707 (nds32_expand_ict_move): New.
18708 (nds32_indirect_call_referenced_p): New.
18709 (nds32_symbol_binds_local_p): Delete.
18710 (nds32_long_call_p): Modify.
18711 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
18712 * config/nds32/nds32-protos.h
18713 (symbolic_reference_mentioned_p): Declare.
18714 (nds32_legitimize_ict_address): Declare.
18715 (nds32_expand_ict_move): Declare.
18716 (nds32_indirect_call_referenced_p): Declare.
18717 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
18718 (nds32_relax_group): Use nds32_ict_const_p as condition.
18719 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
18720 (nds32_asm_file_start): Output ict_model directive in asm code.
18721 (nds32_legitimate_address_p): Consider indirect call.
18722 (nds32_print_operand): Consider indirect call.
18723 (nds32_print_operand_address): Consider indirect call.
18724 (nds32_insert_attributes): Handle "indirect_call" attribute.
18725 (TARGET_LEGITIMATE_ADDRESS_P): Define.
18726 (TARGET_LEGITIMATE_CONSTANT_P): Define.
18727 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
18728 (TARGET_DELEGITIMIZE_ADDRESS): Define.
18729 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
18730 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
18731 (TARGET_ICT_MODEL_SMALL): Define.
18732 (TARGET_ICT_MODEL_LARGE): Define.
18733 * config/nds32/nds32.md (movsi): Consider ict model.
18734 (call, call_value): Consider ict model.
18735 (sibcall, sibcall_value): Consider ict model.
18736 * config/nds32/nds32.opt (mict-model): New option.
18737 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
18740 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
18741 Monk Chiang <sh.chiang04@gmail.com>
18742 Jim Wilson <jimw@sifive.com>
18744 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
18745 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
18746 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
18747 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
18748 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
18749 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
18750 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
18751 compute save_libcall_adjustment properly.
18752 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
18753 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
18754 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
18755 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
18756 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
18757 (ABI_SPEC): Handle mabi=ilp32e.
18758 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
18759 (RVE): Add RVE mask.
18760 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
18761 <-march>: Add rv32e as an example.
18763 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
18766 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
18767 (intra_create_variable_infos): Handle C++ constructors.
18769 2018-05-18 Martin Liska <mliska@suse.cz>
18771 * passes.def: Remove a redundant pass.
18773 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
18776 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
18778 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18780 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
18782 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
18784 (armv2, armv2a, armv3, armv3m): Delete architectures.
18785 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
18786 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
18787 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
18789 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
18790 (*mulsidi3adddi): Likewise.
18791 (mulsidi3): Likewise.
18792 (*mulsidi3_nov6): Likewise.
18793 (umulsidi3): Likewise.
18794 (umulsidi3_nov6): Likewise.
18795 (umaddsidi4): Likewise.
18796 (*umulsidi3adddi): Likewise.
18797 (smulsi3_highpart): Likewise.
18798 (*smulsi3_highpart_nov6): Likewise.
18799 (umulsi3_highpart): Likewise.
18800 (*umulsi3_highpart_nov6): Likewise.
18801 * config/arm/arm.h (arm_arch3m): Delete.
18802 * config/arm/arm.c (arm_arch3m): Delete.
18803 (arm_option_override_internal): Update armv3-related comment.
18804 (arm_configure_build_target): Delete use of isa_bit_mode32.
18805 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
18806 (arm_rtx_costs_internal): Delete check of arm_arch3m.
18807 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
18808 (mulsa3): Likewise.
18809 (mulusa3): Likewise.
18810 * config/arm/arm-protos.h (arm_arch3m): Delete.
18811 * config/arm/arm-tables.opt: Regenerate.
18812 * config/arm/arm-tune.md: Likewise.
18813 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
18814 deleted architectures.
18816 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18818 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
18819 (armv5t, armv5te): New features.
18820 (ARMv5, ARMv5e): Delete fgroups.
18821 (ARMv5t, ARMv5te): Adjust for above changes.
18822 (ARMv6m): Likewise.
18823 (armv5, armv5e): Delete arches.
18824 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
18826 (*call_reg_arm): Likewise.
18827 (*call_value_reg_armv5): Likewise.
18828 (*call_value_reg_arm): Likewise.
18829 (*call_symbol): Likewise.
18830 (*call_value_symbol): Likewise.
18831 (*sibcall_insn): Likewise.
18832 (*sibcall_value_insn): Likewise.
18833 (clzsi2): Likewise.
18834 (prefetch): Likewise.
18835 (define_split and define_peephole2 dependent on arm_arch5):
18837 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
18839 (TARGET_ARM_QBIT): Likewise.
18840 (TARGET_DSP_MULTIPLY): Likewise.
18841 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
18842 (arm_arch5, arm_arch5e): Delete.
18843 (arm_arch5t, arm_arch5te): Declare.
18844 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
18845 (arm_arch5t): Declare.
18846 (arm_option_reconfigure_globals): Update for the above.
18847 (arm_options_perform_arch_sanity_checks): Update comment, replace
18848 use of arm_arch5 with arm_arch5t.
18849 (use_return_insn): Likewise.
18850 (arm_emit_call_insn): Likewise.
18851 (output_return_instruction): Likewise.
18852 (arm_final_prescan_insn): Likewise.
18853 (arm_coproc_builtin_available): Likewise.
18854 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
18855 arm_arch5e with arm_arch5t and arm_arch5te.
18856 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
18857 (arm_arch5t, arm_arch5te): Declare.
18858 * config/arm/arm-tables.opt: Regenerate.
18859 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
18860 * config/arm/t-multilib: Likewise.
18861 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
18862 instead of arm_arch5.
18863 (*call_reg_thumb1): Likewise.
18864 (*call_value_reg_thumb1_v5): Likewise.
18865 (*call_value_reg_thumb1): Likewise.
18866 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
18868 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
18870 2018-05-18 Martin Liska <mliska@suse.cz>
18872 PR gcov-profile/84846
18873 * doc/gcov.texi: Document -t option of gcov tool.
18875 2018-05-18 Martin Liska <mliska@suse.cz>
18877 PR gcov-profile/84846
18878 * gcov.c (print_usage): Add new -t option.
18879 (process_args): Handle the option.
18880 (generate_results): Use stdout as output when requested by
18883 2018-05-18 Martin Liska <mliska@suse.cz>
18885 PR gcov-profile/84846
18886 * coverage.c (coverage_init): Write PWD to .gcno file.
18887 * doc/gcov.texi: Document how working directory is printed.
18888 * gcov-dump.c (dump_gcov_file): Print PWD.
18889 * gcov.c (output_intermediate_file): Likewise.
18890 (read_graph_file): Read PWD string.
18891 (output_lines): Print PWD.
18893 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
18895 PR middle-end/85817
18896 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
18897 for retval and return false if all args to phi are zero.
18899 2018-05-18 Richard Biener <rguenther@suse.de>
18901 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
18903 (evrp_dom_walker::before_dom_children): Call it.
18905 2018-05-18 Richard Biener <rguenther@suse.de>
18907 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
18908 results when processing array refs with variable index.
18910 2018-05-18 Toon Moene <toon@moene.org>
18912 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
18913 directly after that of -floop-interchange. Indicate that both
18914 options are enabled by default when specifying -O3.
18916 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18918 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
18919 iterator. Delete separate integer-mode vec_set<mode> expander.
18920 (aarch64_simd_vec_setv2di): Delete.
18921 (vec_setv2di): Delete.
18922 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
18923 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
18924 the "w, r" alternative.
18926 2018-05-18 Martin Liska <mliska@suse.cz>
18928 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
18929 * tree-pass.h (make_pass_lower_switch_O0): New function.
18930 * tree-switch-conversion.c (node_has_low_bound): Remove.
18931 (node_has_high_bound): Likewise.
18932 (node_is_bounded): Likewise.
18933 (class pass_lower_switch): Make it a template type and create
18935 (pass_lower_switch::execute): Add template argument.
18936 (make_pass_lower_switch): New function.
18937 (make_pass_lower_switch_O0): New function.
18938 (do_jump_if_equal): Remove.
18939 (emit_case_nodes): Simplify to just handle all 3 cases and leave
18940 all the hard work to tree optimization passes.
18942 2018-05-18 Martin Liska <mliska@suse.cz>
18944 * dbgcnt.c (limit_low): Renamed from limit.
18945 (limit_high): New variable.
18946 (dbg_cnt_is_enabled): Check for upper limit.
18947 (dbg_cnt): Adjust dumping.
18948 (dbg_cnt_set_limit_by_index): Add new argument for high
18950 (dbg_cnt_set_limit_by_name): Likewise.
18951 (dbg_cnt_process_single_pair): Parse new format.
18952 (dbg_cnt_process_opt): Use strtok.
18953 (dbg_cnt_list_all_counters): Remove 'value' and add
18955 * doc/invoke.texi: Document changes.
18957 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
18959 * doc/sourcebuild.texi (scalar_all_fma): Document.
18960 * tree.def (FMA_EXPR): Delete.
18961 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
18962 * internal-fn.c (ternary_direct): New macro.
18963 (expand_ternary_optab_fn): Likewise.
18964 (direct_ternary_optab_supported_p): Likewise.
18965 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
18966 * builtins.c (fold_builtin_fma): Delete.
18967 (fold_builtin_3): Don't call it.
18968 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
18969 * expr.c (expand_expr_real_2): Likewise.
18970 * fold-const.c (operand_equal_p): Likewise.
18971 (fold_ternary_loc): Likewise.
18972 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
18973 * gimple.c (DEFTREECODE): Likewise.
18974 * gimplify.c (gimplify_expr): Likewise.
18975 * optabs-tree.c (optab_for_tree_code): Likewise.
18976 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18977 * tree-eh.c (operation_could_trap_p): Likewise.
18978 (stmt_could_throw_1_p): Likewise.
18979 * tree-inline.c (estimate_operator_cost): Likewise.
18980 * tree-pretty-print.c (dump_generic_node): Likewise.
18981 (op_code_prio): Likewise.
18982 * tree-ssa-loop-im.c (stmt_cost): Likewise.
18983 * tree-ssa-operands.c (get_expr_operands): Likewise.
18984 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
18985 * fold-const-call.h (fold_fma): Delete.
18986 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
18987 CFN_FNMA and CFN_FNMS.
18988 (fold_fma): Delete.
18989 * genmatch.c (combined_fn): New enum.
18990 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
18991 (commutative_op): New function.
18992 (commutate): Use it. Handle more than 2 operands.
18993 (dt_operand::gen_gimple_expr): Use commutative_op.
18994 (parser::parse_expr): Allow :c to be used with non-binary
18995 operators if the commutative operand is known.
18996 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
18997 CFN_FMS, CFN_FNMA and CFN_FNMS.
18998 (backprop::process_assign_use): Remove FMA_EXPR handling.
18999 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19000 (gen_hsa_fma): New function.
19001 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19002 IFN_FNMA and IFN_FNMS.
19003 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19004 * gimple-fold.h (follow_all_ssa_edges): Declare.
19005 * gimple-fold.c (follow_all_ssa_edges): New function.
19006 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19007 gimple_build interface and use follow_all_ssa_edges to fold the result.
19008 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19009 instead of checking for optabs directly.
19010 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19011 rather than FMA_EXPRs.
19012 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19013 call to IFN_FMA instead of an FMA_EXPR.
19015 2018-05-17 Jim Wilson <jimw@sifive.com>
19017 * expr.c (do_tablejump): When converting index to Pmode, if we have a
19018 sign extended promoted subreg, and the range does not have the sign bit
19019 set, then do a sign extend.
19021 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19022 test, check for sign extended subreg and/or constant operands, and
19023 do a sign extend in that case.
19025 2018-05-17 Steve Ellcey <sellcey@cavium.com>
19027 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19028 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19030 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19031 Change logics_shift_reg to logics_shift_imm.
19032 (thunderx2t99_fp_loadpair_basic): Delete.
19033 (thunderx2t99_fp_storepair_basic): Delete.
19034 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19035 (thunderx2t99_asimd_polynomial): Delete.
19036 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19037 and neon_fp_mul_d_scalar_q.
19038 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19039 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19040 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19041 (thunderx2t99_asimd_lut): Add missing tbl types.
19042 (thunderx2t99_asimd_ext): Delete.
19043 (thunderx2t99_asimd_load1_1_mult): Delete.
19044 (thunderx2t99_asimd_load1_2_mult): Delete.
19045 (thunderx2t99_asimd_load1_ldp): New.
19046 (thunderx2t99_asimd_load1): New.
19047 (thunderx2t99_asimd_load2): Add missing *load2* types.
19048 (thunderx2t99_asimd_load3): New.
19049 (thunderx2t99_asimd_load4): New.
19050 (thunderx2t99_asimd_store1_1_mult): Delete.
19051 (thunderx2t99_asimd_store1_2_mult): Delete.
19052 (thunderx2t99_asimd_store2_mult): Delete.
19053 (thunderx2t99_asimd_store2_onelane): Delete.
19054 (thunderx2t99_asimd_store_stp): New.
19055 (thunderx2t99_asimd_store1): New.
19056 (thunderx2t99_asimd_store2): New.
19057 (thunderx2t99_asimd_store3): New.
19058 (thunderx2t99_asimd_store4): New.
19060 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
19062 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19063 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
19065 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
19066 Segher Boessenkool <segher@kernel.crashing.org>
19069 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19072 2018-05-17 Richard Biener <rguenther@suse.de>
19074 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19075 for pruning loop and prune defs feeding only already visited PHIs.
19077 2018-05-17 Richard Biener <rguenther@suse.de>
19079 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19081 2018-05-17 Bin Cheng <bin.cheng@arm.com>
19082 Richard Biener <rguenther@suse.de>
19084 PR tree-optimization/85793
19085 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19086 for VMAT_ELEMENTWISE.
19088 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
19090 * internal-fn.h (lookup_internal_fn): Declare
19091 * internal-fn.c (lookup_internal_fn): New function.
19092 * gimple.c (gimple_build_call_from_tree): Handle calls to
19093 internal functions.
19094 * gimple-pretty-print.c (dump_gimple_call): Print "." before
19095 internal function names.
19096 * tree-pretty-print.c (dump_generic_node): Likewise.
19097 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19099 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
19101 * gimple-fold.h (gimple_build): Make the function forms take
19102 combined_fn rather than built_in_function.
19103 (gimple_simplify): Likewise.
19104 * gimple-match-head.c (gimple_simplify): Likewise.
19105 * gimple-fold.c (gimple_build): Likewise.
19106 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19107 rather than gimple_build_call_internal.
19108 (get_initial_defs_for_reduction): Likewise.
19109 (vect_create_epilog_for_reduction): Likewise.
19110 (vectorizable_live_operation): Likewise.
19112 2018-05-17 Martin Liska <mliska@suse.cz>
19114 * gimple-ssa-sprintf.c (format_directive): Do not use
19115 space in between 'G_' and '('.
19117 2018-05-17 Jakub Jelinek <jakub@redhat.com>
19120 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19121 even if the mask is not all ones.
19124 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19126 (ix86_gimple_fold_builtin): Likewise.
19129 * config/i386/i386.c: Include tree-vector-builder.h.
19130 (ix86_vector_shift_count): New function.
19131 (ix86_fold_builtin): Fold shift builtins by scalar count.
19132 (ix86_gimple_fold_builtin): Likewise.
19134 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19135 _mm512_setzero): New intrinsics.
19137 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
19138 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19140 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19141 code generation for cases where splatting a value is not useful.
19142 * simplify-rtx.c (simplify_ternary_operation): Simplify
19143 vec_merge across a vec_duplicate and a paradoxical subreg forming
19144 a vector mode to a vec_concat.
19146 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
19148 * config.gcc: Support "goldmont-plus".
19149 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19151 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19152 PROCESSOR_GOLDMONT_PLUS.
19153 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19154 (processor_target_table): Add "goldmont-plus".
19155 (PTA_GOLDMONT_PLUS): Define.
19156 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19157 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19158 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19159 (fold_builtin_cpu): Add "goldmont-plus".
19160 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19161 (ix86_option_override_internal): Add "goldmont-plus".
19162 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19163 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19164 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19165 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19167 2018-05-17 Richard Biener <rguenther@suse.de>
19169 PR tree-optimization/85757
19170 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19171 remove defs that only feed that PHI from further processing.
19173 2018-05-16 Jim Wilson <jimw@sifive.com>
19175 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19177 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19179 2018-05-16 Mark Wielaard <mark@klomp.org>
19181 * dwarf2out.c (count_index_strings): New function.
19182 (output_indirect_strings): Call count_index_strings and generate
19183 header for dwarf_version >= 5.
19185 2018-05-16 Mark Wielaard <mark@klomp.org>
19187 * dwarf2out.c (dwarf_FORM): New function.
19188 (set_indirect_string): Use dwarf_FORM.
19189 (reset_indirect_string): Likewise.
19190 (size_of_die): Likewise.
19191 (value_format): Likewise.
19192 (output_die): Likewise.
19193 (add_skeleton_AT_string): Likewise.
19194 (output_macinfo_op): Likewise.
19195 (index_string): Likewise.
19196 (output_index_string_offset): Likewise.
19197 (output_index_string): Likewise.
19198 (count_index_strings): Likewise.
19200 2018-05-16 Carl Love <cel@us.ibm.com>
19202 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19203 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19205 2018-05-16 Martin Jambor <mjambor@suse.cz>
19207 * ipa-prop.c (ipa_free_all_edge_args): Remove.
19208 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19210 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
19212 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19213 (fnma<mode>4): Likewise.
19214 (fms<mode>4): Likewise.
19215 (fnms<mode>4): Likewise.
19216 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19217 (aarch64_fnma<mode>4): Likewise.
19218 (aarch64_fms<mode>4): Likewise.
19219 (aarch64_fnms<mode>4): Likewise.
19220 (aarch64_fnmadd<mode>4): Likewise.
19222 2018-05-16 Jason Merrill <jason@redhat.com>
19224 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
19226 2018-05-16 Richard Biener <rguenther@suse.de>
19228 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19229 (dump_stmt_cost): Declare.
19230 (add_stmt_cost): Dump cost we add.
19231 (add_stmt_costs): New function.
19232 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19233 No longer exported.
19234 (vect_analyze_stmt): Adjust prototype.
19235 (vectorizable_condition): Likewise.
19236 (vectorizable_live_operation): Likewise.
19237 (vectorizable_reduction): Likewise.
19238 (vectorizable_induction): Likewise.
19239 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19240 cost vector to pass to vectorizable_ and record afterwards.
19241 (vect_model_reduction_cost): Take cost vector argument and adjust.
19242 (vect_model_induction_cost): Likewise.
19243 (vectorizable_reduction): Likewise.
19244 (vectorizable_induction): Likewise.
19245 (vectorizable_live_operation): Likewise.
19246 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19247 SLP_TREE_NUMBER_OF_VEC_STMTS.
19248 (vect_analyze_slp_cost_1): Remove.
19249 (vect_analyze_slp_cost): Likewise.
19250 (vect_slp_analyze_node_operations): Take visited args and
19251 a target cost vector. Avoid processing already visited stmt sets.
19252 (vect_slp_analyze_operations): Use a local cost vector to gather
19253 costs and register those of non-discarded instances.
19254 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19255 (vect_schedule_slp_instance): Remove copying of
19256 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
19258 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19259 adding cost. Record cost entry location.
19260 (vect_prologue_cost_for_slp_op): Function to compute cost of
19261 a constant or invariant generated for SLP vect in the prologue,
19262 split out from vect_analyze_slp_cost_1.
19263 (vect_model_simple_cost): Make static. Adjust for SLP costing.
19264 (vect_model_promotion_demotion_cost): Likewise.
19265 (vect_model_store_cost): Likewise, make static.
19266 (vect_model_load_cost): Likewise.
19267 (vectorizable_bswap): Add cost vector arg and adjust.
19268 (vectorizable_call): Likewise.
19269 (vectorizable_simd_clone_call): Likewise.
19270 (vectorizable_conversion): Likewise.
19271 (vectorizable_assignment): Likewise.
19272 (vectorizable_shift): Likewise.
19273 (vectorizable_operation): Likewise.
19274 (vectorizable_store): Likewise.
19275 (vectorizable_load): Likewise.
19276 (vectorizable_condition): Likewise.
19277 (vectorizable_comparison): Likewise.
19278 (can_vectorize_live_stmts): Likewise.
19279 (vect_analyze_stmt): Likewise.
19280 (vect_transform_stmt): Adjust calls to vectorizable_*.
19281 * tree-vectorizer.c: Include gimple-pretty-print.h.
19282 (dump_stmt_cost): New function.
19284 2018-05-16 Richard Biener <rguenther@suse.de>
19286 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19287 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19288 * tree-ssa-dse.c: Include tree-ssa-loop.h.
19289 (check_name): New callback.
19290 (dse_classify_store): Track cycles via a visited bitmap of PHI
19291 defs and simplify handling of in-loop and across loop dead stores
19292 and properly fail for loop-variant refs. Handle byte-tracking with
19293 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19296 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
19298 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19299 (vect_get_mask_type_for_stmt): Likewise.
19300 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19302 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
19303 to determine the statement's vector type and the vector type that
19304 should be used for calculating nunits. Deal with cases in which
19305 the type has to be deferred.
19306 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19307 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19308 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19309 (vect_determine_vf_for_stmt): New functions, split out from...
19310 (vect_determine_vectorization_factor): ...here.
19311 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19312 (vect_get_mask_type_for_stmt): New functions, split out from
19313 vect_determine_vectorization_factor.
19315 2018-05-16 Richard Biener <rguenther@suse.de>
19317 * tree-cfg.c (verify_gimple_assign_ternary): Properly
19318 verify the [VEC_]COND_EXPR embedded comparison.
19320 2018-05-15 Martin Sebor <msebor@redhat.com>
19322 PR tree-optimization/85753
19323 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19324 RECORD_TYPE in addition to ARRAY_TYPE.
19326 2018-05-15 Martin Sebor <msebor@redhat.com>
19328 PR middle-end/85643
19329 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19331 2018-05-15 Richard Biener <rguenther@suse.de>
19333 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19334 add by_clobber_p one. Change algorithm to collect all defs
19335 representing uses we need to walk and try reducing them to
19336 a single one before failing.
19337 (dse_dom_walker::dse_optimize_stmt): Adjust.
19339 2018-05-13 Mark Wielaard <mark@klomp.org>
19341 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19342 (size_of_loc_descr): Likewise.
19343 (output_loc_operands): Likewise.
19344 (output_loc_operands_raw): Likewise.
19345 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19346 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19347 (hash_loc_operands): Likewise.
19348 (compare_loc_operands): Likewise.
19350 2018-05-14 Mark Wielaard <mark@klomp.org>
19352 * dwarf2out.c (count_index_addrs): New function.
19353 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19355 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19357 PR tree-optimization/83648
19358 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19359 return value as malloc candidate.
19361 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19364 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19365 param as true in call to suggest_attribute.
19367 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
19369 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19370 -mreadonly-in-sdata.
19372 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19374 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19376 (aarch64_crypto_aesd_fused): Likewise.
19378 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
19380 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19381 (movsi_aarch64): Likewise.
19382 (load_pairsi): Likewise.
19383 (load_pairdi): Likewise.
19384 (store_pairsi): Likewise.
19385 (store_pairdi): Likewise.
19386 (load_pairsf): Likewise.
19387 (load_pairdf): Likewise.
19388 (store_pairsf): Likewise.
19389 (store_pairdf): Likewise.
19390 (zero_extend): Likewise.
19391 (trunc): Swap alternatives.
19392 (fcvt_target): Add '?' to prefer w over r.
19394 2018-05-14 Jakub Jelinek <jakub@redhat.com>
19397 * config/i386/i386.md: Disallow non-commutative arithmetics in
19398 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19399 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
19400 in the peephole2 before it.
19402 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
19404 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19405 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19406 (ix86_handle_option): Handle -mcldemote.
19407 * config.gcc: New header.
19408 * config/i386/cldemoteintrin.h: New file.
19409 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19410 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19412 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19413 OPTION_MASK_ISA_CLDEMOTE.
19414 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19415 (ix86_valid_target_attribute_inner_p): Ditto.
19416 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19417 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19418 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19419 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19420 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19422 * config/i386/i386.opt: Add -mcldemote.
19423 * config/i386/x86intrin.h: New header.
19424 * doc/invoke.texi: Add -mcldemote.
19426 2018-05-14 Richard Biener <rguenther@suse.de>
19428 * doc/match-and-simplify.texi: Adjust :s documentation.
19430 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
19432 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19433 intended memcpy size.
19434 (REORDER_45): Likewise.
19436 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
19438 * sort.cc: New file.
19439 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19440 * vec.c (qsort_chk): Use gcc_qsort.
19441 * Makefile.in (OBJS-libcommon): Add sort.o.
19442 (build/sort.o): New target. Use it...
19443 (BUILD_RTL): ... here, and...
19444 (build/gencfn-macros): ... here, and...
19445 (build/genmatch): ... here.
19447 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19448 Chung-Ju Wu <jasonwucj@gmail.com>
19450 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19451 * config/nds32/nds32-graywolf.md: New file.
19452 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19453 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19455 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19456 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19457 * config/nds32/nds32.md (pipeline_model): Add graywolf.
19458 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19459 * config/nds32/pipelines.md: Include n15 settings.
19461 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19462 Chung-Ju Wu <jasonwucj@gmail.com>
19464 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19465 * config/nds32/nds32-n13.md: New file.
19466 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19467 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19469 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19470 * config/nds32/nds32.md (pipeline_model): Add n13.
19471 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19472 * config/nds32/pipelines.md: Include n13 settings.
19474 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19475 Chung-Ju Wu <jasonwucj@gmail.com>
19477 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19478 * config/nds32/nds32-n10.md: New file.
19479 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19480 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19482 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19483 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19484 * config/nds32/nds32.md (pipeline_model): Add n10.
19485 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19486 * config/nds32/pipelines.md: Include n10 settings.
19488 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
19489 Kito Cheng <kito.cheng@gmail.com>
19490 Chung-Ju Wu <jasonwucj@gmail.com>
19492 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19493 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19494 Add enum values for DSP extension instructions.
19495 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19497 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19498 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19499 New code iterators.
19500 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19501 * config/nds32/nds32-dspext.md: New file for DSP implementation.
19502 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19503 * config/nds32/nds32-intrinsic.md: Likewise.
19504 * config/nds32/nds32_intrinsic.h: Likewise.
19505 * config/nds32/nds32-md-auxiliary.c: Likewise.
19506 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19507 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19508 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19509 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19510 * config/nds32/nds32-protos.h: New declarations for DSP extension.
19511 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19512 TYPE_DMAC in switch statement.
19513 * config/nds32/nds32.c: New checking and implementation for DSP
19514 extension instructions.
19515 * config/nds32/nds32.h: Likewise.
19516 * config/nds32/nds32.md: Likewise.
19517 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19518 * config/nds32/predicates.md: Implement new predicates for DSP
19521 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
19523 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19524 Reformat alternatives and attributes so it is easier to identify
19525 which constraints/attributes go with which instruction.
19526 (mov<mode>_hardfloat32, FMOVE64): Likewise.
19527 (mov<mode>_softfloat32, FMOVE64): Likewise.
19528 (mov<mode>_hardfloat64, FMOVE64): Likewise.
19529 (mov<mode>_softfloat64, FMOVE64): Likewise.
19531 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
19533 * doc/extend.texi (PowerPC Built-in Functions): Rename this
19535 (Basic PowerPC Built-in Functions): The new name of the
19536 subsection previously known as "PowerPC Built-in Functions".
19537 (Basic PowerPC Built-in Functions Available on all Configurations):
19539 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19540 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19541 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19542 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19544 2018-05-11 Martin Jambor <mjambor@suse.cz>
19547 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19550 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
19553 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19555 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
19557 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19558 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19559 (ix86_handle_option): Handle -mwaitpkg.
19560 * config.gcc: New header.
19561 * config/i386/cpuid.h (bit_WAITPKG): New bit.
19562 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19563 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19565 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19566 OPTION_MASK_ISA_WAITPKG.
19567 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19568 (ix86_option_override_internal): Add PTA_WAITPKG.
19569 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19570 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19571 IX86_BUILTIN_TPAUSE.
19572 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19573 __builtin_ia32_umwait and __builtin_ia32_tpause.
19574 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19575 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19576 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19577 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19578 UNSPECV_TPAUSE): New.
19579 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19580 * config/i386/i386.opt: Add -mwaitpkg.
19581 * config/i386/waitpkgintrin.h: New file.
19582 * config/i386/x86intrin.h: New header.
19583 * doc/invoke.texi: Add -mwaitpkg.
19585 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
19588 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
19590 (cortex-m0): Use armv6s-m isa.
19591 (cortex-m0plus): Likewise.
19592 (cortex-m1): Likewise.
19593 (cortex-m0.small-multiply): Likewise.
19594 (cortex-m0plus.small-multiply): Likewise.
19595 (cortex-m1.small-multiply): Likewise.
19597 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
19598 Jakub Jelinek <jakub@redhat.com>
19600 PR tree-optimization/85692
19601 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
19602 source permute as well.
19604 2018-05-11 Martin Liska <mliska@suse.cz>
19607 * doc/extend.texi: Document LLVM style format for no_sanitize
19610 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
19612 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
19613 mode_supports_vsx_dform_quad to mode_supports_dq_form.
19614 (mode_supports_vsx_dform_quad): Likewise.
19615 (mode_supports_vmx_dform): Move these functions to be next to the
19616 other mode_supports functions.
19617 (mode_supports_dq_form): Likewise.
19618 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
19619 mode_supports_dq_form.
19620 (reg_offset_addressing_ok_p): Likewise.
19621 (offsettable_ok_by_alignment): Likewise.
19622 (rs6000_legitimate_offset_address_p): Likewise.
19623 (legitimate_lo_sum_address_p): Likewise.
19624 (rs6000_legitimize_address): Likewise.
19625 (rs6000_legitimize_reload_address): Likewise.
19626 (rs6000_secondary_reload_inner): Likewise.
19627 (rs6000_preferred_reload_class): Likewise.
19628 (rs6000_output_move_128bit): Likewise.
19630 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
19632 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
19633 Generate SImode target register for null target.
19634 <case IX86_BUILTIN_XGETBV>: Ditto.
19635 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
19636 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
19638 2018-05-10 Carl Love <cel@us.ibm.com>
19640 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19641 dcbtt and dcbtstt if operands[2] is 0.
19643 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
19646 * config/i386/sse.md (usadv64qi): New expander.
19648 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
19650 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
19651 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
19652 -maltivec=be support.
19653 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
19654 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
19655 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
19656 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
19657 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
19658 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
19659 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
19660 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
19661 altivec_vsumsws): Adjust.
19662 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
19663 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
19664 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
19666 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
19667 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
19668 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
19669 (altivec_lve<VI_char>x): Delete expand.
19670 (*altivec_lve<VI_char>x_internal): Rename to...
19671 (altivec_lve<VI_char>x): ... this.
19672 (altivec_lvxl_<mode>): Delete expand.
19673 (*altivec_lvxl_<mode>_internal): Rename to ...
19674 (altivec_lvxl_<mode>): ... this.
19675 (altivec_stvxl_<mode>): Delete expand.
19676 (*altivec_stvxl_<mode>_internal): Rename to ...
19677 (altivec_stvxl_<mode>): ... this.
19678 (altivec_stve<VI_char>x): Delete expand.
19679 (*altivec_stve<VI_char>x_internal): Rename to ...
19680 (altivec_stve<VI_char>x): ... this.
19681 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
19682 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
19683 reduc_plus_scal_<mode>): Adjust.
19684 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
19686 (rs6000_cpu_cpp_builtins): Adjust.
19687 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
19688 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
19689 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
19690 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
19691 -maltivec=be support.
19692 (rs6000_split_vec_extract_var): Adjust.
19693 (rs6000_split_v4si_init): Adjust.
19694 (swap_selector_for_mode): Delete.
19695 (altivec_expand_lvx_be, altivec_expand_stvx_be,
19696 altivec_expand_stvex_be): Delete.
19697 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
19698 -maltivec=be support.
19699 (rs6000_gimple_fold_builtin): Ditto.
19700 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
19702 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
19703 (TARGET_DIRECT_MOVE_64BIT): Adjust.
19704 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
19705 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
19706 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
19707 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
19708 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
19709 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
19710 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
19711 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
19712 anonymous split): Adjust.
19713 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
19714 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
19716 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
19718 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
19719 when --with-gxx-include-dir is also specified.
19720 * configure: Regenerate.
19722 2018-05-09 Jim Wilson <jimw@sifive.com>
19725 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
19726 * config/riscv/t-withmultilib: New.
19727 * config/riscv/withmultilib.h: New.
19728 * doc/install.texi: Document RISC-V --with-multilib-list support.
19730 2018-05-09 Richard Biener <rguenther@suse.de>
19732 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
19734 (vect_bb_vectorization_profitable_p): Adjust. Compute
19735 actual scalar cost using the cost vector and the add_stmt_cost
19738 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
19740 PR rtl-optimization/85645
19741 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
19742 in the REG_CFA_REGISTER note for LR, don't leave it empty.
19744 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
19746 PR rtl-optimization/85645
19747 * shrink-wrap.c (spread_components): Return a boolean saying if
19748 anything was changed.
19749 (try_shrink_wrapping_separate): Iterate spread_components until
19750 nothing changes anymore.
19752 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
19754 PR rtl-optimization/85645
19755 * regrename.c (build_def_use): Also kill the chains that include the
19756 destination of a REG_CFA_REGISTER note.
19758 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
19760 PR rtl-optimization/85645
19761 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
19762 insn that has a REG_CFA_REGISTER note.
19764 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
19766 * cfgexpand.c (expand_clobber): New function.
19767 (expand_gimple_stmt_1): Use it.
19768 * tree-vect-stmts.c (vect_clobber_variable): New function,
19770 (vectorizable_simd_clone_call): ...here.
19771 (vectorizable_store): Emit a clobber either side of an
19772 IFN_STORE_LANES sequence.
19773 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
19775 2018-05-09 Tom de Vries <tom@codesourcery.com>
19778 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
19779 (define_insn "trap_if_false"): Add exit after trap.
19781 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
19783 PR rtl-optimization/85638
19784 * bb-reorder.c: Include common/common-target.h.
19785 (create_forwarder_block): New function extracted from...
19786 (fix_up_crossing_landing_pad): ...here. Rename into...
19787 (dw2_fix_up_crossing_landing_pad): ...this.
19788 (sjlj_fix_up_crossing_landing_pad): New function.
19789 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
19790 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
19791 from both partitions and exit the loop after one iteration.
19793 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
19796 * doc/extend.texi (PowerPC Built-in Functions): Rename this
19798 (Basic PowerPC Built-in Functions): The new name of the
19799 subsection previously known as "PowerPC Built-in Functions".
19800 (Basic PowerPC Built-in Functions Available on all Configurations):
19802 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19804 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19806 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19808 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19811 2018-05-08 Jim Wilson <jimw@sifive.com>
19813 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
19814 (LD_EMUL_SUFFIX): New.
19815 (LINK_SPEC): Use it.
19817 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
19819 * doc/extend.texi (PowerPC Built-in Functions): Rename this
19821 (Basic PowerPC Built-in Functions): The new name of the
19822 subsection previously known as "PowerPC Built-in Functions".
19823 (Basic PowerPC Built-in Functions Available on all Configurations):
19825 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
19827 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
19829 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
19831 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
19834 2018-05-08 Jakub Jelinek <jakub@redhat.com>
19837 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
19838 after cmpelim optimization.
19840 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
19842 * config.gcc: Support "goldmont".
19843 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
19844 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19845 PROCESSOR_GOLDMONT.
19846 * config/i386/i386.c (m_GOLDMONT): Define.
19847 (processor_target_table): Add "goldmont".
19848 (PTA_GOLDMONT): Define.
19849 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
19850 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
19851 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
19852 (fold_builtin_cpu): Add "goldmont".
19853 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
19854 (ix86_option_override_internal): Add "goldmont".
19855 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
19856 (processor_type): Add PROCESSOR_GOLDMONT.
19857 * config/i386/i386.md: Add CPU "glm".
19858 * config/i386/glm.md: New file.
19859 * config/i386/x86-tune.def: Add m_GOLDMONT.
19860 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
19862 2018-05-08 Jakub Jelinek <jakub@redhat.com>
19865 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
19867 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
19868 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
19869 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
19872 * config/i386/i386.c (ix86_fold_builtin): Handle
19873 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
19876 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
19877 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
19879 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
19882 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
19883 (check_arch): Likewise.
19884 (check_fpu): Return the result rather than printing it.
19885 (end arch): Fix operator precedence.
19886 (end cpu): Likewise.
19887 (END): Print the result from check_fpu.
19889 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
19890 Alan Hayward <alan.hayward@arm.com>
19891 David Sherwood <david.sherwood@arm.com>
19893 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
19894 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
19895 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
19896 (*fcmuo<mode>_and): New patterns.
19898 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
19900 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
19901 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
19902 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
19903 (cmp_op, sve_imm_con): New code attributes.
19904 (SVE_COND_INT_CMP, imm_con): Delete.
19905 (cmp_op): Remove above unspecs from int attribute.
19906 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
19908 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
19909 comparison-specific unspecs.
19910 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
19911 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
19912 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
19913 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
19914 (*vec_fcm<cmp_op><mode>): Rename to...
19915 (*fcm<cmp_op><mode>): ...this and adjust likewise.
19916 (*vec_fcmuo<mode>): Rename to...
19917 (*fcmuo<mode>): ...this and adjust likewise.
19918 (*pred_fcm<cmp_op><mode>): New pattern.
19919 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
19920 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
19922 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
19924 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
19925 (aarch64_emit_sve_predicated_cond): New function.
19926 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
19927 (aarch64_emit_unspec_cond_or): Replace with...
19928 (aarch64_emit_sve_or_conds): ...this new function. Use
19929 aarch64_emit_sve_ptrue_op for the individual comparisons and
19930 aarch64_emit_binop to OR them together.
19931 (aarch64_emit_inverted_unspec_cond): Replace with...
19932 (aarch64_emit_sve_inverted_cond): ...this new function. Use
19933 aarch64_emit_sve_ptrue_op for the comparison and
19934 aarch64_emit_unop to invert the result.
19935 (aarch64_expand_sve_vec_cmp_float): Update after the above
19936 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
19938 2018-05-07 Nathan Sidwell <nathan@acm.org>
19940 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
19941 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
19942 (Backwards Compatibility): Likewise.
19944 2018-05-07 Luis Machado <luis.machado@linaro.org>
19948 2018-05-07 Luis Machado <luis.machado@linaro.org>
19950 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19951 <prefetch_dynamic_strides>: New const bool field.
19952 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19953 prefetch_dynamic_strides.
19954 (exynosm1_prefetch_tune): Likewise.
19955 (thunderxt88_prefetch_tune): Likewise.
19956 (thunderx_prefetch_tune): Likewise.
19957 (thunderx2t99_prefetch_tune): Likewise.
19958 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
19960 (aarch64_override_options_internal): Update to set
19961 PARAM_PREFETCH_DYNAMIC_STRIDES.
19962 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
19963 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
19964 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
19965 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
19966 prefetch-dynamic-strides setting.
19968 2018-05-07 Luis Machado <luis.machado@linaro.org>
19970 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19971 <minimum_stride>: New const int field.
19972 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19973 minimum_stride field.
19974 (exynosm1_prefetch_tune): Likewise.
19975 (thunderxt88_prefetch_tune): Likewise.
19976 (thunderx_prefetch_tune): Likewise.
19977 (thunderx2t99_prefetch_tune): Likewise.
19978 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
19979 (aarch64_override_options_internal): Update to set
19980 PARAM_PREFETCH_MINIMUM_STRIDE.
19981 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
19982 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
19983 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
19984 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
19985 stride is constant and is below the minimum stride threshold.
19987 2018-05-07 Luis Machado <luis.machado@linaro.org>
19989 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
19992 2018-05-07 Luis Machado <luis.machado@linaro.org>
19994 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
19995 <prefetch_dynamic_strides>: New const bool field.
19996 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
19997 prefetch_dynamic_strides.
19998 (exynosm1_prefetch_tune): Likewise.
19999 (thunderxt88_prefetch_tune): Likewise.
20000 (thunderx_prefetch_tune): Likewise.
20001 (thunderx2t99_prefetch_tune): Likewise.
20002 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20004 (aarch64_override_options_internal): Update to set
20005 PARAM_PREFETCH_DYNAMIC_STRIDES.
20006 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20007 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20008 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20009 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20010 prefetch-dynamic-strides setting.
20012 2018-05-07 Luis Machado <luis.machado@linaro.org>
20014 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20015 <minimum_stride>: New const int field.
20016 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20017 minimum_stride field.
20018 (exynosm1_prefetch_tune): Likewise.
20019 (thunderxt88_prefetch_tune): Likewise.
20020 (thunderx_prefetch_tune): Likewise.
20021 (thunderx2t99_prefetch_tune): Likewise.
20022 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20023 (aarch64_override_options_internal): Update to set
20024 PARAM_PREFETCH_MINIMUM_STRIDE.
20025 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20026 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20027 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20028 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20029 stride is constant and is below the minimum stride threshold.
20031 2018-05-06 Jakub Jelinek <jakub@redhat.com>
20034 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20035 the type is addressable. Don't force op into register if it has
20038 2018-05-05 Roland McGrath <mcgrathr@google.com>
20041 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20042 any section for which we don't know a specific type it should have,
20043 regardless of name. Previously this was done only for the exact
20044 names ".init_array", ".fini_array", and ".preinit_array".
20045 (default_elf_asm_named_section): Add comment about
20046 relationship with default_section_type_flags and SECTION_NOTYPE.
20047 (get_section): Don't consider it a type conflict if one side has
20048 SECTION_NOTYPE and the other doesn't, as long as neither has the
20049 SECTION_BSS et al used in the default_section_type_flags logic.
20051 2018-05-05 Tom de Vries <tom@codesourcery.com>
20054 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20055 (workaround_barsyncs): New function.
20056 (nvptx_reorg): Use workaround_barsyncs.
20057 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20058 (define_expand "nvptx_membar_cta"): New define_expand.
20059 (define_insn "*nvptx_membar_cta"): New insn.
20061 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
20063 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20064 To improve optimization opportunities.
20065 * builtin-types.def: The new needed builtin types for the above.
20067 2018-05-04 Richard Biener <rguenther@suse.de>
20069 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20070 * gimple-ssa-store-merging.c
20071 (imm_store_chain_info::output_merged_store): Remove redundant create,
20072 release split_store vector contents on failure.
20073 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20074 scalar stmt vector on cache hit.
20076 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
20078 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20080 * config.gcc (powerpc-xilinx-eabi*): Remove.
20081 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20083 (fusion_addis_mem_combo_load): Ditto.
20084 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20086 (rs6000_cpu_cpp_builtins): Ditto.
20087 * config/rs6000/rs6000-linux.c
20088 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20089 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20090 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20092 (rs6000_setup_reg_addr_masks): Ditto.
20093 (rs6000_init_hard_regno_mode_ok): Ditto.
20094 (rs6000_option_override_internal): Ditto.
20095 (legitimate_lo_sum_address_p): Ditto.
20096 (rs6000_legitimize_address): Ditto.
20097 (rs6000_legitimize_reload_address): Ditto.
20098 (rs6000_legitimate_address_p): Ditto.
20099 (abi_v4_pass_in_fpr): Ditto.
20100 (setup_incoming_varargs): Ditto.
20101 (rs6000_gimplify_va_arg): Ditto.
20102 (rs6000_split_multireg_move): Ditto.
20103 (rs6000_savres_strategy): Ditto.
20104 (rs6000_emit_prologue_components): Ditto.
20105 (rs6000_emit_epilogue_components): Ditto.
20106 (rs6000_emit_prologue): Ditto.
20107 (rs6000_emit_epilogue): Ditto.
20108 (rs6000_elf_file_end): Ditto.
20109 (rs6000_function_value): Ditto.
20110 (rs6000_libcall_value): Ditto.
20111 * config/rs6000/rs6000.h: Ditto.
20112 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20113 (TARGET_MINMAX): ... this. New.
20114 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20115 * config/rs6000/rs6000.md: Remove Xilinx FP support.
20116 (*movsi_internal1_single): Delete.
20117 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20118 mfpu=, mxilinx-fpu): Delete.
20119 * config/rs6000/singlefp.h: Delete.
20120 * config/rs6000/sysv4.h: Remove Xilinx FP support.
20121 * config/rs6000/t-rs6000: Ditto.
20122 * config/rs6000/t-xilinx: Delete.
20123 * config/rs6000/titan.md: Adjust for fp_type removal.
20124 * config/rs6000/vsx.md: Remove Xilinx FP support.
20125 (VStype_simple): Delete.
20126 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20127 * config/rs6000/xfpu.h: Delete.
20128 * config/rs6000/xfpu.md: Delete.
20129 * config/rs6000/xilinx.h: Delete.
20130 * config/rs6000/xilinx.opt: Delete.
20131 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20132 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20134 2018-05-04 Tom de Vries <tom@codesourcery.com>
20137 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20140 2018-05-04 Richard Biener <rguenther@suse.de>
20142 PR middle-end/85627
20143 * tree-complex.c (update_complex_assignment): We are always in SSA form.
20144 (expand_complex_div_wide): Likewise.
20145 (expand_complex_operations_1): Likewise.
20146 (expand_complex_libcall): Preserve EH info of the original stmt.
20147 (tree_lower_complex): Handle removed blocks.
20148 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20149 on complex multiplication and division libcall builtins.
20151 2018-05-04 Richard Biener <rguenther@suse.de>
20153 PR middle-end/85574
20154 * fold-const.c (negate_expr_p): Restrict negation of operand
20155 zero of a division to when we know that can happen without
20157 (fold_negate_expr_1): Likewise.
20159 2018-05-04 Jakub Jelinek <jakub@redhat.com>
20162 * real.h (real_nextafter): Declare.
20163 * real.c (real_nextafter): New function.
20164 * fold-const-call.c (fold_const_nextafter): New function.
20165 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20166 CASE_CFN_NEXTTOWARD.
20167 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20168 even when arg1_mode is different from arg0_mode.
20170 2018-05-03 Nathan Sidwell <nathan@acm.org>
20172 * doc/extend.texi (Deprecated Features): Remove
20173 -ffriend-injection.
20174 (Backwards Compatibility): Likewise.
20175 * doc/invoke.texi (C++ Language Options): Likewise.
20176 (C++ Dialect Options): Likewise.
20178 2018-05-03 Jakub Jelinek <jakub@redhat.com>
20181 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20182 _mm512_mask_mullox_epi64): New intrinsics.
20184 2018-05-03 Tom de Vries <tom@codesourcery.com>
20187 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20188 dump files): Add offload-tree.
20190 2018-05-03 Richard Biener <rguenther@suse.de>
20192 PR tree-optimization/85615
20193 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20194 to loops not nested in BBs loop father to avoid creating multi-entry
20197 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20199 PR tree-optimization/70291
20200 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20201 arguments. Change return type to tree. Emit libcall as a new
20202 statement rather than replacing existing one when inplace_p is true.
20203 (expand_complex_multiplication_components): New function.
20204 (expand_complex_multiplication): Expand floating-point complex
20205 multiplication using the above.
20206 (expand_complex_division): Rename inner_type parameter to type.
20207 Update expand_complex_libcall call-site.
20208 (expand_complex_operations_1): Update expand_complex_multiplication
20209 and expand_complex_division call-sites.
20211 2018-05-02 Jakub Jelinek <jakub@redhat.com>
20214 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20215 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20216 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20217 the highest significant bit of the shift count mask is clear. In
20218 check whether and[sq]i3 is needed verify that all significant bits
20219 of the shift count other than the highest are set.
20221 2018-05-02 Tom de Vries <tom@codesourcery.com>
20224 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20225 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20226 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20227 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20228 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20229 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20230 __builtin_goacc_parlevel_size.
20232 2018-05-02 Richard Biener <rguenther@suse.de>
20234 PR tree-optimization/85597
20235 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20236 do not use split vect_get_vec_defs call but call vect_get_slp_defs
20239 2018-05-02 Tom de Vries <tom@codesourcery.com>
20242 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20243 dump files): Add ltrans-tree.
20245 2018-05-02 Tom de Vries <tom@codesourcery.com>
20248 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20249 dump files): Add wpa-ipa.
20251 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
20253 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
20254 powerpc*-*-linux*paired* target.
20255 * config/rs6000/750cl.h: Delete.
20256 * config/rs6000/paired.h: Delete.
20257 * config/rs6000/paired.md: Delete.
20258 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20260 * config/rs6000/rs6000-builtin.def: Remove paired float support.
20261 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20262 comment. Remove paired float support.
20263 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20264 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20266 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20267 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20269 * config/rs6000/rs6000.c: Remove paired float support.
20270 (paired_expand_vector_init, paired_expand_vector_move,
20271 paired_emit_vector_compare, paired_emit_vector_cond_expr,
20272 (paired_expand_lv_builtin, paired_expand_stv_builtin,
20273 paired_expand_builtin, paired_expand_predicate_builtin,
20274 paired_init_builtins): Delete.
20275 * config/rs6000/rs6000.h: Remove paired float support.
20276 * config/rs6000/rs6000.md: Remove paired float support.
20277 (move_from_CR_ov_bit): Delete.
20278 * config/rs6000/rs6000.opt (mpaired): Delete.
20279 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20280 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20282 2018-05-02 Richard Biener <rguenther@suse.de>
20284 PR middle-end/85567
20285 * gimplify.c (gimplify_save_expr): When in SSA form allow
20286 SAVE_EXPRs to compute to SSA vars.
20288 2018-05-02 Jakub Jelinek <jakub@redhat.com>
20291 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20292 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20293 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20294 clobber operands[2], instead use a new pseudo. Formatting fixes.
20296 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
20298 PR tree-optimization/85586
20299 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20300 exit early for statements in the same group if the accesses are
20303 2018-05-02 Tom de Vries <tom@codesourcery.com>
20306 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20309 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
20311 PR tree-optimization/85143
20312 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20314 2018-05-01 Tom de Vries <tom@codesourcery.com>
20317 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20318 not found" error message.
20320 2018-05-01 Tom de Vries <tom@codesourcery.com>
20323 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20324 * vec.c (test_ordered_remove_if): New function.
20325 (vec_c_tests): Call test_ordered_remove_if.
20326 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20327 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20328 * tree-vect-patterns.c (vect_pattern_recog_1): Use
20329 VEC_ORDERED_REMOVE_IF.
20331 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20333 PR tree-optimization/82665
20334 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20335 pointer subtraction where arguments come from a memchr call.
20337 2018-05-01 Jakub Jelinek <jakub@redhat.com>
20339 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20340 --push-state --as-needed and --pop-state instead of --as-needed and
20341 --no-as-needed if ld supports it.
20342 * configure: Regenerated.
20345 * doc/install.texi2html: Replace _002d with - and _002a with * in
20346 generated html files using sed.
20348 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20351 * gcc-rich-location.c (blank_line_before_p): New function.
20352 (use_new_line): New function.
20353 (gcc_rich_location::add_fixit_insert_formatted): New function.
20354 * gcc-rich-location.h
20355 (gcc_rich_location::add_fixit_insert_formatted): New function.
20357 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20359 * selftest.c (assert_streq): Rename "expected" and "actual" to
20360 "val1" and "val2". Extend NULL-handling to cover both inputs
20361 symmetrically, while still requiring both to be non-NULL for a pass.
20362 * selftest.h (assert_streq): Rename "expected" and "actual" to
20364 (ASSERT_EQ): Likewise.
20365 (ASSERT_EQ_AT): Likewise.
20366 (ASSERT_KNOWN_EQ): Likewise.
20367 (ASSERT_KNOWN_EQ_AT): Likewise.
20368 (ASSERT_NE): Likewise.
20369 (ASSERT_MAYBE_NE): Likewise.
20370 (ASSERT_MAYBE_NE_AT): Likewise.
20371 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
20372 the assertion to pass.
20373 (ASSERT_STREQ_AT): Likewise.
20375 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
20377 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20378 interaction with -pie.
20380 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20382 * selftest.h: Fix alphabetization of per-source-file selftest
20385 2018-04-30 Jason Merrill <jason@redhat.com>
20387 PR c++/61982 - dead stores to destroyed objects.
20388 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20391 2018-04-30 Jason Merrill <jason@redhat.com>
20393 * tree.c (build_clobber): New.
20394 * tree.h: Declare it.
20395 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20397 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20399 * diagnostic-show-locus.c (layout::layout): Update for
20400 location_get_source_line returning a char_span.
20401 (struct char_span): Move to input.h.
20402 (struct correction): Update for fields in char_span becoming
20404 (struct source_line): Update for location_get_source_line
20405 returning a char_span.
20406 (layout::print_line): Likewise.
20407 * edit-context.c (edited_file::print_content): Likewise.
20408 (edited_file::print_diff_hunk): Likewise.
20409 (edited_file::print_run_of_changed_lines): Likewise.
20410 (edited_file::get_num_lines): Likewise.
20411 (edited_line::edited_line): Likewise.
20412 * final.c (asm_show_source): Likewise.
20413 * input.c (location_get_source_line): Convert return type
20414 from const char * to char_span, losing the final "line_len"
20416 (dump_location_info): Update for the above.
20417 (get_substring_ranges_for_loc): Likewise. Use a char_span
20418 when handling the literal within the line.
20419 (test_reading_source_line): Update for location_get_source_line
20420 returning a char_span.
20421 * input.h (class char_span): Move here from
20422 diagnostic-show-locus.c, converting from a struct to a class.
20423 Make data members private.
20424 (char_span::operator bool): New.
20425 (char_span::length): New.
20426 (char_span::get_buffer): New.
20427 (char_span::operator[]): New.
20428 (char_span::subspan): Make const.
20429 (char_span::xstrdup): New.
20430 (location_get_source_line): Convert return type from const char *
20431 to char_span, losing the final "line_size" param.
20433 2018-04-30 Jan Hubicka <jh@suse.cz>
20435 * lto-wrapper.c (ltrans_priorities): New static var.
20436 (cmp_priority): New.
20437 (run_gcc): Read priorities and if doing parallel build order
20438 the Makefile by them.
20440 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20442 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20444 2018-04-30 Richard Biener <rguenther@suse.de>
20446 * tree-cfg.c (verify_address): Remove base argument, add
20447 flag whether to check TREE_ADDRESSABLE and do that.
20448 (verify_expr): Remove.
20449 (verify_types_in_gimple_reference): Add pieces from verify_expr.
20450 (verify_gimple_assign_single): Likewise.
20451 (verify_gimple_switch): Likewise.
20452 (verify_expr_location_1): Dereference tp once. Add (disabled)
20453 piece from verify_expr.
20454 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20456 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20458 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20460 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20462 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20463 (small_data_pattern): Likewise.
20464 (arc_rewrite_small_data): Likewise.
20465 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20466 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20467 (get_symbol_alignment): New function.
20468 (legitimate_small_data_address_p): Likewise.
20469 (legitimate_scaled_address): Update, call
20470 legitimate_small_data_address_p.
20471 (output_sdata): New static variable.
20472 (arc_print_operand): Update how we handle small data operands.
20473 (arc_print_operand_address): Likewise.
20474 (arc_legitimate_address_p): Update, use
20475 legitimate_small_data_address_p.
20476 (arc_rewrite_small_data_p): Remove.
20477 (arc_rewrite_small_data_1): Likewise.
20478 (arc_rewrite_small_data): Likewise.
20479 (small_data_pattern): Likewise.
20480 (compact_sda_memory_operand): Update to use
20481 legitimate_small_data_address_p and get_symbol_alignment.
20482 (prepare_move_operands): Don't rewite sdata pattern.
20483 (prepare_extend_operands): Remove.
20484 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20486 (zero_extendqisi2): Likewise.
20487 (zero_extendhisi2): Likewise.
20488 (extendqihi2): Likewise.
20489 (extendqisi2): Likewise.
20490 (extendhisi2): Likewise.
20491 (addsi3): Likewise.
20492 (subsi3): Likewise.
20493 (andsi3): Likewise.
20494 * config/arc/constraints.md (Usd): Change it to memory constraint.
20496 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20498 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20499 as source of std instructions.
20500 * config/arc/arc.md (movsi_insn): Update pattern predicate to
20501 allow 6-bit constants as source for store instructions.
20502 (movdi_insn): Update instruction pattern to allow 6-bit constants
20503 as source for store instructions.
20505 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
20507 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20509 2018-04-30 Nathan Sidwell <nathan@acm.org>
20510 Sandra Loosemore <sandra@codesourcery.com>
20512 * dumpfile.c (dump_open): Allow '-' for stdout.
20513 * doc/invoke.texi (Developer Options): Document dump filename
20514 determination early. Document stdin/stdout selection.
20516 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
20518 Microblaze Target: PIC data text relative
20520 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20521 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20523 * config/microblaze/microblaze.h (microblaze_constant_address_p):
20524 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20525 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20526 New addressing mode for data-text relative position indepenedent code.
20527 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20528 'ADDRESS_SYMBOLIC_TXT_REL'.
20529 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20530 (microblaze_legitimate_pic_operand): Exclude function calls from
20531 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20532 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20534 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20535 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20536 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20537 for 'address + offset'.
20538 (microblaze_expand_prologue): Add new function prologue call for
20540 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20541 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20542 table in case of TARGET_PIC_DATA_TEXT_REL.
20543 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20544 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20545 Add new macros 'UNSPEC_TEXT',
20546 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20547 + exclude function calls from 'UNSPEC_PLT' in case of data text
20549 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20550 new target hook for generating address diff vector tables in case of
20552 * doc/tm.texi : Regenerate.
20553 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20554 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20555 of addr diff vector generation.
20556 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20557 target hook definition.
20558 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20559 Add default function for generate_pic_addr_diff_vec -> flag_pic.
20560 * doc/invoke.texi (Add new pic option): Add new microblaze pic
20561 option for data text relative.
20563 2018-04-30 Richard Biener <rguenther@suse.de>
20565 * tree-chrec.h (evolution_function_is_constant_p): Remove
20567 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20569 2018-04-30 Richard Biener <rguenther@suse.de>
20572 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20574 2018-04-30 Richard Biener <rguenther@suse.de>
20576 PR tree-optimization/28364
20577 PR tree-optimization/85275
20578 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20579 copying first exit test.
20581 2018-04-28 Mark Wielaard <mark@klomp.org>
20583 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20584 dwarf_version >= 5.
20585 (dwarf_AT): Handle DW_AT_addr_base.
20586 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20588 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
20591 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
20592 (*ashl<dwi>3_doubleword_mask_1): Ditto.
20593 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
20594 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
20596 2018-04-28 Richard Biener <rguenther@suse.de>
20598 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
20599 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
20600 to reflect use. Only add interesting stmts.
20602 2018-04-27 Martin Jambor <mjambor@suse.cz>
20605 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
20606 the jump function allows for passing through aggregate values.
20608 2018-04-27 David Malcolm <dmalcolm@redhat.com>
20610 * input.h (in_system_header_at): Convert from macro to inline
20612 (from_macro_expansion_at): Likewise.
20613 (from_macro_definition_at): Likewise.
20615 2018-04-27 Jeff Law <law@redhat.com>
20617 * config.gcc: Mark tile* targets as deprecated/obsolete.
20619 2018-04-27 Richard Biener <rguenther@suse.de>
20621 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
20624 2018-04-27 Richard Biener <rguenther@suse.de>
20626 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
20628 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
20630 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
20631 with Yd constraint. Set "preferred_for_speed" attribute from
20632 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
20633 with Yd constraint.
20634 (*movdi_internal): Ditto.
20635 (movti_interunit splitters): Remove
20636 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
20637 (movdi_interunit splitters): Ditto.
20638 * config/i386/constraints.md (Ye): Remove.
20639 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
20641 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20644 * config/aarch64/constraints.md (Usg): Limit to 31.
20645 (Usj): Limit to 63.
20647 2018-04-27 Jakub Jelinek <jakub@redhat.com>
20649 PR tree-optimization/85529
20650 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
20651 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
20652 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
20653 zero extension or masking of the MSB bit.
20654 (optimize_range_tests): Add FIRST_BB argument, pass it through
20655 to optimize_range_tests_var_bound.
20656 (maybe_optimize_range_tests, reassociate_bb): Adjust
20657 optimize_range_tests callers.
20659 2018-04-26 Richard Biener <rguenther@suse.de>
20660 Jakub Jelinek <jakub@redhat.com>
20662 * cgraph.h (symbol_table): Just declare debug method here.
20663 * symtab.c (symbol_table::debug): Define.
20665 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
20667 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
20669 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
20671 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
20672 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
20673 (*movdi_internal): Substitute Yi and Yj constraint with x
20674 and Ym and Yn constraint with y constraint. Update "isa"
20675 attribute and set "preferred_for_speed" attribute from
20676 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
20677 (*movsi_internal): Ditto.
20678 (*movdf_internal): Ditto.
20679 (*movsf_internal): Ditto.
20680 (*zero_extendsidi2): Ditto.
20681 * config/i386/sse.md (vec_set<mode>_0): Ditto.
20682 (sse2_loadld): Ditto.
20683 (*vec_extract<ssevecmodelower>_0): Ditto.
20684 (*vec_extractv4si_0_zext_sse4): Ditto.
20685 (vec_concatv2di): Ditto.
20686 (*vec_dup<mode>): Ditto.
20687 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
20688 * config/i386/constraints.md (Yi): Remove.
20693 2018-04-26 Nathan Sidwell <nathan@acm.org>
20695 * dumpfile.c (dump_open): New.
20696 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
20697 (dump_finish): Detect stdio/stderr by value not name.
20699 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
20701 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
20703 2018-04-26 Tom de Vries <tom@codesourcery.com>
20706 * config/nvptx/nvptx.c (verify_neutering_jumps)
20707 (verify_neutering_labels): New function
20708 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
20710 2018-04-26 Tom de Vries <tom@codesourcery.com>
20713 * config/nvptx/nvptx.c (needs_neutering_p): New function.
20714 (nvptx_single): Use needs_neutering_p to skip over insns that do not
20717 2018-04-26 Richard Biener <rguenther@suse.de>
20718 Tom de Vries <tom@codesourcery.com>
20721 * lto-streamer-out.c (output_function): Fixup loops if required to match
20722 discovery done in the reader.
20724 2018-04-26 Richard Biener <rguenther@suse.de>
20726 PR tree-optimization/85116
20727 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
20728 have a loop exit from the single latch predecessor. Remove
20729 case of header with just condition.
20730 (ch_base::copy_headers): Exclude infinite loops from any
20732 (pass_ch::execute): Record exits.
20734 2018-04-26 Richard Biener <rguenther@suse.de>
20736 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
20737 prologue cost vector and pass it to vect_get_load_cost.
20738 (vect_get_peeling_costs_all_drs): Likewise.
20739 (vect_peeling_hash_get_lowest_cost): Likewise.
20740 (vect_enhance_data_refs_alignment): Likewise.
20742 2018-04-26 Richard Biener <rguenther@suse.de>
20744 PR middle-end/85450
20745 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
20746 checking of integer<->pointer conversions.
20747 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
20748 sign-/zero-extending pointer types.
20749 (expand_omp_for_static_chunk): Likewise.
20751 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
20752 Jean Lee <xiaoyur347@gmail.com>
20754 * config/mips/mips.c (mips_asan_shadow_offset): New function.
20755 (TARGET_ASAN_SHADOW_OFFSET): Define.
20756 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
20757 true for -fsanitize=address.
20759 2018-04-25 Mark Wielaard <mark@klomp.org>
20761 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
20764 2018-04-25 Jakub Jelinek <jakub@redhat.com>
20766 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
20767 than "alu", remove explicit "memory" and "imm_disp" attributes.
20768 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
20770 PR middle-end/85414
20771 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
20772 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
20773 gen_lowpart_no_emit.
20775 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
20778 * config/i386/i386.c (ix86_expand_builtin): Change memory
20779 operand to XI, extend p0 to Pmode.
20780 * config/i386/i386.md: Change unspec volatile and operand
20781 1 mode to XI, change operand 0 mode to P.
20783 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
20785 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
20786 GET_MODE_MASK before any checking.
20787 (nds32_can_use_bset_p): Likewise.
20788 (nds32_can_use_btgl_p): Likewise.
20790 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
20792 * config/nds32/nds32-doubleword.md: New define_split pattern for
20793 illegal register number.
20795 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
20797 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
20799 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
20801 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
20803 2018-04-25 Richard Biener <rguenther@suse.de>
20805 * lto-streamer.h (LTO_major_version): Bump to 8.
20807 2018-04-25 Jakub Jelinek <jakub@redhat.com>
20809 * BASE-VER: Set to 9.0.0.
20811 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
20813 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
20814 in __abskf2 and __powikf2.
20816 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20819 * config/aarch64/constraints.md (Usg, Usj): New constraints.
20820 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
20821 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
20822 Use the above on operand 2. Reindent.
20823 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
20825 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
20828 * common/config/i386/i386-common.c (ix86_handle_option): Don't
20830 * config/i386/i386.opt (mcet): Removed.
20831 * doc/install.texi: Remove -mcet documentation.
20832 * doc/invoke.texi: Likewise.
20834 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
20837 * doc/install.texi: Remove -mcet from bootstrap-cet.
20839 2018-04-24 Jakub Jelinek <jakub@redhat.com>
20842 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
20843 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
20847 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
20848 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
20849 containing a CONST_VECTOR.
20851 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
20853 * doc/install.texi: Update newlib dependency for nvptx.
20855 2018-04-24 Jakub Jelinek <jakub@redhat.com>
20858 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
20859 instead of INTVAL when shifting x left.
20861 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
20863 PR tree-optimization/85478
20864 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
20865 vect_grouped_store_supported for single element vectors.
20867 2018-04-24 Richard Biener <rguenther@suse.de>
20870 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
20871 load cost increase to the case of non-constant step.
20873 2018-04-24 Jakub Jelinek <jakub@redhat.com>
20876 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
20877 destination if any_malformed_asm.
20879 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
20881 PR middle-end/85496
20882 * expr.c (store_field): In the bitfield case, if the value comes from
20883 a function call and is returned in registers by means of a PARALLEL,
20884 do not change the mode of the temporary unless BLKmode and VOIDmode.
20886 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
20888 PR rtl-optimization/85423
20889 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
20890 dependencies to debug insns when the previous insn is non-debug.
20892 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
20894 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
20895 enums into a single definition.
20896 (fls): Fix predicates and printing.
20899 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
20901 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
20902 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
20903 and short u6 immediate.
20904 (check_if_valid_sleep_operand): Remove.
20905 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
20907 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20909 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
20910 flag_always_save_lp condition.
20911 * config/nds32/nds32.opt (malways-save-lp): New option.
20913 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20915 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
20916 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
20917 * config/nds32/nds32.h
20918 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
20919 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
20921 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20923 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
20924 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
20926 2018-04-22 Shiva Chen <shiva0217@gmail.com>
20927 Chung-Ju Wu <jasonwucj@gmail.com>
20929 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
20931 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
20932 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
20934 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20936 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
20938 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20940 * config/nds32/nds32-protos.h (nds32_data_alignment,
20941 nds32_local_alignment): Declare.
20942 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
20943 nds32_local_alignment): New functions.
20944 (TARGET_CONSTANT_ALIGNMENT): Define.
20945 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
20947 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20949 * config/nds32/nds32.c
20950 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
20951 (TARGET_MODES_TIEABLE_P): Likewise.
20953 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
20955 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
20956 level Ofast and Og.
20958 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
20959 Chung-Ju Wu <jasonwucj@gmail.com>
20961 * config/nds32/constants.md (unspec_volatile_element): Add enum values
20962 for unaligned access.
20963 * config/nds32/nds32-intrinsic.c: Implementation of expanding
20965 * config/nds32/nds32-intrinsic.md: Likewise.
20966 * config/nds32/nds32_intrinsic.h: Likewise.
20967 * config/nds32/nds32.h (nds32_builtins): Likewise.
20968 * config/nds32/nds32.opt (munaligned-access): New option.
20969 * config/nds32/nds32.c (nds32_asm_file_start): Display
20970 flag_unaligned_access status.
20972 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
20974 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
20975 -mno-relax is present.
20976 * config/riscv/linux.h (LINK_SPEC): Ditto.
20978 2018-04-20 Martin Sebor <msebor@redhat.com>
20981 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
20983 (gimple_fold_builtin_stxcpy_chk): Same.
20984 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
20986 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
20989 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
20990 __powikf2 when long double is IEEE 128-bit.
20992 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
20994 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
20995 step to make sure stack always aligned.
20997 2018-04-20 Carl Love <cel@us.ibm.com>
21000 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21001 size check for arg0.
21003 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
21004 Tom de Vries <tom@codesourcery.com>
21007 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21008 Emit insns for calls too.
21009 (nvptx_find_par): Always look for worker-level predecessor insn.
21010 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
21012 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21013 (nvptx_process_pars): Propagate frames for calls.
21015 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
21018 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21020 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21021 (ix86_handle_option): Don't handle OPT_mibt.
21022 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21024 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21026 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21027 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21028 (ix86_target_macros): Define __CET__ with flag_cf_protection
21029 for -fcf-protection.
21030 * config/i386/i386.c (isa2_opts): Remove -mibt.
21031 * config/i386/i386.h (TARGET_IBT): Removed.
21032 (TARGET_IBT_P): Likewise.
21033 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21034 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21035 * config/i386/i386.opt (mcet): Update help message.
21036 (mshstk): Likewise.
21038 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
21039 -mcet as an alias for -mshstk.
21041 2018-04-20 Richard Biener <rguenther@suse.de>
21043 PR middle-end/85475
21044 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21045 complexity by forcing a single use of the multiply operand.
21047 2018-04-20 Martin Jambor <mjambor@suse.cz>
21050 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21051 recursion dependency to only apply to non-clones.
21053 2018-04-20 Martin Jambor <mjambor@suse.cz>
21056 * ipa-cp.c (create_specialized_node): Check that clones of
21057 self-recursive edges exist during IPA-CP.
21059 2018-04-19 Toon Moene <toon@moene.org>
21061 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21064 2018-04-19 Jakub Jelinek <jakub@redhat.com>
21066 PR tree-optimization/85467
21067 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21068 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
21069 VECTOR_CST element to type.
21071 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21074 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21075 * config/i386/i386.md (builtin_setjmp_setup): Removed.
21076 (builtin_longjmp): Likewise.
21077 (save_stack_nonlocal): New pattern.
21078 (restore_stack_nonlocal): Likewise.
21080 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21083 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21084 Replace ASM_OUTPUT_LABEL with fprintf.
21086 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21089 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21090 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21091 * config/i386/i386-c.c (ix86_target_macros_internal): Also
21092 define __IBT__ and __SHSTK__ for -fcf-protection.
21093 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21095 (ix86_trampoline_init): Likewise.
21096 (x86_output_mi_thunk): Likewise.
21097 (ix86_notrack_prefixed_insn_p): Likewise.
21098 (ix86_option_override_internal): Don't disallow -fcf-protection.
21099 * config/i386/i386.md (rdssp<mode>): Also enable for
21101 (incssp<mode>): Likewise.
21102 (nop_endbr): Likewise.
21103 * config/i386/i386.opt (mcet): Change help message to built-in
21106 (mshstk): Likewise.
21107 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21108 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
21109 enable CET built-in functions.
21111 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
21113 * common/config/i386/i386-common.c
21114 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21115 OPTION_MASK_ISA_MOVDIRI_UNSET,
21116 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21117 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21118 * config.gcc (movdirintrin.h): New header.
21119 * config/i386/cpuid.h (bit_MOVDIRI,
21120 bit_MOVDIR64B): New bits.
21121 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21123 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21124 (VOID, PVOID, PCVOID)): New function types.
21125 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21126 __builtin_ia32_directstoreu_u64,
21127 __builtin_ia32_movdir64b): New builtins.
21128 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21129 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21131 (ix86_valid_target_attribute_inner_p): Ditto.
21132 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21133 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21134 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21135 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21136 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21137 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21138 (movdiri<mode>, movdir64b_<mode>): New.
21139 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21140 * config/i386/immintrin.h: Include movdirintrin.h.
21141 * config/i386/movdirintrin.h: New file.
21142 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21144 2018-04-19 Richard Biener <rguenther@suse.de>
21146 PR middle-end/85455
21147 * cfg.c (clear_bb_flags): When loop state says we have
21148 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21150 2018-04-19 Richard Biener <rguenther@suse.de>
21152 PR tree-optimization/84737
21153 * tree-vect-data-refs.c (vect_copy_ref_info): New function
21154 copying restrict info.
21155 (vect_setup_realignment): Use it.
21156 * tree-vectorizer.h (vect_copy_ref_info): Declare.
21157 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21158 the first DR to all generated stores.
21159 (vectorizable_load): Likewise for loads.
21161 2018-04-19 Jakub Jelinek <jakub@redhat.com>
21163 PR tree-optimization/85446
21164 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21165 the integral and pointer types to have the same precision.
21167 * doc/install.texi: Document --disable-cet being the default and
21170 2018-04-18 Martin Liska <mliska@suse.cz>
21172 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21175 2018-04-18 Martin Liska <mliska@suse.cz>
21178 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
21181 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21182 arguments if they are comparable.
21184 2018-04-18 Martin Liska <mliska@suse.cz>
21187 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
21190 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21193 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
21196 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21197 ENDBR after calling __morestack.
21199 2018-04-18 David Malcolm <dmalcolm@redhat.com>
21202 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21203 by using gcc_base_ver to generate a gcc_driver_version, and use
21204 it when generating GCC_DRIVER_NAME.
21205 * configure: Regenerate.
21207 2018-04-18 Jakub Jelinek <jakub@redhat.com>
21210 * config.gcc: Obsolete powerpc*-*-*spe*.
21212 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21215 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21216 (stabstr_D): Change type of unum from unsigned int to
21217 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
21220 2018-04-17 Jim Wilson <jimw@sifive.com>
21223 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21224 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21225 Set arg_pointer_offset after using pretend_args_size.
21227 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21229 PR rtl-optimization/85431
21230 * dse.c (record_store): Ignore zero width stores.
21233 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
21234 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21235 __builtin_stack_restore rather than after it.
21236 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21237 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21238 argument instead of virtual_dynamic_stack_rtx.
21240 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
21242 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21244 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21245 Add note to error message to explain internal mapping of overloaded
21246 built-in function name to non-overloaded built-in function name.
21247 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21250 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21253 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21254 where the inputs overlap with the output.
21256 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21258 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21259 (=v, v) alternative and explicit "memory" attribute.
21260 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
21261 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21263 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21264 "sselog1" type instead of "sselog".
21265 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21266 "sselog". Remove explicit "memory" attribute.
21267 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21268 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21270 (vec_extract_hi_v32hi): Merge all alternatives into one, use
21271 "sselog1" type instead of "sselog". Remove explicit "memory"
21273 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21274 use "sselog1" type instead of "sselog". Remove explicit "memory"
21276 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21277 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21279 (vec_extract_hi_v64qi): Merge all alternatives into one, use
21280 "sselog1" type instead of "sselog". Remove explicit "memory"
21282 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21283 use "sselog1" type instead of "sselog". Remove explicit "memory"
21287 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21289 PR middle-end/85414
21290 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21293 2018-04-17 Martin Jambor <mjambor@suse.cz>
21296 * ipa-cp.c (create_specialized_node): Call
21297 expand_all_artificial_thunks if necessary.
21299 2018-04-17 Martin Liska <mliska@suse.cz>
21302 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21303 in message, remote space in between '_G' and '('.
21305 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21308 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21309 avx512f_vmcmp<mode>3<round_saeonly_name>,
21310 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21311 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21312 avx512f_rndscale<mode><round_saeonly_name>,
21313 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21314 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21315 Use %<iptr>2 instead of %2 for -masm=intel.
21316 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21317 avx512f_vcvttss2usi<round_saeonly_name>,
21318 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21320 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21321 avx512f_vcvttsd2usi<round_saeonly_name>,
21322 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21323 Use %q1 instead of %1 for -masm=intel.
21324 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21325 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21326 of %3 for -masm=intel.
21327 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21329 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21331 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21333 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21334 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21336 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21338 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21339 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21340 %g1 and one with %0 and %1.
21341 (avx512er_vmrcp28<mode><round_saeonly_name>,
21342 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21343 %1 for -masm=intel.
21344 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21345 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21346 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21347 of %0 and %{%4%} for -masm=intel.
21348 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21349 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21350 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21351 order of %0 and %{%5%}%{z%} for -masm=intel.
21353 2018-04-17 Jan Hubicka <jh@suse.cz>
21356 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21358 2018-04-17 Martin Liska <mliska@suse.cz>
21361 * multiple_target.c (create_dispatcher_calls): Set apostrophes
21362 for target_clone error message. Make default implementation
21363 clone to be a local declaration.
21364 (separate_attrs): Add new argument and check for an empty
21366 (expand_target_clones): Handle it.
21367 (ipa_target_clone): Make redirection just for target_clones
21370 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
21371 Tom de Vries <tom@codesourcery.com>
21373 PR middle-end/84955
21374 * omp-expand.c (expand_oacc_for): Add dummy false branch for
21375 tiled basic blocks without omp continue statements.
21377 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
21380 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21381 vec_extract expression as having side effects to make sure it gets
21384 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
21387 * config/i386/i386.c (get_builtin_code_for_version): Check
21390 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
21393 * config.gcc: Support "skylake".
21394 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21396 * config/i386/i386.c (m_SKYLAKE): Define.
21397 (processor_target_table): Add "skylake".
21398 (ix86_option_override_internal): Add "skylake".
21399 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21400 PROCESSOR_CANNONLAKE.
21401 (get_builtin_code_for_version): Fix priority for
21402 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21403 PROCESSOR_SKYLAKE-AVX512.
21404 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21405 (processor_type): Add PROCESSOR_SKYLAKE.
21407 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
21408 Jason Merrill <jason@redhat.com>
21411 * convert.c (convert_to_integer_1): Use direct recursion for
21412 enumeral types and types with a precision less than the number
21413 of bits in their mode.
21415 2018-04-16 Julia Koval <julia.koval@intel.com>
21418 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21419 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21421 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
21424 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21425 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21426 and -mno-direct-move.
21428 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
21431 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21432 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21433 Ensure negative shifts result in {0}.
21435 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
21437 PR rtl-optimization/79916
21438 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21439 regs (if any) to define how to gnerate SD moves when LRA is in
21442 2018-04-13 Jakub Jelinek <jakub@redhat.com>
21444 PR rtl-optimization/85393
21445 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21446 * except.c (expand_dw2_landing_pad_for_region): Make static.
21447 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21448 a label and unconditional jump to old_bb, rather than
21449 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21452 PR rtl-optimization/85376
21453 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21454 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21455 instead of a specific value.
21457 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
21458 Bin Cheng <bin.cheng@arm.com>
21460 PR tree-optimization/82965
21461 PR tree-optimization/83991
21462 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21463 by_profile_only parameter.
21464 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21465 information if the loop was predicted to iterate too many times.
21466 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21468 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
21471 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21474 2018-04-13 Martin Liska <mliska@suse.cz>
21475 Jakub Jelinek <jakub@redhat.com>
21477 PR middle-end/81657
21478 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21479 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21480 * builtins.c (expand_builtin_memory_copy_args): Use
21481 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21482 handle dest_addr == pc_rtx.
21484 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
21487 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21488 asked to not generate direct moves.
21489 (fix_trunc<mode>si2_stfiwx): Similar.
21490 (fix_trunc<mode>si2_internal): Similar.
21492 2018-04-12 Jakub Jelinek <jakub@redhat.com>
21495 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21496 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21497 lookup if dest in some wider mode is known to be const0_rtx and
21498 if so, record permanent equivalence for it to be ZERO_EXTEND of
21499 the narrower mode destination.
21501 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
21503 * lto-streamer-out.c (output_function): Revert 259346.
21504 * omp-expand.c (expand_oacc_for): Likewise.
21506 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
21508 PR rtl-optimization/85354
21509 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21510 * sel-sched.c (sel_global_init): ... here.
21512 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
21515 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21516 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21517 mode for PE-COFF targets.
21518 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21519 (i386_pe_asm_lto_end): Likewise.
21520 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21521 (TARGET_ASM_LTO_END): Likewise.
21522 * config/i386/winnt.c (saved_debug_info_level): New static variable.
21523 (i386_pe_asm_lto_start): New function.
21524 (i386_pe_asm_lto_end): Likewise.
21526 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
21527 Richard Biener <rguenther@suse.de>
21529 PR middle-end/84955
21530 * lto-streamer-out.c (output_function): Fix CFG loop state before
21532 * omp-expand.c (expand_oacc_for): Handle calls to internal
21533 functions like regular functions.
21535 2018-04-12 Richard Biener <rguenther@suse.de>
21538 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21539 for the early LTO debug to properly generate references to it
21540 during DIE emission. Do not re-use that for the skeleton for
21542 (dwarf2out_early_finish): Likewise.
21544 2018-04-12 Jakub Jelinek <jakub@redhat.com>
21547 * config/i386/sse.md
21548 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21549 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21550 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21551 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21552 and output is a reg, avoid creating invalid lowpart subreg, but
21553 instead split into a 512-bit move. Don't split if not AVX512VL,
21554 input is xmm16+ reg and output is a mem.
21555 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21556 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21557 xmm16+ reg and output is a mem.
21559 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21561 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21562 also for flag_dwarf2_cfi_asm.
21564 2018-04-12 Jakub Jelinek <jakub@redhat.com>
21566 PR rtl-optimization/85342
21567 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21568 a bool scalar var inside of the loop instead. Don't try to update
21569 recog_data.operand after failed apply_change_group.
21571 2018-04-12 Tom de Vries <tom@codesourcery.com>
21574 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21575 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
21576 array with flexible array member as array without given dimension.
21577 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21578 argument for undefined param to true.
21580 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
21583 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21584 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21585 from PowerPC section.
21586 * config/rs6000/sysv4.opt (mcall-): Improve help text.
21587 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21588 help text that is too long.
21589 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
21590 help text that is too long.
21591 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
21592 help text that is too long.
21594 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
21596 * config/alpha/alpha.md (stack_probe_internal): Rename
21597 from "probe_stack". Update all callers.
21599 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
21601 PR rtl-optimization/84566
21602 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
21603 sched_macro_fuse_insns.
21605 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
21608 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
21609 (compute_block_dependences): ... from here.
21611 2018-04-11 Jakub Jelinek <jakub@redhat.com>
21613 PR tree-optimization/85331
21614 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
21615 from int to HOST_WIDE_INT.
21617 2018-04-11 Martin Jambor <mjambor@suse.cz>
21620 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
21621 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
21622 not the same as the source val.
21623 (cgraph_edge_brings_value_p): New parameter.
21624 (gather_edges_for_value): Pass destination value to
21625 cgraph_edge_brings_value_p.
21626 (perhaps_add_new_callers): Likewise.
21627 (get_info_about_necessary_edges): Likewise and exclude values brought
21628 only by self-recursive edges.
21629 (create_specialized_node): Redirect only clones of self-calling edges.
21630 (+self_recursive_pass_through_p): New function.
21631 (find_more_scalar_values_for_callers_subset): Use it.
21632 (find_aggregate_values_for_callers_subset): Likewise.
21633 (known_aggs_to_agg_replacement_list): Removed.
21634 (decide_whether_version_node): Re-calculate known constants for all
21635 remaining context clones.
21637 2018-04-11 Richard Biener <rguenther@suse.de>
21640 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
21641 from early DWARF output.
21642 (dwarf2out_early_finish): Output line info unconditionally into
21643 early DWARF and add reference to it.
21645 2018-04-11 Jakub Jelinek <jakub@redhat.com>
21648 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
21649 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
21650 other than V2DFmode using iptr mode attribute.
21651 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
21653 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
21655 PR rtl-optimization/84659
21656 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
21658 2018-04-11 Jakub Jelinek <jakub@redhat.com>
21661 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
21663 (output_loc_list): Pass address of a dummy size variable even in the
21664 locview handling loop.
21665 (index_location_lists): Add comment on why skip_loc_list_entry can't
21668 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
21671 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
21674 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
21677 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21678 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
21679 and -mstring-compare-inline-limit.
21681 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
21684 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
21685 for stack clash protection in a register whenever we need it to be in
21688 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
21690 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
21691 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
21693 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
21696 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
21698 (mlong-double-): Ditto.
21699 * config/rs6000/sysv4.opt (msdata=): Ditto.
21700 (mtls-size=): Ditto.
21702 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
21704 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
21705 erroneous entries for
21706 "vector int vec_ldl (int, long int *)", and
21707 "vector unsigned int vec_ldl (int, unsigned long int *)".
21708 Add comments and entries for
21709 "vector bool char vec_ldl (int, bool char *)",
21710 "vector bool short vec_ldl (int, bool short *)",
21711 "vector bool int vec_ldl (int, bool int *)",
21712 "vector bool long long vec_ldl (int, bool long long *)",
21713 "vector pixel vec_ldl (int, pixel *)",
21714 "vector long long vec_ldl (int, long long *)",
21715 "vector unsigned long long vec_ldl (int, unsigned long long *)".
21716 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
21717 type tree bool_long_long_type_node and correct definition of
21718 bool_V2DI_type_node to make reference to this new type tree.
21719 (rs6000_mangle_type): Replace erroneous reference to
21720 bool_long_type_node with bool_long_long_type_node.
21721 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
21722 comments to emphasize sign distinctions for char and int types and
21723 replace RS6000_BTI_bool_long constant with
21724 RS6000_BTI_bool_long_long constant. Also add comment to restrict
21725 use of RS6000_BTI_pixel.
21726 (bool_long_type_node): Remove this macro definition.
21727 (bool_long_long_type_node): New macro definition
21729 2018-04-10 Jakub Jelinek <jakub@redhat.com>
21731 PR rtl-optimization/85300
21732 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
21733 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
21734 simplify_unary_operation fails.
21736 2018-04-10 Martin Liska <mliska@suse.cz>
21738 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
21739 cgraph_edge and ipa_ref.
21741 2018-04-10 Jakub Jelinek <jakub@redhat.com>
21745 * config/i386/sse.md
21746 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
21747 computation of the VEC_MERGE selector from mask.
21748 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
21749 Fix decoding of the VEC_MERGE selector into mask.
21751 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
21753 PR tree-optimization/85286
21754 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
21756 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
21758 * final.c (final_1): Set insn_last_address as well as
21759 insn_current_address.
21761 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21764 * explow.c (emit_stack_probe): Call validize_mem on memory location
21765 before passing it to gen_probe_stack. Create address operand and
21766 legitimize it for the probe_stack_address case.
21768 2018-04-09 Jan Hubicka <jh@suse.cz>
21771 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
21772 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
21773 * tree.c (free_lang_data_in_type): Fix handling of binfos;
21775 (free_lang_data): Rebuild type inheritance graph.
21777 2018-04-09 Martin Sebor <msebor@redhat.com>
21779 * invoke.texi (-finline-small-functions): Mention other optimization
21781 (-findirect-inlining, -fpartial-inlining): Same.
21782 (-finline-functions-called-once): Same.
21783 (-freorder-blocks-and-partition): Same.
21785 2018-04-09 Jan Hubicka <jh@suse.cz>
21788 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
21789 jumps; choose last target that matches the criteria (i.e.
21790 no partition changes for non-crossing jumps).
21791 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
21792 support for redirecting crossing jumps to non-crossing.
21794 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
21796 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
21797 also for naked functions.
21799 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
21801 * config/arc/arc.md (add_shift): New pattern.
21802 (add_shift2): Likewise.
21803 (sub_shift): Likewise.
21804 (sub_shift_cmp0_noout): Likewise.
21805 (compare_si_ashiftsi): Likewise.
21806 (xbfu_cmp0_noout): New combine pattern.
21807 (xbfu_cmp0"): Likewise.
21808 (movsi_set_cc_insn): Place the predicable variant first.
21809 (commutative_binary_cmp0_noout): Remove clobber.
21810 (commutative_binary_cmp0): New pattern.
21811 (noncommutative_binary_cmp0): Likewise.
21812 (noncommutative_binary_cmp0_noout): Likewise.
21813 (noncommutative_binary_comparison_result_used): Removed.
21814 (rsub_cmp0): New pattern.
21815 (rsub_cmp0_noout): Likewise.
21816 (extzvsi): Changed, keep only meaningful variants.
21817 (SQH, SEZ): New iterators.
21818 (SQH_postfix): New mode attribute.
21819 (SEZ_prefix): New code attribute.
21820 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
21821 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
21822 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
21823 of numerical value.
21824 (noncommutative_operator): Check the availability of barrel
21827 2018-04-09 Richard Biener <rguenther@suse.de>
21829 PR tree-optimization/85284
21830 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
21831 Only use the niter constraining form of simple_iv when the exit
21832 is always executed.
21834 2018-04-09 Tom de Vries <tom@codesourcery.com>
21837 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
21838 (define_expand "*memory_barrier"): New define_expand.
21839 (define_insn "memory_barrier"): New insn.
21841 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
21843 PR rtl-optimization/80463
21844 PR rtl-optimization/83972
21845 PR rtl-optimization/83480
21847 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
21848 correct producer for the insn.
21849 (tidy_control_flow): Fixup seqnos in case of debug insns.
21851 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
21853 PR rtl-optimization/83913
21855 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
21856 different sched-times when merging exprs.
21858 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
21860 PR rtl-optimization/83962
21862 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
21863 tidy_fallthru_edge and tidy_control_flow.
21865 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
21867 PR rtl-optimization/83530
21869 * sel-sched.c (force_next_insn): New global variable.
21870 (remove_insn_for_debug): When force_next_insn is true, also leave only
21871 next insn in the ready list.
21872 (sel_sched_region): When the region wasn't scheduled, make another pass
21873 over it with force_next_insn set to 1.
21875 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
21877 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
21879 * config/nds32/constants.md (unspec_volatile_element): Add enum values
21880 for interrupt control.
21881 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21882 functions for interrupt control.
21883 * config/nds32/nds32-intrinsic.md: Likewise.
21884 * config/nds32/nds32_intrinsic.h: Likewise.
21885 * config/nds32/nds32.h (nds32_builtins): Likewise.
21887 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
21889 * config/nds32/nds32.c (nds32_init_machine_status,
21890 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
21891 strict_aligned_p field.
21892 (nds32_expand_to_rtl_hook): New function.
21893 (TARGET_EXPAND_TO_RTL_HOOK): Define.
21894 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
21896 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21897 Chung-Ju Wu <jasonwucj@gmail.com>
21899 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
21900 * config/nds32/nds32-n7.md: New file.
21901 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
21902 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
21904 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
21905 * config/nds32/nds32.md (pipeline_model): Add n7.
21906 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
21907 * config/nds32/pipelines.md: Include n7 settings.
21909 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21910 Chung-Ju Wu <jasonwucj@gmail.com>
21912 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
21913 * config/nds32/nds32-e8.md: New file.
21914 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
21915 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
21917 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
21918 * config/nds32/nds32.md (pipeline_model): Add e8.
21919 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
21920 * config/nds32/pipelines.md: Include e8 settings.
21922 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21923 Chung-Ju Wu <jasonwucj@gmail.com>
21925 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
21926 * config/nds32/nds32-n8.md: New file.
21927 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
21928 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
21930 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
21931 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
21932 * config/nds32/nds32.md (pipeline_model): Add n8.
21933 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
21934 * config/nds32/pipelines.md: Include n8 settings.
21936 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
21937 Chung-Ju Wu <jasonwucj@gmail.com>
21939 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
21940 * config/nds32/nds32-n9-2r1w.md: New file.
21941 * config/nds32/nds32-n9-3r2w.md: New file.
21942 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
21943 nds32_register_ports): New or modify for cpu n9.
21944 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
21946 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
21947 * config/nds32/nds32-utils.c: New file.
21948 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
21949 TARGET_MUL_SLOW): Define.
21950 * config/nds32/nds32.md (pipeline_model): New attribute.
21951 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
21952 New options that support cpu n9.
21953 * config/nds32/pipelines.md: Include n9 settings.
21954 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
21956 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
21958 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
21959 information if necessary.
21960 (output_cond_branch_compare_zero): Likewise.
21961 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
21962 (nds32_target_alignment): Refine for alignment.
21963 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
21964 (FUNCTION_BOUNDARY): Modify.
21965 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
21967 * config/nds32/nds32.opt (malways-align, malign-functions): New.
21969 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
21971 * config/nds32/constants.md (unspec_volatile_element): Add values for
21972 TLB operation and data prefetch.
21973 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
21974 functions for TLB operation and data prefetch.
21975 * config/nds32/nds32-intrinsic.md: Likewise.
21976 * config/nds32/nds32_intrinsic.h: Likewise.
21977 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
21978 (nds32_print_operand): Likewise.
21979 * config/nds32/nds32.h (nds32_builtins): Likewise.
21981 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
21982 Andrew Pinski <pinsika@gcc.gnu.org>
21984 PR middle-end/82976
21985 * match.pd: Use constant_boolean_node of correct type instead of
21986 boolean_true_node or boolean_false_node for simplifying
21987 pointer comparisons to zero.
21989 2018-04-07 Jakub Jelinek <jakub@redhat.com>
21991 PR tree-optimization/80021
21992 * tree.c (verify_type_variant): Make error call in verify_variant_match
21993 translatable and remove final full stop.
21995 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
21997 * config/nds32/constants.md (unspec_volatile_element): Add
21998 UNSPEC_VOLATILE_EH_RETURN.
21999 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22000 nds32_output_stack_pop): Support dwarf exception handling process.
22001 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22002 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22003 exception handling process.
22004 (nds32_compute_stack_frame): Likewise.
22005 (nds32_return_addr_rtx): Likewise.
22006 (nds32_initial_elimination_offset): Likewise.
22007 (nds32_expand_prologue): Likewise.
22008 (nds32_expand_epilogue): Likewise.
22009 (nds32_dynamic_chain_address): New function.
22010 * config/nds32/nds32.h (machine_function): Add fields for dwarf
22011 exception handling.
22012 (DYNAMIC_CHAIN_ADDRESS): Define.
22013 (EH_RETURN_DATA_REGNO): Define.
22014 (EH_RETURN_STACKADJ_RTX): Define.
22015 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22016 patterns for dwarf exception handling.
22018 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22020 * config/nds32/nds32.h: Clean up obsolete macros.
22022 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22024 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22025 Add enum values for particular instructions.
22026 * config/nds32/nds32-intrinsic.c: Implementation of expanding
22027 particular intrinsic functions.
22028 * config/nds32/nds32-intrinsic.md: Likewise.
22029 * config/nds32/nds32_intrinsic.h: Likewise.
22030 * config/nds32/nds32.h (nds32_builtins): Likewise.
22031 * config/nds32/nds32.md (type): Add pbsad and pbsada.
22032 (btst, ave): New patterns for particular instructions.
22034 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22036 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22037 Add enum values for atomic load/store and memory sync.
22038 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22040 * config/nds32/nds32-intrinsic.md: Likewise.
22041 * config/nds32/nds32_intrinsic.h: Likewise.
22042 * config/nds32/nds32.h (nds32_builtins): Likewise.
22044 2018-04-07 Jakub Jelinek <jakub@redhat.com>
22046 PR tree-optimization/85257
22047 * fold-const.c (native_encode_vector): If not all elts could fit
22048 and off is -1, return 0 rather than offset.
22049 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22050 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22051 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
22052 adjust buffer in native_interpret_expr call.
22054 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22056 * config/nds32/constants.md (unspec_volatile_element): Add cache
22057 control enum values.
22058 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22059 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22060 * config/nds32/nds32.c (nds32_cctl_names): New.
22061 (nds32_print_operand): Handle cache control register names.
22062 * config/nds32/nds32.h (nds32_builtins): New enum values.
22063 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22065 * config/nds32/nds32.md (type): Add mmu.
22066 * config/nds32/pipelines.md (simple_insn): Add mmu.
22068 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22070 * config/nds32/nds32.md (type): Remove call.
22071 * config/nds32/pipelines.md (simple_insn): Likewise.
22073 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22075 * config/nds32/constants.md (unspec_volatile_element): Add
22076 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22077 UNSPEC_VOLATILE_FMFCFG.
22078 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22079 description for fmfcfg and fmfcsr.
22080 (bdesc_1arg): Add fmtcsr.
22081 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22082 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22083 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22084 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22085 unspec_fmfcfg): New patterns.
22086 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22087 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22088 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22089 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22090 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22091 __nds32__fmfcfg): Define.
22093 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22095 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22096 intrinsic register names.
22097 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22098 intrinsic register enum values and macros.
22100 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22102 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22103 for load/store addressing form.
22104 (nds32_print_operand_address): Likewise.
22106 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
22109 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22110 based on LABEL_REF. Remove useless assertion.
22111 (pic_address_needs_scratch): Fix formatting.
22112 (sparc_legitimize_pic_address): Minor tweaks.
22113 (sparc_delegitimize_address): Adjust assertion accordingly.
22114 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22115 into symbolic_operand.
22116 (movsi_high_pic_label_ref): Likewise.
22117 (movsi_lo_sum_pic_label_ref): Likewise.
22118 (movdi_pic_label_ref): Likewise.
22119 (movdi_high_pic_label_ref): Likewise.
22120 (movdi_lo_sum_pic_label_ref): Likewise.
22122 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
22124 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22125 custom LIB_SPEC setup.
22127 2018-04-06 Ruslan Bukin <br@bsdpad.com>
22128 Kito Cheng <kito.cheng@gmail.com>
22130 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22131 * config/riscv/freebsd.h: New.
22133 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22135 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22136 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22139 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22140 Kito Cheng <kito.cheng@gmail.com>
22142 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22143 nds32_output_call, nds32_symbol_binds_local_p): New functions.
22144 * config/nds32/nds32-protos.h (nds32_output_call,
22145 nds32_output_return): Declare.
22146 * config/nds32/nds32.md: Refine all the call and return patterns.
22148 2018-04-06 Jakub Jelinek <jakub@redhat.com>
22151 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22152 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22154 PR rtl-optimization/84872
22155 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22156 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22157 EDGE_CROSSING edge.
22159 2018-04-06 Tamar Christina <tamar.christina@arm.com>
22161 * expr.c (copy_blkmode_to_reg): Revert 254862.
22162 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22164 2018-04-06 Richard Biener <rguenther@suse.de>
22166 PR middle-end/85244
22167 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22168 after seeing a component reference with an adjacent field. Treat
22169 refs to arrays at struct end of external decls similar to
22170 refs to unconstrained commons.
22172 2018-04-06 Jakub Jelinek <jakub@redhat.com>
22175 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22176 look through SAVE_EXPRs with non-side-effects argument. Adjust
22178 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22181 2018-04-06 Richard Biener <rguenther@suse.de>
22183 PR middle-end/85180
22184 * alias.c (find_base_term): New wrapper around find_base_term
22185 unwinding CSELIB_VAL_PTR changes.
22186 (find_base_term): Do not restore CSELIB_VAL_PTR during the
22189 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22191 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22193 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22194 constant definitions.
22195 ("nop"): lr 0,0 -> nopr r0
22196 ("nop_lr0", "nop_lr1"): New insn definitions.
22198 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22200 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22201 NDS32_V3PUSH_AVAILABLE_P macro.
22203 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
22204 Chung-Ju Wu <jasonwucj@gmail.com>
22206 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22207 (nds32*-*-*): Add float and fpu_config into supported_defaults.
22208 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22209 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22210 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22211 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22212 * config/nds32/constraints.md: New constraints and checking for hard
22213 float configuration.
22214 * config/nds32/iterators.md: New mode iterator and attribute for hard
22215 float configuration.
22216 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22218 * config/nds32/nds32-fpu.md: New file.
22219 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22220 deal with hard float code generation.
22221 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22223 (abi_type, float_reg_number): New enum type.
22224 * config/nds32/nds32-predicates.c: New predicates for hard float.
22225 * config/nds32/nds32-protos.h: Declare functions for hard float.
22226 * config/nds32/nds32.c: Implementation for hard float configuration.
22227 * config/nds32/nds32.h: Definitions for hard float configuration.
22228 * config/nds32/nds32.md: Include hard float machine description and
22229 modify patterns for hard float configuration.
22230 * config/nds32/nds32.opt: New options for hard float configuration.
22231 * config/nds32/predicates.md: New predicates for hard float
22234 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22236 * common/config/nds32/nds32-common.c
22237 (nds32_option_optimization_table): Enable -mreleax-hint by default.
22239 2018-04-05 Jakub Jelinek <jakub@redhat.com>
22241 PR middle-end/85195
22242 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22243 CONSTRUCTOR_ELT (ctor, ...)->value.
22245 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
22248 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22250 2018-04-05 Tom de Vries <tom@codesourcery.com>
22253 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22256 2018-04-05 Shiva Chen <shiva0217@gmail.com>
22257 Kito Cheng <kito.cheng@gmail.com>
22259 * config/nds32/constraints.md (U33): Fine-tune checking condition.
22260 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22261 * config/nds32/nds32.h (nds32_16bit_address_type): Add
22262 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22264 2018-04-05 Shiva Chen <shiva0217@gmail.com>
22265 Kito Cheng <kito.cheng@gmail.com>
22267 * config/nds32/constraints.md (Ufe): New memory constraint.
22268 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22269 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22270 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22272 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22273 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22275 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22277 * config/nds32/nds32.md: Use optimize_size in the condition for
22278 alu-shift instructions.
22280 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22282 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22284 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22286 * config/nds32/nds32.md (negsi2): Refine pattern.
22288 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
22289 Chung-Ju Wu <jasonwucj@gmail.com>
22291 * config/nds32/iterators.md (shift_rotate): New code iterator.
22292 (shift): New code attribute.
22293 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22294 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22295 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22296 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22297 bit-wise operations.
22298 (andsi3, *andsi3): Ditto.
22299 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22300 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22301 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22302 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22303 nds32_ior_operand, nds32_xor_operand): New predicates.
22305 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22307 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22308 (addsi3, subsi3): ... this.
22310 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22312 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22314 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22316 * config/nds32/nds32.md: Adjust indention.
22318 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
22320 * config/nds32/nds32.md (feature): New attribute.
22322 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22324 * config/nds32/nds32.md (subtype): New attribute.
22326 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
22329 * config/arm/arm-builtins.c (arm_expand_builtin): Change
22330 expansion to perform a bitwise AND of the argument followed by a
22331 boolean negation of the result.
22333 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
22335 PR rtl-optimization/84878
22336 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22337 the basic block. Assert the use reference is not artificial and that
22338 it has an associated insn.
22340 2018-04-04 Michael Matz <matz@suse.de>
22342 * builtins.c (compute_objsize): Pass correct operand
22343 to array_at_struct_end_p.
22345 2018-04-04 Richard Biener <rguenther@suse.de>
22348 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22349 from contexts for DINFO_LEVEL_TERSE and below.
22351 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22353 * config/nds32/nds32-doubleword.md (move_<mode>): Require
22354 resiter_operand condition.
22355 * config/nds32/nds32.md (*move<mode>): Ditto.
22357 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22358 Monk Chiang <sh.chiang04@gmail.com>
22360 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22362 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22364 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22366 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22367 Kito Cheng <kito.cheng@gmail.com>
22369 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22370 nds32_cond_code_str, output_cond_branch,
22371 output_cond_branch_compare_zero, nds32_expand_cbranch,
22372 nds32_expand_cstore, nds32_expand_movcc,
22373 nds32_output_cbranchsi4_equality_zero,
22374 nds32_output_cbranchsi4_equality_reg,
22375 nds32_output_cbranchsi4_equality_reg_or_const_int,
22376 nds32_output_cbranchsi4_greater_less_zero: New functions.
22377 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22378 nds32_expand_cstore, nds32_expand_movcc,
22379 nds32_output_cbranchsi4_equality_zero,
22380 nds32_output_cbranchsi4_equality_reg,
22381 nds32_output_cbranchsi4_equality_reg_or_const_int,
22382 nds32_output_cbranchsi4_greater_less_zero): Declare.
22383 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22384 nds32_rimm11s_operand): New predicates.
22385 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22386 * config/nds32/nds32.md: Rewrite all the branch and conditional move
22389 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22391 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22392 * config/nds32/nds32.md: Ditto.
22393 * config/nds32/pipelines.md: Ditto.
22395 2018-04-04 Richard Biener <rguenther@suse.de>
22397 PR tree-optimization/85168
22398 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22399 propagating abnormals.
22401 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22403 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22405 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22406 Kito Cheng <kito.cheng@gmail.com>
22408 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22409 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22410 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22411 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22412 * config/nds32/nds32.md (sibcall_internal): New.
22413 (sibcall_register): Remove.
22414 (sibcall_immediate): Remove.
22415 (sibcall_value_internal): New.
22416 (sibcall_value_register): Remove.
22417 (sibcall_value_immediate): Remove.
22418 * config/nds32/predicates.md (nds32_general_register_operand): New.
22419 (nds32_call_address_operand): New.
22421 2018-04-03 Jakub Jelinek <jakub@redhat.com>
22423 PR rtl-optimization/85167
22424 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22425 bb_defs if *split_p, instead preinitialize it to NULL.
22427 PR tree-optimization/85156
22428 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22429 evaluating the argument multiple times.
22431 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
22433 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22435 (_mm_cvtpd_ps): Likewise.
22436 (_mm_cvttpd_epi32): Likewise.
22437 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22438 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22439 vector, pixel, and bool following altivec.h include.
22441 2018-04-03 Martin Sebor <msebor@redhat.com>
22443 * doc/extend.texi (Common Function Attributes): Clarify.
22444 (const attribute): Likewise.
22445 (pure attribute): Likewise.
22447 2018-04-03 Jakub Jelinek <jakub@redhat.com>
22450 * config/i386/i386.c (ix86_expand_vector_set): Use
22451 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
22453 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
22455 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22456 instructions when changing rounding bits to preserve precision bits
22457 in the x87 control word.
22459 2018-04-03 Martin Liska <mliska@suse.cz>
22461 PR tree-optimization/82491
22462 * rtl.h (strip_offset_and_add): Replace += suboffset with
22463 poly_uint64 () + suboffset.
22465 2018-03-29 Martin Liska <mliska@suse.cz>
22466 Martin Jambor <mjambor@suse.cz>
22469 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22470 param_type is not an integral or pointer type.
22472 2018-04-03 Richard Biener <rguenther@suse.de>
22474 * sese.h (recompute_all_dominators): Remove.
22476 2018-04-02 Martin Sebor <msebor@redhat.com>
22478 * doc/invoke.texi (-Wrestrict): Fix typos.
22480 2018-04-02 Jim Wilson <jimw@sifive.com>
22482 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22483 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22484 (<optab>di3, <optab>si3_extend): Likewise.
22485 (<optab>si3_mask, <optab>si3_mask_1): New.
22486 (<optab>di3_mask, <optab>di3_mask_1): New.
22487 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22488 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22489 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22491 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
22493 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22496 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
22498 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22499 (nds32_canonicalize_comparison): New function.
22501 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
22502 Kito Cheng <kito.cheng@gmail.com>
22503 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22505 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22506 * config/nds32/constants.md (unspec_volatile_element): Add
22507 UNSPEC_VOLATILE_RELAX_GROUP.
22508 * config/nds32/nds32-relax-opt.c: New file.
22509 * config/nds32/nds32-predicates.c
22510 (nds32_symbol_load_store_p): New function.
22511 * config/nds32/nds32-protos.h
22512 (nds32_symbol_load_store_p): Declare function.
22513 (make_pass_nds32_relax_opt): Declare new rtl pass function.
22514 * config/nds32/nds32.c
22515 (nds32_register_pass): New function to register pass.
22516 (nds32_register_passes): New function to register passes.
22517 * config/nds32/nds32.md (relax_group): New pattern.
22518 * config/nds32/nds32.opt (mrelax-hint): New option.
22519 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22521 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
22523 * config/nds32/t-nds32: Modify files dependency.
22525 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
22527 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22528 (PROFILE_HOOK): Define its implementation.
22530 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
22532 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22533 type and 32-bit size.
22535 2018-04-01 Jakub Jelinek <jakub@redhat.com>
22537 PR middle-end/85090
22538 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22539 (V_128_256): New mode iterator.
22540 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22541 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22542 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22544 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22545 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22547 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
22550 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22551 NaN inputs correctly.
22553 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
22556 * config/rs6000/vsx.md (??r): New mode attribute.
22557 (*vsx_mov<mode>_64bit): Use it.
22558 (*vsx_mov<mode>_32bit): Likewise.
22560 2018-03-30 Martin Sebor <msebor@redhat.com>
22562 PR tree-optimization/84818
22563 * builtins.c (check_access): Use warning_n.
22565 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22568 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22570 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22573 2018-03-30 Julia Koval <julia.koval@intel.com>
22576 * x86-tune.def (movx, partial_reg_dependency): Enable for
22579 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
22581 PR inline-asm/84985
22582 * lra-constraints.c (process_alt_operands): Move setting
22583 this_alternative_matches below.
22585 2018-03-29 Martin Liska <mliska@suse.cz>
22588 * doc/invoke.texi: Document how LTO works with debug info.
22589 Describe auto-load support of binutils. Mention 'x86-64'
22590 as valid option value of -march option.
22592 2018-03-29 Jakub Jelinek <jakub@redhat.com>
22594 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
22597 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
22598 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
22599 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
22602 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
22605 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
22606 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
22607 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
22608 for RS6000_BTM_POWERPC64.
22609 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
22610 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
22611 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
22614 (DIVDEU): Likewise.
22616 2018-03-28 Carl Love <cel@us.ibm.com>
22619 2017-09-27 Carl Love <cel@us.ibm.com>
22621 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
22622 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
22623 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
22626 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22628 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
22629 instead of __vector bool.
22630 (_mm_max_pu8): Likewise.
22631 (_mm_min_pi16): Likewise.
22633 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
22636 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
22637 (DIVWEUO): Likewise.
22638 (DIVDEO): Likewise.
22639 (DIVDEUO): Likewise.
22640 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
22641 DIVWEUO and DIVDEUO.
22642 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
22643 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
22644 (div_extend): Likewise.
22645 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
22647 (__builtin_divweuo): Likewise.
22648 (__builtin_divdeo): Likewise.
22649 (__builtin_divdeuo): Likewise.
22651 2018-03-28 Jakub Jelinek <jakub@redhat.com>
22654 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
22655 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
22657 PR tree-optimization/82004
22658 * gimple-match-head.c (optimize_pow_to_exp): New function.
22659 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
22660 Don't fold to exp if optimize_pow_to_exp is false.
22662 2018-03-28 Martin Liska <mliska@suse.cz>
22665 * calls.c (initialize_argument_information): Fix trailing space.
22666 * common.opt: Fix typo and provide better explanation for
22667 -fsanitize-coverage option.
22668 * config/i386/i386.opt: Fix typo.
22670 2018-03-28 Jakub Jelinek <jakub@redhat.com>
22671 Martin Liska <mliska@suse.cz>
22674 * gimplify.c (asan_poison_variable): Don't do the check for
22675 gimplify_omp_ctxp here.
22676 (gimplify_decl_expr): Do it here.
22677 (gimplify_target_expr): Likewise.
22679 2018-03-28 Martin Liska <mliska@suse.cz>
22682 * config/i386/i386.c (ix86_function_arg_advance): Do not call
22683 chkp_type_bounds_count if MPX is not enabled.
22685 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
22687 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
22689 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
22692 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
22693 function to create the function decl for complex long double
22694 multiply and divide for -mabi=ieeelongdouble.
22695 (init_float128_ieee): Call it.
22697 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
22700 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
22701 -fcf-protection=branch -mibt.
22702 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
22704 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
22707 * config/arm/arm.c (arm_valid_symbolic_address): Handle
22708 arm_word_relocations.
22710 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
22713 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
22714 extern array declarations.
22716 2018-03-27 Richard Biener <rguenther@suse.de>
22718 PR middle-end/84067
22719 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
22720 explicit single_use checks.
22722 2018-03-27 Richard Biener <rguenther@suse.de>
22724 PR tree-optimization/85082
22725 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
22728 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
22730 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
22731 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
22732 Turn on fasynchronous-unwind-tables and funwind-tables.
22734 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
22737 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
22738 (*bmi_blsr_<mode>_ccz): Ditto.
22740 2018-03-26 Tom de Vries <tom@codesourcery.com>
22742 PR tree-optimization/85063
22743 * omp-general.c (offloading_function_p): New function. Factor out
22745 * omp-offload.c (pass_omp_target_link::gate): ... here.
22746 * omp-general.h (offloading_function_p): Declare.
22747 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
22748 with attribute omp declare target for offloading functions.
22750 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
22752 PR tree-optimization/84005
22753 * tree-data-ref.h (get_base_for_alignment): Declare.
22754 * tree-data-ref.c (get_base_for_alignment_1): New function.
22755 (get_base_for_alignment): Likewise.
22756 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
22757 get_base_for_alignment to find a suitable base object, instead
22758 of always using drb->base_address.
22760 2018-03-23 Jakub Jelinek <jakub@redhat.com>
22762 PR inline-asm/85022
22763 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
22764 known size by default.
22766 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
22768 PR inline-asm/85030
22769 * lra-constraints.c (process_alt_operands): Don't match BLKmode
22770 and non BLKmode operands.
22772 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22775 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
22776 Clean up attributes.
22778 2018-03-23 Richard Biener <rguenther@suse.de>
22781 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
22782 we are going to emit early debug for LTO.
22784 2018-03-23 Jakub Jelinek <jakub@redhat.com>
22786 PR inline-asm/85034
22787 * function.c (match_asm_constraints_1): Don't optimize if input
22788 doesn't satisfy general_operand predicate for output's mode.
22790 PR inline-asm/85022
22791 * alias.c (write_dependence_p): Don't require for x_canonicalized
22792 non-VOIDmode if x has VOIDmode.
22795 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
22796 just don't try to optimize it rather than assert it never happens.
22798 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
22800 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
22801 macro expansions for definition of ST_INTERNAL_<mode> and
22802 LD_INTERNAL_<mode> builtins.
22803 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
22805 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
22807 (altivec_expand_st_builtin): Likewise.
22808 (altivec_expand_builtin): Remove calls to deleted functions.
22809 (rs6000_address_for_altivec): Delete this function.
22810 * config/rs6000/vector.md: Remove expands for
22811 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
22813 2018-03-22 Sudakshina Das <sudi.das@arm.com>
22816 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
22817 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
22818 re-computing once computed.
22819 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
22820 (arm_init_machine_status): Initialize
22821 machine->static_chain_stack_bytes.
22823 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
22826 * doc/extend.texi: Add four new prototypes for vec_ld.
22827 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
22828 definitions for more logical presentation.
22829 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
22830 entries for V1TI variants of __builtin_altivec_ld builtin.
22831 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
22832 handling of V1TI variant of LVX icode pattern.
22833 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
22834 (rs6000_gimple_fold_builtin): Likewise.
22835 (altivec_init_builtins): Add code to define
22836 __builtin_altivec_lvx_v1ti function.
22838 2018-03-22 Jakub Jelinek <jakub@redhat.com>
22840 PR inline-asm/84941
22841 * function.c (match_asm_constraints_1): Don't do the optimization
22842 if input isn't a REG, SUBREG, MEM or constant.
22844 2018-03-22 Tom de Vries <tom@codesourcery.com>
22846 PR tree-optimization/84956
22847 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
22848 bb_has_abnormal_pred.
22850 2018-03-22 Jakub Jelinek <jakub@redhat.com>
22853 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
22854 DECL_INITIAL (decl) to decl at the end.
22855 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
22856 adjust the comment.
22858 2018-03-21 Joseph Myers <joseph@codesourcery.com>
22860 * doc/extend.texi (__builtin_tgmath): Document when complex
22861 integer types are treated as _Complex _Float64.
22863 2018-03-21 Tom de Vries <tom@codesourcery.com>
22865 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
22867 2018-03-21 Jakub Jelinek <jakub@redhat.com>
22869 PR tree-optimization/84960
22870 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
22871 if it is ENTRY block, move them into single succ of ENTRY in that case.
22873 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
22875 PR tree-optimization/84811
22876 * poly-int.h (poly_span_traits): Remove the T3 parameter and
22877 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
22878 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
22879 (known_subrange_p): Update accordingly. Cast each value involved
22880 in the size comparison, rather than casting the result of the
22883 2018-03-21 Jakub Jelinek <jakub@redhat.com>
22885 PR tree-optimization/84982
22886 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
22887 by flipping the least significant bit rather than all bits from
22888 bitpos to bitpos + bitsize - 1.
22890 2018-03-21 Nathan Sidwell <nathan@acm.org>
22892 * doc/extend.texi (Deprecated Features): Remove mention of
22893 long-deleted deprecations.
22895 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22898 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
22899 * configure: Regenerate.
22901 2018-03-21 Tom de Vries <tom@codesourcery.com>
22903 PR tree-optimization/83126
22904 * tree-parloops.c (num_phis): New function.
22905 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
22907 2018-03-21 Nathan Sidwell <nathan@acm.org>
22909 * doc/extend.texi (Deprecated Features): Update deprecated flags,
22910 mention anon-struct/union members and trailing attributes.
22912 2018-03-21 Bin Cheng <bin.cheng@arm.com>
22914 PR tree-optimization/84969
22915 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
22916 builtin memset partitions if they set different rhs values.
22918 2018-03-21 Jakub Jelinek <jakub@redhat.com>
22920 PR rtl-optimization/84989
22921 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
22922 VEC_DUPLICATE with scalar result mode.
22924 2018-03-21 Martin Liska <mliska@suse.cz>
22927 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
22928 not intended return statement.
22930 2018-03-21 Martin Liska <mliska@suse.cz>
22933 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
22934 (chkp_find_bound_slots_1): Limit number of iterations.
22936 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
22939 * Minor grammar fixes for x86 options.
22941 2018-03-20 Jakub Jelinek <jakub@redhat.com>
22944 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
22945 holding REG_CFA_RESTORE notes, instead turn them into a USE.
22947 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
22950 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
22951 (altivec_lvx_<mode>_1op): Likewise.
22952 (altivec_stvx_<mode>_2op): Likewise.
22953 (altivec_stvx_<mode>_1op): Likewise.
22954 (altivec_lvx_<VM2:mode>): New define_expand.
22955 (altivec_stvx_<VM2:mode>): Likewise.
22956 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
22957 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22958 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
22959 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
22960 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
22961 (rs6000_gen_lvx): Likewise.
22962 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
22963 (altivec_expand_stv_builtin): Likewise.
22964 (altivec_expand_builtin): Likewise.
22965 * config/rs6000/vector.md: Likewise.
22967 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22970 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
22973 2018-03-20 Richard Biener <rguenther@suse.de>
22976 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
22977 sign-conversions as zero, fall back to standard scalar_stmt
22980 2018-03-20 Martin Liska <mliska@suse.cz>
22983 * predict.c (rebuild_frequencies): Handle case when we have
22984 PROFILE_ABSENT, but flag_guess_branch_prob is false.
22986 2018-03-20 Jakub Jelinek <jakub@redhat.com>
22989 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
22990 flag_section_anchors.
22991 * varasm.c (use_blocks_for_decl_p): Remove hack for
22992 dw2_force_const_mem.
22995 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
22997 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
22998 be created, use lowpart_subreg of operands[0] rather than operands[0]
23000 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23001 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23002 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23003 and n constraint instead of aarch64_shift_imm_di and Usd.
23004 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23005 (*aarch64_<optab>_reg_minus<mode>3): ... this.
23007 2018-03-20 Sudakshina Das <sudi.das@arm.com>
23010 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23011 to favor GPR over NEON registers.
23012 (<shift>di3_neon): Likewise.
23014 2018-03-20 Tom de Vries <tom@codesourcery.com>
23017 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23018 (nvptx_process_pars): Emit bar.sync asap and alap.
23020 2018-03-20 Tom de Vries <tom@codesourcery.com>
23023 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23024 seen_label if seen_label is already set.
23026 2018-03-20 Jakub Jelinek <jakub@redhat.com>
23029 * config/i386/i386.c (fold_builtin_cpu): For features above 31
23030 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23031 Use 1U instead of 1. Formatting fixes.
23034 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23035 instead of TREE_TYPE (s1) for the return value.
23037 2018-03-19 Jakub Jelinek <jakub@redhat.com>
23039 PR tree-optimization/84946
23040 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23041 bitsize + bitsize in poly_uint64 rather than poly_int64.
23044 * dwarf2asm.c: Include fold-const.c.
23045 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23046 of decl rather than decl itself.
23048 PR rtl-optimization/84643
23049 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23051 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
23054 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23055 calling assemble_variable.
23057 2018-03-19 Sudakshina Das <sudi.das@arm.com>
23060 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23061 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23063 2018-03-19 Jim Wilson <jimw@sifive.com>
23066 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23067 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23068 (riscv_first_stack_step): Likewise.
23069 (riscv_option_override): Use STACK_BOUNDARY instead of
23070 MIN_STACK_BOUNDARY.
23071 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23072 MIN_STACK_BOUNDARY.
23073 (BIGGEST_ALIGNMENT): Set to 128.
23074 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23075 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23078 2018-03-19 Richard Biener <rguenther@suse.de>
23080 PR tree-optimization/84933
23081 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23082 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23084 2018-03-19 Richard Biener <rguenther@suse.de>
23086 PR tree-optimization/84859
23087 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23088 (cond_if_else_store_replacement): Perform sinking operation on
23089 single-store BBs regardless of MAX_STORES_TO_SINK setting.
23090 Generalize what a BB with a single eligible store is.
23092 2018-03-19 Richard Biener <rguenther@suse.de>
23094 PR tree-optimization/84929
23095 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23096 chrec_is_positive against non-chrec arg.
23098 2018-03-19 Tamar Christina <tamar.christina@arm.com>
23101 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23103 2018-03-18 Martin Liska <mliska@suse.cz>
23105 PR rtl-optimization/84635
23106 * regrename.c (build_def_use): Use matches_mode only when
23109 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
23111 PR tree-optimization/84913
23112 * tree-vect-loop.c (vectorizable_reduction): Don't try to
23113 vectorize chains of COND_EXPRs.
23115 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23117 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23119 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23121 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23123 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23125 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23127 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
23128 Kito Cheng <kito.cheng@gmail.com>
23130 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23131 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23132 (nds32_adjust_reg_alloc_order): New function.
23133 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23135 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
23137 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23138 nds32_print_operand, nds32_print_operand_address): Use
23139 HOST_WIDE_INT_PRINT_DEC instead.
23141 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
23143 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23145 2018-03-17 Jakub Jelinek <jakub@redhat.com>
23148 * config/i386/i386.c (initial_ix86_tune_features,
23149 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23150 unsigned long long.
23151 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23152 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23153 rather than 1u << ix86_tune. Formatting fix.
23154 (ix86_option_override_internal): Change ix86_arch_mask from
23155 unsigned int to unsigned HOST_WIDE_INT, initialize to
23156 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23157 (ix86_function_specific_restore): Likewise.
23159 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23162 * postreload.c (reload_combine_recognize_pattern): Perform
23163 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23164 truncate_int_for_mode the result for the destination's mode.
23167 * hsa-gen.c (mem_type_for_type): Fix comment typo.
23168 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23170 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23173 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
23176 * lra-assigns.c (lra_split_hard_reg_for): Don't use
23177 regno_allocno_class_array and sorted_pseudos.
23178 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23179 insns where regno is used.
23181 2018-03-16 Martin Liska <mliska@suse.cz>
23184 * multiple_target.c (create_dispatcher_calls): Redirect
23185 reference in the symbol table.
23187 2018-03-16 Martin Liska <mliska@suse.cz>
23190 * multiple_target.c (create_dispatcher_calls): Redirect also
23193 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23197 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23198 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23199 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23201 2018-03-16 Julia Koval <julia.koval@intel.com>
23203 * doc/invoke.texi (Skylake Server): Add CLWB.
23204 Cannonlake): Remove CLWB.
23206 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23208 PR tree-optimization/84841
23209 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23211 (FLOAT_ONE_CONST_TYPE): Define.
23212 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23213 (sort_by_operand_rank): Put entries with higher constant_type last
23214 rather than first to match comments.
23216 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
23218 * config/nios2/nios2.md (movsi_internal): Fix thinko in
23221 2018-03-15 Jakub Jelinek <jakub@redhat.com>
23224 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23225 check and use address of target always.
23227 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
23230 * config/i386/i386.c (indirect_thunk_needed): Update comments.
23231 (indirect_thunk_bnd_needed): Likewise.
23232 (indirect_thunks_used): Likewise.
23233 (indirect_thunks_bnd_used): Likewise.
23234 (indirect_return_needed): New.
23235 (indirect_return_bnd_needed): Likewise.
23236 (output_indirect_thunk_function): Add a bool argument for
23238 (output_indirect_thunk_function): Don't generate alias for
23239 function return thunk.
23240 (ix86_code_end): Call output_indirect_thunk_function to generate
23241 function return thunks.
23242 (ix86_output_function_return): Set indirect_return_bnd_needed
23243 and indirect_return_needed instead of indirect_thunk_bnd_needed
23244 and indirect_thunk_needed.
23246 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
23248 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23249 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23250 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23252 2018-03-15 David Malcolm <dmalcolm@redhat.com>
23253 Paul Hua <paul.hua.gm@gmail.com>
23256 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23258 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
23260 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23261 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23262 resp. SFmode cases.
23264 2018-03-15 Tamar Christina <tamar.christina@arm.com>
23267 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23268 instead of GET_MODE_SIZE when comparing Units.
23270 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
23273 * varasm.c (hash_section): Return an unchangeble hash value
23274 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23275 Return !aarch64_can_use_per_function_literal_pools_p ().
23277 2018-03-15 Jakub Jelinek <jakub@redhat.com>
23280 * optabs.c (emit_conditional_move): Pass address of cmode's copy
23281 rather than address of cmode as last argument to prepare_cmp_insn.
23283 2018-03-15 Julia Koval <julia.koval@intel.com>
23285 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23286 F_AVX512VNNI, F_AVX512BITALG): New.
23288 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
23291 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
23292 insn for floating-point loads and stores.
23294 2018-03-14 Carl Love <cel@us.ibm.com>
23296 * config/rs6000/rs6000-c.c: Add macro definitions for
23297 ALTIVEC_BUILTIN_VEC_PERMXOR.
23298 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23299 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23300 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23301 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23303 * config/doc/extend.texi: Add prototypes for vec_permxor.
23305 2018-03-14 David Malcolm <dmalcolm@redhat.com>
23308 * diagnostic-show-locus.c (class layout_point): Convert m_line
23309 from int to linenum_type.
23310 (line_span::comparator): Use linenum "compare" function when
23311 comparing line numbers.
23312 (test_line_span): New function.
23313 (layout_range::contains_point): Convert param "row" from int to
23315 (layout_range::intersects_line_p): Likewise.
23316 (layout::will_show_line_p): Likewise.
23317 (layout::print_source_line): Likewise.
23318 (layout::should_print_annotation_line_p): Likewise.
23319 (layout::print_annotation_line): Likewise.
23320 (layout::print_leading_fixits): Likewise.
23321 (layout::annotation_line_showed_range_p): Likewise.
23322 (struct line_corrections): Likewise for field m_row.
23323 (line_corrections::line_corrections): Likewise for param "row".
23324 (layout::print_trailing_fixits): Likewise.
23325 (layout::get_state_at_point): Likewise.
23326 (layout::get_x_bound_for_row): Likewise.
23327 (layout::print_line): Likewise.
23328 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23329 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23330 * input.c (selftest::test_linenum_comparisons): New function.
23331 (selftest::input_c_tests): Call it.
23332 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23333 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23334 * selftest.h (ASSERT_GT): New macro.
23335 (ASSERT_GT_AT): New macro.
23336 (ASSERT_LT): New macro.
23337 (ASSERT_LT_AT): New macro.
23339 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
23341 PR rtl-optimization/84780
23342 * combine.c (distribute_links): Don't make a link based on pc_rtx.
23344 2018-03-14 Martin Liska <mliska@suse.cz>
23346 * tree.c (record_node_allocation_statistics): Use
23347 get_stats_node_kind.
23348 (get_stats_node_kind): New function extracted from
23349 record_node_allocation_statistics.
23350 (free_node): Use get_stats_node_kind.
23352 2018-03-14 Richard Biener <rguenther@suse.de>
23354 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23355 that the value-set of ANTIC_IN doesn't grow.
23358 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23360 (BB_VISITED_WITH_VISITED_SUCCS): New define.
23361 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23363 2018-03-14 Julia Koval <julia.koval@intel.com>
23365 * config.gcc (icelake-client, icelake-server): New.
23367 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23368 (initial_ix86_arch_features): Ditto.
23369 (PTA_SKYLAKE): Add SGX.
23370 (PTA_ICELAKE): Remove.
23371 (PTA_ICELAKE_CLIENT): New.
23372 (PTA_ICELAKE_SERVER): New.
23373 (ix86_option_override_internal): Split up icelake on icelake client and
23375 (get_builtin_code_for_version): Ditto.
23376 (fold_builtin_cpu): Ditto.
23377 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23378 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23379 * config/i386/i386.h (processor_type): Ditto.
23380 * doc/invoke.texi: Ditto.
23382 2018-03-14 Jakub Jelinek <jakub@redhat.com>
23385 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23386 INTEGER_CST offset, add it together with bitpos / 8 and
23387 sign extend based on POINTER_SIZE.
23391 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
23394 * config/i386/constraints.md (Yc): New register constraint.
23395 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23396 Use Yc constraint for alternative 2 of operand 0. Remove
23397 preferred_for_speed attribute.
23399 2018-03-14 Richard Biener <rguenther@suse.de>
23401 PR tree-optimization/84830
23402 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23403 with the old one to avoid oscillations.
23405 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
23408 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23410 (assign_by_spills): Return a flag of reload assignment failure.
23411 Do not process the reload assignment failures. Do not spill other
23412 reload pseudos if they has the same reg class. Update n if
23414 (lra_assign): Add a return arg. Set up from the result of
23415 assign_by_spills call.
23416 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23417 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
23418 usage_insns if it is not NULL.
23419 (spill_hard_reg_in_range): New function.
23420 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23421 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23422 function prototypes.
23423 (lra_assign): Change prototype.
23424 * lra.c (lra): Add code to deal with fails by splitting hard reg
23427 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
23429 * config/riscv/riscv.opt (mrelax): New option.
23430 * config/riscv/riscv.c (riscv_file_start): Emit ".option
23431 "norelax" when riscv_mrelax is disabled.
23432 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23434 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23437 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23438 reassociation for int modes.
23440 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23442 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23443 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23445 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23446 * config/aarch64/aarch64-sve.md
23447 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23448 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23449 (*extend<mode><Vwide>2): Rename to...
23450 (aarch64_sve_extend<mode><Vwide>2): ...this.
23451 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23452 renaming the old pattern to...
23453 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
23455 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23456 define_expand, renaming the old pattern to...
23457 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23458 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23459 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23460 account when deciding which SVE instruction the optab should use.
23461 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23463 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23465 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23466 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23467 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23468 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23469 (tlsdesc_small_<mode>): Turn a define_expand and use
23470 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
23471 (tlsdesc_small_advsimd_<mode>): ...this.
23472 (tlsdesc_small_sve_<mode>): New pattern.
23474 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23476 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23477 (UNSPEC_UMUL_HIGHPART): New constants.
23478 (MUL_HIGHPART): New int iteraor.
23479 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23480 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23482 (*<su>mul<mode>3_highpart): New define_insn.
23484 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
23487 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23490 2018-03-13 Martin Liska <mliska@suse.cz>
23493 * (sem_item_optimizer::sem_item_optimizer): Initialize new
23495 (sem_item_optimizer::~sem_item_optimizer): Release it.
23496 (sem_item_optimizer::merge_classes): Register variable aliases.
23497 (sem_item_optimizer::fixup_pt_set): New function.
23498 (sem_item_optimizer::fixup_points_to_sets): Likewise.
23499 * ipa-icf.h: Declare new variables and functions.
23501 2018-03-13 Jakub Jelinek <jakub@redhat.com>
23503 PR middle-end/84834
23504 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23505 integer_pow2p@2 and test integer_pow2p in condition.
23506 (A < 0 ? C : 0): Similarly for @1.
23508 PR middle-end/84831
23509 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23510 characters starting at p contain '\0' character, don't look beyond
23514 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23515 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23518 * reg-stack.c (change_stack): Change update_end var from int to
23519 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23520 also call set_block_for_insn on the newly added insns and rescan.
23523 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23524 on the last operand.
23527 * tree.c (stabilize_reference_1): Return save_expr (e) for
23528 STATEMENT_LIST even if it doesn't have side-effects.
23530 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
23532 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23534 2018-03-12 Renlin Li <renlin.li@arm.com>
23536 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23537 aarch64_output_scalar_simd_mov_immediate.
23539 2018-03-12 Martin Sebor <msebor@redhat.com>
23541 PR tree-optimization/83456
23542 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23543 for perfectly overlapping calls to memcpy.
23544 (gimple_fold_builtin_memory_chk): Same.
23545 (gimple_fold_builtin_strcpy): Handle no-warning.
23546 (gimple_fold_builtin_stxcpy_chk): Same.
23547 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23549 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
23551 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23552 parameter. Use it for SFmode.
23553 (rs6000_function_arg_advance_1): Adjust.
23554 (rs6000_function_arg): Adjust.
23555 (rs6000_gimplify_va_arg): Pass false for that new parameter.
23557 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
23559 PR rtl-optimization/84169
23560 PR rtl-optimization/84780
23561 * combine.c (can_combine_p): Check for a 2-insn combination whether
23562 the destination register is used between the two insns, too.
23564 2018-03-12 Richard Biener <rguenther@suse.de>
23566 PR tree-optimization/84803
23567 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23568 for refs DR analysis didn't process.
23570 2018-03-12 Richard Biener <rguenther@suse.de>
23572 PR tree-optimization/84777
23573 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23574 force-vectorize loops ignore whether we are optimizing for size.
23576 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
23578 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23579 (TARGET_MD_ASM_ADJUST): Define.
23581 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
23582 Kito Cheng <kito.cheng@gmail.com>
23583 Chung-Ju Wu <jasonwucj@gmail.com>
23585 * config/nds32/nds32.c (nds32_compute_stack_frame,
23586 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23587 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23588 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
23589 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
23590 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
23591 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
23592 * config/nds32/nds32.md (prologue, epilogue): Use macro
23593 NDS32_V3PUSH_AVAILABLE_P to do checking.
23595 2018-03-11 Jakub Jelinek <jakub@redhat.com>
23598 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
23599 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
23600 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
23601 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
23602 addition of most attributes on !orig_type_die or the attribute not
23603 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
23605 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
23606 Chung-Ju Wu <jasonwucj@gmail.com>
23608 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
23609 __NDS32_VH__ macro.
23610 * config/nds32/nds32.opt (mvh): New option.
23612 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
23613 Chung-Ju Wu <jasonwucj@gmail.com>
23615 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
23617 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
23618 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
23621 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
23622 Chung-Ju Wu <jasonwucj@gmail.com>
23624 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
23626 * config/nds32/nds32-multiple.md (strlensi): New pattern.
23627 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
23629 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
23630 Kito Cheng <kito.cheng@gmail.com>
23631 Chung-Ju Wu <jasonwucj@gmail.com>
23633 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
23634 UNSPEC_FFMISM and UNSPEC_FLMISM.
23635 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
23636 for ffb, ffmism and flmism.
23637 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
23638 (unspec_ffmism): Ditto.
23639 (unspec_flmism): Ditto.
23640 (nds32_expand_builtin_impl): Check if string extension is available.
23641 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
23642 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
23644 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
23647 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
23650 * lra-assigns.c (assign_by_spills): Return a flag of reload
23651 assignment failure. Do not process the reload assignment
23652 failures. Do not spill other reload pseudos if they has the same
23654 (lra_assign): Add a return arg. Set up from the result of
23655 assign_by_spills call.
23656 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23657 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
23658 usage_insns if it is not NULL.
23659 (spill_hard_reg_in_range): New function.
23660 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23661 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23662 function prototypes.
23663 (lra_assign): Change prototype.
23664 * lra.c (lra): Add code to deal with fails by splitting hard reg
23667 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
23670 * config/i386/i386.opt: Replace Enforcment with Enforcement.
23672 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
23675 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
23676 (dw_val_node): Add val_symbolic_view.
23677 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
23678 (symview_upper_bound): New.
23679 (new_line_info_table): Initialize symviews_since_reset.
23680 (dwarf2out_source_line): Count symviews_since_reset and set
23681 symview_upper_bound.
23682 (dw_val_equal_p): Handle symview.
23683 (add_AT_symview): New.
23684 (print_dw_val): Handle symview.
23685 (attr_checksum, attr_checksum_ordered): Likewise.
23686 (same_dw_val_p, size_of_die): Likewise.
23687 (value_format, output_die): Likewise.
23688 (add_high_low_attributes): Use add_AT_symview for entry_view.
23689 (dwarf2out_finish): Reset symview_upper_bound, clear
23692 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
23695 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
23696 Add strict argument and use it.
23697 (rs6000_split_multireg_move): Update for new strict argument.
23698 (mem_operand_gpr): Disallow all non-offsettable addresses.
23699 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
23701 2018-03-09 Jakub Jelinek <jakub@redhat.com>
23704 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
23705 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
23706 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
23709 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
23710 decl, use remap_type if we want to use the type.
23712 2018-03-09 Martin Sebor <msebor@redhat.com>
23714 PR tree-optimization/84526
23715 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23717 (builtin_access::generic_overlap): Be prepared to handle non-array
23720 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
23722 PR rtl-optimization/84682
23723 * lra-constraints.c (process_address_1): Check is_address flag
23724 for address constraints.
23725 (process_alt_operands): Likewise.
23726 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
23727 preprocess_constraints.
23728 * recog.h (preprocess_constraints): Add oploc parameter.
23730 * recog.c (preprocess_constraints): Test address_operand for
23731 CT_ADDRESS constraints.
23733 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
23736 * lra-assigns.c (assign_by_spills): Return a flag of reload
23737 assignment failure. Do not process the reload assignment
23738 failures. Do not spill other reload pseudos if they has the same
23740 (lra_assign): Add a return arg. Set up from the result of
23741 assign_by_spills call.
23742 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23743 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
23744 usage_insns if it is not NULL.
23745 (spill_hard_reg_in_range): New function.
23746 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23747 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23748 function prototypes.
23749 (lra_assign): Change prototype.
23750 * lra.c (lra): Add code to deal with fails by splitting hard reg
23753 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23756 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
23757 Accept complain bool parameter. Only emit errors if it is true.
23758 (arm_parse_cpu_option_name): Likewise.
23759 (arm_target_thumb_only): Adjust callers of the above.
23760 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
23761 prototype to take a default true bool parameter.
23762 (arm_parse_arch_option_name): Likewise.
23764 2018-03-09 David Malcolm <dmalcolm@redhat.com>
23765 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
23769 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
23770 * configure: Regenerate.
23771 * configure.ac ("linker --version-script option"): New.
23772 ("linker soname option"): New.
23774 2018-03-09 Richard Biener <rguenther@suse.de>
23776 PR tree-optimization/84775
23777 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
23778 immediate uses of predicate stmts and mark them modified.
23781 PR tree-optimization/84178
23782 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23784 (version_loop_for_if_conversion): Delay update_ssa call.
23785 (tree_if_conversion): Delay update_ssa until after predicate
23788 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
23791 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
23792 when the function accesses prior frames.
23794 2018-03-08 Jakub Jelinek <jakub@redhat.com>
23797 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
23798 gen_llsym, otherwise call maybe_gen_llsym.
23800 PR inline-asm/84742
23801 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
23802 has ',' character inside of it.
23804 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23807 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
23808 as clobbering CC_REGNUM.
23810 2018-03-08 Richard Biener <rguenther@suse.de>
23812 PR middle-end/84552
23813 * tree-scalar-evolution.c: Include tree-into-ssa.h.
23814 (follow_copies_to_constant): Do not follow SSA names registered
23817 2018-03-08 Richard Biener <rguenther@suse.de>
23819 PR tree-optimization/84178
23820 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
23822 (version_loop_for_if_conversion): Delay update_ssa call.
23823 (tree_if_conversion): Delay update_ssa until after predicate
23826 2018-03-08 David Malcolm <dmalcolm@redhat.com>
23828 PR tree-optimization/84178
23829 * tree-if-conv.c (release_bb_predicate): Remove the
23830 the assertion that the stmts have NULL use_ops.
23831 Discard the statements, asserting that they haven't
23832 yet been added to a BB.
23834 2018-03-08 Richard Biener <rguenther@suse.de>
23836 PR tree-optimization/84746
23837 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
23838 (phi_translate): Pass in destination ANTIC_OUT set.
23839 (phi_translate_1): Likewise. For a simplified result lookup
23840 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
23841 (phi_translate_set): Adjust.
23842 (do_pre_regular_insertion): Likewise.
23843 (do_pre_partial_partial_insertion): Likewise.
23845 2018-03-08 Martin Liska <mliska@suse.cz>
23847 PR gcov-profile/84735
23848 * doc/gcov.texi: Document usage of profile files.
23849 * gcov-io.h: Document changes in the format.
23851 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
23855 * dwarf2out.c (struct dw_line_info_table): Update comments for
23857 (FORCE_RESET_NEXT_VIEW): New.
23858 (FORCE_RESETTING_VIEW_P): New.
23859 (RESETTING_VIEW_P): Check for -1 too.
23860 (ZERO_VIEW_P): Likewise.
23861 (new_line_info_table): Force-reset next view.
23862 (dwarf2out_begin_function): Likewise.
23863 (dwarf2out_source_line): Simplify zero_view_p initialization.
23864 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
23865 view directly. Omit view when omitting .loc at line 0.
23867 2018-03-08 Jakub Jelinek <jakub@redhat.com>
23869 PR tree-optimization/84740
23870 * tree-switch-conversion.c (process_switch): Call build_constructors
23871 only if info.phi_count is non-zero.
23873 PR tree-optimization/84739
23874 * tree-tailcall.c (find_tail_calls): Check call arguments against
23875 DECL_ARGUMENTS (current_function_decl) rather than
23876 DECL_ARGUMENTS (func) when checking for tail recursion.
23878 2018-03-07 Jakub Jelinek <jakub@redhat.com>
23880 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
23881 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
23882 Volker Reichelt's entry and add entries for people that perform
23883 GCC fuzzy testing and report numerous bugs.
23885 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
23888 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23889 readonly data in sdata, if that is disabled.
23890 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
23891 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
23892 -mreadonly-in-sdata option.
23894 2018-03-07 Martin Sebor <msebor@redhat.com>
23896 PR tree-optimization/84468
23897 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
23898 basic block when looking for nul assignment.
23900 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
23903 * except.h (output_function_exception_table): Adjust prototype.
23904 * except.c (output_function_exception_table): Remove FNNAME parameter
23905 and add SECTION parameter. Ouput one part of the table at a time.
23906 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
23907 the first part of the exception table and emit unwind directives.
23908 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
23909 (i386_pe_seh_cold_init): Likewise.
23910 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
23911 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
23912 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
23913 (ix86_output_call_insn): Emit a nop in one more case for SEH.
23914 * config/i386/winnt.c: Include except.h.
23915 (struct seh_frame_state): Add reg_offset, after_prologue and
23916 in_cold_section fields.
23917 (i386_pe_seh_end_prologue): Set seh->after_prologue.
23918 (i386_pe_seh_cold_init): New function.
23919 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
23920 to seh->in_cold_section.
23921 (seh_emit_push): Record the offset of the push.
23922 (seh_emit_save): Record the offet of the save.
23923 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
23924 Test seh->after_prologue to disregard the epilogue.
23925 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
23926 (i386_pe_end_cold_function): New function.
23928 2018-03-07 Jakub Jelinek <jakub@redhat.com>
23931 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
23932 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
23935 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
23937 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
23938 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
23940 PR middle-end/84723
23941 * multiple_target.c: Include tree-inline.h and intl.h.
23942 (expand_target_clones): Diagnose and fail if node->definition and
23943 !tree_versionable_function_p (node->decl).
23945 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
23947 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
23949 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
23950 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
23951 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
23953 2018-03-06 Jakub Jelinek <jakub@redhat.com>
23956 * combine.c (try_combine): Use reg_or_subregno instead of handling
23957 just paradoxical SUBREGs and REGs.
23959 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
23961 * config/arc/arc.c (arc_finalize_pic): Remove function.
23962 (arc_must_save_register): We use single base PIC register, remove
23963 checks to save/restore the PIC register.
23964 (arc_expand_prologue): Likewise.
23965 * config/arc/arc-protos.h (arc_set_default_type_attributes):
23967 (arc_verify_short): Likewise.
23968 (arc_attr_type): Likewise.
23969 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
23970 (walk_stores): Likewise.
23971 (arc_address_cost): Make it static.
23972 (arc_verify_short): Likewise.
23973 (branch_dest): Likewise.
23974 (arc_attr_type): Likewise.
23975 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
23976 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
23977 (arc_final_prescan_insn): Remove inserting the nops due to
23978 hardware hazards. It is done in reorg step.
23979 (insn_length_variant_t): Remove.
23980 (insn_length_parameters_t): Likewise.
23981 (arc_insn_length_parameters): Likewise.
23982 (arc_get_insn_variants): Likewise.
23983 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
23985 2018-03-06 Jakub Jelinek <jakub@redhat.com>
23987 PR inline-asm/84683
23988 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
23991 PR tree-optimization/84687
23992 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
23994 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
23996 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23998 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
23999 Rename to ppc_speculation_barrier.
24000 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24001 __builtin_ppc_speculation_barrier.
24003 2018-03-05 Jakub Jelinek <jakub@redhat.com>
24006 * combine.c (combine_simplify_rtx): Don't try to simplify if
24007 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24010 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
24012 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24013 to 32 bytes when compiling for POWER9.
24015 2018-03-05 Jakub Jelinek <jakub@redhat.com>
24018 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24019 regparm >= 3 with no arg reg available also for calls with
24020 flag_force_indirect_call. Pass decl to ix86_function_regparm.
24023 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24025 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
24027 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
24030 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24032 2018-03-05 Richard Biener <rguenther@suse.de>
24034 PR tree-optimization/84486
24035 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24036 When inserting a __builtin_assume_aligned call set the LHS
24037 SSA name alignment info accordingly.
24039 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
24041 PR tree-optimization/84114
24042 * config/aarch64/aarch64.c (aarch64_reassociation_width)
24043 Avoid reassociation of FLOAT_MODE addition.
24045 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
24047 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24048 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24049 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24050 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24051 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24052 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24053 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24055 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24056 __builtin_ia32_wbinvd): New builtins.
24057 (SPECIAL_ARGS2): New.
24058 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24059 (SPECIAL_ARGS2): New.
24060 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24061 (ix86_valid_target_attribute_inner_p): Ditto.
24062 (ix86_init_mmx_sse_builtins): Add special_args2.
24063 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24064 TARGET_WBNOINVD_P): New.
24065 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24066 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24067 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24068 * config/i386/immintrin.h (_wbinvd): New intrinsic.
24069 * config/i386/pconfigintrin.h: New file.
24070 * config/i386/wbnoinvdintrin.h: Ditto.
24071 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24073 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24075 2018-03-05 Richard Biener <rguenther@suse.de>
24077 PR tree-optimization/84670
24078 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24080 (BB_VISITED_WITH_VISITED_SUCCS): New define.
24081 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24082 (compute_antic_aux): Only assert the number of values in ANTIC_IN
24083 doesn't grow if all successors (recursively) were visited at least
24086 2018-03-05 Richard Biener <rguenther@suse.de>
24088 PR tree-optimization/84650
24089 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24090 if executed in the loop pipeline.
24092 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
24094 * doc/configfiles.texi (Configuration Files): Move info about
24095 conditionalizing $target-protos.h to...
24096 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
24097 differs from $target-protos.h.
24099 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
24100 Chung-Ju Wu <jasonwucj@gmail.com>
24102 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24103 * config/nds32/nds32-multiple.md (setmemsi): Define.
24104 * config/nds32/nds32-memory-manipulation.c
24105 (nds32_gen_dup_4_byte_to_word_value): New.
24106 (emit_setmem_word_loop): New.
24107 (emit_setmem_byte_loop): New.
24108 (nds32_expand_setmem_loop): New.
24109 (nds32_expand_setmem_loop_v3m): New.
24110 (nds32_expand_setmem_unroll): New.
24111 (nds32_expand_setmem): New.
24113 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24114 Chung-Ju Wu <jasonwucj@gmail.com>
24116 * config/nds32/nds32-memory-manipulation.c
24117 (nds32_emit_load_store): New.
24118 (nds32_emit_post_inc_load_store): New.
24119 (nds32_emit_mem_move): New.
24120 (nds32_emit_mem_move_block): New.
24121 (nds32_expand_movmemsi_loop_unknown_size): New.
24122 (nds32_expand_movmemsi_loop_known_size): New.
24123 (nds32_expand_movmemsi_loop): New.
24124 (nds32_expand_movmemsi_unroll): New.
24125 (nds32_expand_movmemqi): Rename ...
24126 (nds32_expand_movmemsi): ... to this.
24127 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24128 (movmemsi): ... to this.
24129 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24130 (nds32_expand_movmemsi): ... to this.
24132 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24133 Monk Chiang <sh.chiang04@gmail.com>
24134 Chung-Ju Wu <jasonwucj@gmail.com>
24136 * config/nds32/nds32-protos.h
24137 (nds32_expand_load_multiple): New arguments.
24138 (nds32_expand_store_multiple): Ditto.
24139 (nds32_valid_multiple_load_store): Rename ...
24140 (nds32_valid_multiple_load_store_p): ... to this.
24141 * config/nds32/nds32-memory-manipulation.c
24142 (nds32_expand_load_multiple): Refine implementation.
24143 (nds32_expand_store_multiple): Ditto.
24144 * config/nds32/nds32-multiple.md
24145 (load_multiple): Update nds32_expand_load_multiple interface.
24146 (store_multiple): Update nds32_expand_store_multiple interface.
24147 * config/nds32/nds32-predicates.c
24148 (nds32_valid_multiple_load_store): Rename ...
24149 (nds32_valid_multiple_load_store_p): ... to this and refine
24151 * config/nds32/predicates.md
24152 (nds32_load_multiple_and_update_address_operation): New predicate.
24153 (nds32_store_multiple_and_update_address_operation): New predicate.
24155 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24156 Chung-Ju Wu <jasonwucj@gmail.com>
24158 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24159 (combo): New attribute.
24160 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24162 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
24164 * config/nds32/nds32.opt: Change -mcmodel= default value.
24166 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
24167 Monk Chiang <sh.chiang04@gmail.com>
24168 Chung-Ju Wu <jasonwucj@gmail.com>
24170 * config/nds32/constants.md (unspec_element): New enum.
24171 * config/nds32/constraints.md (Umw): New constraint.
24172 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24173 * config/nds32/nds32-intrinsic.md: Likewise.
24174 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24175 (nds32_valid_smw_lwm_base_p): New.
24176 (nds32_output_smw_single_word): New.
24177 (nds32_output_lmw_single_word): New.
24178 (nds32_expand_unaligned_load): New.
24179 (nds32_expand_unaligned_store): New.
24180 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24181 (nds32_output_smw_single_word): Declare.
24182 (nds32_output_lmw_single_word): Declare.
24183 (nds32_expand_unaligned_load): Declare.
24184 (nds32_expand_unaligned_store): Declare.
24185 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24186 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24187 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24188 NDS32_BUILTIN_UASTORE_DW.
24189 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24192 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
24193 Kito Cheng <kito.cheng@gmail.com>
24194 Chung-Ju Wu <jasonwucj@gmail.com>
24196 * config/nds32/nds32-intrinsic.c
24197 (nds32_expand_builtin_null_ftype_reg): Delete.
24198 (nds32_expand_builtin_reg_ftype_imm): Ditto.
24199 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24200 (nds32_read_argument): New.
24201 (nds32_legitimize_target): Ditto.
24202 (nds32_legitimize_argument): Ditto.
24203 (nds32_check_constant_argument): Ditto.
24204 (nds32_expand_unop_builtin): Ditto.
24205 (nds32_expand_unopimm_builtin): Ditto.
24206 (nds32_expand_binop_builtin): Ditto.
24207 (nds32_builtin_decl_impl): Ditto.
24208 (builtin_description): Ditto.
24209 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24210 (nds32_init_builtins_impl): Ditto.
24211 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24212 (nds32_builtin_decl): New.
24213 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24214 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24216 2018-03-02 Jeff Law <law@redhat.com>
24218 * reorg.c (stop_search_p): Handle DEBUG_INSN.
24219 (redundant_insn, fill_simple_delay_slots): Likewise.
24220 (fill_slots_from_thread): Likewise.
24221 * resource.c (mark_referenced_resources): Likewise.
24222 (mark_set_resources, find_dead_or_set_registers): Likewise.
24224 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24226 * substring-locations.h (format_warning_va): Formatting fix for
24227 ATTRIBUTE_GCC_DIAG.
24228 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24230 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24231 * substring-locations.c: Include intl.h.
24232 (format_warning_va): Turned into small wrapper around
24233 format_warning_n_va, renamed to ...
24234 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24235 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24237 (format_warning_at_substring_n): New function.
24238 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24239 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
24240 format_warning_at_substring with just a shorter name instead of
24241 const function pointer.
24242 (fmtwarn_n): New function.
24243 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24244 appropriate, get rid of all the fmtstr temporaries, move conditionals
24245 with G_() wrapped string literals directly into fmtwarn arguments,
24246 cast dir.len to (int), formatting fixes.
24248 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
24250 * doc/invoke.texi: Remove "Cilk Plus" references.
24252 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24253 Richard Biener <rguenther@suse.de>
24256 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24257 for error or warning attributes if CALL_FROM_THUNK_P is set.
24260 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24263 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24264 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24267 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24268 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24270 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24271 instead of -1U in last predictors element's probability member.
24273 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
24276 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24277 arguments if they are comparable.
24279 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
24281 PR tree-optimization/84634
24282 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24283 masks and masked_loop_p with a single loop_masks, making sure it's
24284 null for bb vectorization.
24286 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
24288 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24289 (vect_analyze_data_ref_access): Use loop->safe_len rather than
24290 loop->force_vectorize to check whether there is no alias.
24292 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24295 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24297 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24299 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24300 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24301 instead of a loop around prev_real_insn.
24302 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24305 PR inline-asm/84625
24306 * config/i386/i386.c (ix86_print_operand): Use conditional
24307 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24310 2018-03-02 Richard Biener <rguenther@suse.de>
24312 PR tree-optimization/84427
24313 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24314 (bitmap_set_subtract_values): Rewrite to handle multiple
24317 (prune_clobbered_mems): Likewise.
24318 (phi_translate): Take edge instead of pred/phiblock.
24319 (phi_translate_1): Likewise.
24320 (phi_translate_set): Likewise. Insert all translated
24321 exprs for a value into the set, keeping possibly multiple
24322 expressions per value.
24323 (compute_antic_aux): Adjust for phi_translate changes.
24324 When intersecting union the expressions and prune those
24325 not in the final value set, keeping possibly multiple
24326 expressions per value. Do not use value-insertion
24327 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24328 all expressions. Add verification that the value-sets
24329 only shrink during iteration.
24330 (compute_partial_antic_aux): Adjust for the phi_translate changes.
24331 (do_pre_regular_insertion): Likewise.
24332 (do_pre_partial_partial_insertion): Likewise.
24334 2018-03-02 Richard Biener <rguenther@suse.de>
24337 * config/darwin.c (saved_debug_info_level): New static global.
24338 (darwin_asm_lto_start): Disable debug info generation for LTO out.
24339 (darwin_asm_lto_end): Restore debug info generation settings.
24341 2018-03-01 Martin Liska <mliska@suse.cz>
24344 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24345 volatile arguments.
24347 2018-03-01 Richard Biener <rguenther@suse.de>
24350 * dwarf2out.c (gen_variable_die): Properly handle late VLA
24351 type annotation with LTO when debug was disabled at compile-time.
24353 2018-03-01 Matthew Fortune <mfortune@gmail.com>
24355 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24357 (mips_final_postscan_insn): Likewise.
24359 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
24361 PR rtl-optimization/84528
24362 * alias.c (init_alias_target): Add commentary.
24363 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24364 a unique base value if the frame pointer is not eliminated
24365 to the stack pointer.
24367 2018-03-01 Tom de Vries <tom@codesourcery.com>
24369 PR rtl-optimization/83327
24370 * lra-int.h (hard_regs_spilled_into): Declare.
24371 * lra.c (hard_regs_spilled_into): Define.
24372 (init_reg_info): Init hard_regs_spilled_into.
24373 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24374 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24375 (process_bb_lives): Handle hard_regs_spilled_into.
24376 (lra_create_live_ranges_1): Before doing liveness propagation, clear
24377 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24379 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
24381 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24382 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24383 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24384 * config/rs6000/aix72.h: New file.
24386 2018-02-28 Jakub Jelinek <jakub@redhat.com>
24388 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24389 instead of warning_at with conditional singular and plural messages
24393 * stor-layout.c (update_alignment_for_field): For
24394 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24395 && !DECL_PACKED (field), do the alignment update, just use
24396 only desired_align instead of MAX (type_align, desired_align)
24398 (place_field): Don't do known_align < desired_align handling
24399 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24400 is non-NULL, instead do it after rli->prev_field handling and
24401 only if not within a bitfield word. For DECL_PACKED (field)
24402 use type_align of BITS_PER_UNIT.
24404 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
24406 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24407 superfluous parentheses and trailing spaces.
24409 2018-02-28 Richard Biener <rguenther@suse.de>
24411 PR tree-optimization/84584
24412 * graphite-scop-detection.c (scop_detection::add_scop): Discard
24413 SCoPs with fake exit edge.
24415 2018-02-28 Martin Liska <mliska@suse.cz>
24418 * timevar.c (timer::print): Fix format to properly print 100%
24421 2018-02-28 Richard Biener <rguenther@suse.de>
24423 PR middle-end/84607
24424 * genmatch.c (capture_info::walk_match): Do not mark
24425 captured expressions without operands as expr_p given
24426 they act more like predicates and should be subject to
24427 "lost tail" side-effect preserving.
24429 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
24431 PR rtl-optimization/81611
24432 * auto-inc-dec.c (attempt_change): Move dead note from
24433 mem_insn if it's the next use of regno
24434 (find_address): Take address use of reg holding
24435 non-incremented value. Add parm to limit search to the named
24437 (merge_in_block): Attempt to use a mem insn that is the next
24438 use of the original regno.
24440 2018-02-27 Martin Sebor <msebor@redhat.com>
24443 * doc/invoke.texi (-Wmissing-attributes): New option.
24444 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24446 2018-02-27 Martin Sebor <msebor@redhat.com>
24448 PR translation/84207
24449 * diagnostic-core.h (warning_n, error_n, inform_n): Change
24450 n argument to unsigned HOST_WIDE_INT.
24451 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24452 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
24453 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24454 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24456 2018-02-27 Richard Biener <rguenther@suse.de>
24458 PR tree-optimization/84512
24459 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24460 Do not use the estimate returned from record_stmt_cost for
24461 the scalar iteration cost but sum properly using add_stmt_cost.
24463 2018-02-27 Richard Biener <rguenther@suse.de>
24465 PR tree-optimization/84466
24466 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24467 Adjust last change to less strictly validate use operands.
24469 2018-02-27 Martin Liska <mliska@suse.cz>
24471 PR gcov-profile/84548
24472 * gcov.c (process_file): Allow partial overlap and consider it
24473 also as group functions.
24474 (output_lines): Properly calculate range of lines for a group.
24476 2018-02-27 Martin Liska <mliska@suse.cz>
24478 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24479 'ggc' suffixes. Change first column width.
24480 (timer::print): Fix formatting of the column.
24482 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
24484 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24485 preserve inline entry blocks for the sake of debug inline
24486 entry point markers alone.
24487 (remove_unused_locals): Suggest in comments a better place to
24488 force the preservation of inline entry blocks that are
24489 otherwise unused, but do not preserve them.
24491 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
24493 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24495 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
24498 * config/i386/constraints.md (Bs): Replace
24499 ix86_indirect_branch_register with
24500 TARGET_INDIRECT_BRANCH_REGISTER.
24502 * config/i386/i386.md (indirect_jump): Likewise.
24503 (tablejump): Likewise.
24504 (*sibcall_memory): Likewise.
24505 (*sibcall_value_memory): Likewise.
24506 Peepholes of indirect call and jump via memory: Likewise.
24507 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24508 (*sibcall_value_GOT_32): Likewise.
24509 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24510 (GOT_memory_operand): Likewise.
24511 (call_insn_operand): Likewise.
24512 (sibcall_insn_operand): Likewise.
24513 (GOT32_symbol_operand): Likewise.
24514 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24516 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
24518 PR rtl-optimization/83496
24519 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24520 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
24521 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24522 redundant insn, if any.
24523 (relax_delay_slots): Likewise.
24524 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24526 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
24528 PR tree-optimization/83965
24529 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24530 that grouped statements are part of a reduction chain. Return
24531 true if the statement is not marked as a reduction itself but
24532 is part of a group.
24533 (vect_recog_dot_prod_pattern): Don't check whether the statement
24534 is part of a group here.
24535 (vect_recog_sad_pattern): Likewise.
24536 (vect_recog_widen_sum_pattern): Likewise.
24538 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
24541 * final.c (rest_of_clean_state): Also look for calls inside sequences.
24543 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
24546 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24548 (ix86_output_indirect_function_return): New prototype.
24549 (ix86_split_simple_return_pop_internal): Likewise.
24550 * config/i386/i386.c (indirect_return_via_cx): New.
24551 (indirect_return_via_cx_bnd): Likewise.
24552 (indirect_thunk_name): Handle return va CX_REG.
24553 (output_indirect_thunk_function): Create alias for
24554 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24555 (ix86_output_indirect_jmp): Remove the bool argument.
24556 (ix86_output_indirect_function_return): New function.
24557 (ix86_split_simple_return_pop_internal): Likewise.
24558 * config/i386/i386.md (*indirect_jump): Don't pass false
24559 to ix86_output_indirect_jmp.
24560 (*tablejump_1): Likewise.
24561 (simple_return_pop_internal): Change it to define_insn_and_split.
24562 Call ix86_split_simple_return_pop_internal to split it for
24563 -mfunction-return=.
24564 (simple_return_indirect_internal): Call
24565 ix86_output_indirect_function_return instead of
24566 ix86_output_indirect_jmp.
24568 2018-02-26 Jakub Jelinek <jakub@redhat.com>
24571 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24572 memset and value initialization afterwards.
24574 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
24576 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24578 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
24581 * common/config/aarch64/aarch64-common.c
24582 (aarch_option_optimization_table[]): Switch
24583 off fomit-frame-pointer
24585 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
24586 Chung-Ju Wu <jasonwucj@gmail.com>
24588 * config/nds32/nds32-multiple.md (load_multiple): Disallow
24590 (store_multiple): Ditto.
24592 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
24594 * config.gcc: Add --with-cpu support for nds32 target.
24595 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
24596 * config/nds32/nds32.opt: Add -mcpu= option.
24598 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
24600 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
24601 isel=yes): Warn for these deprecated options.
24603 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
24605 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
24606 ISA_2_5_MASKS_EMBEDDED.
24608 2018-02-23 Jakub Jelinek <jakub@redhat.com>
24610 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
24611 p->max as pointers rather than using iterative_hash_expr.
24613 2018-02-23 Carl Love <cel@us.ibm.com>
24615 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
24616 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
24618 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
24619 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
24620 P8V_BUILTIN_VEC_VUNSIGNED2.
24622 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
24625 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
24626 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
24628 * lra-constraints.c (curr_insn_transform): Set up
24629 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
24631 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
24632 * lra-eliminations.c (spill_pseudos): Ditto.
24633 (process_insn_for_elimination): Ditto.
24634 * lra-lives.c (reg_early_clobber_p): Use the new macros.
24635 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
24636 LRA_NON_CLOBBERED_ALT.
24638 2018-02-22 Martin Sebor <msebor@redhat.com>
24640 PR tree-optimization/84480
24641 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
24642 to maybe_diag_stxncpy_trunc. Call it.
24643 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
24644 from gimple_fold_builtin_strcpy. Print inlining stack.
24645 (handle_builtin_stxncpy): Print inlining stack.
24646 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
24648 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
24651 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
24652 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
24653 and -fcheck-pointer-bounds are used together.
24654 (indirect_thunk_prefix): New enum.
24655 (indirect_thunk_need_prefix): New function.
24656 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
24657 "_nt" instead of "_bnd" for NOTRACK prefix.
24658 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
24659 (output_indirect_thunk_function): Likewise.
24661 (ix86_code_end): Update output_indirect_thunk_function calls.
24662 (ix86_output_indirect_branch_via_reg): Replace
24663 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
24664 (ix86_output_indirect_branch_via_push): Likewise.
24665 (ix86_output_function_return): Likewise.
24666 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
24667 incompatible with -fcf-protection=branch and
24668 -fcheck-pointer-bounds.
24670 2018-02-22 Steve Ellcey <sellcey@cavium.com>
24673 * config/aarch64/aarch64.c (aarch64_print_address_internal):
24674 Change gcc_assert call to output_operand_lossage.
24676 2018-02-22 Steve Ellcey <sellcey@cavium.com>
24678 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
24680 2018-02-22 DJ Delorie <dj@redhat.com>
24681 Sebastian Perta <sebastian.perta@renesas.com>
24682 Oleg Endo <olegendo@gcc.gnu.org>
24684 * config/rx/rx.c (rx_rtx_costs): New function.
24685 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
24687 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
24689 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
24691 2018-02-22 Martin Liska <mliska@suse.cz>
24694 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
24695 Add "native" as a possible value.
24697 2018-02-22 Martin Liska <mliska@suse.cz>
24700 * config/i386/i386.c (ix86_option_override_internal):
24701 Add "native" as a possible value for -march and -mtune.
24703 2018-02-22 Jakub Jelinek <jakub@redhat.com>
24706 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
24707 to all type variants.
24709 PR tree-optimization/84503
24710 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
24711 width as info->bitpos + info->bitsize - start.
24712 (merged_store_group::merge_overlapping): Simplify width computation.
24713 (check_no_overlap): New function.
24714 (imm_store_chain_info::try_coalesce_bswap): Compute expected
24715 start + width and last_order of the group, fail if check_no_overlap
24717 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
24718 to group if check_no_overlap fails.
24720 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
24722 * config/rs6000/altivec.md: Delete contraint arguments to
24723 define_expand, define_split, and define_peephole2, and in
24724 define_insn_and_split if always unused.
24725 * config/rs6000/darwin.md: Ditto.
24726 * config/rs6000/dfp.md: Ditto.
24727 * config/rs6000/rs6000.md: Ditto.
24728 * config/rs6000/sync.md: Ditto.
24729 * config/rs6000/vector.md: Ditto.
24730 * config/rs6000/vsx.md: Ditto.
24732 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
24734 * config/rs6000/altivec.md: Write output control strings as braced
24735 blocks instead of double-quoted strings.
24736 * config/rs6000/darwin.md: Ditto.
24737 * config/rs6000/rs6000.md: Ditto.
24738 * config/rs6000/vector.md: Ditto.
24739 * config/rs6000/vsx.md: Ditto.
24741 2018-02-21 Jason Merrill <jason@redhat.com>
24743 PR c++/84314 - ICE with templates and fastcall attribute.
24744 * attribs.c (build_type_attribute_qual_variant): Remove assert.
24746 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
24748 * ipa-cp.c (determine_versionability): Fix comment typos.
24750 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
24753 * ipa-cp.c (determine_versionability): Do not version functions caling
24756 2018-02-21 Martin Liska <mliska@suse.cz>
24759 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
24760 Add "native" as a possible value.
24761 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
24762 the macro when native cpu detection is available.
24764 2018-02-21 Martin Liska <mliska@suse.cz>
24767 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
24768 Add "native" as a possible value.
24769 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
24770 when native cpu detection is available.
24772 2018-02-21 Jakub Jelinek <jakub@redhat.com>
24773 Martin Sebor <msebor@redhat.com>
24775 PR tree-optimization/84478
24776 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
24778 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
24779 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
24780 support which is conservatively correct, for 2 only stay conservative
24781 for maxlen. Formatting and comment capitalization fixes. Add STRICT
24782 argument to the 2 argument get_range_strlen, adjust 6 arg
24783 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
24785 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
24786 (gimple_fold_builtin_strlen): Pass true as last argument to
24789 2018-02-20 Martin Sebor <msebor@redhat.com>
24791 PR middle-end/84095
24792 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
24793 (builtin_memref::set_base_and_offset): Same. Handle inner references.
24794 (builtin_memref::builtin_memref): Factor out parts into
24795 set_base_and_offset and call it.
24797 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
24799 PR middle-end/84406
24800 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
24801 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
24802 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
24803 search at the associated MODE_INT.
24805 2018-02-20 Jeff Law <law@redhat.com>
24807 PR middle-end/82123
24808 PR tree-optimization/81592
24809 PR middle-end/79257
24810 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
24811 for range data rather than using global data.
24812 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
24813 range data rather than using global data.
24814 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
24815 pass it to children as needed.
24816 (struct directive::fmtresult): Similarly.
24817 (struct directive::set_width): Similarly.
24818 (struct directive::set_precision): Similarly.
24819 (format_integer, format_directive, parse_directive): Similarly.
24820 (format_none): Accept unnamed vr_values parameter.
24821 (format_percent, format_floating, format_character): Similarly.
24822 (format_string, format_plain): Similarly.
24823 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
24824 the EVRP range analyzer for range data rather than using global data.
24825 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
24826 gimple-ssa-evrp-analyze.h
24827 (class sprintf_dom_walker): Add after_dom_children member function.
24828 Add evrp_range_analyzer member.
24829 (sprintf_dom_walker::before_dom_children): Call into the EVRP
24830 range analyzer as needed.
24831 (sprintf_dom_walker::after_dom_children): New member function.
24832 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
24834 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
24835 (evrp_range_analyzer::pop_to_marker): Likewise.
24837 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
24839 PR tree-optimization/84419
24840 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
24841 with the required type if its current type is compatible but
24844 2018-02-20 Jakub Jelinek <jakub@redhat.com>
24846 PR middle-end/82004
24847 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
24848 after vectorization.
24850 2018-02-20 Martin Liska <mliska@suse.cz>
24853 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
24854 possible values if we don't have a hint.
24856 2018-02-20 Martin Liska <mliska@suse.cz>
24860 * final.c (shorten_branches): Build align_tab array with one
24862 * opts.c (finish_options): Add alignment option limit check.
24863 (MAX_CODE_ALIGN): Likewise.
24864 (MAX_CODE_ALIGN_VALUE): Likewise.
24865 * doc/invoke.texi: Document maximum allowed option value for
24866 all -falign-* options.
24868 2018-02-19 Jakub Jelinek <jakub@redhat.com>
24871 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
24872 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
24873 * var-tracking.c (emit_note_insn_var_location): Remove all references
24874 to NOTE_INSN_CALL_ARG_LOCATION.
24875 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
24876 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
24877 Use copy_rtx_if_shared.
24878 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
24879 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
24880 (dwarf2out_var_location): Remove handling of
24881 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
24883 * final.c (final_scan_insn): Remove all references to
24884 NOTE_INSN_CALL_ARG_LOCATION.
24885 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
24886 before dumping final insns.
24887 * except.c (emit_note_eh_region_end): Remove all references to
24888 NOTE_INSN_CALL_ARG_LOCATION.
24889 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
24890 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
24891 * config/arc/arc.c (hwloop_optimize): Likewise.
24892 * config/arm/arm.c (create_fix_barrier): Likewise.
24893 * config/s390/s390.c (s390_chunkify_start): Likewise.
24894 * config/sh/sh.c (find_barrier): Likewise.
24895 * config/i386/i386.c (rest_of_insert_endbranch,
24896 ix86_seh_fixup_eh_fallthru): Likewise.
24897 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
24898 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
24899 * config/frv/frv.c (frv_function_prologue): Likewise.
24900 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
24902 (note_outside_basic_block_p): Remove all references to
24903 NOTE_INSN_CALL_ARG_LOCATION.
24904 * gengtype.c (adjust_field_rtx_def): Likewise.
24905 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
24907 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
24908 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
24911 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
24914 PR tree-optimization/84452
24915 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
24916 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
24919 2018-02-19 Martin Liska <mliska@suse.cz>
24922 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
24924 2018-02-19 Martin Liska <mliska@suse.cz>
24925 Richard Sandiford <richard.sandiford@linaro.org>
24927 PR tree-optimization/82491
24928 * gimple-fold.c (get_base_constructor): Make earlier bail out
24931 2018-02-19 Carl Love <cel@us.ibm.com>
24933 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
24934 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
24936 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
24937 P8V_BUILTIN_VEC_NEG.
24939 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
24941 * config/rl78/rl78.md (movdf): New define expand.
24943 2018-02-19 Martin Liska <mliska@suse.cz>
24946 * doc/invoke.texi: Fix typo.
24947 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
24949 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
24951 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
24952 handle rs6000_single_float and rs6000_double_float specially for
24955 2018-02-16 Jeff Law <law@redhat.com>
24957 * config/rx/rx.c (add_pop_cfi_notes): New function.;
24958 (pop_regs): Use it.
24960 2018-02-16 Jakub Jelinek <jakub@redhat.com>
24963 * ipa-inline.c (inline_small_functions): Fix a typo.
24965 2018-02-16 Nathan Sidwell <nathan@acm.org>
24967 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
24969 2018-02-16 Carl Love <cel@us.ibm.com>
24971 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
24972 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
24973 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
24974 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
24975 expansion to P8V_BUILTIN_VEC_FLOAT2.
24977 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
24979 PR rtl-optimization/70023
24980 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
24981 src_regno into account.
24983 2018-02-16 Carl Love <cel@us.ibm.com>
24985 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
24986 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
24987 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
24988 * config/rs6000/rs6000.c: Remove case statements for
24989 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
24990 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
24991 and P9V_BUILTIN_VEC_VINSERT4B.
24992 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
24993 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
24994 * config/rs6000/vsx.md:
24995 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
24998 2018-02-16 Carl Love <cel@us.ibm.com>
25000 * config/rs6000/altivec.h: Add builtin names vec_extract4b
25002 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25004 * config/rs6000/rs6000-c.c: Add the definitions for
25005 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25006 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25007 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25008 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
25009 definition for insert4b and define insn *insert3b_internal.
25010 * doc/extend.texi: Add documentation for vec_extract4b.
25012 2018-02-16 Nathan Sidwell <nathan@acm.org>
25014 * doc/extend.texi (Backwards Compatibility): Mention friend
25015 injection. Note for-scope is deprecated.
25016 * doc/invoke.texi (-ffriend-injection): Deprecate.
25018 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
25020 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25021 that moved to I2, also allow destinations that are a paradoxical
25022 subreg (instead of a normal reg).
25024 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
25027 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25030 2018-02-16 Richard Biener <rguenther@suse.de>
25032 PR tree-optimization/84037
25033 PR tree-optimization/84016
25035 * config/i386/i386.c (ix86_builtin_vectorization_cost):
25036 Adjust vec_construct for the fact we need additional higher latency
25037 128bit inserts for AVX256 and AVX512 vector builds.
25038 (ix86_add_stmt_cost): Scale vector construction cost for
25041 2018-02-16 Richard Biener <rguenther@suse.de>
25043 PR tree-optimization/84417
25044 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25045 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25046 (non_rewritable_lvalue_p): Likewise, use poly-ints.
25048 2018-02-16 Martin Liska <mliska@suse.cz>
25051 * internal-fn.def (ASAN_CHECK): Set proper flags.
25052 (ASAN_MARK): Likewise.
25054 2018-02-16 Julia Koval <julia.koval@intel.com>
25056 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25057 from PTA_CANNONLAKE.
25059 2018-02-16 Jakub Jelinek <jakub@redhat.com>
25062 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25063 Use ++iter rather than iter++ for std::list iterators.
25064 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
25065 defer deleting them until all nodes in the forest are processed. Do
25066 free even leaf nodes. Change to_process into auto_vec.
25069 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25070 * vec.h (vec_default_construct): Use memset instead of placement new
25071 if BROKEN_VALUE_INITIALIZATION is defined.
25072 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25073 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25076 PR rtl-optimization/83723
25077 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25078 * lra.c (lra_substitute_pseudo): Likewise. If true, use
25079 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
25081 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25083 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25085 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
25087 PR rtl-optimization/81443
25088 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25089 from inner REGs to paradoxical SUBREGs.
25091 2018-02-16 Richard Biener <rguenther@suse.de>
25093 PR tree-optimization/84399
25094 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25095 For operands we can analyze at their definition make sure we can
25096 analyze them at each use as well.
25098 2018-02-16 Richard Biener <rguenther@suse.de>
25100 PR tree-optimization/84190
25101 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25102 volatile accesses if the decl isn't volatile.
25104 2018-02-15 Jason Merrill <jason@redhat.com>
25106 PR c++/84314 - ICE with templates and fastcall attribute.
25107 * attribs.c (build_type_attribute_qual_variant): Don't clobber
25108 TYPE_CANONICAL on an existing type.
25110 2018-02-15 Jakub Jelinek <jakub@redhat.com>
25112 PR tree-optimization/84383
25113 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25114 dstoff nor call operand_equal_p if dstbase is NULL.
25116 PR tree-optimization/84334
25117 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25118 also a CONSTANT_CLASS_P, punt.
25120 2018-02-14 Jim Wilson <jimw@sifive.com>
25122 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25123 first SMALL_OPERAND check. New local min_second_step. Move assert
25124 to where locals are set. Add TARGET_RVC support.
25125 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25127 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
25129 * doc/invoke.texi: Correct -Wformat-overflow code sample.
25131 2018-02-14 Martin Sebor <msebor@redhat.com>
25133 PR tree-optimization/83698
25134 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25135 arrays constrain the offset range to their bounds.
25136 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25137 (builtin_access::overlap): Avoid setting the size of overlap if it's
25139 (maybe_diag_overlap): Also consider arrays when deciding what values
25140 of offsets to include in diagnostics.
25142 2018-02-14 Martin Sebor <msebor@redhat.com>
25145 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25146 that correspond to the kind of a declaration.
25148 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
25151 * config/pa/pa.md: Load address of PIC label using the linkage table
25152 if the label is nonlocal.
25154 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
25156 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25157 warning message if user requests -maltivec=be.
25158 * doc/invoke.texi: Document deprecation of -maltivec=be.
25160 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
25163 * config/rs6000/rs6000-c.c: Update definitions for
25164 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25165 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25167 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
25170 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25171 add _get_ssp intrinsics. Remove argument from
25172 __builtin_ia32_rdssp[d|q].
25173 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25174 * config/i386/i386-builtin.def: Remove argument from
25175 __builtin_ia32_rdssp[d|q].
25176 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25177 ix86_expand_special_args_builtin for _rdssp[d|q].
25178 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25179 Clear register before usage.
25180 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25181 Add documentation for new _get_ssp and _inc_ssp intrinsics.
25183 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
25185 PR tree-optimization/84357
25186 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25187 operand 1 of an ARRAY_REF too.
25189 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
25192 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25193 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25195 (set_of_reg): New struct.
25196 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25197 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25198 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25200 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25201 Split into bitclr, bitset, bitinvert patterns if appropriate.
25202 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25203 use rx_fuse_in_memory_bitop.
25204 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25205 to named insn, correct maximum insn length.
25207 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
25210 * machmode.def: Define a complex mode for PARTIAL_INT.
25211 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25213 * doc/rtl.texi: Document CSPImode.
25214 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25216 (msp430_hard_regno_nregs_with_padding): Likewise.
25218 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
25221 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25223 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
25225 PR rtl-optimization/84169
25226 * combine.c (try_combine): New variable split_i2i3. Set it to true if
25227 we generated a parallel as new i3 and we split that to new i2 and i3
25228 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
25229 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
25230 those to i2, not i1. Partially rewrite this scan code.
25232 2018-02-13 Jakub Jelinek <jakub@redhat.com>
25235 * stor-layout.c (place_field): For variable length fields, adjust
25236 offset_align afterwards not just based on the field's alignment,
25237 but also on the size.
25239 PR middle-end/84309
25240 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25241 of exps and logs in the use_exp2 case.
25243 2018-02-13 Jeff Law <law@redhat.com>
25245 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25246 entry for "vector".
25248 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25251 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
25255 * common.opt (gas-loc-support, gas-locview-support): New.
25256 (ginline-points, ginternal-reset-location-views): New.
25257 * doc/invoke.texi: Document them. Use @itemx where intended.
25258 (gvariable-location-views): Adjust.
25259 * target.def (reset_location_view): New.
25260 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25261 (TARGET_RESET_LOCATION_VIEW): New.
25262 * doc/tm.texi: Rebuilt.
25263 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25264 (dwarf2out_default_as_locview_support): New.
25265 (output_asm_line_debug_info): Use option variables.
25266 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25267 (output_loc_list): Likewise.
25268 (add_high_low_attributes): Check option variables.
25269 Don't output entry view attribute in strict mode.
25270 (gen_inlined_subroutine_die): Check option variables.
25271 (dwarf2out_inline_entry): Likewise.
25272 (init_sections_and_labels): Likewise.
25273 (dwarf2out_early_finish): Likewise.
25274 (maybe_reset_location_view): New, from...
25275 (dwarf2out_var_location): ... here. Call it.
25276 * debug.h (dwarf2out_default_as_loc_support): Declare.
25277 (dwarf2out_default_as_locview_support): Declare.
25278 * hooks.c (hook_int_rtx_insn_0): New.
25279 * hooks.h (hook_int_rtx_insn_0): Declare.
25280 * toplev.c (process_options): Take -gas-loc-support and
25281 -gas-locview-support from dwarf2out. Enable
25282 -gvariable-location-views by default only with locview
25283 assembler support. Enable -ginternal-reset-location-views by
25284 default only if the target defines the corresponding hook.
25285 Enable -ginline-points by default if location views are
25286 enabled; force it disabled if statement frontiers are
25288 * tree-inline.c (expand_call_inline): Check option variables.
25289 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25291 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
25293 PR tree-optimization/84321
25294 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25295 handling. Also check whether the anti-range contains any values
25296 that satisfy the mask; switch to a VR_RANGE if not.
25298 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
25301 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25303 2018-02-13 Martin Jambor <mjambor@suse.cz>
25306 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25307 of call statements, also set location of a load to a temporary.
25309 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
25311 * config/rl78/rl78.c (add_vector_labels): New function.
25312 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25313 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25314 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25315 which checks that no arguments are passed.
25316 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25317 * doc/extend.texi: Documentation for the new attribute.
25319 2018-02-13 Andreas Schwab <schwab@suse.de>
25321 * config/riscv/linux.h (CPP_SPEC): Define.
25323 2018-02-13 Jakub Jelinek <jakub@redhat.com>
25326 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25327 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25328 OPTION_MASK_ISA_AES as first argument to def_builtin_const
25329 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25330 instead of OPTION_MASK_ISA_PCLMUL as first argument to
25331 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25332 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25333 temporarily for AES and PCLMUL builtins.
25335 PR tree-optimization/84339
25336 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25337 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25340 PR middle-end/84309
25341 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25342 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25343 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25345 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25347 * omp-simd-clone.h: New file.
25348 * omp-simd-clone.c: Include omp-simd-clone.h.
25349 (expand_simd_clones): No longer static.
25350 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25351 cgraph.h and omp-simd-clone.h.
25352 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25353 (vect_recog_widen_shift_pattern): Formatting fix.
25354 (vect_pattern_recog_1): Don't check optab for calls.
25357 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25358 operands[2] into a REG before using gen_lowpart on it.
25360 2018-02-12 Jeff Law <law@redhat.com>
25363 * config/sh/sh.c (find_barrier): Consider a sibling call
25366 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25367 successfully back substituting a reg.
25369 2018-02-12 Richard Biener <rguenther@suse.de>
25371 PR tree-optimization/84037
25372 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25373 parameter, move visited init to caller.
25374 (vect_slp_analyze_operations): Separate cost from validity
25375 check, initialize visited once for all instances.
25376 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25378 * tree-vect-stmts.c (vect_model_simple_cost): Make early
25380 (vect_model_promotion_demotion_cost): Likewise.
25381 (vectorizable_bswap): Guard cost modeling with !slp_node
25382 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25384 (vectorizable_call): Likewise.
25385 (vectorizable_conversion): Likewise.
25386 (vectorizable_assignment): Likewise.
25387 (vectorizable_shift): Likewise.
25388 (vectorizable_operation): Likewise.
25389 (vectorizable_store): Likewise.
25390 (vectorizable_load): Likewise.
25391 (vectorizable_condition): Likewise.
25392 (vectorizable_comparison): Likewise.
25394 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
25397 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25398 (ASAN_MARK): Fix fnspec to account for return value, change pointer
25399 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25401 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
25403 PR middle-end/83665
25404 * params.def (inline-min-speedup): Increase from 8 to 15.
25405 (max-inline-insns-auto): Decrease from 40 to 30.
25406 * ipa-split.c (consider_split): Add some buffer for function to
25407 be considered inlining candidate.
25408 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25411 2018-02-12 Richard Biener <rguenther@suse.de>
25413 PR tree-optimization/84037
25414 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25415 matched stmts if we cannot swap the non-matched ones.
25417 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
25419 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25420 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25421 _mm_maskz_scalef_round_ss): New intrinsics.
25422 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25423 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25424 __builtin_ia32_scalefss_round): Remove.
25425 (__builtin_ia32_scalefsd_mask_round,
25426 __builtin_ia32_scalefss_mask_round): New intrinsics.
25427 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25428 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25429 ((match_operand:VF_128 2 "<round_nimm_predicate>"
25430 "<round_constraint>")): Changed to ...
25431 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25432 "<round_scalar_constraint>")): ... this.
25433 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25434 %0, %1, %2<round_op3>}"): Changed to ...
25435 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25436 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25437 %2<round_scalar_mask_op3>}"): ... this.
25438 * config/i386/subst.md (round_scalar_nimm_predicate): New.
25440 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
25442 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25443 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25444 (_mm_maskz_sqrt_round_ss): New intrinsics.
25445 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25446 (__builtin_ia32_sqrtsd_mask_round)
25447 (__builtin_ia32_sqrtss_mask_round): New builtins.
25448 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25449 (__builtin_ia32_sqrtss_round): Remove.
25450 (__builtin_ia32_sqrtsd_mask_round)
25451 (__builtin_ia32_sqrtss_mask_round): New builtins.
25452 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25453 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25454 ((match_operand:VF_128 1 "vector_operand"
25455 "xBm,<round_constraint>")): Changed to ...
25456 ((match_operand:VF_128 1 "vector_operand"
25457 "xBm,<round_scalar_constraint>")): ... this.
25458 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25459 %0, %2, %<iptr>1<round_op3>}): Changed to ...
25460 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25461 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25462 %<iptr>1<round_scalar_mask_op3>}): ... this.
25463 ((set_attr "prefix" "<round_prefix>")): Changed to ...
25464 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25466 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
25469 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25470 Cast vec_cmpeq result to correct type.
25471 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25472 Cast vec_cmpgt result to correct type.
25474 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
25476 * final.c (final_scan_insn_1): Renamed from...
25477 (final_scan_insn): ... this. New wrapper, to recover
25478 seen from the outermost call in recursive ones.
25479 * config/sparc/sparc.c (output_return): Drop seen from call.
25480 (output_sibcall): Likewise.
25481 * config/visium/visium.c (output_branch): Likewise.
25483 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
25485 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25488 2018-02-10 Alan Modra <amodra@gmail.com>
25491 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25492 Specify LR as an input.
25494 2018-02-10 Jakub Jelinek <jakub@redhat.com>
25497 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25498 remove_member_access_dummy_vars): New functions.
25499 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25500 lower_omp_1, execute_lower_omp): Use them.
25502 PR rtl-optimization/84308
25503 * shrink-wrap.c (spread_components): Release todo vector.
25505 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
25507 PR rtl-optimization/57193
25508 * ira-color.c (struct allocno_color_data): Add member
25509 conflict_allocno_hard_prefs.
25510 (update_conflict_allocno_hard_prefs): New.
25511 (bucket_allocno_compare_func): Add a preference based on
25512 conflict_allocno_hard_prefs.
25513 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25514 (color_allocnos): Remove a dead code. Initiate
25515 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
25517 2018-02-09 Jakub Jelinek <jakub@redhat.com>
25520 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25521 constraint from =wa to wa. Avoid a subreg on the output operand,
25522 instead use a pseudo and subreg it in a move.
25523 (p9_xxbrd_<mode>): Changed to ...
25524 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25525 (p9_xxbrd_v2df): New expander.
25526 (p9_xxbrw_<mode>): Changed to ...
25527 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25528 (p9_xxbrw_v4sf): New expander.
25530 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
25532 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25534 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
25537 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25538 multiply in 32-bit mode.
25539 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25540 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25543 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
25545 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25546 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25547 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25548 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25550 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
25553 * dwarf2out.c (is_trivial_indirect_ref): New function.
25554 (dwarf2out_late_global_decl): Do not generate a location
25555 attribute for variables that have a non-trivial DECL_VALUE_EXPR
25556 and that are not defined in the current unit.
25558 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
25560 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25561 instead of a libcall for UNORDERED.
25563 2018-02-09 Tamar Christina <tamar.christina@arm.com>
25566 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25567 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25569 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25572 * config/s390/s390.c (s390_set_current_function): Invoke
25573 s390_indirect_branch_settings also if fndecl didn't change.
25575 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
25577 * config/rs6000/rs6000.md (blockage): Set length to zero.
25579 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
25581 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25583 2018-02-09 Jakub Jelinek <jakub@redhat.com>
25586 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25587 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25591 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
25592 PARALLEL incoming that failed vt_get_decl_and_offset check.
25594 PR middle-end/84237
25595 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
25596 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
25598 (get_variable_section): For decls in named .bss* sections pass true as
25599 second argument to bss_initializer_p.
25601 2018-02-09 Marek Polacek <polacek@redhat.com>
25602 Jakub Jelinek <jakub@redhat.com>
25605 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
25606 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
25607 Sync some changes from cxx_fold_indirect_ref.
25609 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
25611 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
25613 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
25614 (BLOCK_INLINE_ENTRY_LABEL): New.
25615 (dwarf2out_var_location): Disregard inline entry markers.
25616 (inline_entry_data): New struct.
25617 (inline_entry_data_hasher): New hashtable type.
25618 (inline_entry_data_hasher::hash): New.
25619 (inline_entry_data_hasher::equal): New.
25620 (inline_entry_data_table): New variable.
25621 (add_high_low_attributes): Add DW_AT_entry_pc and
25622 DW_AT_GNU_entry_view attributes if a pending entry is found
25623 in inline_entry_data_table. Add old entry_pc attribute only
25624 if debug nonbinding markers are disabled.
25625 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
25626 markers are enabled.
25627 (block_within_block_p, dwarf2out_inline_entry): New.
25628 (dwarf2out_finish): Check that no entries remained in
25629 inline_entry_data_table.
25630 * final.c (reemit_insn_block_notes): Handle inline entry notes.
25631 (final_scan_insn, notice_source_line): Likewise.
25632 (rest_of_clean_state): Skip inline entry markers.
25633 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
25635 * gimple.c (gimple_build_debug_inline_entry): New.
25636 * gimple.h (enum gimple_debug_subcode): Add
25637 GIMPLE_DEBUG_INLINE_ENTRY.
25638 (gimple_build_debug_inline_entry): Declare.
25639 (gimple_debug_inline_entry_p): New.
25640 (gimple_debug_nonbind_marker_p): Adjust.
25641 * insn-notes.def (INLINE_ENTRY): New.
25642 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
25643 inline entry marker notes.
25644 (print_insn): Likewise.
25645 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
25646 (INSN_DEBUG_MARKER_KIND): Likewise.
25647 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
25648 * tree-inline.c (expand_call_inline): Build and insert
25649 debug_inline_entry stmt.
25650 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
25651 inline entry blocks early, if nonbind markers are enabled.
25652 (dump_scope_block): Dump fragment info.
25653 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
25654 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
25655 (gimple_build_debug_inline_entry): New.
25656 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
25657 Enable/disable inline entry points too.
25658 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
25659 (DEBUG_INSN): Describe inline entry markers.
25661 * common.opt (gvariable-location-views): New.
25662 (gvariable-location-views=incompat5): New.
25663 * config.in: Rebuilt.
25664 * configure: Rebuilt.
25665 * configure.ac: Test assembler for view support.
25666 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
25667 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
25668 * dwarf2out.c (var_loc_view): New typedef.
25669 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
25670 (dwarf2out_locviews_in_attribute): New.
25671 (dwarf2out_locviews_in_loclist): New.
25672 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
25673 (enum dw_line_info_opcode): Add LI_adv_address.
25674 (struct dw_line_info_table): Add view.
25675 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
25676 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
25677 (zero_view_p): New variable.
25678 (ZERO_VIEW_P): New macro.
25679 (output_asm_line_debug_info): New.
25680 (struct var_loc_node): Add view.
25681 (add_AT_view_list, AT_loc_list): New.
25682 (add_var_loc_to_decl): Add view param. Test it against last.
25683 (new_loc_list): Add view params. Record them.
25684 (AT_loc_list_ptr): Handle loc and view lists.
25685 (view_list_to_loc_list_val_node): New.
25686 (print_dw_val): Handle dw_val_class_view_list.
25687 (size_of_die): Likewise.
25688 (value_format): Likewise.
25689 (loc_list_has_views): New.
25690 (gen_llsym): Set vl_symbol too.
25691 (maybe_gen_llsym, skip_loc_list_entry): New.
25692 (dwarf2out_maybe_output_loclist_view_pair): New.
25693 (output_loc_list): Output view list or entries too.
25694 (output_view_list_offset): New.
25695 (output_die): Handle dw_val_class_view_list.
25696 (output_dwarf_version): New.
25697 (output_compilation_unit_header): Use it.
25698 (output_skeleton_debug_sections): Likewise.
25699 (output_rnglists, output_line_info): Likewise.
25700 (output_pubnames, output_aranges): Update version comments.
25701 (output_one_line_info_table): Output view numbers in asm comments.
25702 (dw_loc_list): Determine current endview, pass it to new_loc_list.
25703 Call maybe_gen_llsym.
25704 (loc_list_from_tree_1): Adjust.
25705 (add_AT_location_description): Create view list attribute if
25706 needed, check it's absent otherwise.
25707 (convert_cfa_to_fb_loc_list): Adjust.
25708 (maybe_emit_file): Call output_asm_line_debug_info for test.
25709 (dwarf2out_var_location): Reset views as needed. Precompute
25710 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
25711 attribute. Set view.
25712 (new_line_info_table): Reset next view.
25713 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
25714 (dwarf2out_source_line): Likewise. Output view resets and labels to
25715 the assembler, or select appropriate line info opcodes.
25716 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
25717 (optimize_string_length): Catch it. Adjust.
25718 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
25719 dw_val_class_view_list, and remove it if no longer needed.
25720 (hash_loc_list): Hash view numbers.
25721 (loc_list_hasher::equal): Compare them.
25722 (optimize_location_lists): Check whether a view list symbol is
25723 needed, and whether the locview attribute is present, and
25724 whether they match. Remove the locview attribute if no longer
25726 (index_location_lists): Call skip_loc_list_entry for test.
25727 (dwarf2out_finish): Call output_asm_line_debug_info for test.
25728 Use output_dwarf_version.
25729 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
25730 (struct dw_val_node): Add val_view_list.
25731 * final.c (SEEN_NEXT_VIEW): New.
25732 (set_next_view_needed): New.
25733 (clear_next_view_needed): New.
25734 (maybe_output_next_view): New.
25735 (final_start_function): Rename to...
25736 (final_start_function_1): ... this. Take pointer to FIRST,
25737 add SEEN parameter. Emit param bindings in the initial view.
25738 (final_start_function): Reintroduce SEEN-less interface.
25739 (final): Rename to...
25740 (final_1): ... this. Take SEEN parameter. Output final pending
25741 next view at the end.
25742 (final): Reintroduce seen-less interface.
25743 (final_scan_insn): Output pending next view before switching
25744 sections or ending a block. Mark the next view as needed when
25745 outputting variable locations. Notify debug backend of section
25746 changes, and of location view changes.
25747 (rest_of_handle_final): Adjust.
25748 * toplev.c (process_options): Autodetect value for debug variable
25749 location views option. Warn on incompat5 without -gdwarf-5.
25750 * doc/invoke.texi (gvariable-location-views): New.
25751 (gvariable-location-views=incompat5): New.
25752 (gno-variable-location-views): New.
25754 2018-02-08 David Malcolm <dmalcolm@redhat.com>
25756 PR tree-optimization/84136
25757 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
25758 that the result of find_edge is non-NULL.
25760 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
25763 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
25764 storing integer register in SImode. Fix cost of 256 and 512
25765 byte aligned SSE register store.
25767 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
25769 * config/i386/i386.c (ix86_multiplication_cost): Fix
25770 multiplication cost for TARGET_AVX512DQ.
25772 2018-02-08 Marek Polacek <polacek@redhat.com>
25774 PR tree-optimization/84238
25775 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
25778 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
25780 PR tree-optimization/84265
25781 * tree-vect-stmts.c (vectorizable_store): Don't treat
25782 VMAT_CONTIGUOUS accesses as grouped.
25783 (vectorizable_load): Likewise.
25785 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
25787 PR tree-optimization/81635
25788 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
25789 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
25790 (test_round_for_mask): New functions.
25791 (wide_int_cc_tests): Call test_round_for_mask.
25792 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
25793 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
25794 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
25795 range returned by get_range_info.
25797 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
25800 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
25801 * symtab.c: Include builtins.h
25802 (symtab_node::output_to_lto_symbol_table_p): Move here
25803 from lto-streamer-out.c:output_symbol_p.
25804 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
25805 (output_symbol_p): Move all logic to symtab.c
25806 (produce_symtab): Update.
25808 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
25810 * config/s390/s390-opts.h (enum indirect_branch): Define.
25811 * config/s390/s390-protos.h (s390_return_addr_from_memory)
25812 (s390_indirect_branch_via_thunk)
25813 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
25814 (enum s390_indirect_branch_type): Define.
25815 * config/s390/s390.c (struct s390_frame_layout, struct
25816 machine_function): Remove.
25817 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
25818 (indirect_branch_table_label_no, indirect_branch_table_name):
25820 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
25821 (enum s390_indirect_branch_option): Define.
25822 (s390_return_addr_from_memory): New function.
25823 (s390_handle_string_attribute): New function.
25824 (s390_attribute_table): Add new attribute handler.
25825 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
25826 (s390_indirect_branch_via_thunk): New function.
25827 (s390_indirect_branch_via_inline_thunk): New function.
25828 (s390_function_ok_for_sibcall): When jumping via thunk disallow
25829 sibling call optimization for non z10 compiles.
25830 (s390_emit_call): Force indirect branch target to be a single
25831 register. Add r1 clobber for non-z10 compiles.
25832 (s390_emit_epilogue): Emit return jump via return_use expander.
25833 (s390_reorg): Handle JUMP_INSNs as execute targets.
25834 (s390_option_override_internal): Perform validity checks for the
25835 new command line options.
25836 (s390_indirect_branch_attrvalue): New function.
25837 (s390_indirect_branch_settings): New function.
25838 (s390_set_current_function): Invoke s390_indirect_branch_settings.
25839 (s390_output_indirect_thunk_function): New function.
25840 (s390_code_end): Implement target hook.
25841 (s390_case_values_threshold): Implement target hook.
25842 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
25844 * config/s390/s390.h (struct s390_frame_layout)
25845 (struct machine_function): Move here from s390.c.
25846 (TARGET_INDIRECT_BRANCH_NOBP_RET)
25847 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
25848 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
25849 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
25850 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
25851 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
25852 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
25853 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
25854 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
25855 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
25856 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
25857 (mnemonic attribute): Add values which aren't recognized
25859 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
25860 pattern for branch conversion. Fix mnemonic attribute.
25861 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
25862 indirect branch via thunk if requested.
25863 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
25864 ("*indirect_jump"): Disable for branch conversion using out of
25866 ("indirect_jump_via_thunk<mode>_z10")
25867 ("indirect_jump_via_thunk<mode>")
25868 ("indirect_jump_via_inlinethunk<mode>_z10")
25869 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
25870 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
25871 ("casesi_jump_via_inlinethunk<mode>_z10")
25872 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
25873 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
25874 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
25875 ("*indirect2_jump"): Disable for branch conversion.
25876 ("casesi_jump"): Turn into expander and expand patterns for branch
25878 ("return_use"): New expander.
25879 ("*return"): Emit return via thunk and rename it to ...
25880 ("*return<mode>"): ... this one.
25881 * config/s390/s390.opt: Add new options and and enum for the
25884 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
25886 * lra-constraints.c (match_reload): Unconditionally use
25887 gen_lowpart_SUBREG, rather than selecting between that
25888 and equivalent gen_rtx_SUBREG code.
25890 2018-02-08 Richard Biener <rguenther@suse.de>
25892 PR tree-optimization/84233
25893 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
25894 changed flag instead of boguously re-using phi_inserted.
25896 2018-02-08 Martin Jambor <mjambor@suse.cz>
25898 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
25899 static local variables.
25901 2018-02-08 Richard Biener <rguenther@suse.de>
25903 PR tree-optimization/84278
25904 * tree-vect-stmts.c (vectorizable_store): When looking for
25905 smaller vector types to perform grouped strided loads/stores
25906 make sure the mode is supported by the target.
25907 (vectorizable_load): Likewise.
25909 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
25911 * config/aarch64/aarch64.c (aarch64_components_for_bb):
25912 Increase LDP/STP opportunities by adding adjacent callee-saves.
25914 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
25916 PR rtl-optimization/84068
25917 PR rtl-optimization/83459
25918 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
25920 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
25922 PR tree-optimization/84224
25923 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
25924 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
25925 non-zero arguments.
25927 2018-02-07 Iain Sandoe <iain@codesourcery.com>
25930 * config/rs6000/altivec.md (*restore_world): Remove LR use.
25931 * config/rs6000/predicates.md (restore_world_operation): Adjust op
25932 count, remove one USE.
25934 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
25936 * doc/install.texi (Configuration): Document the
25937 --with-long-double-format={ibm,ieee} PowerPC configuration
25941 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
25942 Convert from define_expand to be define_insn_and_split. Rework
25943 float/double/_Float128 conversions to QI/HI/SImode to work with
25944 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
25945 conversions to QI/HImode types did a store and then a load to
25946 truncate the value. For conversions to VSX registers, don't split
25947 the insn, instead emit the code directly. Use the code iterator
25948 any_fix to combine signed and unsigned conversions.
25949 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
25950 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
25951 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25952 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
25953 (fix_<mode>di2_hw): Likewise.
25954 (fixuns_<mode>di2_hw): Likewise.
25955 (fix_<mode>si2_hw): Likewise.
25956 (fixuns_<mode>si2_hw): Likewise.
25957 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
25958 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
25959 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
25960 fix<uns>_trunc<SFDF:mode>si2_p8.
25961 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
25962 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
25963 (fix<uns>_<mode>_mem): Likewise.
25964 (fctiw<u>z_<mode>_mem): Likewise.
25965 (fix<uns>_<mode>_mem): Likewise.
25966 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
25967 the register allocator from doing a direct move to the GPRs to do
25968 a store, and instead use the ISA 3.0 store byte/half-word from
25969 vector register instruction. For IEEE 128-bit floating point,
25970 also optimize stores of 32-bit ints.
25971 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
25973 2018-02-07 Alan Hayward <alan.hayward@arm.com>
25975 * genextract.c (push_pathstr_operand): New function to support
25977 (walk_rtx): Call push_pathstr_operand.
25978 (print_path): Support [a-zA-Z].
25980 2018-02-07 Richard Biener <rguenther@suse.de>
25982 PR tree-optimization/84037
25983 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
25984 (cse_and_gimplify_to_preheader): Declare.
25985 (vect_get_place_in_interleaving_chain): Likewise.
25986 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
25988 (_loop_vec_info::~_loop_vec_info): Delete it.
25989 (cse_and_gimplify_to_preheader): New function.
25990 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
25991 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
25992 (vectorizable_load): Likewise. For grouped stores always base
25993 the IV on the first element.
25994 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
25995 condition before gimplifying.
25997 2018-02-07 Jakub Jelinek <jakub@redhat.com>
25999 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26000 *DIV_EXPR and *MOD_EXPR.
26002 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
26005 * config/i386/i386.c (ix86_option_override_internal): Mask out
26006 the CF_SET bit when checking -fcf-protection.
26008 2018-02-07 Tom de Vries <tom@codesourcery.com>
26011 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26014 2018-02-07 Richard Biener <rguenther@suse.de>
26016 PR tree-optimization/84204
26017 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26020 PR tree-optimization/84205
26021 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26022 special-case isl_ast_op_zdiv_r.
26024 PR tree-optimization/84223
26025 * graphite-scop-detection.c (gather_bbs::before_dom_children):
26026 Only add conditions from within the region.
26027 (gather_bbs::after_dom_children): Adjust.
26029 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
26032 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26033 * config/avr/avr.md: Only post-reload split REG-REG moves if
26034 either register is GENERAL_REG_P.
26036 2018-02-07 Jakub Jelinek <jakub@redhat.com>
26038 PR tree-optimization/84235
26039 * tree-ssa-scopedtables.c
26040 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26041 if the subtraction is performed in floating point type where NaNs are
26042 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26043 build 1. Formatting fix.
26045 2018-02-06 Jakub Jelinek <jakub@redhat.com>
26048 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26049 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26050 and skip it regardless of bb boundaries. Use CALL_P macro,
26051 don't test INSN_P (insn) together with CALL_P or JUMP_P check
26052 unnecessarily, formatting fix.
26054 2018-02-06 Michael Collison <michael.collison@arm.com>
26056 * config/arm/thumb2.md:
26057 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26058 (*thumb_mov_notscc): Ditto.
26060 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
26063 * config/rs6000/rs6000.md (su code attribute): Use "u" for
26064 unsigned_fix, not "s".
26066 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26068 * configure.ac (gcc_fn_eh_frame_ro): New function.
26069 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26070 correct .eh_frame permissions.
26071 * configure: Regenerate.
26073 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
26075 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26076 irrelevant options.
26078 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26080 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26081 Display warning message for -mno-speculate-indirect-jumps.
26083 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
26085 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26087 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26089 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
26091 PR tree-optimization/84225
26092 * tree-eh.c (find_trapping_overflow): Only call
26093 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26095 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
26098 * config/i386/i386.c: Reimplement the check of possible options
26099 -mibt/-mshstk conbination. Change error messages.
26100 * doc/invoke.texi: Fix a typo: remove extra '='.
26102 2018-02-06 Marek Polacek <polacek@redhat.com>
26104 PR tree-optimization/84228
26105 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26107 2018-02-06 Tamar Christina <tamar.christina@arm.com>
26110 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26111 emitted arch directives.
26112 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26113 __ARM_FEATURE_COPROC before changing architectures.
26115 2018-02-06 Richard Biener <rguenther@suse.de>
26117 * config/i386/i386.c (print_reg): Fix typo.
26118 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26120 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
26122 * configure: Regenerate.
26124 2018-02-05 Martin Sebor <msebor@redhat.com>
26126 PR tree-optimization/83369
26127 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26130 2018-02-05 Martin Liska <mliska@suse.cz>
26132 * doc/invoke.texi: Cherry-pick upstream r323995.
26134 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
26136 * ira.c (ira_init_register_move_cost): Adjust comment.
26138 2018-02-05 Martin Liska <mliska@suse.cz>
26140 PR gcov-profile/84137
26141 * doc/gcov.texi: Fix typo in documentation.
26143 2018-02-05 Martin Liska <mliska@suse.cz>
26145 PR gcov-profile/83879
26146 * doc/gcov.texi: Document necessity of --dynamic-list-data when
26147 using dlopen functionality.
26149 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
26151 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26152 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26153 _mm_maskz_range_ss, _mm_mask_range_round_ss,
26154 _mm_maskz_range_round_ss): New intrinsics.
26155 (__builtin_ia32_rangesd128_round)
26156 (__builtin_ia32_rangess128_round): Remove.
26157 (__builtin_ia32_rangesd128_mask_round,
26158 __builtin_ia32_rangess128_mask_round): New builtins.
26159 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26160 __builtin_ia32_rangess128_round): Remove.
26161 (__builtin_ia32_rangesd128_mask_round,
26162 __builtin_ia32_rangess128_mask_round): New builtins.
26163 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26164 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26165 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26166 "<round_saeonly_constraint>")): Changed to ...
26167 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26168 "<round_saeonly_scalar_constraint>")): ... this.
26169 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26170 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26171 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26172 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26173 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26175 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
26177 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26179 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26180 Remove all values except native, 8540 and 8548.
26182 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
26184 * config/i386/i386.c (ix86_output_function_return): Pass
26185 INVALID_REGNUM, instead of -1, as invalid register number to
26186 indirect_thunk_name and output_indirect_thunk.
26188 2018-02-02 Julia Koval <julia.koval@intel.com>
26190 * config.gcc: Add -march=icelake.
26191 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26192 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26193 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26194 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26195 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26196 (processor_target_table): Add icelake.
26197 (ix86_option_override_internal): Handle new PTAs.
26198 (get_builtin_code_for_version): Handle icelake.
26199 (M_INTEL_COREI7_ICELAKE): New.
26200 (fold_builtin_cpu): Handle icelake.
26201 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26202 * doc/invoke.texi: Add -march=icelake.
26204 2018-02-02 Julia Koval <julia.koval@intel.com>
26206 * config/i386/i386.c (ix86_option_override_internal): Change flags type
26207 to wide_int_bitmask.
26208 * wide-int-bitmask.h: New.
26210 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
26213 * config/i386/i386.md: Replace Pmode with word_mode in
26214 builtin_setjmp_setup and builtin_longjmp to support x32.
26216 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
26220 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26222 (rs6000_supported_cpu_names): New static variable.
26223 (linux_cpu_translation_table): Likewise.
26224 (elf_platform) <cpu>: Define new static variable and use it.
26225 Translate kernel AT_PLATFORM name to canonical name if needed.
26226 Error if platform name is unknown.
26228 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
26231 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26233 2018-02-01 Jeff Law <law@redhat.com>
26236 * config/i386/i386.c (release_scratch_register_on_entry): Add new
26237 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
26238 the scratch if RELEASE_VIA_POP is false.
26239 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26240 If we have to save a temporary register, decrement SIZE appropriately.
26241 Pass new arguments to release_scratch_register_on_entry.
26242 (ix86_adjust_stack_and_probe): Likewise.
26243 (ix86_emit_probe_stack_range): Pass new arguments to
26244 release_scratch_register_on_entry.
26246 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
26248 PR rtl-optimization/84157
26249 * combine.c (change_zero_ext): Use REG_P predicate in
26250 front of HARD_REGISTER_P predicate.
26252 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
26254 * config/avr/avr.c (avr_option_override): Move disabling of
26255 -fdelete-null-pointer-checks to...
26256 * common/config/avr/avr-common.c (avr_option_optimization_table):
26259 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26261 PR tree-optimization/81635
26262 * tree-data-ref.c (split_constant_offset_1): For types that
26263 wrap on overflow, try to use range info to prove that wrapping
26266 2018-02-01 Renlin Li <renlin.li@arm.com>
26269 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26270 TAILCALL_ADDR_REGS.
26271 (aarch64_register_move_cost): Likewise.
26272 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26273 TAILCALL_ADDR_REGS.
26274 (REG_CLASS_NAMES): Likewise.
26275 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26276 TAILCALL_ADDR_REGS. Remove IP registers.
26277 * config/aarch64/aarch64.md (Ucs): Update register constraint.
26279 2018-02-01 Richard Biener <rguenther@suse.de>
26281 * domwalk.h (dom_walker::dom_walker): Add additional constructor
26282 for specifying RPO order and allow NULL for that.
26283 * domwalk.c (dom_walker::dom_walker): Likewise.
26284 (dom_walker::walk): Handle NULL RPO order.
26285 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26287 (rewrite_update_dom_walker): Likewise.
26288 (mark_def_dom_walker): Likewise.
26290 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26292 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26293 (aarch64_maybe_expand_sve_subreg_move): Declare.
26294 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26295 * config/aarch64/predicates.md (aarch64_any_register_operand): New
26297 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26298 that are semantically a reverse operation.
26299 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26300 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26301 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26303 (aarch64_can_change_mode_class): For big-endian, forbid changes
26304 between two SVE modes if they have different element sizes.
26306 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26308 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26309 the TImode handling for big-endian targets.
26311 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26313 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26314 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
26316 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26317 Remove BSWAP handing for big-endian targets and use the form of
26318 LD1RQ appropariate for the mode.
26320 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26322 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26323 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26324 duplicated element.
26326 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26329 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26330 check for operands that need to go through aarch64_sve_reload_be.
26332 2018-02-01 Jakub Jelinek <jakub@redhat.com>
26334 PR tree-optimization/81661
26335 PR tree-optimization/84117
26336 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26337 * tree-eh.c: Include gimplify.h.
26338 (find_trapping_overflow, replace_trapping_overflow,
26339 rewrite_to_non_trapping_overflow): New functions.
26340 * tree-vect-loop.c: Include tree-eh.h.
26341 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26342 * tree-data-ref.c: Include tree-eh.h.
26343 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26345 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
26347 PR rtl-optimization/84123
26348 * combine.c (change_zero_ext): Check if hard register satisfies
26349 can_change_dest_mode before calling gen_lowpart_SUBREG.
26351 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
26354 * ira.c (ira_init_register_move_cost): Remove assert.
26356 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26358 PR rtl-optimization/84071
26359 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26360 * doc/tm.texi: Regenerate.
26362 2018-01-31 Richard Biener <rguenther@suse.de>
26364 PR tree-optimization/84132
26365 * tree-data-ref.c (analyze_miv_subscript): Properly
26366 check whether evolution_function_is_affine_multivariate_p
26367 before calling gcd_of_steps_may_divide_p.
26369 2018-01-31 Julia Koval <julia.koval@intel.com>
26372 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26373 * config/i386/i386.md (rdpid_rex64) New.
26374 (rdpid): Make 32bit only.
26376 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
26379 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26380 an IDENTIFIER_NODE for FUNCTION_TYPE's.
26382 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26385 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
26387 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26389 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26391 PR rtl-optimization/84071
26392 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26393 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26395 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
26397 * config/arc/arc.c (arc_handle_aux_attribute): New function.
26398 (arc_attribute_table): Add 'aux' attribute.
26399 (arc_in_small_data_p): Consider aux like variables.
26400 (arc_is_aux_reg_p): New function.
26401 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26402 (arc_get_aux_arg): New function.
26403 (prepare_move_operands): Handle aux-register access.
26404 (arc_handle_aux_attribute): New function.
26405 * doc/extend.texi (ARC Variable attributes): Add subsection.
26407 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
26409 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26410 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26411 (arc_attribute_table): Add 'uncached' attribute.
26412 (arc_print_operand): Print '.di' flag for uncached memory
26414 (arc_in_small_data_p): Do not consider for small data the uncached
26416 (arc_is_uncached_mem_p): New function.
26417 * config/arc/predicates.md (compact_store_memory_operand): Check
26418 for uncached memory accesses.
26419 (nonvol_nonimm_operand): Likewise.
26420 * doc/extend.texi (ARC Type Attribute): New subsection.
26422 2018-01-31 Jakub Jelinek <jakub@redhat.com>
26425 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26426 falign-loops=): Add Optimization flag.
26428 2018-01-30 Jeff Law <law@redhat.com>
26431 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26432 INT_REGISTERS_SAVED. Check it prior to calling
26433 get_scratch_register_on_entry.
26434 (ix86_adjust_stack_and_probe): Similarly.
26435 (ix86_emit_probe_stack_range): Similarly.
26436 (ix86_expand_prologue): Corresponding changes.
26438 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26441 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26442 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26444 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
26447 * lra-constraints.c (curr_insn_transform): Process AND in the
26450 2018-01-30 Jakub Jelinek <jakub@redhat.com>
26452 PR rtl-optimization/83986
26453 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26454 dependence against last_pending_memory_flush in addition to
26455 pending_jump_insns.
26457 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
26459 PR tree-optimization/81611
26460 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26463 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26466 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26469 2018-01-30 Richard Biener <rguenther@suse.de>
26470 Jakub Jelinek <jakub@redhat.com>
26472 PR tree-optimization/84111
26473 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26474 inner loops added during recursion, as they don't have up-to-date
26477 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
26480 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26481 (can_inline_edge_by_limits_p): ... here.
26482 (can_early_inline_edge_p, check_callers,
26483 update_caller_keys, update_callee_keys, recursive_inlining,
26484 add_new_edges_to_heap, speculation_useful_p,
26485 inline_small_functions,
26486 inline_small_functions, flatten_function,
26487 inline_to_all_callers_1): Update.
26489 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
26491 * profile-count.c (profile_count::combine_with_ipa_count): Handle
26494 2018-01-30 Richard Biener <rguenther@suse.de>
26496 PR tree-optimization/83008
26497 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26498 invariant and constant vector uses in stmts when they need
26499 more than one stmt.
26501 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26504 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26505 * configure: Regenerate.
26507 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
26509 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26511 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26512 Use gen_rtx_REG rather than gen_lowpart.
26514 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
26516 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26517 rather than 0 when creating partial subregs.
26519 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
26521 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26524 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
26527 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26528 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26529 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26530 flags. This restores the settings used before the 2017-07-24.
26531 Turning off pre increment/decrement/modify allows IVOPTS to
26532 optimize DF/SF loops where the index is an int.
26534 2018-01-29 Richard Biener <rguenther@suse.de>
26535 Kelvin Nilsen <kelvin@gcc.gnu.org>
26538 * tree-vect-stmts.c (vectorizable_call): Don't call
26539 targetm.vectorize_builtin_md_vectorized_function if callee is
26542 2018-01-22 Carl Love <cel@us.ibm.com>
26544 * doc/extend.tex: Fix typo in second arg in
26545 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26547 2018-01-29 Richard Biener <rguenther@suse.de>
26549 PR tree-optimization/84086
26550 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26551 (flush_ssaname_freelist): When SSA names were released reset
26552 the SCEV hash table.
26554 2018-01-29 Richard Biener <rguenther@suse.de>
26556 PR tree-optimization/84057
26557 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26558 removed paths when removing edges.
26560 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
26562 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26563 -mfunction-return=@var{choice}.
26565 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
26567 PR diagnostic/84034
26568 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26569 Handle CR like TAB.
26570 (layout::print_source_line): Likewise.
26571 (test_get_line_width_without_trailing_whitespace): Add test cases.
26573 2018-01-27 Jakub Jelinek <jakub@redhat.com>
26575 PR middle-end/84040
26576 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26579 2018-01-26 Jim Wilson <jimw@sifive.com>
26581 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26583 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26586 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26588 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
26589 and CMP + SUB-immediate -> SUBS.
26591 2018-01-26 Martin Sebor <msebor@redhat.com>
26593 PR tree-optimization/83896
26594 * tree-ssa-strlen.c (get_string_len): Rename...
26595 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
26596 Avoid assuming length is constant.
26597 (handle_char_store): Use HOST_WIDE_INT for string length.
26599 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
26602 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
26603 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
26605 2018-01-26 Richard Biener <rguenther@suse.de>
26607 PR rtl-optimization/84003
26608 * dse.c (record_store): Only record redundant stores when
26609 the earlier store aliases at least all accesses the later one does.
26611 2018-01-26 Jakub Jelinek <jakub@redhat.com>
26613 PR rtl-optimization/83985
26614 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
26615 REG_CFA_RESTORE insns.
26616 (delete_unmarked_insns): Don't ignore separate shrink wrapping
26617 REG_CFA_RESTORE insns here.
26620 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
26621 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
26623 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
26625 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
26626 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
26627 (arc_init): Likewise.
26628 (arc_override_options): Likewise.
26629 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
26631 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
26633 * config/arc/arc.h (TARGET_DBNZ): Define.
26634 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
26635 properly set the tune attribute.
26636 (dbnz): Use TARGET_DBNZ guard.
26637 * config/arc/arc.opt (mtune): Add core3 option.
26639 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
26641 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
26642 recognize new pic like addresses.
26643 (arc_delegitimize_address): Clean up.
26645 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
26647 * config/arc/arc-arches.def: Option mrf16 valid for all
26649 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
26650 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
26651 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
26652 * config/arc/arc-tables.opt: Regenerate.
26653 * config/arc/arc.c (arc_conditional_register_usage): Handle
26654 reduced register file case.
26655 (arc_file_start): Set must have build attributes.
26656 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
26657 mrf16 option value.
26658 * config/arc/arc.opt (mrf16): Add new option.
26659 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
26660 * config/arc/genmultilib.awk: Handle new mrf16 option.
26661 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
26662 * config/arc/t-multilib: Regenerate.
26663 * doc/invoke.texi (ARC Options): Document mrf16 option.
26665 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
26667 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
26668 * config/arc/arc.c (arc_handle_secure_attribute): New function.
26669 (arc_attribute_table): Add 'secure_call' attribute.
26670 (arc_print_operand): Print secure call operand.
26671 (arc_function_ok_for_sibcall): Don't optimize tail calls when
26673 (arc_is_secure_call_p): New function. * config/arc/arc.md
26674 (call_i): Add support for sjli instruction.
26675 (call_value_i): Likewise.
26676 * config/arc/constraints.md (Csc): New constraint.
26678 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
26679 John Eric Martin <John.Martin@emmicro-us.com>
26681 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
26682 * config/arc/arc.c (_arc_jli_section): New struct.
26683 (arc_jli_section): New type.
26684 (rc_jli_sections): New static variable.
26685 (arc_handle_jli_attribute): New function.
26686 (arc_attribute_table): Add jli_always and jli_fixed attribute.
26687 (arc_file_end): New function.
26688 (TARGET_ASM_FILE_END): Define.
26689 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
26690 (arc_add_jli_section): New function.
26691 (jli_call_scan): Likewise.
26692 (arc_reorg): Call jli_call_scan.
26693 (arc_output_addsi): Remove 'S' from printing asm operand.
26694 (arc_is_jli_call_p): New function.
26695 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
26697 (movhi_insn): Likewise.
26698 (movsi_insn): Likewise.
26699 (movsi_set_cc_insn): Likewise.
26700 (loadqi_update): Likewise.
26701 (load_zeroextendqisi_update): Likewise.
26702 (load_signextendqisi_update): Likewise.
26703 (loadhi_update): Likewise.
26704 (load_zeroextendhisi_update): Likewise.
26705 (load_signextendhisi_update): Likewise.
26706 (loadsi_update): Likewise.
26707 (loadsf_update): Likewise.
26708 (movsicc_insn): Likewise.
26709 (bset_insn): Likewise.
26710 (bxor_insn): Likewise.
26711 (bclr_insn): Likewise.
26712 (bmsk_insn): Likewise.
26713 (bicsi3_insn): Likewise.
26714 (cmpsi_cc_c_insn): Likewise.
26715 (movsi_ne): Likewise.
26716 (movsi_cond_exec): Likewise.
26717 (clrsbsi2): Likewise.
26718 (norm_f): Likewise.
26728 (call_i): Remove 'S' asm letter, add jli instruction.
26729 (call_value_i): Likewise.
26730 * config/arc/arc.op (mjli-always): New option.
26731 * config/arc/constraints.md (Cji): New constraint.
26732 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
26734 (subsf3_fpx): Likewise.
26735 (mulsf3_fpx): Likewise.
26736 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
26738 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
26739 function attrbutes.
26740 * doc/invoke.texi (ARC): Document mjli-always option.
26742 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
26744 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
26745 avoid addition with 0 and use incw and decw where possible.
26747 2018-01-26 Richard Biener <rguenther@suse.de>
26749 PR tree-optimization/81082
26750 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
26751 association if it requires casting to unsigned.
26752 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
26753 from fold_plusminus_mult_expr to catch important cases late when
26754 range info is available.
26756 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26758 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
26759 * configure.ac (hidden_linkonce): New test.
26760 * configure: Regenerate.
26761 * config.in: Regenerate.
26763 2018-01-26 Julia Koval <julia.koval@intel.com>
26765 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
26766 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
26767 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
26768 _mm_mask_bitshuffle_epi64_mask): Fix type.
26769 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
26770 USI_FTYPE_V4DI_V4DI_USI): Remove.
26771 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
26772 __builtin_ia32_vpshufbitqmb256_mask,
26773 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
26774 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
26775 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
26777 2018-01-26 Alan Modra <amodra@gmail.com>
26780 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
26781 UNSPEC_VBPERMQ. Sort other unspecs.
26783 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
26785 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
26787 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
26789 PR middle-end/83055
26790 * predict.c (drop_profile): Do not push/pop cfun; update also
26792 (handle_missing_profiles): Fix logic looking for zero profiles.
26794 2018-01-25 Jakub Jelinek <jakub@redhat.com>
26796 PR middle-end/83977
26797 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
26798 on functions with #pragma omp declare simd or functions with simd
26800 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
26801 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
26802 Remove trailing \n from warning_at calls.
26804 2018-01-25 Tom de Vries <tom@codesourcery.com>
26807 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26808 for neutered workers.
26810 2018-01-24 Joseph Myers <joseph@codesourcery.com>
26813 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
26814 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
26816 2018-01-24 Jeff Law <law@redhat.com>
26819 * i386.c (get_probe_interval): Move to earlier point.
26820 (ix86_compute_frame_layout): If -fstack-clash-protection and
26821 the frame is larger than the probe interval, then use pushes
26822 to save registers rather than reg->mem moves.
26823 (ix86_expand_prologue): Remove conditional for int_registers_saved
26826 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
26829 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
26830 min/max for never referenced object.
26832 2018-01-24 Jakub Jelinek <jakub@redhat.com>
26834 PR middle-end/83977
26835 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
26837 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
26838 attributes from DECL_ATTRIBUTES (decl) without affecting
26839 DECL_ATTRIBUTES (current_function_decl).
26840 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
26841 functions with non-NULL DECL_ABSTRACT_ORIGIN.
26843 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
26845 PR tree-optimization/83979
26846 * fold-const.c (fold_comparison): Use constant_boolean_node
26847 instead of boolean_{true,false}_node.
26849 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
26851 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
26854 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26856 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
26857 Simplify the clause that sets the length attribute.
26858 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
26859 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
26860 clause that sets the length attribute.
26861 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
26863 2018-01-24 Tom de Vries <tom@codesourcery.com>
26866 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
26867 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
26868 Add strict parameter.
26869 (prevent_branch_around_nothing): Insert dummy insn between branch to
26870 label and label with no ptx insn inbetween.
26871 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
26873 2018-01-24 Tom de Vries <tom@codesourcery.com>
26876 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
26877 for neutered threads in warp.
26878 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
26880 2018-01-24 Richard Biener <rguenther@suse.de>
26882 PR tree-optimization/83176
26883 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
26886 2018-01-24 Richard Biener <rguenther@suse.de>
26888 PR tree-optimization/82819
26889 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
26890 code generating pluses that are no-ops in the target precision.
26892 2018-01-24 Richard Biener <rguenther@suse.de>
26894 PR middle-end/84000
26895 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
26897 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
26899 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
26900 to merge probabilities.
26901 * predict.c (probably_never_executed): Also mark as cold functions
26902 with global 0 profile and guessed local profile.
26903 * profile-count.c (profile_probability::combine_with_count): New
26905 * profile-count.h (profile_probability::operator*,
26906 profile_probability::operator*=, profile_probability::operator/,
26907 profile_probability::operator/=): Reduce precision to adjusted
26908 and set value to guessed on contradictory divisions.
26909 (profile_probability::combine_with_freq): Remove.
26910 (profile_probability::combine_wiht_count): Declare.
26911 (profile_count::force_nonzero):: Set to adjusted.
26912 (profile_count::probability_in):: Set quality to adjusted.
26913 * tree-ssa-tail-merge.c (replace_block_by): Use
26914 combine_with_count.
26916 2018-01-23 Andrew Waterman <andrew@sifive.com>
26917 Jim Wilson <jimw@sifive.com>
26919 * config/riscv/riscv.c (riscv_stack_boundary): New.
26920 (riscv_option_override): Set riscv_stack_boundary. Handle
26921 riscv_preferred_stack_boundary_arg.
26922 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
26923 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
26924 (STACK_BOUNDARY): Set to riscv_stack_boundary.
26925 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
26926 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
26927 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
26929 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
26932 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
26933 of struct ix86_frame.
26934 (ix86_expand_epilogue): Likewise. Add a local variable for
26935 the reg_save_offset field in struct ix86_frame.
26937 2018-01-23 Bin Cheng <bin.cheng@arm.com>
26939 PR tree-optimization/82604
26940 * tree-loop-distribution.c (enum partition_kind): New enum item
26941 PKIND_PARTIAL_MEMSET.
26942 (partition_builtin_p): Support above new enum item.
26943 (generate_code_for_partition): Ditto.
26944 (compute_access_range): Differentiate cases that equality can be
26945 proven at all loops, the innermost loops or no loops.
26946 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
26947 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
26948 (finalize_partitions, distribute_loop): Don't fuse partition of
26949 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
26950 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
26951 parloop is enabled.
26953 2018-01-23 Martin Liska <mliska@suse.cz>
26955 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
26956 order to ignore the predictor.
26957 (PRED_POLYMORPHIC_CALL): Likewise.
26958 (PRED_RECURSIVE_CALL): Likewise.
26960 2018-01-23 Martin Liska <mliska@suse.cz>
26962 * tree-profile.c (tree_profiling): Print function header to
26963 aware reader which function we are working on.
26964 * value-prof.c (gimple_find_values_to_profile): Do not print
26965 not interesting value histograms.
26967 2018-01-23 Martin Liska <mliska@suse.cz>
26969 * profile-count.h (enum profile_quality): Add
26970 profile_uninitialized as the first value. Do not number values
26971 as they are zero based.
26972 (profile_count::verify): Update sanity check.
26973 (profile_probability::verify): Likewise.
26975 2018-01-23 Nathan Sidwell <nathan@acm.org>
26977 * doc/invoke.texi (ffor-scope): Deprecate.
26979 2018-01-23 David Malcolm <dmalcolm@redhat.com>
26981 PR tree-optimization/83510
26982 * domwalk.c (set_all_edges_as_executable): New function.
26983 (dom_walker::dom_walker): Convert bool param
26984 "skip_unreachable_blocks" to enum reachability. Move setup of
26985 edge flags to set_all_edges_as_executable and only do it when
26986 reachability is REACHABLE_BLOCKS.
26987 * domwalk.h (enum dom_walker::reachability): New enum.
26988 (dom_walker::dom_walker): Convert bool param
26989 "skip_unreachable_blocks" to enum reachability.
26990 (set_all_edges_as_executable): New decl.
26991 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
26992 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
26994 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
26995 but converting true to REACHABLE_BLOCKS.
26996 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
26998 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
26999 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27000 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27002 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27003 if check_all_array_refs will be called.
27005 2018-01-23 David Malcolm <dmalcolm@redhat.com>
27007 * tree.c (selftest::test_location_wrappers): Add more test
27010 2018-01-23 David Malcolm <dmalcolm@redhat.com>
27012 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27013 (selftest::test_bit_in_range): Likewise.
27015 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
27018 * doc/sourcebuild.texi (vect_float): Say that the selector
27019 only describes the situation when -funsafe-math-optimizations is on.
27020 (vect_float_strict): Document.
27022 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
27024 PR tree-optimization/83965
27025 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27026 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27027 instead of checking only for a reduction.
27028 (vect_recog_widen_sum_pattern): Likewise.
27030 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
27032 * predict.c (probably_never_executed): Only use precise profile info.
27033 (compute_function_frequency): Skip after inlining hack since we now
27034 have quality checking.
27036 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
27038 * profile-count.h (profile_probability::very_unlikely,
27039 profile_probability::unlikely, profile_probability::even): Set
27040 precision to guessed.
27042 2018-01-23 Richard Biener <rguenther@suse.de>
27044 PR tree-optimization/83963
27045 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27046 Properly terminate dominator walk when crossing the exit edge not
27047 when visiting its source block.
27049 2018-01-23 Jakub Jelinek <jakub@redhat.com>
27052 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27053 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27055 2018-01-22 Jakub Jelinek <jakub@redhat.com>
27057 PR tree-optimization/83957
27058 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
27059 semicolon after for body surrounded by braces.
27061 PR tree-optimization/83081
27062 * profile-count.h (profile_probability::split): New method.
27063 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27064 Use profile_probability::split.
27065 (do_compare_rtx_and_jump): Fix adjustment of probabilities
27066 when splitting a single conditional jump into 2.
27068 2018-01-22 David Malcolm <dmalcolm@redhat.com>
27070 PR tree-optimization/69452
27071 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27074 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27076 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27077 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27078 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27080 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27082 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27084 * config/rl78/rl78.md (movdi): New define_expand.
27085 * config/rl78/rl78.c (rl78_split_movdi): New function.
27087 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
27090 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27092 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27093 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27094 128-bit to produce an UNSPEC move to get the double word with the
27095 signbit and then a shift directly to do signbit.
27096 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27097 implementation with a new version that just does either a direct
27098 move or a regular move. Move memory interface to separate insns.
27099 Move insns so they are next to the expander.
27100 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27101 with signbit move. Split big and little endian case.
27102 (signbit<mode>2_dm_mem_le): Likewise.
27103 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27104 (signbit<mode>2_dm2): Likewise.
27106 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27108 * config/rl78/rl78.md (anddi3): New define_expand.
27110 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27112 * config/rl78/rl78.md (umindi3): New define_expand.
27114 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27116 * config/rl78/rl78.md (smindi3): New define_expand.
27118 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27120 * config/rl78/rl78.md (smaxdi3): New define_expand.
27122 2018-01-22 Carl Love <cel@us.ibm.com>
27124 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27125 LVX_V1TI): Add macro expansion.
27126 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27127 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27128 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27129 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27130 Change check to determine if the instruction is a byte reversing
27131 entry. Fix typo in comment.
27132 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27133 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27134 Add def_builtin calls for new builtins.
27135 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27136 Add define_insn expansion.
27138 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27140 * config/rl78/rl78.md (umaxdi3): New define_expand.
27142 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27144 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27145 for non-QImode registers.
27147 2018-01-22 Richard Biener <rguenther@suse.de>
27149 PR tree-optimization/83963
27150 * graphite-scop-detection.c (scop_detection::get_sese): Delay
27151 including the loop exit block.
27152 (scop_detection::merge_sese): Likewise.
27153 (scop_detection::add_scop): Do it here instead.
27155 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27157 * doc/sourcebuild.texi (arm_softfloat): Document.
27159 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
27162 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27163 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27164 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27166 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27167 David Edelsohn <dje.gcc@gmail.com>
27170 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27171 Change "crset eq" to "crset 2".
27172 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27173 (*call_indirect_aix<mode>_nospec): Likewise.
27174 (*call_value_indirect_aix<mode>_nospec): Likewise.
27175 (*call_indirect_elfv2<mode>_nospec): Likewise.
27176 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27177 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27178 change assembly output from . to $.
27179 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27180 (indirect_jump<mode>_nospec): Change assembly output from . to $.
27181 (*tablejump<mode>_internal1_nospec): Likewise.
27183 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
27186 * config/sh/sh_optimize_sett_clrt.cc:
27187 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27189 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
27191 PR tree-optimization/83940
27192 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27193 offset_dt to vect_constant_def rather than vect_unknown_def_type.
27194 (vect_check_load_store_mask): Add a mask_dt_out parameter and
27195 use it to pass back the definition type.
27196 (vect_check_store_rhs): Likewise rhs_dt_out.
27197 (vect_build_gather_load_calls): Add a mask_dt argument and use
27198 it instead of a call to vect_is_simple_use.
27199 (vectorizable_store): Update calls to vect_check_load_store_mask
27200 and vect_check_store_rhs. Use the dt returned by the latter instead
27201 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
27202 instead of calls to vect_is_simple_use. Pass the scalar rather
27203 than the vector operand to vect_is_simple_use when handling
27204 second and subsequent copies of an rhs value.
27205 (vectorizable_load): Update calls to vect_check_load_store_mask
27206 and vect_build_gather_load_calls. Use the cached mask_dt and
27207 gs_info.offset_dt instead of calls to vect_is_simple_use.
27209 2018-01-20 Jakub Jelinek <jakub@redhat.com>
27211 PR middle-end/83945
27212 * tree-emutls.c: Include gimplify.h.
27213 (lower_emutls_2): New function.
27214 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27215 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27216 it before further processing.
27219 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27220 UINTVAL (trueop1) instead of INTVAL (op1).
27222 2018-01-19 Jakub Jelinek <jakub@redhat.com>
27226 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27227 INCOMING_FRAME_SP_OFFSET if not defined.
27228 (scan_trace): Add ENTRY argument. If true and
27229 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27230 emit a note to adjust the CFA offset.
27231 (create_cfi_notes): Adjust scan_trace callers.
27232 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27233 INCOMING_FRAME_SP_OFFSET in the CIE.
27234 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27235 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27237 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27238 * doc/tm.texi: Regenerated.
27240 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
27242 PR rtl-optimization/83147
27243 * lra-constraints.c (remove_inheritance_pseudos): Use
27244 lra_substitute_pseudo_within_insn.
27246 2018-01-19 Tom de Vries <tom@codesourcery.com>
27247 Cesar Philippidis <cesar@codesourcery.com>
27250 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27252 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
27255 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27256 spaces for function labels.
27258 2018-01-19 Martin Liska <mliska@suse.cz>
27260 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27261 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27262 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27263 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27264 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27265 (PRED_CONST_RETURN): Change from 69 to 65.
27266 (PRED_NULL_RETURN): Change from 91 to 71.
27267 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27268 (PRED_LOOP_GUARD): Change from 66 to 73.
27270 2018-01-19 Martin Liska <mliska@suse.cz>
27272 * predict.c (predict_insn_def): Add new assert.
27273 (struct branch_predictor): Change type to signed integer.
27274 (test_prediction_value_range): Amend test to cover
27275 PROB_UNINITIALIZED.
27276 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27277 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27278 (PRED_LOOP_ITERATIONS_MAX): Likewise.
27279 (PRED_LOOP_IV_COMPARE): Likewise.
27280 * predict.h (PROB_UNINITIALIZED): Define new constant.
27282 2018-01-19 Martin Liska <mliska@suse.cz>
27284 * predict.c (dump_prediction): Add new format for
27285 analyze_brprob.py script which is enabled with -details
27287 * profile-count.h (precise_p): New function.
27289 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
27291 PR tree-optimization/83922
27292 * tree-vect-loop.c (vect_verify_full_masking): Return false if
27293 there are no statements that need masking.
27294 (vect_active_double_reduction_p): New function.
27295 (vect_analyze_loop_operations): Use it when handling phis that
27296 are not in the loop header.
27298 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
27300 PR tree-optimization/83914
27301 * tree-vect-loop.c (vectorizable_induction): Don't convert
27302 init_expr or apply the peeling adjustment for inductions
27303 that are nested within the vectorized loop.
27305 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27307 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27310 2018-01-18 Jakub Jelinek <jakub@redhat.com>
27314 * function.h (gimplify_parameters): Add gimple_seq * argument.
27315 * function.c: Include gimple.h and options.h.
27316 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27317 for the added local temporaries if needed.
27318 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27319 if there are any parameter cleanups, wrap whole body into a
27320 try/finally with the cleanups.
27322 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
27325 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27326 Use GET_MODE_CLASS for scalar floating point.
27328 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
27332 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27333 Fix call of call_cgraph_insertion_hooks.
27335 2018-01-18 Martin Sebor <msebor@redhat.com>
27337 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27339 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
27342 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27345 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
27348 * common.opt: (-ffile-prefix-map): New option.
27349 * opts.c (common_handle_option): Defer it.
27350 * opts-global.c (handle_common_deferred_options): Handle it.
27351 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27352 * file-prefix-map.h: New file.
27353 (remap_debug_filename, add_debug_prefix_map): ...here.
27354 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27355 * final.c (debug_prefix_map, add_debug_prefix_map
27356 remap_debug_filename): Move to...
27357 * file-prefix-map.c: New file.
27358 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27359 generalize, get rid of alloca(), use strrchr() instead of strchr().
27360 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27361 Implement in terms of add_prefix_map().
27362 (remap_macro_filename, remap_debug_filename): Implement in term of
27364 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27365 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27366 * dbxout.c: Include file-prefix-map.h.
27367 * varasm.c: Likewise.
27368 * vmsdbgout.c: Likewise.
27369 * xcoffout.c: Likewise.
27370 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27371 * doc/cppopts.texi (-fmacro-prefix-map): Document.
27372 * doc/invoke.texi (-ffile-prefix-map): Document.
27373 (-fdebug-prefix-map): Update description.
27375 2018-01-18 Martin Liska <mliska@suse.cz>
27377 * config/i386/i386.c (indirect_thunk_name): Document that also
27379 (output_indirect_thunk): Document why both instructions
27380 (pause and lfence) are generated.
27382 2018-01-18 Richard Biener <rguenther@suse.de>
27384 PR tree-optimization/83887
27385 * graphite-scop-detection.c
27386 (scop_detection::get_nearest_dom_with_single_entry): Remove.
27387 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27388 (scop_detection::merge_sese): Re-implement with a flood-fill
27389 algorithm that properly finds a SESE region if it exists.
27391 2018-01-18 Jakub Jelinek <jakub@redhat.com>
27394 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27395 pointer_diff optimizations use view_convert instead of convert.
27397 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27399 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27400 Generate different code for -mno-speculate-indirect-jumps.
27401 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27402 (*call_indirect_aix<mode>): Disable for
27403 -mno-speculate-indirect-jumps.
27404 (*call_indirect_aix<mode>_nospec): New define_insn.
27405 (*call_value_indirect_aix<mode>): Disable for
27406 -mno-speculate-indirect-jumps.
27407 (*call_value_indirect_aix<mode>_nospec): New define_insn.
27408 (*sibcall_nonlocal_sysv<mode>): Generate different code for
27409 -mno-speculate-indirect-jumps.
27410 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27412 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
27414 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27415 long double type, set the flags for noting the default long double
27416 type, even if we don't pass or return a long double type.
27418 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
27421 * ipa-inline.c (flatten_function): Do not overwrite final inlining
27424 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
27426 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27427 support for merge[hl].
27428 (fold_mergehl_helper): New helper function.
27429 (tree-vector-builder.h): New #include for tree_vector_builder usage.
27430 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27431 (altivec_vmrglw_direct): Add xxmrglw insn.
27433 2018-01-17 Andrew Waterman <andrew@sifive.com>
27435 * config/riscv/riscv.c (riscv_conditional_register_usage): If
27436 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27438 2018-01-17 David Malcolm <dmalcolm@redhat.com>
27441 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27442 call the lto_location_cache before reading the
27443 DECL_SOURCE_LOCATION of the types.
27445 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
27446 Richard Sandiford <richard.sandiford@linaro.org>
27448 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27449 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27450 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27451 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27452 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27454 * config/aarch64/constraints.md (aarch64_movti_operand):
27456 * config/aarch64/predicates.md (Uti): Add new constraint.
27458 2018-01-17 Carl Love <cel@us.ibm.com>
27460 * config/rs6000/vsx.md (define_expand xl_len_r,
27461 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27462 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27464 (define_expand, define_insn): Move the shift left from the
27465 define_insn to the define_expand for lxvl and stxvl instructions.
27466 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27467 and XL_LEN_R definitions to PURE.
27469 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
27471 * config/i386/i386.c (indirect_thunk_name): Declare regno
27472 as unsigned int. Compare regno with INVALID_REGNUM.
27473 (output_indirect_thunk): Ditto.
27474 (output_indirect_thunk_function): Ditto.
27475 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
27476 in the call to output_indirect_thunk_function.
27478 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
27480 PR middle-end/83884
27481 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27482 rather than the size of inner_type to determine the stack slot size
27483 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27485 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
27488 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27491 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
27493 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27494 endian Linux systems to optionally enable multilibs for selecting
27495 the long double type if the user configured an explicit type.
27496 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27497 have no long double multilibs if not defined.
27498 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27499 warn if the user used -mabi={ieee,ibm}longdouble and we built
27500 multilibs for long double.
27501 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27502 appropriate multilib option.
27503 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27505 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27506 for building long double multilibs.
27507 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27509 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
27511 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27514 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27516 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27519 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27522 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27525 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
27527 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27528 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27530 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
27532 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27533 different rtl trees depending on TARGET_64BIT.
27534 (rs6000_gen_lvx): Likewise.
27536 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
27538 * config/visium/visium.md (nop): Tweak comment.
27539 (hazard_nop): Likewise.
27541 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27543 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27544 -mspeculate-indirect-jumps.
27545 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27546 for -mno-speculate-indirect-jumps.
27547 (*call_indirect_elfv2<mode>_nospec): New define_insn.
27548 (*call_value_indirect_elfv2<mode>): Disable for
27549 -mno-speculate-indirect-jumps.
27550 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27551 (indirect_jump): Emit different RTL for
27552 -mno-speculate-indirect-jumps.
27553 (*indirect_jump<mode>): Disable for
27554 -mno-speculate-indirect-jumps.
27555 (*indirect_jump<mode>_nospec): New define_insn.
27556 (tablejump): Emit different RTL for
27557 -mno-speculate-indirect-jumps.
27558 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27559 (tablejumpsi_nospec): New define_expand.
27560 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27561 (tablejumpdi_nospec): New define_expand.
27562 (*tablejump<mode>_internal1): Disable for
27563 -mno-speculate-indirect-jumps.
27564 (*tablejump<mode>_internal1_nospec): New define_insn.
27565 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27568 2018-01-16 Artyom Skrobov tyomitch@gmail.com
27570 * caller-save.c (insert_save): Drop unnecessary parameter. All
27573 2018-01-16 Jakub Jelinek <jakub@redhat.com>
27574 Richard Biener <rguenth@suse.de>
27577 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27578 return early, inline manually is_gimple_sizepos. Make sure if we
27579 call gimplify_expr we don't end up with a gimple constant.
27580 * tree.c (variably_modified_type_p): Don't return true for
27581 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
27582 * gimplify.h (is_gimple_sizepos): Remove.
27584 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
27586 PR tree-optimization/83857
27587 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27588 vectorizable_live_operation for pure SLP statements.
27589 (vectorizable_live_operation): Handle PHIs.
27591 2018-01-16 Richard Biener <rguenther@suse.de>
27593 PR tree-optimization/83867
27594 * tree-vect-stmts.c (vect_transform_stmt): Precompute
27595 nested_in_vect_loop_p since the scalar stmt may get invalidated.
27597 2018-01-16 Jakub Jelinek <jakub@redhat.com>
27600 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
27601 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
27602 If off is not INTEGER_CST, issue a may not be aligned warning
27603 rather than isn't aligned. Use isn%'t rather than isn't.
27604 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
27606 <case MULT_EXPR>: Improve the case when bottom and one of the
27607 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
27608 operand, in that case check if the other operand is multiple of
27609 bottom divided by the INTEGER_CST operand.
27611 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
27614 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
27615 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
27616 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
27617 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
27618 * config/pa/pa.c (pa_function_arg_advance): Likewise.
27619 (pa_function_arg, pa_arg_partial_bytes): Likewise.
27620 (pa_function_arg_size): New function.
27622 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
27624 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
27625 in a separate statement.
27627 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
27629 PR tree-optimization/83847
27630 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
27631 group gathers and scatters.
27633 2018-01-16 Jakub Jelinek <jakub@redhat.com>
27635 PR rtl-optimization/86620
27636 * params.def (max-sched-ready-insns): Bump minimum value to 1.
27638 PR rtl-optimization/83213
27639 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
27640 to last if both are JUMP_INSNs.
27642 PR tree-optimization/83843
27643 * gimple-ssa-store-merging.c
27644 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
27645 store_immediate_info for bswap/nop orig_stores.
27647 2018-01-15 Andrew Waterman <andrew@sifive.com>
27649 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
27651 <UDIV>: Increase cost if !TARGET_DIV.
27653 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
27655 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
27656 (define_attr "cr_logical_3op"): New.
27657 (cceq_ior_compare): Adjust.
27658 (cceq_ior_compare_complement): Adjust.
27659 (*cceq_rev_compare): Adjust.
27660 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
27661 (is_cracked_insn): Adjust.
27662 (insn_must_be_first_in_group): Adjust.
27663 * config/rs6000/40x.md: Adjust.
27664 * config/rs6000/440.md: Adjust.
27665 * config/rs6000/476.md: Adjust.
27666 * config/rs6000/601.md: Adjust.
27667 * config/rs6000/603.md: Adjust.
27668 * config/rs6000/6xx.md: Adjust.
27669 * config/rs6000/7450.md: Adjust.
27670 * config/rs6000/7xx.md: Adjust.
27671 * config/rs6000/8540.md: Adjust.
27672 * config/rs6000/cell.md: Adjust.
27673 * config/rs6000/e300c2c3.md: Adjust.
27674 * config/rs6000/e500mc.md: Adjust.
27675 * config/rs6000/e500mc64.md: Adjust.
27676 * config/rs6000/e5500.md: Adjust.
27677 * config/rs6000/e6500.md: Adjust.
27678 * config/rs6000/mpc.md: Adjust.
27679 * config/rs6000/power4.md: Adjust.
27680 * config/rs6000/power5.md: Adjust.
27681 * config/rs6000/power6.md: Adjust.
27682 * config/rs6000/power7.md: Adjust.
27683 * config/rs6000/power8.md: Adjust.
27684 * config/rs6000/power9.md: Adjust.
27685 * config/rs6000/rs64.md: Adjust.
27686 * config/rs6000/titan.md: Adjust.
27688 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
27690 * config/i386/predicates.md (indirect_branch_operand): Rewrite
27691 ix86_indirect_branch_register logic.
27693 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
27695 * config/i386/constraints.md (Bs): Update
27696 ix86_indirect_branch_register check. Don't check
27697 ix86_indirect_branch_register with GOT_memory_operand.
27699 * config/i386/predicates.md (GOT_memory_operand): Don't check
27700 ix86_indirect_branch_register here.
27701 (GOT32_symbol_operand): Likewise.
27703 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
27705 * config/i386/predicates.md (constant_call_address_operand):
27706 Rewrite ix86_indirect_branch_register logic.
27707 (sibcall_insn_operand): Likewise.
27709 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
27711 * config/i386/constraints.md (Bs): Replace
27712 ix86_indirect_branch_thunk_register with
27713 ix86_indirect_branch_register.
27715 * config/i386/i386.md (indirect_jump): Likewise.
27716 (tablejump): Likewise.
27717 (*sibcall_memory): Likewise.
27718 (*sibcall_value_memory): Likewise.
27719 Peepholes of indirect call and jump via memory: Likewise.
27720 * config/i386/i386.opt: Likewise.
27721 * config/i386/predicates.md (indirect_branch_operand): Likewise.
27722 (GOT_memory_operand): Likewise.
27723 (call_insn_operand): Likewise.
27724 (sibcall_insn_operand): Likewise.
27725 (GOT32_symbol_operand): Likewise.
27727 2018-01-15 Jakub Jelinek <jakub@redhat.com>
27729 PR middle-end/83837
27730 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
27731 type rather than type addr's type points to.
27732 (expand_omp_atomic_mutex): Likewise.
27733 (expand_omp_atomic): Likewise.
27735 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
27738 * config/i386/i386.c (output_indirect_thunk_function): Use
27739 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
27740 for __x86_return_thunk.
27742 2018-01-15 Richard Biener <rguenther@suse.de>
27744 PR middle-end/83850
27745 * expmed.c (extract_bit_field_1): Fix typo.
27747 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27750 * config/arm/iterators.md (VF): New mode iterator.
27751 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
27752 Remove integer-related logic from pattern.
27753 (neon_vabd<mode>_3): Likewise.
27755 2018-01-15 Jakub Jelinek <jakub@redhat.com>
27757 PR middle-end/82694
27758 * common.opt (fstrict-overflow): No longer an alias.
27759 (fwrapv-pointer): New option.
27760 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
27761 also for pointer types based on flag_wrapv_pointer.
27762 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
27763 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
27764 opts->x_flag_wrapv got set.
27765 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
27766 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
27767 POINTER_TYPE_OVERFLOW_UNDEFINED.
27768 * match.pd: Likewise in address comparison pattern.
27769 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
27771 2018-01-15 Richard Biener <rguenther@suse.de>
27774 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
27775 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
27776 Reset type names to their identifier if their TYPE_DECL doesn't
27777 have linkage (and thus is used for ODR and devirt).
27778 (save_debug_info_for_decl): Remove.
27779 (save_debug_info_for_type): Likewise.
27780 (add_tree_to_fld_list): Adjust.
27781 * tree-pretty-print.c (dump_generic_node): Make dumping of
27782 type names more robust.
27784 2018-01-15 Richard Biener <rguenther@suse.de>
27786 * BASE-VER: Bump to 8.0.1.
27788 2018-01-14 Martin Sebor <msebor@redhat.com>
27791 * builtins.c (check_access): Avoid warning when the no-warning bit
27794 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
27796 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
27797 * ira-color (allocno_hard_regs_compare): Likewise.
27799 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
27802 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
27803 Use .pushsection/.popsection.
27805 2018-01-14 Martin Sebor <msebor@redhat.com>
27808 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
27810 2018-01-14 Jakub Jelinek <jakub@redhat.com>
27812 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
27813 entry from extra_headers.
27814 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
27815 extra_headers, make the list bitwise identical to the i?86-*-* one.
27817 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
27819 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
27820 -mcmodel=large with -mindirect-branch=thunk,
27821 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
27822 -mfunction-return=thunk-extern.
27823 * doc/invoke.texi: Document -mcmodel=large is incompatible with
27824 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
27825 -mfunction-return=thunk and -mfunction-return=thunk-extern.
27827 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
27829 * config/i386/i386.c (print_reg): Print the name of the full
27830 integer register without '%'.
27831 (ix86_print_operand): Handle 'V'.
27832 * doc/extend.texi: Document 'V' modifier.
27834 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
27836 * config/i386/constraints.md (Bs): Disallow memory operand for
27837 -mindirect-branch-register.
27839 * config/i386/predicates.md (indirect_branch_operand): Likewise.
27840 (GOT_memory_operand): Likewise.
27841 (call_insn_operand): Likewise.
27842 (sibcall_insn_operand): Likewise.
27843 (GOT32_symbol_operand): Likewise.
27844 * config/i386/i386.md (indirect_jump): Call convert_memory_address
27845 for -mindirect-branch-register.
27846 (tablejump): Likewise.
27847 (*sibcall_memory): Likewise.
27848 (*sibcall_value_memory): Likewise.
27849 Disallow peepholes of indirect call and jump via memory for
27850 -mindirect-branch-register.
27851 (*call_pop): Replace m with Bw.
27852 (*call_value_pop): Likewise.
27853 (*sibcall_pop_memory): Replace m with Bs.
27854 * config/i386/i386.opt (mindirect-branch-register): New option.
27855 * doc/invoke.texi: Document -mindirect-branch-register option.
27857 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
27859 * config/i386/i386-protos.h (ix86_output_function_return): New.
27860 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
27861 set function_return_type.
27862 (indirect_thunk_name): Add ret_p to indicate thunk for function
27864 (output_indirect_thunk_function): Pass false to
27865 indirect_thunk_name.
27866 (ix86_output_indirect_branch_via_reg): Likewise.
27867 (ix86_output_indirect_branch_via_push): Likewise.
27868 (output_indirect_thunk_function): Create alias for function
27869 return thunk if regno < 0.
27870 (ix86_output_function_return): New function.
27871 (ix86_handle_fndecl_attribute): Handle function_return.
27872 (ix86_attribute_table): Add function_return.
27873 * config/i386/i386.h (machine_function): Add
27874 function_return_type.
27875 * config/i386/i386.md (simple_return_internal): Use
27876 ix86_output_function_return.
27877 (simple_return_internal_long): Likewise.
27878 * config/i386/i386.opt (mfunction-return=): New option.
27879 (indirect_branch): Mention -mfunction-return=.
27880 * doc/extend.texi: Document function_return function attribute.
27881 * doc/invoke.texi: Document -mfunction-return= option.
27883 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
27885 * config/i386/i386-opts.h (indirect_branch): New.
27886 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
27887 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
27888 with local indirect jump when converting indirect call and jump.
27889 (ix86_set_indirect_branch_type): New.
27890 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
27891 (indirectlabelno): New.
27892 (indirect_thunk_needed): Likewise.
27893 (indirect_thunk_bnd_needed): Likewise.
27894 (indirect_thunks_used): Likewise.
27895 (indirect_thunks_bnd_used): Likewise.
27896 (INDIRECT_LABEL): Likewise.
27897 (indirect_thunk_name): Likewise.
27898 (output_indirect_thunk): Likewise.
27899 (output_indirect_thunk_function): Likewise.
27900 (ix86_output_indirect_branch_via_reg): Likewise.
27901 (ix86_output_indirect_branch_via_push): Likewise.
27902 (ix86_output_indirect_branch): Likewise.
27903 (ix86_output_indirect_jmp): Likewise.
27904 (ix86_code_end): Call output_indirect_thunk_function if needed.
27905 (ix86_output_call_insn): Call ix86_output_indirect_branch if
27907 (ix86_handle_fndecl_attribute): Handle indirect_branch.
27908 (ix86_attribute_table): Add indirect_branch.
27909 * config/i386/i386.h (machine_function): Add indirect_branch_type
27910 and has_local_indirect_jump.
27911 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
27913 (tablejump): Likewise.
27914 (*indirect_jump): Use ix86_output_indirect_jmp.
27915 (*tablejump_1): Likewise.
27916 (simple_return_indirect_internal): Likewise.
27917 * config/i386/i386.opt (mindirect-branch=): New option.
27918 (indirect_branch): New.
27921 (thunk-inline): Likewise.
27922 (thunk-extern): Likewise.
27923 * doc/extend.texi: Document indirect_branch function attribute.
27924 * doc/invoke.texi: Document -mindirect-branch= option.
27926 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
27929 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
27931 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
27933 * ipa-inline.c (want_inline_small_function_p): Return false if
27934 inlining has already failed with CIF_FINAL_ERROR.
27935 (update_caller_keys): Call want_inline_small_function_p before
27937 (update_callee_keys): Likewise.
27939 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
27941 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
27943 (rs6000_quadword_masked_address_p): Likewise.
27944 (quad_aligned_load_p): Likewise.
27945 (quad_aligned_store_p): Likewise.
27946 (const_load_sequence_p): Add comment to describe the outer-most loop.
27947 (mimic_memory_attributes_and_flags): New function.
27948 (rs6000_gen_stvx): Likewise.
27949 (replace_swapped_aligned_store): Likewise.
27950 (rs6000_gen_lvx): Likewise.
27951 (replace_swapped_aligned_load): Likewise.
27952 (replace_swapped_load_constant): Capitalize argument name in
27953 comment describing this function.
27954 (rs6000_analyze_swaps): Add a third pass to search for vector loads
27955 and stores that access quad-word aligned addresses and replace
27956 with stvx or lvx instructions when appropriate.
27957 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
27958 New function prototype.
27959 (rs6000_quadword_masked_address_p): Likewise.
27960 (rs6000_gen_lvx): Likewise.
27961 (rs6000_gen_stvx): Likewise.
27962 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
27963 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
27964 when memory address is aligned.
27965 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
27966 this split to select lvx instruction when memory address is aligned.
27967 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
27968 instruction when memory address is aligned.
27969 (*vsx_le_perm_load_v16qi): Likewise.
27970 (four unnamed splitters): Modify to select the stvx instruction
27971 when memory is aligned.
27973 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
27975 * predict.c (determine_unlikely_bbs): Handle correctly BBs
27976 which appears in the queue multiple times.
27978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
27979 Alan Hayward <alan.hayward@arm.com>
27980 David Sherwood <david.sherwood@arm.com>
27982 * tree-vectorizer.h (vec_lower_bound): New structure.
27983 (_loop_vec_info): Add check_nonzero and lower_bounds.
27984 (LOOP_VINFO_CHECK_NONZERO): New macro.
27985 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
27986 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
27987 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
27988 fields. Make seg_len the distance travelled, not including the
27990 (dr_direction_indicator): Declare.
27991 (dr_zero_step_indicator): Likewise.
27992 (dr_known_forward_stride_p): Likewise.
27993 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
27995 (runtime_alias_check_p): Allow runtime alias checks with
27997 (operator ==): Compare access_size and align.
27998 (prune_runtime_alias_test_list): Rework for new distinction between
27999 the access_size and seg_len.
28000 (create_intersect_range_checks_index): Likewise. Cope with polynomial
28002 (get_segment_min_max): New function.
28003 (create_intersect_range_checks): Use it.
28004 (dr_step_indicator): New function.
28005 (dr_direction_indicator): Likewise.
28006 (dr_zero_step_indicator): Likewise.
28007 (dr_known_forward_stride_p): Likewise.
28008 * tree-loop-distribution.c (data_ref_segment_size): Return
28009 DR_STEP * (niters - 1).
28010 (compute_alias_check_pairs): Update call to the dr_with_seg_len
28012 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28013 (vect_preserves_scalar_order_p): New function, split out from...
28014 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
28015 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28016 (vect_vfa_access_size): New function.
28017 (vect_vfa_align): Likewise.
28018 (vect_compile_time_alias): Take access_size_a and access_b arguments.
28019 (dump_lower_bound): New function.
28020 (vect_check_lower_bound): Likewise.
28021 (vect_small_gap_p): Likewise.
28022 (vectorizable_with_step_bound_p): Likewise.
28023 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28024 depencies if the vectorization factor is 1. Convert the checks
28025 for nonzero steps into checks on the bounds of DR_STEP. Try using
28026 a bunds check for variable steps if the minimum required step is
28027 relatively small. Update calls to the dr_with_seg_len
28028 constructor and to vect_compile_time_alias.
28029 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28031 (vect_loop_versioning): Call it.
28032 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28034 (vect_estimate_min_profitable_iters): Account for any bounds checks.
28036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28037 Alan Hayward <alan.hayward@arm.com>
28038 David Sherwood <david.sherwood@arm.com>
28040 * doc/sourcebuild.texi (vect_scatter_store): Document.
28041 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28043 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28045 * genopinit.c (main): Add supports_vec_scatter_store and
28046 supports_vec_scatter_store_cached to target_optabs.
28047 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28048 IFN_MASK_SCATTER_STORE.
28049 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28051 * internal-fn.h (internal_store_fn_p): Declare.
28052 (internal_fn_stored_value_index): Likewise.
28053 * internal-fn.c (scatter_store_direct): New macro.
28054 (expand_scatter_store_optab_fn): New function.
28055 (direct_scatter_store_optab_supported_p): New macro.
28056 (internal_store_fn_p): New function.
28057 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28058 IFN_MASK_SCATTER_STORE.
28059 (internal_fn_mask_index): Likewise.
28060 (internal_fn_stored_value_index): New function.
28061 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28062 for scatter stores.
28063 * optabs-query.h (supports_vec_scatter_store_p): Declare.
28064 * optabs-query.c (supports_vec_scatter_store_p): New function.
28065 * tree-vectorizer.h (vect_get_store_rhs): Declare.
28066 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28067 true for scatter stores.
28068 (vect_gather_scatter_fn_p): Handle scatter stores too.
28069 (vect_check_gather_scatter): Consider using scatter stores if
28070 supports_vec_scatter_store_p.
28071 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28072 scatter stores too.
28073 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28074 internal_fn_stored_value_index.
28075 (check_load_store_masking): Handle scatter stores too.
28076 (vect_get_store_rhs): Make public.
28077 (vectorizable_call): Use internal_store_fn_p.
28078 (vectorizable_store): Handle scatter store internal functions.
28079 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28080 when deciding whether the end of the group has been reached.
28081 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28082 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28083 (mask_scatter_store<mode>): New insns.
28085 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28086 Alan Hayward <alan.hayward@arm.com>
28087 David Sherwood <david.sherwood@arm.com>
28089 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28090 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28091 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28093 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28094 Use vect_truncate_gather_scatter_offset if we can't treat the
28095 operation as a normal gather load or scatter store.
28096 (get_group_load_store_type): Take the gather_scatter_info
28097 as argument. Try using a gather load or scatter store for
28098 single-element groups.
28099 (get_load_store_type): Update calls to get_group_load_store_type
28100 and vect_use_strided_gather_scatters_p.
28102 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28103 Alan Hayward <alan.hayward@arm.com>
28104 David Sherwood <david.sherwood@arm.com>
28106 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28107 optional tree argument.
28108 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28110 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28111 but continue to use the current value as a fallback.
28112 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28113 to compare the updates.
28114 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28115 (get_load_store_type): Use it when handling a strided access.
28116 (vect_get_strided_load_store_ops): New function.
28117 (vect_get_data_ptr_increment): Likewise.
28118 (vectorizable_load): Handle strided gather loads. Always pass
28119 a step to vect_create_data_ref_ptr and bump_vector_ptr.
28121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28122 Alan Hayward <alan.hayward@arm.com>
28123 David Sherwood <david.sherwood@arm.com>
28125 * doc/md.texi (gather_load@var{m}): Document.
28126 (mask_gather_load@var{m}): Likewise.
28127 * genopinit.c (main): Add supports_vec_gather_load and
28128 supports_vec_gather_load_cached to target_optabs.
28129 * optabs-tree.c (init_tree_optimization_optabs): Use
28130 ggc_cleared_alloc to allocate target_optabs.
28131 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28132 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28134 * internal-fn.h (internal_load_fn_p): Declare.
28135 (internal_gather_scatter_fn_p): Likewise.
28136 (internal_fn_mask_index): Likewise.
28137 (internal_gather_scatter_fn_supported_p): Likewise.
28138 * internal-fn.c (gather_load_direct): New macro.
28139 (expand_gather_load_optab_fn): New function.
28140 (direct_gather_load_optab_supported_p): New macro.
28141 (direct_internal_fn_optab): New function.
28142 (internal_load_fn_p): Likewise.
28143 (internal_gather_scatter_fn_p): Likewise.
28144 (internal_fn_mask_index): Likewise.
28145 (internal_gather_scatter_fn_supported_p): Likewise.
28146 * optabs-query.c (supports_at_least_one_mode_p): New function.
28147 (supports_vec_gather_load_p): Likewise.
28148 * optabs-query.h (supports_vec_gather_load_p): Declare.
28149 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28150 and memory_type field.
28151 (NUM_PATTERNS): Bump to 15.
28152 * tree-vect-data-refs.c: Include internal-fn.h.
28153 (vect_gather_scatter_fn_p): New function.
28154 (vect_describe_gather_scatter_call): Likewise.
28155 (vect_check_gather_scatter): Try using internal functions for
28156 gather loads. Recognize existing calls to a gather load function.
28157 (vect_analyze_data_refs): Consider using gather loads if
28158 supports_vec_gather_load_p.
28159 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28160 (vect_get_gather_scatter_offset_type): Likewise.
28161 (vect_convert_mask_for_vectype): Likewise.
28162 (vect_add_conversion_to_patterm): Likewise.
28163 (vect_try_gather_scatter_pattern): Likewise.
28164 (vect_recog_gather_scatter_pattern): New pattern recognizer.
28165 (vect_vect_recog_func_ptrs): Add it.
28166 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28167 internal_fn_mask_index and internal_gather_scatter_fn_p.
28168 (check_load_store_masking): Take the gather_scatter_info as an
28169 argument and handle gather loads.
28170 (vect_get_gather_scatter_ops): New function.
28171 (vectorizable_call): Check internal_load_fn_p.
28172 (vectorizable_load): Likewise. Handle gather load internal
28174 (vectorizable_store): Update call to check_load_store_masking.
28175 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28176 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28177 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28178 (aarch64_gather_scale_operand_d): New predicates.
28179 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28180 (mask_gather_load<mode>): New insns.
28182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28183 Alan Hayward <alan.hayward@arm.com>
28184 David Sherwood <david.sherwood@arm.com>
28186 * optabs.def (fold_left_plus_optab): New optab.
28187 * doc/md.texi (fold_left_plus_@var{m}): Document.
28188 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28189 * internal-fn.c (fold_left_direct): Define.
28190 (expand_fold_left_optab_fn): Likewise.
28191 (direct_fold_left_optab_supported_p): Likewise.
28192 * fold-const-call.c (fold_const_fold_left): New function.
28193 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28194 * tree-parloops.c (valid_reduction_p): New function.
28195 (gather_scalar_reductions): Use it.
28196 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28197 (vect_finish_replace_stmt): Declare.
28198 * tree-vect-loop.c (fold_left_reduction_fn): New function.
28199 (needs_fold_left_reduction_p): New function, split out from...
28200 (vect_is_simple_reduction): ...here. Accept reductions that
28201 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28202 (vect_force_simple_reduction): Also store the reduction type in
28203 the assignment's STMT_VINFO_REDUC_TYPE.
28204 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28205 (merge_with_identity): New function.
28206 (vect_expand_fold_left): Likewise.
28207 (vectorize_fold_left_reduction): Likewise.
28208 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
28209 scalar phi in place for it. Check for target support and reject
28210 cases that would reassociate the operation. Defer the transform
28211 phase to vectorize_fold_left_reduction.
28212 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28213 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28214 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28218 * tree-if-conv.c (predicate_mem_writes): Remove redundant
28219 call to ifc_temp_var.
28221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28222 Alan Hayward <alan.hayward@arm.com>
28223 David Sherwood <david.sherwood@arm.com>
28225 * target.def (legitimize_address_displacement): Take the original
28226 offset as a poly_int.
28227 * targhooks.h (default_legitimize_address_displacement): Update
28229 * targhooks.c (default_legitimize_address_displacement): Likewise.
28230 * doc/tm.texi: Regenerate.
28231 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28232 as an argument, moving assert of ad->disp == ad->disp_term to...
28233 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
28234 Try calling targetm.legitimize_address_displacement before expanding
28235 the address rather than afterwards, and adjust for the new interface.
28236 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28237 Match the new hook interface. Handle SVE addresses.
28238 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28239 new hook interface.
28241 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28243 * Makefile.in (OBJS): Add early-remat.o.
28244 * target.def (select_early_remat_modes): New hook.
28245 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28246 * doc/tm.texi: Regenerate.
28247 * targhooks.h (default_select_early_remat_modes): Declare.
28248 * targhooks.c (default_select_early_remat_modes): New function.
28249 * timevar.def (TV_EARLY_REMAT): New timevar.
28250 * passes.def (pass_early_remat): New pass.
28251 * tree-pass.h (make_pass_early_remat): Declare.
28252 * early-remat.c: New file.
28253 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28255 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28258 Alan Hayward <alan.hayward@arm.com>
28259 David Sherwood <david.sherwood@arm.com>
28261 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28262 vfm1 with a bound_epilog parameter.
28263 (vect_do_peeling): Update calls accordingly, and move the prologue
28264 call earlier in the function. Treat the base bound_epilog as 0 for
28265 fully-masked loops and retain vf - 1 for other loops. Add 1 to
28266 this base when peeling for gaps.
28267 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28268 with fully-masked loops.
28269 (vect_estimate_min_profitable_iters): Handle the single peeled
28270 iteration in that case.
28272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28273 Alan Hayward <alan.hayward@arm.com>
28274 David Sherwood <david.sherwood@arm.com>
28276 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28277 single-element interleaving even if the size is not a power of 2.
28278 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28279 accesses for single-element interleaving if the group size is
28282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28283 Alan Hayward <alan.hayward@arm.com>
28284 David Sherwood <david.sherwood@arm.com>
28286 * doc/md.texi (fold_extract_last_@var{m}): Document.
28287 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28288 * optabs.def (fold_extract_last_optab): New optab.
28289 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28290 * internal-fn.c (fold_extract_direct): New macro.
28291 (expand_fold_extract_optab_fn): Likewise.
28292 (direct_fold_extract_optab_supported_p): Likewise.
28293 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28294 * tree-vect-loop.c (vect_model_reduction_cost): Handle
28295 EXTRACT_LAST_REDUCTION.
28296 (get_initial_def_for_reduction): Do not create an initial vector
28297 for EXTRACT_LAST_REDUCTION reductions.
28298 (vectorizable_reduction): Leave the scalar phi in place for
28299 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
28300 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
28301 epilogue code for EXTRACT_LAST_REDUCTION and defer the
28302 transform phase to vectorizable_condition.
28303 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28305 (vect_finish_stmt_generation): ...here.
28306 (vect_finish_replace_stmt): New function.
28307 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28308 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28310 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28312 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28313 Alan Hayward <alan.hayward@arm.com>
28314 David Sherwood <david.sherwood@arm.com>
28316 * doc/md.texi (extract_last_@var{m}): Document.
28317 * optabs.def (extract_last_optab): New optab.
28318 * internal-fn.def (EXTRACT_LAST): New internal function.
28319 * internal-fn.c (cond_unary_direct): New macro.
28320 (expand_cond_unary_optab_fn): Likewise.
28321 (direct_cond_unary_optab_supported_p): Likewise.
28322 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28323 loops using EXTRACT_LAST.
28324 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28325 (extract_last_<mode>): ...this optab.
28326 (vec_extract<mode><Vel>): Update accordingly.
28328 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28329 Alan Hayward <alan.hayward@arm.com>
28330 David Sherwood <david.sherwood@arm.com>
28332 * target.def (empty_mask_is_expensive): New hook.
28333 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28334 * doc/tm.texi: Regenerate.
28335 * targhooks.h (default_empty_mask_is_expensive): Declare.
28336 * targhooks.c (default_empty_mask_is_expensive): New function.
28337 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28338 if the target says that empty masks are expensive.
28339 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28341 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28344 Alan Hayward <alan.hayward@arm.com>
28345 David Sherwood <david.sherwood@arm.com>
28347 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28348 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28349 (vect_use_loop_mask_for_alignment_p): New function.
28350 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28351 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28352 niters_skip argument. Make sure that the first niters_skip elements
28353 of the first iteration are inactive.
28354 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28355 Update call to vect_set_loop_masks_directly.
28356 (get_misalign_in_elems): New function, split out from...
28357 (vect_gen_prolog_loop_niters): ...here.
28358 (vect_update_init_of_dr): Take a code argument that specifies whether
28359 the adjustment should be added or subtracted.
28360 (vect_update_init_of_drs): Likewise.
28361 (vect_prepare_for_masked_peels): New function.
28362 (vect_do_peeling): Skip prologue peeling if we're using a mask
28363 instead. Update call to vect_update_inits_of_drs.
28364 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28366 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28367 alignment. Do not include the number of peeled iterations in
28368 the minimum threshold in that case.
28369 (vectorizable_induction): Adjust the start value down by
28370 LOOP_VINFO_MASK_SKIP_NITERS iterations.
28371 (vect_transform_loop): Call vect_prepare_for_masked_peels.
28372 Take the number of skipped iterations into account when calculating
28374 * tree-vect-stmts.c (vect_gen_while_not): New function.
28376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28377 Alan Hayward <alan.hayward@arm.com>
28378 David Sherwood <david.sherwood@arm.com>
28380 * doc/sourcebuild.texi (vect_fully_masked): Document.
28381 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28382 default value to 0.
28383 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28385 (vect_analyze_loop_2): ...here. Don't check the vectorization
28386 factor against the number of loop iterations if the loop is
28389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28390 Alan Hayward <alan.hayward@arm.com>
28391 David Sherwood <david.sherwood@arm.com>
28393 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28394 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28395 (dump_groups): Update accordingly.
28396 (iv_use::mem_type): New member variable.
28397 (address_p): New function.
28398 (record_use): Add a mem_type argument and initialize the new
28400 (record_group_use): Add a mem_type argument. Use address_p.
28401 Remove obsolete null checks of base_object. Update call to record_use.
28402 (find_interesting_uses_op): Update call to record_group_use.
28403 (find_interesting_uses_cond): Likewise.
28404 (find_interesting_uses_address): Likewise.
28405 (get_mem_type_for_internal_fn): New function.
28406 (find_address_like_use): Likewise.
28407 (find_interesting_uses_stmt): Try find_address_like_use before
28408 calling find_interesting_uses_op.
28409 (addr_offset_valid_p): Use the iv mem_type field as the type
28410 of the addressed memory.
28411 (add_autoinc_candidates): Likewise.
28412 (get_address_cost): Likewise.
28413 (split_small_address_groups_p): Use address_p.
28414 (split_address_groups): Likewise.
28415 (add_iv_candidate_for_use): Likewise.
28416 (autoinc_possible_for_pair): Likewise.
28417 (rewrite_groups): Likewise.
28418 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28419 (determine_group_iv_cost): Update after split of USE_ADDRESS.
28420 (get_alias_ptr_type_for_ptr_address): New function.
28421 (rewrite_use_address): Rewrite address uses in calls that were
28422 identified by find_address_like_use.
28424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28425 Alan Hayward <alan.hayward@arm.com>
28426 David Sherwood <david.sherwood@arm.com>
28428 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28430 * gimple-expr.h (is_gimple_addressable: Likewise.
28431 * gimple-expr.c (is_gimple_address): Likewise.
28432 * internal-fn.c (expand_call_mem_ref): New function.
28433 (expand_mask_load_optab_fn): Use it.
28434 (expand_mask_store_optab_fn): Likewise.
28436 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28437 Alan Hayward <alan.hayward@arm.com>
28438 David Sherwood <david.sherwood@arm.com>
28440 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28441 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28442 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28443 (cond_umax@var{mode}): Document.
28444 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28445 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28446 (cond_umin_optab, cond_umax_optab): New optabs.
28447 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28448 (COND_IOR, COND_XOR): New internal functions.
28449 * internal-fn.h (get_conditional_internal_fn): Declare.
28450 * internal-fn.c (cond_binary_direct): New macro.
28451 (expand_cond_binary_optab_fn): Likewise.
28452 (direct_cond_binary_optab_supported_p): Likewise.
28453 (get_conditional_internal_fn): New function.
28454 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28455 Cope with reduction statements that are vectorized as calls rather
28457 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28458 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28459 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28460 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28461 (UNSPEC_COND_EOR): New unspecs.
28462 (optab): Add mappings for them.
28463 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28464 (sve_int_op, sve_fp_op): New int attributes.
28466 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28467 Alan Hayward <alan.hayward@arm.com>
28468 David Sherwood <david.sherwood@arm.com>
28470 * optabs.def (while_ult_optab): New optab.
28471 * doc/md.texi (while_ult@var{m}@var{n}): Document.
28472 * internal-fn.def (WHILE_ULT): New internal function.
28473 * internal-fn.h (direct_internal_fn_supported_p): New override
28474 that takes two types as argument.
28475 * internal-fn.c (while_direct): New macro.
28476 (expand_while_optab_fn): New function.
28477 (convert_optab_supported_p): Likewise.
28478 (direct_while_optab_supported_p): New macro.
28479 * wide-int.h (wi::udiv_ceil): New function.
28480 * tree-vectorizer.h (rgroup_masks): New structure.
28481 (vec_loop_masks): New typedef.
28482 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28483 and fully_masked_p.
28484 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28485 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28486 (vect_max_vf): New function.
28487 (slpeel_make_loop_iterate_ntimes): Delete.
28488 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28489 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28490 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28491 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28492 internal-fn.h, stor-layout.h and optabs-query.h.
28493 (vect_set_loop_mask): New function.
28494 (add_preheader_seq): Likewise.
28495 (add_header_seq): Likewise.
28496 (interleave_supported_p): Likewise.
28497 (vect_maybe_permute_loop_masks): Likewise.
28498 (vect_set_loop_masks_directly): Likewise.
28499 (vect_set_loop_condition_masked): Likewise.
28500 (vect_set_loop_condition_unmasked): New function, split out from
28501 slpeel_make_loop_iterate_ntimes.
28502 (slpeel_make_loop_iterate_ntimes): Rename to..
28503 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
28504 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28505 (vect_do_peeling): Update call accordingly.
28506 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28508 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28509 mask_compare_type, can_fully_mask_p and fully_masked_p.
28510 (release_vec_loop_masks): New function.
28511 (_loop_vec_info): Use it to free the loop masks.
28512 (can_produce_all_loop_masks_p): New function.
28513 (vect_get_max_nscalars_per_iter): Likewise.
28514 (vect_verify_full_masking): Likewise.
28515 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28516 retries, and free the mask rgroups before retrying. Check loop-wide
28517 reasons for disallowing fully-masked loops. Make the final decision
28518 about whether use a fully-masked loop or not.
28519 (vect_estimate_min_profitable_iters): Do not assume that peeling
28520 for the number of iterations will be needed for fully-masked loops.
28521 (vectorizable_reduction): Disable fully-masked loops.
28522 (vectorizable_live_operation): Likewise.
28523 (vect_halve_mask_nunits): New function.
28524 (vect_double_mask_nunits): Likewise.
28525 (vect_record_loop_mask): Likewise.
28526 (vect_get_loop_mask): Likewise.
28527 (vect_transform_loop): Handle the case in which the final loop
28528 iteration might handle a partial vector. Call vect_set_loop_condition
28529 instead of slpeel_make_loop_iterate_ntimes.
28530 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28531 (check_load_store_masking): New function.
28532 (prepare_load_store_mask): Likewise.
28533 (vectorizable_store): Handle fully-masked loops.
28534 (vectorizable_load): Likewise.
28535 (supportable_widening_operation): Use vect_halve_mask_nunits for
28537 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28538 (vect_gen_while): New function.
28539 * config/aarch64/aarch64.md (umax<mode>3): New expander.
28540 (aarch64_uqdec<mode>): New insn.
28542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28543 Alan Hayward <alan.hayward@arm.com>
28544 David Sherwood <david.sherwood@arm.com>
28546 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28547 (reduc_xor_scal_optab): New optabs.
28548 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28549 (reduc_xor_scal_@var{m}): Document.
28550 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28551 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28552 internal functions.
28553 * fold-const-call.c (fold_const_call): Handle them.
28554 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28555 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28556 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28557 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28558 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28559 (UNSPEC_XORV): New unspecs.
28560 (optab): Add entries for them.
28561 (BITWISEV): New int iterator.
28562 (bit_reduc_op): New int attributes.
28564 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28565 Alan Hayward <alan.hayward@arm.com>
28566 David Sherwood <david.sherwood@arm.com>
28568 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28569 * internal-fn.def (VEC_SHL_INSERT): New internal function.
28570 * optabs.def (vec_shl_insert_optab): New optab.
28571 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28572 (duplicate_and_interleave): Likewise.
28573 * tree-vect-loop.c: Include internal-fn.h.
28574 (neutral_op_for_slp_reduction): New function, split out from
28575 get_initial_defs_for_reduction.
28576 (get_initial_def_for_reduction): Handle option 2 for variable-length
28577 vectors by loading the neutral value into a vector and then shifting
28578 the initial value into element 0.
28579 (get_initial_defs_for_reduction): Replace the code argument with
28580 the neutral value calculated by neutral_op_for_slp_reduction.
28581 Use gimple_build_vector for constant-length vectors.
28582 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28583 but the first group_size elements have a neutral value.
28584 Use duplicate_and_interleave otherwise.
28585 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28586 Update call to get_initial_defs_for_reduction. Handle SLP
28587 reductions for variable-length vectors by creating one vector
28588 result for each scalar result, with the elements associated
28589 with other scalar results stubbed out with the neutral value.
28590 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
28591 Require IFN_VEC_SHL_INSERT for double reductions on
28592 variable-length vectors, or SLP reductions that have
28593 a neutral value. Require can_duplicate_and_interleave_p
28594 support for variable-length unchained SLP reductions if there
28595 is no neutral value, such as for MIN/MAX reductions. Also require
28596 the number of vector elements to be a multiple of the number of
28597 SLP statements when doing variable-length unchained SLP reductions.
28598 Update call to vect_create_epilog_for_reduction.
28599 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
28600 and remove initial values.
28601 (duplicate_and_interleave): Make public.
28602 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
28603 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
28605 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28606 Alan Hayward <alan.hayward@arm.com>
28607 David Sherwood <david.sherwood@arm.com>
28609 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
28610 (can_duplicate_and_interleave_p): New function.
28611 (vect_get_and_check_slp_defs): Take the vector of statements
28612 rather than just the current one. Remove excess parentheses.
28613 Restriction rejectinon of vect_constant_def and vect_external_def
28614 for variable-length vectors to boolean types, or types for which
28615 can_duplicate_and_interleave_p is false.
28616 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
28617 (duplicate_and_interleave): New function.
28618 (vect_get_constant_vectors): Use gimple_build_vector for
28619 constant-length vectors and suitable variable-length constant
28620 vectors. Use duplicate_and_interleave for other variable-length
28621 vectors. Don't defer the update when inserting new statements.
28623 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28624 Alan Hayward <alan.hayward@arm.com>
28625 David Sherwood <david.sherwood@arm.com>
28627 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
28628 min_profitable_iters doesn't go negative.
28630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28631 Alan Hayward <alan.hayward@arm.com>
28632 David Sherwood <david.sherwood@arm.com>
28634 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
28635 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
28636 * optabs.def (vec_mask_load_lanes_optab): New optab.
28637 (vec_mask_store_lanes_optab): Likewise.
28638 * internal-fn.def (MASK_LOAD_LANES): New internal function.
28639 (MASK_STORE_LANES): Likewise.
28640 * internal-fn.c (mask_load_lanes_direct): New macro.
28641 (mask_store_lanes_direct): Likewise.
28642 (expand_mask_load_optab_fn): Handle masked operations.
28643 (expand_mask_load_lanes_optab_fn): New macro.
28644 (expand_mask_store_optab_fn): Handle masked operations.
28645 (expand_mask_store_lanes_optab_fn): New macro.
28646 (direct_mask_load_lanes_optab_supported_p): Likewise.
28647 (direct_mask_store_lanes_optab_supported_p): Likewise.
28648 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
28650 (vect_load_lanes_supported): Likewise.
28651 * tree-vect-data-refs.c (strip_conversion): New function.
28652 (can_group_stmts_p): Likewise.
28653 (vect_analyze_data_ref_accesses): Use it instead of checking
28654 for a pair of assignments.
28655 (vect_store_lanes_supported): Take a masked_p parameter.
28656 (vect_load_lanes_supported): Likewise.
28657 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
28658 vect_store_lanes_supported and vect_load_lanes_supported.
28659 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
28660 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
28661 parameter. Don't allow gaps for masked accesses.
28662 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
28663 and vect_load_lanes_supported.
28664 (get_load_store_type): Take a masked_p parameter and update
28665 call to get_group_load_store_type.
28666 (vectorizable_store): Update call to get_load_store_type.
28667 Handle IFN_MASK_STORE_LANES.
28668 (vectorizable_load): Update call to get_load_store_type.
28669 Handle IFN_MASK_LOAD_LANES.
28671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28672 Alan Hayward <alan.hayward@arm.com>
28673 David Sherwood <david.sherwood@arm.com>
28675 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
28677 * config/aarch64/aarch64-protos.h
28678 (aarch64_sve_struct_memory_operand_p): Declare.
28679 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
28680 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
28681 (VPRED, vpred): Handle SVE structure modes.
28682 * config/aarch64/constraints.md (Utx): New constraint.
28683 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
28684 (aarch64_sve_struct_nonimmediate_operand): New predicates.
28685 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
28686 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
28687 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
28688 structure modes. Split into pieces after RA.
28689 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
28690 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
28692 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
28693 SVE structure modes.
28694 (aarch64_classify_address): Likewise.
28695 (sizetochar): Move earlier in file.
28696 (aarch64_print_operand): Handle SVE register lists.
28697 (aarch64_array_mode): New function.
28698 (aarch64_sve_struct_memory_operand_p): Likewise.
28699 (TARGET_ARRAY_MODE): Redefine.
28701 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28702 Alan Hayward <alan.hayward@arm.com>
28703 David Sherwood <david.sherwood@arm.com>
28705 * target.def (array_mode): New target hook.
28706 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
28707 * doc/tm.texi: Regenerate.
28708 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
28709 * hooks.c (hook_optmode_mode_uhwi_none): New function.
28710 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
28711 targetm.array_mode.
28712 * stor-layout.c (mode_for_array): Likewise. Support polynomial
28715 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28716 Alan Hayward <alan.hayward@arm.com>
28717 David Sherwood <david.sherwood@arm.com>
28719 * fold-const.c (fold_binary_loc): Check the argument types
28720 rather than the result type when testing for a vector operation.
28722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28724 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
28725 * doc/tm.texi: Regenerate.
28727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28728 Alan Hayward <alan.hayward@arm.com>
28729 David Sherwood <david.sherwood@arm.com>
28731 * doc/invoke.texi (-msve-vector-bits=): Document new option.
28732 (sve): Document new AArch64 extension.
28733 * doc/md.texi (w): Extend the description of the AArch64
28734 constraint to include SVE vectors.
28735 (Upl, Upa): Document new AArch64 predicate constraints.
28736 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
28738 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
28739 (msve-vector-bits=): New option.
28740 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
28741 SVE when these are disabled.
28742 (sve): New extension.
28743 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
28744 modes. Adjust their number of units based on aarch64_sve_vg.
28745 (MAX_BITSIZE_MODE_ANY_MODE): Define.
28746 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
28747 aarch64_addr_query_type.
28748 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
28749 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
28750 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
28751 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
28752 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
28753 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
28754 (aarch64_simd_imm_zero_p): Delete.
28755 (aarch64_check_zero_based_sve_index_immediate): Declare.
28756 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28757 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28758 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28759 (aarch64_sve_float_mul_immediate_p): Likewise.
28760 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28761 rather than an rtx.
28762 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
28763 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
28764 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
28765 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
28766 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
28767 (aarch64_regmode_natural_size): Likewise.
28768 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
28769 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
28771 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
28772 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
28773 for VG and the SVE predicate registers.
28774 (V_ALIASES): Add a "z"-prefixed alias.
28775 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
28776 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
28777 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
28778 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
28779 (REG_CLASS_NAMES): Add entries for them.
28780 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
28781 and the predicate registers.
28782 (aarch64_sve_vg): Declare.
28783 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
28784 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
28785 (REGMODE_NATURAL_SIZE): Define.
28786 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
28788 * config/aarch64/aarch64.c: Include cfgrtl.h.
28789 (simd_immediate_info): Add a constructor for series vectors,
28790 and an associated step field.
28791 (aarch64_sve_vg): New variable.
28792 (aarch64_dbx_register_number): Handle VG and the predicate registers.
28793 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
28794 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
28795 (VEC_ANY_DATA, VEC_STRUCT): New constants.
28796 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
28797 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
28798 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
28799 (aarch64_get_mask_mode): New functions.
28800 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
28801 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
28802 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
28803 predicate modes and predicate registers. Explicitly restrict
28804 GPRs to modes of 16 bytes or smaller. Only allow FP registers
28805 to store a vector mode if it is recognized by
28806 aarch64_classify_vector_mode.
28807 (aarch64_regmode_natural_size): New function.
28808 (aarch64_hard_regno_caller_save_mode): Return the original mode
28810 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
28811 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
28812 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
28813 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
28815 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
28816 does not overlap dest if the function is frame-related. Handle
28818 (aarch64_split_add_offset): New function.
28819 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
28820 them aarch64_add_offset.
28821 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
28822 and update call to aarch64_sub_sp.
28823 (aarch64_add_cfa_expression): New function.
28824 (aarch64_expand_prologue): Pass extra temporary registers to the
28825 functions above. Handle the case in which we need to emit new
28826 DW_CFA_expressions for registers that were originally saved
28827 relative to the stack pointer, but now have to be expressed
28828 relative to the frame pointer.
28829 (aarch64_output_mi_thunk): Pass extra temporary registers to the
28831 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
28832 IP0 and IP1 values for SVE frames.
28833 (aarch64_expand_vec_series): New function.
28834 (aarch64_expand_sve_widened_duplicate): Likewise.
28835 (aarch64_expand_sve_const_vector): Likewise.
28836 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
28837 Handle SVE constants. Use emit_move_insn to move a force_const_mem
28838 into the register, rather than emitting a SET directly.
28839 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
28840 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
28841 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
28842 (offset_9bit_signed_scaled_p): New functions.
28843 (aarch64_replicate_bitmask_imm): New function.
28844 (aarch64_bitmask_imm): Use it.
28845 (aarch64_cannot_force_const_mem): Reject expressions involving
28846 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
28847 (aarch64_classify_index): Handle SVE indices, by requiring
28848 a plain register index with a scale that matches the element size.
28849 (aarch64_classify_address): Handle SVE addresses. Assert that
28850 the mode of the address is VOIDmode or an integer mode.
28851 Update call to aarch64_classify_symbol.
28852 (aarch64_classify_symbolic_expression): Update call to
28853 aarch64_classify_symbol.
28854 (aarch64_const_vec_all_in_range_p): New function.
28855 (aarch64_print_vector_float_operand): Likewise.
28856 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
28857 "vN" for FP registers with SVE modes. Handle (const ...) vectors
28858 and the FP immediates 1.0 and 0.5.
28859 (aarch64_print_address_internal): Handle SVE addresses.
28860 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
28861 (aarch64_regno_regclass): Handle predicate registers.
28862 (aarch64_secondary_reload): Handle big-endian reloads of SVE
28864 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
28865 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
28866 (aarch64_convert_sve_vector_bits): New function.
28867 (aarch64_override_options): Use it to handle -msve-vector-bits=.
28868 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
28869 rather than an rtx.
28870 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
28871 Handle SVE vector and predicate modes. Accept VL-based constants
28872 that need only one temporary register, and VL offsets that require
28873 no temporary registers.
28874 (aarch64_conditional_register_usage): Mark the predicate registers
28875 as fixed if SVE isn't available.
28876 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
28877 Return true for SVE vector and predicate modes.
28878 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
28879 rather than an unsigned int. Handle SVE modes.
28880 (aarch64_preferred_simd_mode): Update call accordingly. Handle
28882 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
28884 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
28885 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
28886 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
28887 (aarch64_sve_float_mul_immediate_p): New functions.
28888 (aarch64_sve_valid_immediate): New function.
28889 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
28890 Explicitly reject structure modes. Check for INDEX constants.
28891 Handle PTRUE and PFALSE constants.
28892 (aarch64_check_zero_based_sve_index_immediate): New function.
28893 (aarch64_simd_imm_zero_p): Delete.
28894 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
28895 vector modes. Accept constants in the range of CNT[BHWD].
28896 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
28897 ask for an Advanced SIMD mode.
28898 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
28899 (aarch64_simd_vector_alignment): Handle SVE predicates.
28900 (aarch64_vectorize_preferred_vector_alignment): New function.
28901 (aarch64_simd_vector_alignment_reachable): Use it instead of
28903 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
28904 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
28906 (MAX_VECT_LEN): Delete.
28907 (expand_vec_perm_d): Add a vec_flags field.
28908 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
28909 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
28910 (aarch64_evpc_ext): Don't apply a big-endian lane correction
28912 (aarch64_evpc_rev): Rename to...
28913 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
28914 (aarch64_evpc_rev_global): New function.
28915 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
28916 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
28918 (aarch64_evpc_sve_tbl): New function.
28919 (aarch64_expand_vec_perm_const_1): Update after rename of
28920 aarch64_evpc_rev. Handle SVE permutes too, trying
28921 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
28922 than aarch64_evpc_tbl.
28923 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
28924 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
28925 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
28926 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
28927 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
28928 (aarch64_expand_sve_vcond): New functions.
28929 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
28930 of aarch64_vector_mode_p.
28931 (aarch64_dwarf_poly_indeterminate_value): New function.
28932 (aarch64_compute_pressure_classes): Likewise.
28933 (aarch64_can_change_mode_class): Likewise.
28934 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
28935 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
28936 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
28937 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
28938 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
28939 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
28940 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
28941 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
28943 (Dn, Dl, Dr): Accept const as well as const_vector.
28944 (Dz): Likewise. Compare against CONST0_RTX.
28945 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
28946 of "vector" where appropriate.
28947 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
28948 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
28949 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
28950 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
28951 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
28952 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
28953 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
28954 (v_int_equiv): Extend to SVE modes.
28955 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
28957 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
28958 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
28959 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
28960 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
28961 (SVE_COND_FP_CMP): New int iterators.
28962 (perm_hilo): Handle the new unpack unspecs.
28963 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
28965 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
28966 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
28967 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
28968 (aarch64_equality_operator, aarch64_constant_vector_operand)
28969 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
28970 (aarch64_sve_nonimmediate_operand): Likewise.
28971 (aarch64_sve_general_operand): Likewise.
28972 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
28973 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
28974 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
28975 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
28976 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
28977 (aarch64_sve_float_arith_immediate): Likewise.
28978 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
28979 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
28980 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
28981 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
28982 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
28983 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
28984 (aarch64_sve_float_arith_operand): Likewise.
28985 (aarch64_sve_float_arith_with_sub_operand): Likewise.
28986 (aarch64_sve_float_mul_operand): Likewise.
28987 (aarch64_sve_vec_perm_operand): Likewise.
28988 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
28989 (aarch64_mov_operand): Accept const_poly_int and const_vector.
28990 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
28991 as well as const_vector.
28992 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
28993 in file. Use CONST0_RTX and CONSTM1_RTX.
28994 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
28995 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
28996 Use aarch64_simd_imm_zero.
28997 * config/aarch64/aarch64-sve.md: New file.
28998 * config/aarch64/aarch64.md: Include it.
28999 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29000 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29001 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29002 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29003 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29004 (sve): New attribute.
29005 (enabled): Disable instructions with the sve attribute unless
29007 (movqi, movhi): Pass CONST_POLY_INT operaneds through
29008 aarch64_expand_mov_immediate.
29009 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29010 CNT[BHSD] immediates.
29011 (movti): Split CONST_POLY_INT moves into two halves.
29012 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29013 Split additions that need a temporary here if the destination
29014 is the stack pointer.
29015 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29016 (*add<mode>3_poly_1): New instruction.
29017 (set_clobber_cc): New expander.
29019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29021 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29022 parameter and use it instead of GET_MODE_SIZE (innermode). Use
29023 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29024 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29025 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
29026 Change innermode from fixed_mode_size to machine_mode.
29027 (simplify_subreg): Update call accordingly. Handle a constant-sized
29028 subreg of a variable-length CONST_VECTOR.
29030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29031 Alan Hayward <alan.hayward@arm.com>
29032 David Sherwood <david.sherwood@arm.com>
29034 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29035 (add_offset_to_base): New function, split out from...
29036 (create_mem_ref): ...here. When handling a scale other than 1,
29037 check first whether the address is valid without the offset.
29038 Add it into the base if so, leaving the index and scale as-is.
29040 2018-01-12 Jakub Jelinek <jakub@redhat.com>
29043 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29044 fold_for_warn before checking if arg2 is INTEGER_CST.
29046 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
29048 * config/rs6000/predicates.md (load_multiple_operation): Delete.
29049 (store_multiple_operation): Delete.
29050 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29051 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29052 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29053 guarded by TARGET_STRING.
29054 (rs6000_output_load_multiple): Delete.
29055 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29056 OPTION_MASK_STRING / TARGET_STRING handling.
29057 (print_operand) <'N', 'O'>: Add comment that these are unused now.
29058 (const rs6000_opt_masks) <"string">: Change mask to 0.
29059 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29060 (MASK_STRING): Delete.
29061 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29063 (load_multiple): Delete.
29070 (store_multiple): Delete.
29077 (movmemsi_8reg): Delete.
29078 (corresponding unnamed define_insn): Delete.
29079 (movmemsi_6reg): Delete.
29080 (corresponding unnamed define_insn): Delete.
29081 (movmemsi_4reg): Delete.
29082 (corresponding unnamed define_insn): Delete.
29083 (movmemsi_2reg): Delete.
29084 (corresponding unnamed define_insn): Delete.
29085 (movmemsi_1reg): Delete.
29086 (corresponding unnamed define_insn): Delete.
29087 * config/rs6000/rs6000.opt (mno-string): New.
29088 (mstring): Replace by deprecation warning stub.
29089 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29091 2018-01-12 Jakub Jelinek <jakub@redhat.com>
29093 * regrename.c (regrename_do_replace): If replacing the same
29094 reg multiple times, try to reuse last created gen_raw_REG.
29097 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29098 main to workaround a bug in GDB.
29100 2018-01-12 Tom de Vries <tom@codesourcery.com>
29103 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29105 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
29107 PR rtl-optimization/80481
29108 * ira-color.c (get_cap_member): New function.
29109 (allocnos_conflict_by_live_ranges_p): Use it.
29110 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29111 (setup_slot_coalesced_allocno_live_ranges): Ditto.
29113 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
29116 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29117 (*saddl_se_1): Ditto.
29118 (*ssubsi_1): Ditto.
29119 (*ssubl_se_1): Ditto.
29121 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29123 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29124 rather than wi::to_widest for DR_INITs.
29125 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29126 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29127 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29129 (vect_analyze_group_access_1): Note that here.
29131 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29133 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29134 polynomial type sizes.
29136 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29138 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29139 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29140 (gimple_add_tmp_var): Likewise.
29142 2018-01-12 Martin Liska <mliska@suse.cz>
29144 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29145 (gimple_alloc_sizes): Likewise.
29146 (dump_gimple_statistics): Use PRIu64 in printf format.
29147 * gimple.h: Change uint64_t to int.
29149 2018-01-12 Martin Liska <mliska@suse.cz>
29151 * tree-core.h: Use uint64_t instead of int.
29152 * tree.c (tree_node_counts): Likewise.
29153 (tree_node_sizes): Likewise.
29154 (dump_tree_statistics): Use PRIu64 in printf format.
29156 2018-01-12 Martin Liska <mliska@suse.cz>
29158 * Makefile.in: As qsort_chk is implemented in vec.c, add
29159 vec.o to linkage of gencfn-macros.
29160 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29161 passing the info to record_node_allocation_statistics.
29162 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29164 * ggc-common.c (struct ggc_usage): Add operator== and use
29165 it in operator< and compare function.
29166 * mem-stats.h (struct mem_usage): Likewise.
29167 * vec.c (struct vec_usage): Remove operator< and compare
29168 function. Can be simply inherited.
29170 2018-01-12 Martin Jambor <mjambor@suse.cz>
29173 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29174 * tree-ssa-math-opts.c: Include domwalk.h.
29175 (convert_mult_to_fma_1): New function.
29176 (fma_transformation_info): New type.
29177 (fma_deferring_state): Likewise.
29178 (cancel_fma_deferring): New function.
29179 (result_of_phi): Likewise.
29180 (last_fma_candidate_feeds_initial_phi): Likewise.
29181 (convert_mult_to_fma): Added deferring logic, split actual
29182 transformation to convert_mult_to_fma_1.
29183 (math_opts_dom_walker): New type.
29184 (math_opts_dom_walker::after_dom_children): New method, body moved
29185 here from pass_optimize_widening_mul::execute, added deferring logic
29187 (pass_optimize_widening_mul::execute): Moved most of code to
29188 math_opts_dom_walker::after_dom_children.
29189 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29190 * config/i386/i386.c (ix86_option_override_internal): Added
29191 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29193 2018-01-12 Richard Biener <rguenther@suse.de>
29196 * dwarf2out.c (gen_variable_die): Do not reset old_die for
29197 inline instance vars.
29199 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
29202 * config/rx/rx.c (rx_is_restricted_memory_address):
29203 Handle SUBREG case.
29205 2018-01-12 Richard Biener <rguenther@suse.de>
29207 PR tree-optimization/80846
29208 * target.def (split_reduction): New target hook.
29209 * targhooks.c (default_split_reduction): New function.
29210 * targhooks.h (default_split_reduction): Declare.
29211 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29212 target requests first reduce vectors by combining low and high
29214 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29215 (get_vectype_for_scalar_type_and_size): Export.
29216 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29217 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29218 * doc/tm.texi: Regenerate.
29219 * config/i386/i386.c (ix86_split_reduction): Implement
29220 TARGET_VECTORIZE_SPLIT_REDUCTION.
29222 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
29225 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29226 in PIC mode except for TARGET_VXWORKS_RTP.
29227 * config/sparc/sparc.c: Include cfgrtl.h.
29228 (TARGET_INIT_PIC_REG): Define.
29229 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29230 (sparc_pic_register_p): New predicate.
29231 (sparc_legitimate_address_p): Use it.
29232 (sparc_legitimize_pic_address): Likewise.
29233 (sparc_delegitimize_address): Likewise.
29234 (sparc_mode_dependent_address_p): Likewise.
29235 (gen_load_pcrel_sym): Remove 4th parameter.
29236 (load_got_register): Adjust call to above. Remove obsolete stuff.
29237 (sparc_expand_prologue): Do not call load_got_register here.
29238 (sparc_flat_expand_prologue): Likewise.
29239 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29240 (sparc_use_pseudo_pic_reg): New function.
29241 (sparc_init_pic_reg): Likewise.
29242 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29243 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29245 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
29247 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29248 Add item for branch_cost.
29250 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
29252 PR rtl-optimization/83565
29253 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29254 not extend the result to a larger mode for rotate operations.
29255 (num_sign_bit_copies1): Likewise.
29257 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29260 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29262 Use values-Xc.o for -pedantic.
29263 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29265 2018-01-12 Martin Liska <mliska@suse.cz>
29268 * ipa-devirt.c (final_warning_record::grow_type_warnings):
29270 (possible_polymorphic_call_targets): Use it.
29271 (ipa_devirt): Likewise.
29273 2018-01-12 Martin Liska <mliska@suse.cz>
29275 * profile-count.h (enum profile_quality): Use 0 as invalid
29276 enum value of profile_quality.
29278 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
29280 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29281 -mext-string options.
29283 2018-01-12 Richard Biener <rguenther@suse.de>
29285 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29286 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29287 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29289 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29291 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
29293 * configure.ac (--with-long-double-format): Add support for the
29294 configuration option to change the default long double format on
29296 * config.gcc (powerpc*-linux*-*): Likewise.
29297 * configure: Regenerate.
29298 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29299 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29300 used without modification.
29302 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29304 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29305 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29306 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29307 MISC_BUILTIN_SPEC_BARRIER.
29308 (rs6000_init_builtins): Likewise.
29309 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29311 (speculation_barrier): New define_insn.
29312 * doc/extend.texi: Document __builtin_speculation_barrier.
29314 2018-01-11 Jakub Jelinek <jakub@redhat.com>
29317 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29318 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29319 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29321 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
29322 integral modes instead of "ss" and "sd".
29323 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29324 vectors with 32-bit and 64-bit elements.
29325 (vecdupssescalarmodesuffix): New mode attribute.
29326 (vec_dup<mode>): Use it.
29328 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
29331 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29332 frame if argument is passed on stack.
29334 2018-01-11 Jakub Jelinek <jakub@redhat.com>
29337 * ree.c (combine_reaching_defs): Optimize also
29338 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29339 reg2=any_extend(exp); reg1=reg2;, formatting fix.
29341 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
29343 PR middle-end/83189
29344 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29346 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
29348 PR middle-end/83718
29349 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29350 after they are computed.
29352 2018-01-11 Bin Cheng <bin.cheng@arm.com>
29354 PR tree-optimization/83695
29355 * gimple-loop-linterchange.cc
29356 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29357 reset cached scev information after interchange.
29358 (pass_linterchange::execute): Remove call to scev_reset_htab.
29360 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29362 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29363 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29364 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29365 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29366 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29367 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29368 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29369 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29370 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29371 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29372 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29373 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29374 (V_lane_reg): Likewise.
29375 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29377 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29378 (vfmal_lane_low<mode>_intrinsic,
29379 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29380 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29381 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29382 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29383 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29384 vfmsl_lane_high<mode>_intrinsic): New define_insns.
29386 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29388 * config/arm/arm-cpus.in (fp16fml): New feature.
29389 (ALL_SIMD): Add fp16fml.
29390 (armv8.2-a): Add fp16fml as an option.
29391 (armv8.3-a): Likewise.
29392 (armv8.4-a): Add fp16fml as part of fp16.
29393 * config/arm/arm.h (TARGET_FP16FML): Define.
29394 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29396 * config/arm/arm-modes.def (V2HF): Define.
29397 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29398 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29399 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29400 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29401 vfmsl_low, vfmsl_high): New set of builtins.
29402 * config/arm/iterators.md (PLUSMINUS): New code iterator.
29403 (vfml_op): New code attribute.
29404 (VFMLHALVES): New int iterator.
29405 (VFML, VFMLSEL): New mode attributes.
29406 (V_reg): Define mapping for V2HF.
29407 (V_hi, V_lo): New mode attributes.
29408 (VF_constraint): Likewise.
29409 (vfml_half, vfml_half_selector): New int attributes.
29410 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29412 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29413 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29415 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29416 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29417 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29418 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
29420 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29421 Document new effective target and option set.
29423 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29425 * config/arm/arm-cpus.in (armv8_4): New feature.
29426 (ARMv8_4a): New fgroup.
29427 (armv8.4-a): New arch.
29428 * config/arm/arm-tables.opt: Regenerate.
29429 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29430 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29431 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29432 Add matching rules for -march=armv8.4-a and extensions.
29433 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29435 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
29438 * config/rx/rx.md (BW): New mode attribute.
29439 (sync_lock_test_and_setsi): Add mode suffix to insn output.
29441 2018-01-11 Richard Biener <rguenther@suse.de>
29443 PR tree-optimization/83435
29444 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29445 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29446 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29448 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29449 Alan Hayward <alan.hayward@arm.com>
29450 David Sherwood <david.sherwood@arm.com>
29452 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29454 (aarch64_classify_address): Initialize it. Track polynomial offsets.
29455 (aarch64_print_address_internal): Use it to check for a zero offset.
29457 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29458 Alan Hayward <alan.hayward@arm.com>
29459 David Sherwood <david.sherwood@arm.com>
29461 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29462 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29463 Return a poly_int64 rather than a HOST_WIDE_INT.
29464 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29465 rather than a HOST_WIDE_INT.
29466 * config/aarch64/aarch64.h (aarch64_frame): Protect with
29467 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
29468 hard_fp_offset, frame_size, initial_adjust, callee_offset and
29469 final_offset from HOST_WIDE_INT to poly_int64.
29470 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29471 to_constant when getting the number of units in an Advanced SIMD
29473 (aarch64_builtin_vectorized_function): Check for a constant number
29475 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29477 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29478 attribute instead of GET_MODE_NUNITS.
29479 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29480 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29481 GET_MODE_SIZE for fixed-size registers.
29482 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29483 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29484 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29485 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29486 (aarch64_print_operand, aarch64_print_address_internal)
29487 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29488 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29489 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29490 Handle polynomial GET_MODE_SIZE.
29491 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
29492 wider than SImode without modification.
29493 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29494 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29495 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29496 passing and returning SVE modes.
29497 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29498 rather than GEN_INT.
29499 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29500 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29501 (aarch64_allocate_and_probe_stack_space): Likewise.
29502 (aarch64_layout_frame): Cope with polynomial offsets.
29503 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29504 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
29505 polynomial offsets.
29506 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29507 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29508 poly_int64 rather than a HOST_WIDE_INT.
29509 (aarch64_get_separate_components, aarch64_process_components)
29510 (aarch64_expand_prologue, aarch64_expand_epilogue)
29511 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29512 (aarch64_anchor_offset): New function, split out from...
29513 (aarch64_legitimize_address): ...here.
29514 (aarch64_builtin_vectorization_cost): Handle polynomial
29515 TYPE_VECTOR_SUBPARTS.
29516 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29518 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29519 number of elements from the PARALLEL rather than the mode.
29520 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29521 rather than GET_MODE_BITSIZE.
29522 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29523 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29524 (aarch64_expand_vec_perm_const_1): Handle polynomial
29525 d->perm.length () and d->perm elements.
29526 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
29527 Apply to_constant to d->perm elements.
29528 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29529 polynomial CONST_VECTOR_NUNITS.
29530 (aarch64_move_pointer): Take amount as a poly_int64 rather
29532 (aarch64_progress_pointer): Avoid temporary variable.
29533 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29534 the mode attribute instead of GET_MODE.
29536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29537 Alan Hayward <alan.hayward@arm.com>
29538 David Sherwood <david.sherwood@arm.com>
29540 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29541 x exists before using it.
29542 (aarch64_add_constant_internal): Rename to...
29543 (aarch64_add_offset_1): ...this. Replace regnum with separate
29544 src and dest rtxes. Handle the case in which they're different,
29545 including when the offset is zero. Replace scratchreg with an rtx.
29546 Use 2 additions if there is no spare register into which we can
29547 move a 16-bit constant.
29548 (aarch64_add_constant): Delete.
29549 (aarch64_add_offset): Replace reg with separate src and dest
29550 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
29551 Use aarch64_add_offset_1.
29552 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29553 an rtx rather than an int. Take the delta as a poly_int64
29554 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
29555 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29556 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29557 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29558 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29559 and aarch64_add_sp.
29560 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29561 aarch64_add_constant.
29563 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29565 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29566 Use scalar_float_mode.
29568 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29570 * config/aarch64/aarch64-simd.md
29571 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29572 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29573 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29574 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29575 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29576 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29577 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29578 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29579 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29580 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29582 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
29585 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29586 targ_options->x_arm_arch_string is non NULL.
29588 2018-01-11 Tamar Christina <tamar.christina@arm.com>
29590 * config/aarch64/aarch64.h
29591 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
29593 2018-01-11 Sudakshina Das <sudi.das@arm.com>
29596 * expmed.c (emit_store_flag_force): Swap if const op0
29597 and change VOIDmode to mode of op0.
29599 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29601 PR rtl-optimization/83761
29602 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
29603 than bytes to mode_for_size.
29605 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
29607 PR middle-end/83189
29608 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
29609 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
29612 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
29614 PR middle-end/83575
29615 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
29616 when in layout mode.
29617 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
29618 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
29621 2018-01-10 Michael Collison <michael.collison@arm.com>
29623 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
29624 * config/aarch64/aarch64-option-extension.def: Add
29625 AARCH64_OPT_EXTENSION of 'fp16fml'.
29626 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29627 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
29628 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
29629 * config/aarch64/constraints.md (Ui7): New constraint.
29630 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
29631 (VFMLA_SEL_W): Ditto.
29634 (VFMLA16_LOW): New int iterator.
29635 (VFMLA16_HIGH): Ditto.
29636 (UNSPEC_FMLAL): New unspec.
29637 (UNSPEC_FMLSL): Ditto.
29638 (UNSPEC_FMLAL2): Ditto.
29639 (UNSPEC_FMLSL2): Ditto.
29640 (f16mac): New code attribute.
29641 * config/aarch64/aarch64-simd-builtins.def
29642 (aarch64_fmlal_lowv2sf): Ditto.
29643 (aarch64_fmlsl_lowv2sf): Ditto.
29644 (aarch64_fmlalq_lowv4sf): Ditto.
29645 (aarch64_fmlslq_lowv4sf): Ditto.
29646 (aarch64_fmlal_highv2sf): Ditto.
29647 (aarch64_fmlsl_highv2sf): Ditto.
29648 (aarch64_fmlalq_highv4sf): Ditto.
29649 (aarch64_fmlslq_highv4sf): Ditto.
29650 (aarch64_fmlal_lane_lowv2sf): Ditto.
29651 (aarch64_fmlsl_lane_lowv2sf): Ditto.
29652 (aarch64_fmlal_laneq_lowv2sf): Ditto.
29653 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
29654 (aarch64_fmlalq_lane_lowv4sf): Ditto.
29655 (aarch64_fmlsl_lane_lowv4sf): Ditto.
29656 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
29657 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
29658 (aarch64_fmlal_lane_highv2sf): Ditto.
29659 (aarch64_fmlsl_lane_highv2sf): Ditto.
29660 (aarch64_fmlal_laneq_highv2sf): Ditto.
29661 (aarch64_fmlsl_laneq_highv2sf): Ditto.
29662 (aarch64_fmlalq_lane_highv4sf): Ditto.
29663 (aarch64_fmlsl_lane_highv4sf): Ditto.
29664 (aarch64_fmlalq_laneq_highv4sf): Ditto.
29665 (aarch64_fmlsl_laneq_highv4sf): Ditto.
29666 * config/aarch64/aarch64-simd.md:
29667 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
29668 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29669 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
29670 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
29671 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
29672 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
29673 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
29674 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
29675 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
29676 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
29677 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
29678 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
29679 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
29680 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
29681 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
29682 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
29683 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
29684 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
29685 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
29686 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
29687 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
29688 (vfmlsl_low_u32): Ditto.
29689 (vfmlalq_low_u32): Ditto.
29690 (vfmlslq_low_u32): Ditto.
29691 (vfmlal_high_u32): Ditto.
29692 (vfmlsl_high_u32): Ditto.
29693 (vfmlalq_high_u32): Ditto.
29694 (vfmlslq_high_u32): Ditto.
29695 (vfmlal_lane_low_u32): Ditto.
29696 (vfmlsl_lane_low_u32): Ditto.
29697 (vfmlal_laneq_low_u32): Ditto.
29698 (vfmlsl_laneq_low_u32): Ditto.
29699 (vfmlalq_lane_low_u32): Ditto.
29700 (vfmlslq_lane_low_u32): Ditto.
29701 (vfmlalq_laneq_low_u32): Ditto.
29702 (vfmlslq_laneq_low_u32): Ditto.
29703 (vfmlal_lane_high_u32): Ditto.
29704 (vfmlsl_lane_high_u32): Ditto.
29705 (vfmlal_laneq_high_u32): Ditto.
29706 (vfmlsl_laneq_high_u32): Ditto.
29707 (vfmlalq_lane_high_u32): Ditto.
29708 (vfmlslq_lane_high_u32): Ditto.
29709 (vfmlalq_laneq_high_u32): Ditto.
29710 (vfmlslq_laneq_high_u32): Ditto.
29711 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
29712 (AARCH64_FL_FOR_ARCH8_4): New.
29713 (AARCH64_ISA_F16FML): New ISA flag.
29714 (TARGET_F16FML): New feature flag for fp16fml.
29715 (doc/invoke.texi): Document new fp16fml option.
29717 2018-01-10 Michael Collison <michael.collison@arm.com>
29719 * config/aarch64/aarch64-builtins.c:
29720 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
29721 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29722 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
29723 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
29724 (AARCH64_ISA_SHA3): New ISA flag.
29725 (TARGET_SHA3): New feature flag for sha3.
29726 * config/aarch64/iterators.md (sha512_op): New int attribute.
29727 (CRYPTO_SHA512): New int iterator.
29728 (UNSPEC_SHA512H): New unspec.
29729 (UNSPEC_SHA512H2): Ditto.
29730 (UNSPEC_SHA512SU0): Ditto.
29731 (UNSPEC_SHA512SU1): Ditto.
29732 * config/aarch64/aarch64-simd-builtins.def
29733 (aarch64_crypto_sha512hqv2di): New builtin.
29734 (aarch64_crypto_sha512h2qv2di): Ditto.
29735 (aarch64_crypto_sha512su0qv2di): Ditto.
29736 (aarch64_crypto_sha512su1qv2di): Ditto.
29737 (aarch64_eor3qv8hi): Ditto.
29738 (aarch64_rax1qv2di): Ditto.
29739 (aarch64_xarqv2di): Ditto.
29740 (aarch64_bcaxqv8hi): Ditto.
29741 * config/aarch64/aarch64-simd.md:
29742 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
29743 (aarch64_crypto_sha512su0qv2di): Ditto.
29744 (aarch64_crypto_sha512su1qv2di): Ditto.
29745 (aarch64_eor3qv8hi): Ditto.
29746 (aarch64_rax1qv2di): Ditto.
29747 (aarch64_xarqv2di): Ditto.
29748 (aarch64_bcaxqv8hi): Ditto.
29749 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
29750 (vsha512h2q_u64): Ditto.
29751 (vsha512su0q_u64): Ditto.
29752 (vsha512su1q_u64): Ditto.
29753 (veor3q_u16): Ditto.
29754 (vrax1q_u64): Ditto.
29755 (vxarq_u64): Ditto.
29756 (vbcaxq_u16): Ditto.
29757 * config/arm/types.md (crypto_sha512): New type attribute.
29758 (crypto_sha3): Ditto.
29759 (doc/invoke.texi): Document new sha3 option.
29761 2018-01-10 Michael Collison <michael.collison@arm.com>
29763 * config/aarch64/aarch64-builtins.c:
29764 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
29765 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29766 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
29767 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
29768 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
29769 (AARCH64_ISA_SM4): New ISA flag.
29770 (TARGET_SM4): New feature flag for sm4.
29771 * config/aarch64/aarch64-simd-builtins.def
29772 (aarch64_sm3ss1qv4si): Ditto.
29773 (aarch64_sm3tt1aq4si): Ditto.
29774 (aarch64_sm3tt1bq4si): Ditto.
29775 (aarch64_sm3tt2aq4si): Ditto.
29776 (aarch64_sm3tt2bq4si): Ditto.
29777 (aarch64_sm3partw1qv4si): Ditto.
29778 (aarch64_sm3partw2qv4si): Ditto.
29779 (aarch64_sm4eqv4si): Ditto.
29780 (aarch64_sm4ekeyqv4si): Ditto.
29781 * config/aarch64/aarch64-simd.md:
29782 (aarch64_sm3ss1qv4si): Ditto.
29783 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
29784 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
29785 (aarch64_sm4eqv4si): Ditto.
29786 (aarch64_sm4ekeyqv4si): Ditto.
29787 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
29788 (sm3part_op): Ditto.
29789 (CRYPTO_SM3TT): Ditto.
29790 (CRYPTO_SM3PART): Ditto.
29791 (UNSPEC_SM3SS1): New unspec.
29792 (UNSPEC_SM3TT1A): Ditto.
29793 (UNSPEC_SM3TT1B): Ditto.
29794 (UNSPEC_SM3TT2A): Ditto.
29795 (UNSPEC_SM3TT2B): Ditto.
29796 (UNSPEC_SM3PARTW1): Ditto.
29797 (UNSPEC_SM3PARTW2): Ditto.
29798 (UNSPEC_SM4E): Ditto.
29799 (UNSPEC_SM4EKEY): Ditto.
29800 * config/aarch64/constraints.md (Ui2): New constraint.
29801 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
29802 * config/arm/types.md (crypto_sm3): New type attribute.
29803 (crypto_sm4): Ditto.
29804 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
29805 (vsm3tt1aq_u32): Ditto.
29806 (vsm3tt1bq_u32): Ditto.
29807 (vsm3tt2aq_u32): Ditto.
29808 (vsm3tt2bq_u32): Ditto.
29809 (vsm3partw1q_u32): Ditto.
29810 (vsm3partw2q_u32): Ditto.
29811 (vsm4eq_u32): Ditto.
29812 (vsm4ekeyq_u32): Ditto.
29813 (doc/invoke.texi): Document new sm4 option.
29815 2018-01-10 Michael Collison <michael.collison@arm.com>
29817 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
29818 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
29819 (AARCH64_FL_FOR_ARCH8_4): New.
29820 (AARCH64_FL_V8_4): New flag.
29821 (doc/invoke.texi): Document new armv8.4-a option.
29823 2018-01-10 Michael Collison <michael.collison@arm.com>
29825 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
29826 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
29827 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
29828 * config/aarch64/aarch64-option-extension.def: Add
29829 AARCH64_OPT_EXTENSION of 'sha2'.
29830 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
29831 (crypto): Disable sha2 and aes if crypto disabled.
29832 (crypto): Enable aes and sha2 if enabled.
29833 (simd): Disable sha2 and aes if simd disabled.
29834 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
29836 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
29837 (TARGET_SHA2): New feature flag for sha2.
29838 (TARGET_AES): New feature flag for aes.
29839 * config/aarch64/aarch64-simd.md:
29840 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
29841 conditional on TARGET_AES.
29842 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
29843 (aarch64_crypto_sha1hsi): Make pattern conditional
29845 (aarch64_crypto_sha1hv4si): Ditto.
29846 (aarch64_be_crypto_sha1hv4si): Ditto.
29847 (aarch64_crypto_sha1su1v4si): Ditto.
29848 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
29849 (aarch64_crypto_sha1su0v4si): Ditto.
29850 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
29851 (aarch64_crypto_sha256su0v4si): Ditto.
29852 (aarch64_crypto_sha256su1v4si): Ditto.
29853 (doc/invoke.texi): Document new aes and sha2 options.
29855 2018-01-10 Martin Sebor <msebor@redhat.com>
29857 PR tree-optimization/83781
29858 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
29861 2018-01-11 Martin Sebor <msebor@gmail.com>
29862 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
29864 PR tree-optimization/83501
29865 PR tree-optimization/81703
29867 * tree-ssa-strlen.c (get_string_cst): Rename...
29868 (get_string_len): ...to this. Handle global constants.
29869 (handle_char_store): Adjust.
29871 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
29872 Jim Wilson <jimw@sifive.com>
29874 * config/riscv/riscv-protos.h (riscv_output_return): New.
29875 * config/riscv/riscv.c (struct machine_function): New naked_p field.
29876 (riscv_attribute_table, riscv_output_return),
29877 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
29878 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
29879 (riscv_compute_frame_info): Only compute frame->mask if not a naked
29881 (riscv_expand_prologue): Add early return for naked function.
29882 (riscv_expand_epilogue): Likewise.
29883 (riscv_function_ok_for_sibcall): Return false for naked function.
29884 (riscv_set_current_function): New.
29885 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
29886 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
29887 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
29888 * doc/extend.texi (RISC-V Function Attributes): New.
29890 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
29892 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
29893 check for 128-bit long double before checking TCmode.
29894 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
29895 128-bit long doubles before checking TFmode or TCmode.
29896 (FLOAT128_IBM_P): Likewise.
29898 2018-01-10 Martin Sebor <msebor@redhat.com>
29900 PR tree-optimization/83671
29901 * builtins.c (c_strlen): Unconditionally return zero for the empty
29903 Use -Warray-bounds for warnings.
29904 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
29905 for non-constant array indices with COMPONENT_REF, arrays of
29906 arrays, and pointers to arrays.
29907 (gimple_fold_builtin_strlen): Determine and set length range for
29908 non-constant character arrays.
29910 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
29912 PR middle-end/81897
29913 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
29916 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
29918 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
29920 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
29923 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
29924 VECTOR_MEM_ALTIVEC_OR_VSX_P.
29925 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
29926 indexed_or_indirect_operand predicate.
29927 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
29928 (*vsx_le_perm_load_v8hi): Likewise.
29929 (*vsx_le_perm_load_v16qi): Likewise.
29930 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
29931 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
29932 (*vsx_le_perm_store_v8hi): Likewise.
29933 (*vsx_le_perm_store_v16qi): Likewise.
29934 (eight unnamed splitters): Likewise.
29936 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
29938 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
29939 * config/rs6000/emmintrin.h: Likewise.
29940 * config/rs6000/mmintrin.h: Likewise.
29941 * config/rs6000/xmmintrin.h: Likewise.
29943 2018-01-10 David Malcolm <dmalcolm@redhat.com>
29946 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
29948 * tree.c (tree_nop_conversion): Return true for location wrapper
29950 (maybe_wrap_with_location): New function.
29951 (selftest::check_strip_nops): New function.
29952 (selftest::test_location_wrappers): New function.
29953 (selftest::tree_c_tests): Call it.
29954 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
29955 (maybe_wrap_with_location): New decl.
29956 (EXPR_LOCATION_WRAPPER_P): New macro.
29957 (location_wrapper_p): New inline function.
29958 (tree_strip_any_location_wrapper): New inline function.
29960 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
29963 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
29964 stack_realign_offset for the largest alignment of stack slot
29966 (ix86_find_max_used_stack_alignment): New function.
29967 (ix86_finalize_stack_frame_flags): Use it. Set
29968 max_used_stack_alignment if we don't realign stack.
29969 * config/i386/i386.h (machine_function): Add
29970 max_used_stack_alignment.
29972 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
29974 * config/arm/arm.opt (-mbranch-cost): New option.
29975 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
29978 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
29981 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
29982 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
29984 2018-01-10 Richard Biener <rguenther@suse.de>
29987 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
29988 early out so it also covers the case where we have a non-NULL
29991 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
29993 PR tree-optimization/83753
29994 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
29995 for non-strided grouped accesses if the number of elements is 1.
29997 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
30000 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30001 * i386.h (TARGET_USE_GATHER): Define.
30002 * x86-tune.def (X86_TUNE_USE_GATHER): New.
30004 2018-01-10 Martin Liska <mliska@suse.cz>
30007 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30008 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30010 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30011 CLEANUP_NO_PARTITIONING is not set.
30013 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
30015 * doc/rtl.texi: Remove documentation of (const ...) wrappers
30016 for vectors, as a partial revert of r254296.
30017 * rtl.h (const_vec_p): Delete.
30018 (const_vec_duplicate_p): Don't test for vector CONSTs.
30019 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30020 * expmed.c (make_tree): Likewise.
30023 * common.md (E, F): Use CONSTANT_P instead of checking for
30025 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30026 checking for CONST_VECTOR.
30028 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30030 PR middle-end/83575
30031 * predict.c (force_edge_cold): Handle in more sane way edges
30032 with no prediction.
30034 2018-01-09 Carl Love <cel@us.ibm.com>
30036 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30038 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30039 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30040 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30041 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
30042 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30043 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
30044 * config/rs6000/rs6000-protos.h: Add extern defition for
30045 rs6000_generate_float2_double_code.
30046 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30048 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30049 (float2_v2df): Add define_expand.
30051 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
30054 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30055 op_mode in the force_to_mode call.
30057 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30059 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30060 instead of checking each element individually.
30061 (aarch64_evpc_uzp): Likewise.
30062 (aarch64_evpc_zip): Likewise.
30063 (aarch64_evpc_ext): Likewise.
30064 (aarch64_evpc_rev): Likewise.
30065 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30066 instead of checking each element individually. Return true without
30068 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30069 whether all selected elements come from the same input, instead of
30070 checking each element individually. Remove calls to gen_rtx_REG,
30071 start_sequence and end_sequence and instead assert that no rtl is
30074 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30076 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30077 order of HIGH and CONST checks.
30079 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30081 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30082 if the destination isn't an SSA_NAME.
30084 2018-01-09 Richard Biener <rguenther@suse.de>
30086 PR tree-optimization/83668
30087 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30089 (canonicalize_loop_form): ... here, renamed from ...
30090 (canonicalize_loop_closed_ssa_form): ... this and amended to
30091 swap successor edges for loop exit blocks to make us use
30092 the RPO order we need for initial schedule generation.
30094 2018-01-09 Joseph Myers <joseph@codesourcery.com>
30096 PR tree-optimization/64811
30097 * match.pd: When optimizing comparisons with Inf, avoid
30098 introducing or losing exceptions from comparisons with NaN.
30100 2018-01-09 Martin Liska <mliska@suse.cz>
30103 * asan.c (shadow_mem_size): Add gcc_assert.
30105 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
30107 Don't save registers in main().
30110 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30111 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30112 * config/avr/avr.c (avr_set_current_function): Don't error if
30113 naked, OS_task or OS_main are specified at the same time.
30114 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30116 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30118 * common/config/avr/avr-common.c (avr_option_optimization_table):
30119 Switch on -mmain-is-OS_task for optimizing compilations.
30121 2018-01-09 Richard Biener <rguenther@suse.de>
30123 PR tree-optimization/83572
30124 * graphite.c: Include cfganal.h.
30125 (graphite_transform_loops): Connect infinite loops to exit
30126 and remove fake edges at the end.
30128 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30130 * ipa-inline.c (edge_badness): Revert accidental checkin.
30132 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30135 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30136 symbols; not inline clones.
30138 2018-01-09 Jakub Jelinek <jakub@redhat.com>
30141 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30142 hard registers. Formatting fixes.
30144 PR preprocessor/83722
30145 * gcc.c (try_generate_repro): Pass
30146 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30147 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30150 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
30151 Kito Cheng <kito.cheng@gmail.com>
30153 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30154 (riscv_leaf_function_p): Delete.
30155 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30157 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30159 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30161 (do_ifelse): New function.
30162 (do_isel): New function.
30163 (do_sub3): New function.
30164 (do_add3): New function.
30165 (do_load_mask_compare): New function.
30166 (do_overlap_load_compare): New function.
30167 (expand_compare_loop): New function.
30168 (expand_block_compare): Call expand_compare_loop() when appropriate.
30169 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30170 option description.
30171 (-mblock-compare-inline-loop-limit): New option.
30173 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30176 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30177 Reverse order of second and third operands in first alternative.
30178 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30179 of first and second elements in UNSPEC_VPERMR vector.
30180 (altivec_expand_vec_perm_le): Likewise.
30182 2018-01-08 Jeff Law <law@redhat.com>
30184 PR rtl-optimizatin/81308
30185 * tree-switch-conversion.c (cfg_altered): New file scoped static.
30186 (process_switch): If group_case_labels makes a change, then set
30188 (pass_convert_switch::execute): If a switch is converted, then
30189 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
30191 PR rtl-optimization/81308
30192 * recog.c (split_all_insns): Conditionally cleanup the CFG after
30195 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
30197 PR target/83663 - Revert r255946
30198 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30199 generation for cases where splatting a value is not useful.
30200 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30201 across a vec_duplicate and a paradoxical subreg forming a vector
30202 mode to a vec_concat.
30204 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30206 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30207 -march=armv8.3-a variants.
30208 * config/arm/t-multilib: Likewise.
30209 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
30211 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30213 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30215 (cceq_ior_compare_complement): Give it a name so I can use it, and
30216 change boolean_or_operator predicate to boolean_operator so it can
30217 be used to generate a crand.
30218 (eqne): New code iterator.
30219 (bd/bd_neg): New code_attrs.
30220 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30221 a single define_insn.
30222 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30223 decrement (bdnzt/bdnzf/bdzt/bdzf).
30224 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30225 with the new names of the branch decrement patterns, and added the
30226 names of the branch decrement conditional patterns.
30228 2018-01-08 Richard Biener <rguenther@suse.de>
30230 PR tree-optimization/83563
30231 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30234 2018-01-08 Richard Biener <rguenther@suse.de>
30236 PR middle-end/83713
30237 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30239 2018-01-08 Richard Biener <rguenther@suse.de>
30241 PR tree-optimization/83685
30242 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30243 references to abnormals.
30245 2018-01-08 Richard Biener <rguenther@suse.de>
30248 * dwarf2out.c (output_indirect_strings): Handle empty
30249 skeleton_debug_str_hash.
30250 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30252 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
30254 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30255 (emit_store_direct): Likewise.
30256 (arc_trampoline_adjust_address): Likewise.
30257 (arc_asm_trampoline_template): New function.
30258 (arc_initialize_trampoline): Use asm_trampoline_template.
30259 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30260 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30261 * config/arc/arc.md (flush_icache): Delete pattern.
30263 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
30265 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30266 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30269 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
30272 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30273 by not USED_FOR_TARGET.
30274 (make_pass_resolve_sw_modes): Likewise.
30276 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
30278 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30281 2018-01-08 Richard Biener <rguenther@suse.de>
30283 PR middle-end/83580
30284 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30286 2018-01-08 Richard Biener <rguenther@suse.de>
30288 PR middle-end/83517
30289 * match.pd ((t * 2) / 2) -> t): Add missing :c.
30291 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
30293 PR middle-end/81897
30294 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30295 basic blocks with a small number of successors.
30296 (convert_control_dep_chain_into_preds): Improve handling of
30298 (dump_predicates): Split apart into...
30299 (dump_pred_chain): ...here...
30300 (dump_pred_info): ...and here.
30301 (can_one_predicate_be_invalidated_p): Add debugging printfs.
30302 (can_chain_union_be_invalidated_p): Improve check for invalidation
30304 (uninit_uses_cannot_happen): Avoid unnecessary if
30305 convert_control_dep_chain_into_preds yielded nothing.
30307 2018-01-06 Martin Sebor <msebor@redhat.com>
30309 PR tree-optimization/83640
30310 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30311 subtracting negative offset from size.
30312 (builtin_access::overlap): Adjust offset bounds of the access to fall
30313 within the size of the object if possible.
30315 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
30317 PR rtl-optimization/83699
30318 * expmed.c (extract_bit_field_1): Restrict the vector usage of
30319 extract_bit_field_as_subreg to cases in which the extracted
30320 value is also a vector.
30322 * lra-constraints.c (process_alt_operands): Test for the equivalence
30323 substitutions when detecting a possible reload cycle.
30325 2018-01-06 Jakub Jelinek <jakub@redhat.com>
30328 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30329 by default if flag_selective_schedling{,2}. Formatting fixes.
30331 PR rtl-optimization/83682
30332 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30333 if it has non-VECTOR_MODE element mode.
30334 (vec_duplicate_p): Likewise.
30336 PR middle-end/83694
30337 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30338 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30340 2018-01-05 Jakub Jelinek <jakub@redhat.com>
30343 * config/i386/i386-builtin.def
30344 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30345 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30346 Require also OPTION_MASK_ISA_AVX512F in addition to
30347 OPTION_MASK_ISA_GFNI.
30348 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30349 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30350 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30351 to OPTION_MASK_ISA_GFNI.
30352 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30353 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30354 OPTION_MASK_ISA_AVX512BW.
30355 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30356 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30357 addition to OPTION_MASK_ISA_GFNI.
30358 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30359 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30360 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30361 to OPTION_MASK_ISA_GFNI.
30362 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30363 a requirement for all ISAs rather than any of them with a few
30365 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30367 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30368 bitmasks to be enabled with 3 exceptions, instead of requiring any
30369 enabled ISA with lots of exceptions.
30370 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30371 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30372 Change avx512bw in isa attribute to avx512f.
30373 * config/i386/sgxintrin.h: Add license boilerplate.
30374 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
30375 to __AVX512F__ and __AVX512VL to __AVX512VL__.
30376 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30377 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30379 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30380 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30381 temporarily sse2 rather than sse if not enabled already.
30384 * config/i386/sse.md (VI248_VLBW): Rename to ...
30385 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
30386 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30387 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30388 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30389 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30390 mode iterator instead of VI248_VLBW.
30392 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
30394 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30395 (record_modified): Skip clobbers; add debug output.
30396 (param_change_prob): Use sreal frequencies.
30398 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
30400 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30401 punt for user-aligned variables.
30403 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
30405 * tree-chrec.c (chrec_contains_symbols): Return true for
30408 2018-01-05 Sudakshina Das <sudi.das@arm.com>
30411 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30412 of (x|y) == x for BICS pattern.
30414 2018-01-05 Jakub Jelinek <jakub@redhat.com>
30416 PR tree-optimization/83605
30417 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30418 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30421 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
30423 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30424 * config/epiphany/rtems.h: New file.
30426 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30427 Uros Bizjak <ubizjak@gmail.com>
30430 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30431 QIreg_operand instead of register_operand predicate.
30432 * config/i386/i386.c (ix86_rop_should_change_byte_p,
30433 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30434 comments instead of -fmitigate[-_]rop.
30436 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30439 * cgraphunit.c (symbol_table::compile): Switch to text_section
30440 before calling assembly_start debug hook.
30441 * run-rtl-passes.c (run_rtl_passes): Likewise.
30444 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30446 * tree-vrp.c (extract_range_from_binary_expr_1): Check
30447 range_int_cst_p rather than !symbolic_range_p before calling
30448 extract_range_from_multiplicative_op_1.
30450 2018-01-04 Jeff Law <law@redhat.com>
30452 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30453 redundant test in assertion.
30455 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30457 * doc/rtl.texi: Document machine_mode wrapper classes.
30459 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30461 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30462 using tree_to_uhwi.
30464 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30466 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30467 the VEC_PERM_EXPR fold to fail.
30469 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30472 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30473 to switched_sections.
30475 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30478 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30479 test for d.testing.
30481 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
30484 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30485 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30487 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30490 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30491 is BLKmode and bitpos not zero or mode change is needed.
30493 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30496 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30499 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
30502 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30503 instead of MULT rtx. Update all corresponding splitters.
30504 (*saddl_se): Ditto.
30505 (*ssub<modesuffix>): Ditto.
30506 (*ssubl_se): Ditto.
30507 (*cmp_sadd_di): Update split patterns.
30508 (*cmp_sadd_si): Ditto.
30509 (*cmp_sadd_sidi): Ditto.
30510 (*cmp_ssub_di): Ditto.
30511 (*cmp_ssub_si): Ditto.
30512 (*cmp_ssub_sidi): Ditto.
30513 * config/alpha/predicates.md (const23_operand): New predicate.
30514 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30515 Look for ASHIFT, not MULT inner operand.
30516 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30518 2018-01-04 Martin Liska <mliska@suse.cz>
30520 PR gcov-profile/83669
30521 * gcov.c (output_intermediate_file): Add version to intermediate
30523 * doc/gcov.texi: Document new field 'version' in intermediate
30524 file format. Fix location of '-k' option of gcov command.
30526 2018-01-04 Martin Liska <mliska@suse.cz>
30529 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30531 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30533 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30535 2018-01-03 Martin Sebor <msebor@redhat.com>
30537 PR tree-optimization/83655
30538 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30539 checking calls with invalid arguments.
30541 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30543 * tree-vect-stmts.c (vect_get_store_rhs): New function.
30544 (vectorizable_mask_load_store): Delete.
30545 (vectorizable_call): Return false for masked loads and stores.
30546 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
30547 instead of gimple_assign_rhs1.
30548 (vectorizable_load): Handle IFN_MASK_LOAD.
30549 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30553 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30555 (vectorizable_mask_load_store): ...here.
30556 (vectorizable_load): ...and here.
30558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30560 * tree-vect-stmts.c (vect_build_all_ones_mask)
30561 (vect_build_zero_merge_argument): New functions, split out from...
30562 (vectorizable_load): ...here.
30564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30566 * tree-vect-stmts.c (vect_check_store_rhs): New function,
30568 (vectorizable_mask_load_store): ...here.
30569 (vectorizable_store): ...and here.
30571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30573 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30575 (vectorizable_mask_load_store): ...here.
30577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30579 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30580 (vect_model_store_cost): Take a vec_load_store_type instead of a
30582 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30583 (vect_model_store_cost): Take a vec_load_store_type instead of a
30585 (vectorizable_mask_load_store): Update accordingly.
30586 (vectorizable_store): Likewise.
30587 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
30589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30591 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
30592 IFN_MASK_LOAD calls here rather than...
30593 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
30595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30596 Alan Hayward <alan.hayward@arm.com>
30597 David Sherwood <david.sherwood@arm.com>
30599 * expmed.c (extract_bit_field_1): For vector extracts,
30600 fall back to extract_bit_field_as_subreg if vec_extract
30603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30604 Alan Hayward <alan.hayward@arm.com>
30605 David Sherwood <david.sherwood@arm.com>
30607 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
30608 they are variable or constant sized.
30609 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
30610 slots for constant-sized data.
30612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30613 Alan Hayward <alan.hayward@arm.com>
30614 David Sherwood <david.sherwood@arm.com>
30616 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
30617 handling COND_EXPRs with boolean comparisons, try to find a better
30618 basis for the mask type than the boolean itself.
30620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30622 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
30623 is calculated and how it can be overridden.
30624 * genmodes.c (max_bitsize_mode_any_mode): New variable.
30625 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
30627 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
30630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30631 Alan Hayward <alan.hayward@arm.com>
30632 David Sherwood <david.sherwood@arm.com>
30634 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
30635 Remove the mode argument.
30636 (aarch64_simd_valid_immediate): Remove the mode and inverse
30638 * config/aarch64/iterators.md (bitsize): New iterator.
30639 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
30640 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
30641 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
30642 aarch64_simd_valid_immediate.
30643 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
30644 (aarch64_reg_or_bic_imm): Likewise.
30645 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
30646 with an insn_type enum and msl with a modifier_type enum.
30647 Replace element_width with a scalar_mode. Change the shift
30648 to unsigned int. Add constructors for scalar_float_mode and
30649 scalar_int_mode elements.
30650 (aarch64_vect_float_const_representable_p): Delete.
30651 (aarch64_can_const_movi_rtx_p)
30652 (aarch64_simd_scalar_immediate_valid_for_move)
30653 (aarch64_simd_make_constant): Update call to
30654 aarch64_simd_valid_immediate.
30655 (aarch64_advsimd_valid_immediate_hs): New function.
30656 (aarch64_advsimd_valid_immediate): Likewise.
30657 (aarch64_simd_valid_immediate): Remove mode and inverse
30658 arguments. Rewrite to use the above. Use const_vec_duplicate_p
30659 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
30660 and aarch64_float_const_representable_p on the result.
30661 (aarch64_output_simd_mov_immediate): Remove mode argument.
30662 Update call to aarch64_simd_valid_immediate and use of
30663 simd_immediate_info.
30664 (aarch64_output_scalar_simd_mov_immediate): Update call
30667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30668 Alan Hayward <alan.hayward@arm.com>
30669 David Sherwood <david.sherwood@arm.com>
30671 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
30672 (mode_nunits): Likewise CONST_MODE_NUNITS.
30673 * machmode.def (ADJUST_NUNITS): Document.
30674 * genmodes.c (mode_data::need_nunits_adj): New field.
30675 (blank_mode): Update accordingly.
30676 (adj_nunits): New variable.
30677 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
30679 (emit_mode_size_inline): Set need_bytesize_adj for all modes
30680 listed in adj_nunits.
30681 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
30682 listed in adj_nunits. Don't emit case statements for such modes.
30683 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
30684 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
30685 nothing if adj_nunits is nonnull.
30686 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
30687 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
30688 (emit_mode_fbit): Update use of print_maybe_const_decl.
30689 (emit_move_size): Likewise. Treat the array as non-const
30691 (emit_mode_adjustments): Handle adj_nunits.
30693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30695 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
30696 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
30697 (VECTOR_MODES): Use it.
30698 (make_vector_modes): Take the prefix as an argument.
30700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30701 Alan Hayward <alan.hayward@arm.com>
30702 David Sherwood <david.sherwood@arm.com>
30704 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
30705 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
30706 for MODE_VECTOR_BOOL.
30707 * machmode.def (VECTOR_BOOL_MODE): Document.
30708 * genmodes.c (VECTOR_BOOL_MODE): New macro.
30709 (make_vector_bool_mode): New function.
30710 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
30712 * lto-streamer-in.c (lto_input_mode_table): Likewise.
30713 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
30715 * stor-layout.c (int_mode_for_mode): Likewise.
30716 * tree.c (build_vector_type_for_mode): Likewise.
30717 * varasm.c (output_constant_pool_2): Likewise.
30718 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
30719 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
30720 for MODE_VECTOR_BOOL.
30721 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
30722 of mode class checks.
30723 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
30724 instead of a list of mode class checks.
30725 (expand_vector_scalar_condition): Likewise.
30726 (type_for_widest_vector_mode): Handle BImode as an inner mode.
30728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30729 Alan Hayward <alan.hayward@arm.com>
30730 David Sherwood <david.sherwood@arm.com>
30732 * machmode.h (mode_size): Change from unsigned short to
30734 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
30735 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30736 or if measurement_type is not polynomial.
30737 (fixed_size_mode::includes_p): Check for constant-sized modes.
30738 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
30739 return a poly_uint16 rather than an unsigned short.
30740 (emit_mode_size): Change the type of mode_size from unsigned short
30741 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
30742 (emit_mode_adjustments): Cope with polynomial vector sizes.
30743 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30745 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30747 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
30748 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
30749 * caller-save.c (setup_save_areas): Likewise.
30750 (replace_reg_with_saved_mem): Likewise.
30751 * calls.c (emit_library_call_value_1): Likewise.
30752 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
30753 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
30754 (gen_lowpart_for_combine): Likewise.
30755 * convert.c (convert_to_integer_1): Likewise.
30756 * cse.c (equiv_constant, cse_insn): Likewise.
30757 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
30758 (cselib_subst_to_values): Likewise.
30759 * dce.c (word_dce_process_block): Likewise.
30760 * df-problems.c (df_word_lr_mark_ref): Likewise.
30761 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
30762 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
30763 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
30764 (rtl_for_decl_location): Likewise.
30765 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
30766 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
30767 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
30768 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
30769 (expand_expr_real_1): Likewise.
30770 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
30771 (pad_below): Likewise.
30772 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30773 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
30774 * ira.c (get_subreg_tracking_sizes): Likewise.
30775 * ira-build.c (ira_create_allocno_objects): Likewise.
30776 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
30777 (ira_sort_regnos_for_alter_reg): Likewise.
30778 * ira-costs.c (record_operand_costs): Likewise.
30779 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
30780 (resolve_simple_move): Likewise.
30781 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
30782 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
30783 (lra_constraints): Likewise.
30784 (CONST_POOL_OK_P): Reject variable-sized modes.
30785 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
30786 (add_pseudo_to_slot, lra_spill): Likewise.
30787 * omp-low.c (omp_clause_aligned_alignment): Likewise.
30788 * optabs-query.c (get_best_extraction_insn): Likewise.
30789 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
30790 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
30791 (expand_mult_highpart, valid_multiword_target_p): Likewise.
30792 * recog.c (offsettable_address_addr_space_p): Likewise.
30793 * regcprop.c (maybe_mode_change): Likewise.
30794 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
30795 * regrename.c (build_def_use): Likewise.
30796 * regstat.c (dump_reg_info): Likewise.
30797 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
30798 (find_reloads, find_reloads_subreg_address): Likewise.
30799 * reload1.c (eliminate_regs_1): Likewise.
30800 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
30801 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
30802 (simplify_binary_operation_1, simplify_subreg): Likewise.
30803 * targhooks.c (default_function_arg_padding): Likewise.
30804 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
30805 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
30806 (verify_gimple_assign_ternary): Likewise.
30807 * tree-inline.c (estimate_move_cost): Likewise.
30808 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30809 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
30810 (get_address_cost_ainc): Likewise.
30811 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
30812 (vect_supportable_dr_alignment): Likewise.
30813 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30814 (vectorizable_reduction): Likewise.
30815 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
30816 (vectorizable_operation, vectorizable_load): Likewise.
30817 * tree.c (build_same_sized_truth_vector_type): Likewise.
30818 * valtrack.c (cleanup_auto_inc_dec): Likewise.
30819 * var-tracking.c (emit_note_insn_var_location): Likewise.
30820 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
30821 (ADDR_VEC_ALIGN): Likewise.
30823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30824 Alan Hayward <alan.hayward@arm.com>
30825 David Sherwood <david.sherwood@arm.com>
30827 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
30829 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
30830 or if measurement_type is polynomial.
30831 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
30832 * combine.c (make_extraction): Likewise.
30833 * dse.c (find_shift_sequence): Likewise.
30834 * dwarf2out.c (mem_loc_descriptor): Likewise.
30835 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
30836 (extract_bit_field, extract_low_bits): Likewise.
30837 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
30838 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
30839 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
30840 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
30841 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
30842 * reload.c (find_reloads): Likewise.
30843 * reload1.c (alter_reg): Likewise.
30844 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
30845 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
30846 * tree-if-conv.c (predicate_mem_writes): Likewise.
30847 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
30848 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
30849 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
30850 * valtrack.c (dead_debug_insert_temp): Likewise.
30851 * varasm.c (mergeable_constant_section): Likewise.
30852 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
30854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30855 Alan Hayward <alan.hayward@arm.com>
30856 David Sherwood <david.sherwood@arm.com>
30858 * expr.c (expand_assignment): Cope with polynomial mode sizes
30859 when assigning to a CONCAT.
30861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30862 Alan Hayward <alan.hayward@arm.com>
30863 David Sherwood <david.sherwood@arm.com>
30865 * machmode.h (mode_precision): Change from unsigned short to
30867 (mode_to_precision): Return a poly_uint16 rather than an unsigned
30869 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
30870 or if measurement_type is not polynomial.
30871 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
30872 in which the mode is already known to be a scalar_int_mode.
30873 * genmodes.c (emit_mode_precision): Change the type of mode_precision
30874 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
30876 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
30877 for GET_MODE_PRECISION.
30878 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
30879 for GET_MODE_PRECISION.
30880 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
30882 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
30883 (expand_field_assignment, make_extraction): Likewise.
30884 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
30885 (get_last_value): Likewise.
30886 * convert.c (convert_to_integer_1): Likewise.
30887 * cse.c (cse_insn): Likewise.
30888 * expr.c (expand_expr_real_1): Likewise.
30889 * lra-constraints.c (simplify_operand_subreg): Likewise.
30890 * optabs-query.c (can_atomic_load_p): Likewise.
30891 * optabs.c (expand_atomic_load): Likewise.
30892 (expand_atomic_store): Likewise.
30893 * ree.c (combine_reaching_defs): Likewise.
30894 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
30895 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
30896 * tree.h (type_has_mode_precision_p): Likewise.
30897 * ubsan.c (instrument_si_overflow): Likewise.
30899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30900 Alan Hayward <alan.hayward@arm.com>
30901 David Sherwood <david.sherwood@arm.com>
30903 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
30904 polynomial numbers of units.
30905 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
30906 (valid_vector_subparts_p): New function.
30907 (build_vector_type): Remove temporary shim and take the number
30908 of units as a poly_uint64 rather than an int.
30909 (build_opaque_vector_type): Take the number of units as a
30910 poly_uint64 rather than an int.
30911 * tree.c (build_vector_from_ctor): Handle polynomial
30912 TYPE_VECTOR_SUBPARTS.
30913 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
30914 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
30915 (build_vector_from_val): If the number of units is variable,
30916 use build_vec_duplicate_cst for constant operands and
30917 VEC_DUPLICATE_EXPR otherwise.
30918 (make_vector_type): Remove temporary is_constant ().
30919 (build_vector_type, build_opaque_vector_type): Take the number of
30920 units as a poly_uint64 rather than an int.
30921 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
30923 * cfgexpand.c (expand_debug_expr): Likewise.
30924 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
30925 (store_constructor, expand_expr_real_1): Likewise.
30926 (const_scalar_mask_from_tree): Likewise.
30927 * fold-const-call.c (fold_const_reduction): Likewise.
30928 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
30929 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
30930 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
30931 (fold_relational_const): Likewise.
30932 (native_interpret_vector): Likewise. Change the size from an
30933 int to an unsigned int.
30934 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
30935 TYPE_VECTOR_SUBPARTS.
30936 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
30937 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
30938 duplicating a non-constant operand into a variable-length vector.
30939 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
30940 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
30941 * ipa-icf.c (sem_variable::equals): Likewise.
30942 * match.pd: Likewise.
30943 * omp-simd-clone.c (simd_clone_subparts): Likewise.
30944 * print-tree.c (print_node): Likewise.
30945 * stor-layout.c (layout_type): Likewise.
30946 * targhooks.c (default_builtin_vectorization_cost): Likewise.
30947 * tree-cfg.c (verify_gimple_comparison): Likewise.
30948 (verify_gimple_assign_binary): Likewise.
30949 (verify_gimple_assign_ternary): Likewise.
30950 (verify_gimple_assign_single): Likewise.
30951 * tree-pretty-print.c (dump_generic_node): Likewise.
30952 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
30953 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
30954 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
30955 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
30956 (vect_shift_permute_load_chain): Likewise.
30957 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
30958 (expand_vector_condition, optimize_vector_constructor): Likewise.
30959 (lower_vec_perm, get_compute_type): Likewise.
30960 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
30961 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
30962 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
30963 (vect_recog_mask_conversion_pattern): Likewise.
30964 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
30965 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
30966 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
30967 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
30968 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
30969 (vectorizable_shift, vectorizable_operation, vectorizable_store)
30970 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
30971 (supportable_widening_operation): Likewise.
30972 (supportable_narrowing_operation): Likewise.
30973 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
30975 * varasm.c (output_constant): Likewise.
30977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30978 Alan Hayward <alan.hayward@arm.com>
30979 David Sherwood <david.sherwood@arm.com>
30981 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
30982 so that both the length == 3 and length != 3 cases set up their
30983 own permute vectors. Add comments explaining why we know the
30984 number of elements is constant.
30985 (vect_permute_load_chain): Likewise.
30987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
30988 Alan Hayward <alan.hayward@arm.com>
30989 David Sherwood <david.sherwood@arm.com>
30991 * machmode.h (mode_nunits): Change from unsigned char to
30993 (ONLY_FIXED_SIZE_MODES): New macro.
30994 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
30995 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
30996 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
30998 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
30999 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31000 or if measurement_type is not polynomial.
31001 * genmodes.c (ZERO_COEFFS): New macro.
31002 (emit_mode_nunits_inline): Make mode_nunits_inline return a
31004 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31005 Use ZERO_COEFFS when emitting initializers.
31006 * data-streamer.h (bp_pack_poly_value): New function.
31007 (bp_unpack_poly_value): Likewise.
31008 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31009 for GET_MODE_NUNITS.
31010 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31011 for GET_MODE_NUNITS.
31012 * tree.c (make_vector_type): Remove temporary shim and make
31013 the real function take the number of units as a poly_uint64
31014 rather than an int.
31015 (build_vector_type_for_mode): Handle polynomial nunits.
31016 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31017 * emit-rtl.c (const_vec_series_p_1): Likewise.
31018 (gen_rtx_CONST_VECTOR): Likewise.
31019 * fold-const.c (test_vec_duplicate_folding): Likewise.
31020 * genrecog.c (validate_pattern): Likewise.
31021 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31022 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31023 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31024 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31025 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31026 * rtlanal.c (subreg_get_info): Likewise.
31027 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31028 (vect_grouped_load_supported): Likewise.
31029 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31030 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31031 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31032 (simplify_const_unary_operation, simplify_binary_operation_1)
31033 (simplify_const_binary_operation, simplify_ternary_operation)
31034 (test_vector_ops_duplicate, test_vector_ops): Likewise.
31035 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31036 instead of CONST_VECTOR_NUNITS.
31037 * varasm.c (output_constant_pool_2): Likewise.
31038 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31039 explicit-encoded elements in the XVEC for variable-length vectors.
31041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31043 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31046 Alan Hayward <alan.hayward@arm.com>
31047 David Sherwood <david.sherwood@arm.com>
31049 * coretypes.h (fixed_size_mode): Declare.
31050 (fixed_size_mode_pod): New typedef.
31051 * builtins.h (target_builtins::x_apply_args_mode)
31052 (target_builtins::x_apply_result_mode): Change type to
31053 fixed_size_mode_pod.
31054 * builtins.c (apply_args_size, apply_result_size, result_vector)
31055 (expand_builtin_apply_args_1, expand_builtin_apply)
31056 (expand_builtin_return): Update accordingly.
31058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31060 * cse.c (hash_rtx_cb): Hash only the encoded elements.
31061 * cselib.c (cselib_hash_rtx): Likewise.
31062 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31063 CONST_VECTOR encoding.
31065 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31066 Jeff Law <law@redhat.com>
31069 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31070 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31071 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31072 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31075 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31076 explicitly probe *sp in a noreturn function if there were any callee
31077 register saves or frame pointer is needed.
31079 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31082 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31083 BLKmode for ternary, binary or unary expressions.
31086 * var-tracking.c (delete_vta_debug_insn): New inline function.
31087 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31088 insns from get_insns () to NULL instead of each bb separately.
31089 Use delete_vta_debug_insn. No longer static.
31090 (vt_debug_insns_local, variable_tracking_main_1): Adjust
31091 delete_vta_debug_insns callers.
31092 * rtl.h (delete_vta_debug_insns): Declare.
31093 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31094 instead of variable_tracking_main.
31096 2018-01-03 Martin Sebor <msebor@redhat.com>
31098 PR tree-optimization/83603
31099 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31100 arguments past the endof the argument list in functions declared
31101 without a prototype.
31102 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31103 Avoid checking when arguments are null.
31105 2018-01-03 Martin Sebor <msebor@redhat.com>
31108 * doc/extend.texi (attribute const): Fix a typo.
31109 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31110 issuing -Wsuggest-attribute for void functions.
31112 2018-01-03 Martin Sebor <msebor@redhat.com>
31114 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31115 offset_int::from instead of wide_int::to_shwi.
31116 (maybe_diag_overlap): Remove assertion.
31117 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31118 * gimple-ssa-sprintf.c (format_directive): Same.
31119 (parse_directive): Same.
31120 (sprintf_dom_walker::compute_format_length): Same.
31121 (try_substitute_return_value): Same.
31123 2018-01-03 Jeff Law <law@redhat.com>
31125 PR middle-end/83654
31126 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31127 non-constant residual for zero at runtime and avoid probing in
31128 that case. Reorganize code for trailing problem to mirror handling
31131 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31133 PR tree-optimization/83501
31134 * tree-ssa-strlen.c (get_string_cst): New.
31135 (handle_char_store): Call get_string_cst.
31137 2018-01-03 Martin Liska <mliska@suse.cz>
31139 PR tree-optimization/83593
31140 * tree-ssa-strlen.c: Include tree-cfg.h.
31141 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31142 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31143 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31145 (strlen_dom_walker::before_dom_children): Call
31146 gimple_purge_dead_eh_edges. Dump tranformation with details
31148 (strlen_dom_walker::before_dom_children): Update call by adding
31149 new argument cleanup_eh.
31150 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31152 2018-01-03 Martin Liska <mliska@suse.cz>
31155 * cif-code.def (VARIADIC_THUNK): New enum value.
31156 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31159 2018-01-03 Jan Beulich <jbeulich@suse.com>
31161 * sse.md (mov<mode>_internal): Tighten condition for when to use
31162 vmovdqu<ssescalarsize> for TI and OI modes.
31164 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31166 Update copyright years.
31168 2018-01-03 Martin Liska <mliska@suse.cz>
31171 * ipa-visibility.c (function_and_variable_visibility): Skip
31172 functions with noipa attribure.
31174 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31176 * gcc.c (process_command): Update copyright notice dates.
31177 * gcov-dump.c (print_version): Ditto.
31178 * gcov.c (print_version): Ditto.
31179 * gcov-tool.c (print_version): Ditto.
31180 * gengtype.c (create_file): Ditto.
31181 * doc/cpp.texi: Bump @copying's copyright year.
31182 * doc/cppinternals.texi: Ditto.
31183 * doc/gcc.texi: Ditto.
31184 * doc/gccint.texi: Ditto.
31185 * doc/gcov.texi: Ditto.
31186 * doc/install.texi: Ditto.
31187 * doc/invoke.texi: Ditto.
31189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31191 * vector-builder.h (vector_builder::m_full_nelts): Change from
31192 unsigned int to poly_uint64.
31193 (vector_builder::full_nelts): Update prototype accordingly.
31194 (vector_builder::new_vector): Likewise.
31195 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31196 (vector_builder::operator ==): Likewise.
31197 (vector_builder::finalize): Likewise.
31198 * int-vector-builder.h (int_vector_builder::int_vector_builder):
31199 Take the number of elements as a poly_uint64 rather than an
31201 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31202 from unsigned int to poly_uint64.
31203 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31204 (vec_perm_indices::new_vector): Likewise.
31205 (vec_perm_indices::length): Likewise.
31206 (vec_perm_indices::nelts_per_input): Likewise.
31207 (vec_perm_indices::input_nelts): Likewise.
31208 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31209 number of elements per input as a poly_uint64 rather than an
31210 unsigned int. Use the original encoding for variable-length
31211 vectors, rather than clamping each individual element.
31212 For the second and subsequent elements in each pattern,
31213 clamp the step and base before clamping their sum.
31214 (vec_perm_indices::series_p): Handle polynomial element counts.
31215 (vec_perm_indices::all_in_range_p): Likewise.
31216 (vec_perm_indices_to_tree): Likewise.
31217 (vec_perm_indices_to_rtx): Likewise.
31218 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31219 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31220 (tree_vector_builder::new_binary_operation): Handle polynomial
31221 element counts. Return false if we need to know the number
31222 of elements at compile time.
31223 * fold-const.c (fold_vec_perm): Punt if the number of elements
31224 isn't known at compile time.
31226 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31228 * vec-perm-indices.h (vec_perm_builder): Change element type
31229 from HOST_WIDE_INT to poly_int64.
31230 (vec_perm_indices::element_type): Update accordingly.
31231 (vec_perm_indices::clamp): Handle polynomial element_types.
31232 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31233 (vec_perm_indices::all_in_range_p): Likewise.
31234 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31236 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31237 polynomial vec_perm_indices element types.
31238 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31239 * fold-const.c (fold_vec_perm): Likewise.
31240 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31241 * tree-vect-generic.c (lower_vec_perm): Likewise.
31242 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31243 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31244 element type to HOST_WIDE_INT.
31246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31247 Alan Hayward <alan.hayward@arm.com>
31248 David Sherwood <david.sherwood@arm.com>
31250 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31251 rather than an int. Use plus_constant.
31252 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31253 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31256 Alan Hayward <alan.hayward@arm.com>
31257 David Sherwood <david.sherwood@arm.com>
31259 * calls.c (emit_call_1, expand_call): Change struct_value_size from
31260 a HOST_WIDE_INT to a poly_int64.
31262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31263 Alan Hayward <alan.hayward@arm.com>
31264 David Sherwood <david.sherwood@arm.com>
31266 * calls.c (load_register_parameters): Cope with polynomial
31267 mode sizes. Require a constant size for BLKmode parameters
31268 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
31269 forces a parameter to be padded at the lsb end in order to
31270 fill a complete number of words, require the parameter size
31271 to be ordered wrt UNITS_PER_WORD.
31273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31274 Alan Hayward <alan.hayward@arm.com>
31275 David Sherwood <david.sherwood@arm.com>
31277 * reload1.c (spill_stack_slot_width): Change element type
31278 from unsigned int to poly_uint64_pod.
31279 (alter_reg): Treat mode sizes as polynomial.
31281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31282 Alan Hayward <alan.hayward@arm.com>
31283 David Sherwood <david.sherwood@arm.com>
31285 * reload.c (complex_word_subreg_p): New function.
31286 (reload_inner_reg_of_subreg, push_reload): Use it.
31288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31289 Alan Hayward <alan.hayward@arm.com>
31290 David Sherwood <david.sherwood@arm.com>
31292 * lra-constraints.c (process_alt_operands): Reject matched
31293 operands whose sizes aren't ordered.
31294 (match_reload): Refer to this check here.
31296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31297 Alan Hayward <alan.hayward@arm.com>
31298 David Sherwood <david.sherwood@arm.com>
31300 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31301 that the mode size is in the set {1, 2, 4, 8, 16}.
31303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31304 Alan Hayward <alan.hayward@arm.com>
31305 David Sherwood <david.sherwood@arm.com>
31307 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31308 Use plus_constant instead of gen_rtx_PLUS.
31310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31311 Alan Hayward <alan.hayward@arm.com>
31312 David Sherwood <david.sherwood@arm.com>
31314 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31315 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31316 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31317 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31318 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31319 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31320 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31321 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31322 * config/i386/i386.c (ix86_push_rounding): ...this new function.
31323 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31325 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31326 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31327 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31328 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31329 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31330 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31331 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31332 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31333 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31334 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31336 * expr.c (emit_move_resolve_push): Treat the input and result
31337 of PUSH_ROUNDING as a poly_int64.
31338 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31339 (emit_push_insn): Likewise.
31340 * lra-eliminations.c (mark_not_eliminable): Likewise.
31341 * recog.c (push_operand): Likewise.
31342 * reload1.c (elimination_effects): Likewise.
31343 * rtlanal.c (nonzero_bits1): Likewise.
31344 * calls.c (store_one_arg): Likewise. Require the padding to be
31345 known at compile time.
31347 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31348 Alan Hayward <alan.hayward@arm.com>
31349 David Sherwood <david.sherwood@arm.com>
31351 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31352 Use plus_constant instead of gen_rtx_PLUS.
31354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31355 Alan Hayward <alan.hayward@arm.com>
31356 David Sherwood <david.sherwood@arm.com>
31358 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31359 rather than an int.
31361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31362 Alan Hayward <alan.hayward@arm.com>
31363 David Sherwood <david.sherwood@arm.com>
31365 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31366 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31367 via stack temporaries. Treat the mode size as polynomial too.
31369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31370 Alan Hayward <alan.hayward@arm.com>
31371 David Sherwood <david.sherwood@arm.com>
31373 * expr.c (expand_expr_real_2): When handling conversions involving
31374 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31375 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
31376 as a poly_uint64 too.
31378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31379 Alan Hayward <alan.hayward@arm.com>
31380 David Sherwood <david.sherwood@arm.com>
31382 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31385 Alan Hayward <alan.hayward@arm.com>
31386 David Sherwood <david.sherwood@arm.com>
31388 * combine.c (can_change_dest_mode): Handle polynomial
31389 REGMODE_NATURAL_SIZE.
31390 * expmed.c (store_bit_field_1): Likewise.
31391 * expr.c (store_constructor): Likewise.
31392 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31393 and polynomial REGMODE_NATURAL_SIZE.
31394 (gen_lowpart_common): Likewise.
31395 * reginfo.c (record_subregs_of_mode): Likewise.
31396 * rtlanal.c (read_modify_subreg_p): Likewise.
31398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31399 Alan Hayward <alan.hayward@arm.com>
31400 David Sherwood <david.sherwood@arm.com>
31402 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31403 numbers of elements.
31405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31406 Alan Hayward <alan.hayward@arm.com>
31407 David Sherwood <david.sherwood@arm.com>
31409 * match.pd: Cope with polynomial numbers of vector elements.
31411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31412 Alan Hayward <alan.hayward@arm.com>
31413 David Sherwood <david.sherwood@arm.com>
31415 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31416 in a POINTER_PLUS_EXPR.
31418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31419 Alan Hayward <alan.hayward@arm.com>
31420 David Sherwood <david.sherwood@arm.com>
31422 * omp-simd-clone.c (simd_clone_subparts): New function.
31423 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31424 (ipa_simd_modify_function_body): Likewise.
31426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31427 Alan Hayward <alan.hayward@arm.com>
31428 David Sherwood <david.sherwood@arm.com>
31430 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31431 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31432 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31433 (expand_vector_condition, vector_element): Likewise.
31434 (subparts_gt): New function.
31435 (get_compute_type): Use subparts_gt.
31436 (count_type_subparts): Delete.
31437 (expand_vector_operations_1): Use subparts_gt instead of
31438 count_type_subparts.
31440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31441 Alan Hayward <alan.hayward@arm.com>
31442 David Sherwood <david.sherwood@arm.com>
31444 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31445 (vect_compile_time_alias): ...this new function. Do the calculation
31446 on poly_ints rather than trees.
31447 (vect_prune_runtime_alias_test_list): Update call accordingly.
31449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31450 Alan Hayward <alan.hayward@arm.com>
31451 David Sherwood <david.sherwood@arm.com>
31453 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31455 (vect_schedule_slp_instance): Likewise.
31457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31458 Alan Hayward <alan.hayward@arm.com>
31459 David Sherwood <david.sherwood@arm.com>
31461 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31462 constant and extern definitions for variable-length vectors.
31463 (vect_get_constant_vectors): Note that the number of units
31464 is known to be constant.
31466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31467 Alan Hayward <alan.hayward@arm.com>
31468 David Sherwood <david.sherwood@arm.com>
31470 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31471 of units as polynomial. Choose between WIDE and NARROW based
31474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31475 Alan Hayward <alan.hayward@arm.com>
31476 David Sherwood <david.sherwood@arm.com>
31478 * tree-vect-stmts.c (simd_clone_subparts): New function.
31479 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31482 Alan Hayward <alan.hayward@arm.com>
31483 David Sherwood <david.sherwood@arm.com>
31485 * tree-vect-stmts.c (vectorizable_call): Treat the number of
31486 vectors as polynomial. Use build_index_vector for
31487 IFN_GOMP_SIMD_LANE.
31489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31490 Alan Hayward <alan.hayward@arm.com>
31491 David Sherwood <david.sherwood@arm.com>
31493 * tree-vect-stmts.c (get_load_store_type): Treat the number of
31494 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31495 for variable-length vectors.
31496 (vectorizable_mask_load_store): Treat the number of units as
31497 polynomial, asserting that it is constant if the condition has
31498 already been enforced.
31499 (vectorizable_store, vectorizable_load): Likewise.
31501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31502 Alan Hayward <alan.hayward@arm.com>
31503 David Sherwood <david.sherwood@arm.com>
31505 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31506 of units as polynomial. Punt if we can't tell at compile time
31507 which vector contains the final result.
31509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31510 Alan Hayward <alan.hayward@arm.com>
31511 David Sherwood <david.sherwood@arm.com>
31513 * tree-vect-loop.c (vectorizable_induction): Treat the number
31514 of units as polynomial. Punt on SLP inductions. Use an integer
31515 VEC_SERIES_EXPR for variable-length integer reductions. Use a
31516 cast of such a series for variable-length floating-point
31519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31520 Alan Hayward <alan.hayward@arm.com>
31521 David Sherwood <david.sherwood@arm.com>
31523 * tree.h (build_index_vector): Declare.
31524 * tree.c (build_index_vector): New function.
31525 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31526 of units as polynomial, forcibly converting it to a constant if
31527 vectorizable_reduction has already enforced the condition.
31528 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
31529 to create a {1,2,3,...} vector.
31530 (vectorizable_reduction): Treat the number of units as polynomial.
31531 Choose vectype_in based on the largest scalar element size rather
31532 than the smallest number of units. Enforce the restrictions
31535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31536 Alan Hayward <alan.hayward@arm.com>
31537 David Sherwood <david.sherwood@arm.com>
31539 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31540 number of units as polynomial.
31542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31543 Alan Hayward <alan.hayward@arm.com>
31544 David Sherwood <david.sherwood@arm.com>
31546 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31547 * target.def (autovectorize_vector_sizes): Return the vector sizes
31548 by pointer, using vector_sizes rather than a bitmask.
31549 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31550 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31551 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31553 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31554 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31555 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31556 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31557 * omp-general.c (omp_max_vf): Likewise.
31558 * omp-low.c (omp_clause_aligned_alignment): Likewise.
31559 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31560 * tree-vect-loop.c (vect_analyze_loop): Likewise.
31561 * tree-vect-slp.c (vect_slp_bb): Likewise.
31562 * doc/tm.texi: Regenerate.
31563 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31565 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31566 the vector size as a poly_uint64 rather than an unsigned int.
31567 (current_vector_size): Change from an unsigned int to a poly_uint64.
31568 (get_vectype_for_scalar_type): Update accordingly.
31569 * tree.h (build_truth_vector_type): Take the size and number of
31570 units as a poly_uint64 rather than an unsigned int.
31571 (build_vector_type): Add a temporary overload that takes
31572 the number of units as a poly_uint64 rather than an unsigned int.
31573 * tree.c (make_vector_type): Likewise.
31574 (build_truth_vector_type): Take the number of units as a poly_uint64
31575 rather than an unsigned int.
31577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31578 Alan Hayward <alan.hayward@arm.com>
31579 David Sherwood <david.sherwood@arm.com>
31581 * target.def (get_mask_mode): Take the number of units and length
31582 as poly_uint64s rather than unsigned ints.
31583 * targhooks.h (default_get_mask_mode): Update accordingly.
31584 * targhooks.c (default_get_mask_mode): Likewise.
31585 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31586 * doc/tm.texi: Regenerate.
31588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31589 Alan Hayward <alan.hayward@arm.com>
31590 David Sherwood <david.sherwood@arm.com>
31592 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
31593 * omp-general.c (omp_max_vf): Likewise.
31594 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
31595 (expand_omp_simd): Handle polynomial safelen.
31596 * omp-low.c (omplow_simd_context): Add a default constructor.
31597 (omplow_simd_context::max_vf): Change from int to poly_uint64.
31598 (lower_rec_simd_input_clauses): Update accordingly.
31599 (lower_rec_input_clauses): Likewise.
31601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31602 Alan Hayward <alan.hayward@arm.com>
31603 David Sherwood <david.sherwood@arm.com>
31605 * tree-vectorizer.h (vect_nunits_for_cost): New function.
31606 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
31607 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
31608 (vect_analyze_slp_cost): Likewise.
31609 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
31610 (vect_model_load_cost): Likewise.
31612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31613 Alan Hayward <alan.hayward@arm.com>
31614 David Sherwood <david.sherwood@arm.com>
31616 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
31617 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
31618 from an unsigned int * to a poly_uint64_pod *.
31619 (calculate_unrolling_factor): New function.
31620 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
31622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31623 Alan Hayward <alan.hayward@arm.com>
31624 David Sherwood <david.sherwood@arm.com>
31626 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
31627 from an unsigned int to a poly_uint64.
31628 (_loop_vec_info::slp_unrolling_factor): Likewise.
31629 (_loop_vec_info::vectorization_factor): Change from an int
31631 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
31632 (vect_get_num_vectors): New function.
31633 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
31634 (vect_get_num_copies): Use vect_get_num_vectors.
31635 (vect_analyze_data_ref_dependences): Change max_vf from an int *
31636 to an unsigned int *.
31637 (vect_analyze_data_refs): Change min_vf from an int * to a
31639 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31640 than an unsigned HOST_WIDE_INT.
31641 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
31642 (vect_analyze_data_ref_dependence): Change max_vf from an int *
31643 to an unsigned int *.
31644 (vect_analyze_data_ref_dependences): Likewise.
31645 (vect_compute_data_ref_alignment): Handle polynomial vf.
31646 (vect_enhance_data_refs_alignment): Likewise.
31647 (vect_prune_runtime_alias_test_list): Likewise.
31648 (vect_shift_permute_load_chain): Likewise.
31649 (vect_supportable_dr_alignment): Likewise.
31650 (dependence_distance_ge_vf): Take the vectorization factor as a
31651 poly_uint64 rather than an unsigned HOST_WIDE_INT.
31652 (vect_analyze_data_refs): Change min_vf from an int * to a
31654 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
31655 vfm1 as a poly_uint64 rather than an int. Make the same change
31656 for the returned bound_scalar.
31657 (vect_gen_vector_loop_niters): Handle polynomial vf.
31658 (vect_do_peeling): Likewise. Update call to
31659 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
31660 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
31662 * tree-vect-loop.c (vect_determine_vectorization_factor)
31663 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
31664 (vect_get_known_peeling_cost): Likewise.
31665 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
31666 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
31667 (vect_transform_loop): Likewise. Use the lowest possible VF when
31668 updating the upper bounds of the loop.
31669 (vect_min_worthwhile_factor): Make static. Return an unsigned int
31670 rather than an int.
31671 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
31672 polynomial unroll factors.
31673 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
31674 (vect_make_slp_decision): Likewise.
31675 (vect_supported_load_permutation_p): Likewise, and polynomial
31677 (vect_analyze_slp_cost): Handle polynomial vf.
31678 (vect_slp_analyze_node_operations): Likewise.
31679 (vect_slp_analyze_bb_1): Likewise.
31680 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
31681 than an unsigned HOST_WIDE_INT.
31682 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
31683 (vectorizable_load): Handle polynomial vf.
31684 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
31686 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
31688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31689 Alan Hayward <alan.hayward@arm.com>
31690 David Sherwood <david.sherwood@arm.com>
31692 * match.pd: Handle bit operations involving three constants
31693 and try to fold one pair.
31695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31697 * tree-vect-loop-manip.c: Include gimple-fold.h.
31698 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
31699 niters_maybe_zero parameters. Handle other cases besides a step of 1.
31700 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
31701 Add a path that uses a step of VF instead of 1, but disable it
31703 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
31704 and niters_no_overflow parameters. Update calls to
31705 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
31706 Create a new SSA name if the latter choses to use a ste other
31707 than zero, and return it via niters_vector_mult_vf_var.
31708 * tree-vect-loop.c (vect_transform_loop): Update calls to
31709 vect_do_peeling, vect_gen_vector_loop_niters and
31710 slpeel_make_loop_iterate_ntimes.
31711 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
31712 (vect_gen_vector_loop_niters): Update declarations after above changes.
31714 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
31716 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
31717 128-bit round to integer instructions.
31718 (ceil<mode>2): Likewise.
31719 (btrunc<mode>2): Likewise.
31720 (round<mode>2): Likewise.
31722 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
31724 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
31725 unaligned VSX load/store on P8/P9.
31726 (expand_block_clear): Allow the use of unaligned VSX
31727 load/store on P8/P9.
31729 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
31731 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
31733 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
31734 swap associated with both a load and a store.
31736 2018-01-02 Andrew Waterman <andrew@sifive.com>
31738 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
31739 * config/riscv/riscv.md (clear_cache): Use it.
31741 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
31743 * web.c: Remove out-of-date comment.
31745 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31747 * expr.c (fixup_args_size_notes): Check that any existing
31748 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
31749 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
31750 (emit_single_push_insn): ...here.
31752 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31754 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
31755 (const_vector_encoded_nelts): New function.
31756 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
31757 (const_vector_int_elt, const_vector_elt): Declare.
31758 * emit-rtl.c (const_vector_int_elt_1): New function.
31759 (const_vector_elt): Likewise.
31760 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
31761 of CONST_VECTOR_ELT.
31763 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31765 * expr.c: Include rtx-vector-builder.h.
31766 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
31767 directly on the tree encoding.
31768 (const_vector_from_tree): Likewise.
31769 * optabs.c: Include rtx-vector-builder.h.
31770 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
31771 sequence of "u" values.
31772 * vec-perm-indices.c: Include rtx-vector-builder.h.
31773 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
31774 directly on the vec_perm_indices encoding.
31776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31778 * doc/rtl.texi (const_vector): Describe new encoding scheme.
31779 * Makefile.in (OBJS): Add rtx-vector-builder.o.
31780 * rtx-vector-builder.h: New file.
31781 * rtx-vector-builder.c: Likewise.
31782 * rtl.h (rtx_def::u2): Add a const_vector field.
31783 (CONST_VECTOR_NPATTERNS): New macro.
31784 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
31785 (CONST_VECTOR_DUPLICATE_P): Likewise.
31786 (CONST_VECTOR_STEPPED_P): Likewise.
31787 (CONST_VECTOR_ENCODED_ELT): Likewise.
31788 (const_vec_duplicate_p): Check for a duplicated vector encoding.
31789 (unwrap_const_vec_duplicate): Likewise.
31790 (const_vec_series_p): Check for a non-duplicated vector encoding.
31791 Say that the function only returns true for integer vectors.
31792 * emit-rtl.c: Include rtx-vector-builder.h.
31793 (gen_const_vec_duplicate_1): Delete.
31794 (gen_const_vector): Call gen_const_vec_duplicate instead of
31795 gen_const_vec_duplicate_1.
31796 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
31797 (gen_const_vec_duplicate): Use rtx_vector_builder.
31798 (gen_const_vec_series): Likewise.
31799 (gen_rtx_CONST_VECTOR): Likewise.
31800 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
31801 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31802 Build a new vector rather than modifying a CONST_VECTOR in-place.
31803 (handle_special_swappables): Update call accordingly.
31804 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
31805 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
31806 Build a new vector rather than modifying a CONST_VECTOR in-place.
31807 (handle_special_swappables): Update call accordingly.
31809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31811 * simplify-rtx.c (simplify_const_binary_operation): Use
31812 CONST_VECTOR_ELT instead of XVECEXP.
31814 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31816 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
31817 the selector elements to be different from the data elements
31818 if the selector is a VECTOR_CST.
31819 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
31820 ssizetype for the selector.
31822 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31824 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
31825 before testing each element individually.
31826 * tree-vect-generic.c (lower_vec_perm): Likewise.
31828 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31830 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
31831 * selftest-run-tests.c (selftest::run_tests): Call it.
31832 * vector-builder.h (vector_builder::operator ==): New function.
31833 (vector_builder::operator !=): Likewise.
31834 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
31835 (vec_perm_indices::all_from_input_p): New function.
31836 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31837 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
31838 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
31839 instead of reading the VECTOR_CST directly. Detect whether both
31840 vector inputs are the same before constructing the vec_perm_indices,
31841 and update the number of inputs argument accordingly. Use the
31842 utility functions added above. Only construct sel2 if we need to.
31844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31846 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
31847 the broadcast of the low byte.
31848 (expand_mult_highpart): Use an explicit encoding for the permutes.
31849 * optabs-query.c (can_mult_highpart_p): Likewise.
31850 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
31851 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31852 (vectorizable_bswap): Likewise.
31853 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
31854 explicit encoding for the power-of-2 permutes.
31855 (vect_permute_store_chain): Likewise.
31856 (vect_grouped_load_supported): Likewise.
31857 (vect_permute_load_chain): Likewise.
31859 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31861 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
31862 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
31863 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
31864 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31865 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
31866 (vect_gen_perm_mask_any): Likewise.
31868 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31870 * int-vector-builder.h: New file.
31871 * vec-perm-indices.h: Include int-vector-builder.h.
31872 (vec_perm_indices): Redefine as an int_vector_builder.
31873 (auto_vec_perm_indices): Delete.
31874 (vec_perm_builder): Redefine as a stand-alone class.
31875 (vec_perm_indices::vec_perm_indices): New function.
31876 (vec_perm_indices::clamp): Likewise.
31877 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
31878 (vec_perm_indices::new_vector): New function.
31879 (vec_perm_indices::new_expanded_vector): Update for new
31880 vec_perm_indices class.
31881 (vec_perm_indices::rotate_inputs): New function.
31882 (vec_perm_indices::all_in_range_p): Operate directly on the
31883 encoded form, without computing elided elements.
31884 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
31885 encoding. Update for new vec_perm_indices class.
31886 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
31887 the given vec_perm_builder.
31888 (expand_vec_perm_var): Update vec_perm_builder constructor.
31889 (expand_mult_highpart): Use vec_perm_builder instead of
31890 auto_vec_perm_indices.
31891 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
31892 vec_perm_indices instead of auto_vec_perm_indices. Use a single
31893 or double series encoding as appropriate.
31894 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
31895 vec_perm_indices instead of auto_vec_perm_indices.
31896 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31897 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31898 (vect_permute_store_chain): Likewise.
31899 (vect_grouped_load_supported): Likewise.
31900 (vect_permute_load_chain): Likewise.
31901 (vect_shift_permute_load_chain): Likewise.
31902 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
31903 (vect_transform_slp_perm_load): Likewise.
31904 (vect_schedule_slp_instance): Likewise.
31905 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31906 (vectorizable_mask_load_store): Likewise.
31907 (vectorizable_bswap): Likewise.
31908 (vectorizable_store): Likewise.
31909 (vectorizable_load): Likewise.
31910 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
31911 vec_perm_indices instead of auto_vec_perm_indices. Use
31912 tree_to_vec_perm_builder to read the vector from a tree.
31913 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
31914 vec_perm_builder instead of a vec_perm_indices.
31915 (have_whole_vector_shift): Use vec_perm_builder and
31916 vec_perm_indices instead of auto_vec_perm_indices. Leave the
31917 truncation to calc_vec_perm_mask_for_shift.
31918 (vect_create_epilog_for_reduction): Likewise.
31919 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
31920 from auto_vec_perm_indices to vec_perm_indices.
31921 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31922 instead of changing individual elements.
31923 (aarch64_vectorize_vec_perm_const): Use new_vector to install
31924 the vector in d.perm.
31925 * config/arm/arm.c (expand_vec_perm_d::perm): Change
31926 from auto_vec_perm_indices to vec_perm_indices.
31927 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
31928 instead of changing individual elements.
31929 (arm_vectorize_vec_perm_const): Use new_vector to install
31930 the vector in d.perm.
31931 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
31932 Update vec_perm_builder constructor.
31933 (rs6000_expand_interleave): Likewise.
31934 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
31935 (rs6000_expand_interleave): Likewise.
31937 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31939 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
31940 to qimode could truncate the indices.
31941 * optabs.c (expand_vec_perm_var): Likewise.
31943 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
31945 * Makefile.in (OBJS): Add vec-perm-indices.o.
31946 * vec-perm-indices.h: New file.
31947 * vec-perm-indices.c: Likewise.
31948 * target.h (vec_perm_indices): Replace with a forward class
31950 (auto_vec_perm_indices): Move to vec-perm-indices.h.
31951 * optabs.h: Include vec-perm-indices.h.
31952 (expand_vec_perm): Delete.
31953 (selector_fits_mode_p, expand_vec_perm_var): Declare.
31954 (expand_vec_perm_const): Declare.
31955 * target.def (vec_perm_const_ok): Replace with...
31956 (vec_perm_const): ...this new hook.
31957 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
31958 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
31959 * doc/tm.texi: Regenerate.
31960 * optabs.def (vec_perm_const): Delete.
31961 * doc/md.texi (vec_perm_const): Likewise.
31962 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
31963 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
31964 expand_vec_perm for constant permutation vectors. Assert that
31965 the mode of variable permutation vectors is the integer equivalent
31966 of the mode that is being permuted.
31967 * optabs-query.h (selector_fits_mode_p): Declare.
31968 * optabs-query.c: Include vec-perm-indices.h.
31969 (selector_fits_mode_p): New function.
31970 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
31971 is defined, instead of checking whether the vec_perm_const_optab
31972 exists. Use targetm.vectorize.vec_perm_const instead of
31973 targetm.vectorize.vec_perm_const_ok. Check whether the indices
31974 fit in the vector mode before using a variable permute.
31975 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
31976 vec_perm_indices instead of an rtx.
31977 (expand_vec_perm): Replace with...
31978 (expand_vec_perm_const): ...this new function. Take the selector
31979 as a vec_perm_indices rather than an rtx. Also take the mode of
31980 the selector. Update call to shift_amt_for_vec_perm_mask.
31981 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
31982 Use vec_perm_indices::new_expanded_vector to expand the original
31983 selector into bytes. Check whether the indices fit in the vector
31984 mode before using a variable permute.
31985 (expand_vec_perm_var): Make global.
31986 (expand_mult_highpart): Use expand_vec_perm_const.
31987 * fold-const.c: Includes vec-perm-indices.h.
31988 * tree-ssa-forwprop.c: Likewise.
31989 * tree-vect-data-refs.c: Likewise.
31990 * tree-vect-generic.c: Likewise.
31991 * tree-vect-loop.c: Likewise.
31992 * tree-vect-slp.c: Likewise.
31993 * tree-vect-stmts.c: Likewise.
31994 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
31996 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
31997 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
31998 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
31999 (aarch64_vectorize_vec_perm_const): ...this new function.
32000 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32001 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32002 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32003 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32004 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32005 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32006 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32008 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
32009 check for NEON modes.
32010 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32011 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32012 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32013 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32015 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
32016 the old VEC_PERM_CONST conditions.
32017 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32018 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32019 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32020 (ia64_vectorize_vec_perm_const_ok): Merge into...
32021 (ia64_vectorize_vec_perm_const): ...this new function.
32022 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32023 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32024 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32025 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32026 * config/mips/mips.c (mips_expand_vec_perm_const)
32027 (mips_vectorize_vec_perm_const_ok): Merge into...
32028 (mips_vectorize_vec_perm_const): ...this new function.
32029 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32030 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32031 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32032 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32033 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32034 (rs6000_expand_vec_perm_const): Delete.
32035 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32037 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32038 (altivec_expand_vec_perm_const_le): Take each operand individually.
32039 Operate on constant selectors rather than rtxes.
32040 (altivec_expand_vec_perm_const): Likewise. Update call to
32041 altivec_expand_vec_perm_const_le.
32042 (rs6000_expand_vec_perm_const): Delete.
32043 (rs6000_vectorize_vec_perm_const_ok): Delete.
32044 (rs6000_vectorize_vec_perm_const): New function.
32045 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32046 an element count and rtx array.
32047 (rs6000_expand_extract_even): Update call accordingly.
32048 (rs6000_expand_interleave): Likewise.
32049 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32050 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32051 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32052 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32053 (rs6000_expand_vec_perm_const): Delete.
32054 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32055 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32056 (altivec_expand_vec_perm_const_le): Take each operand individually.
32057 Operate on constant selectors rather than rtxes.
32058 (altivec_expand_vec_perm_const): Likewise. Update call to
32059 altivec_expand_vec_perm_const_le.
32060 (rs6000_expand_vec_perm_const): Delete.
32061 (rs6000_vectorize_vec_perm_const_ok): Delete.
32062 (rs6000_vectorize_vec_perm_const): New function. Remove stray
32063 reference to the SPE evmerge intructions.
32064 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32065 an element count and rtx array.
32066 (rs6000_expand_extract_even): Update call accordingly.
32067 (rs6000_expand_interleave): Likewise.
32068 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32069 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32071 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32073 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32075 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32076 vector mode and that that mode matches the mode of the data
32078 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32079 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
32080 directly using expand_vec_perm_1 when forcing selectors into
32082 (expand_vec_perm_var): New function, split out from expand_vec_perm.
32084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32086 * optabs-query.h (can_vec_perm_p): Delete.
32087 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32088 * optabs-query.c (can_vec_perm_p): Split into...
32089 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32090 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32091 particular selector is valid.
32092 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32093 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32094 (vect_grouped_load_supported): Likewise.
32095 (vect_shift_permute_load_chain): Likewise.
32096 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32097 (vect_transform_slp_perm_load): Likewise.
32098 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32099 (vectorizable_bswap): Likewise.
32100 (vect_gen_perm_mask_checked): Likewise.
32101 * fold-const.c (fold_ternary_loc): Likewise. Don't take
32102 implementations of variable permutation vectors into account
32103 when deciding which selector to use.
32104 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32105 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32106 with a false third argument.
32107 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32108 to test whether the constant selector is valid and can_vec_perm_var_p
32109 to test whether a variable selector is valid.
32111 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32113 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32114 * optabs-query.c (can_vec_perm_p): Likewise.
32115 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32116 instead of vec_perm_indices.
32117 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32118 (vect_gen_perm_mask_checked): Likewise,
32119 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32120 (vect_gen_perm_mask_checked): Likewise,
32122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32124 * optabs-query.h (qimode_for_vec_perm): Declare.
32125 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32126 (qimode_for_vec_perm): ...this new function.
32127 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32129 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
32131 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32132 does not have a conditional at the top.
32134 2018-01-02 Richard Biener <rguenther@suse.de>
32136 * ipa-inline.c (big_speedup_p): Fix expression.
32138 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
32141 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32144 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
32148 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32149 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32150 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32151 cond_taken_branch_cost 3->4.
32153 2018-01-01 Jakub Jelinek <jakub@redhat.com>
32155 PR tree-optimization/83581
32156 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32157 TODO_cleanup_cfg if any changes have been made.
32159 PR middle-end/83608
32160 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32161 convert_modes if target mode has the right side, but different mode
32164 PR middle-end/83609
32165 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32166 last argument when extracting from CONCAT. If either from_real or
32167 from_imag is NULL, use expansion through memory. If result is not
32168 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32169 the parts directly to inner mode, if even that fails, use expansion
32172 PR middle-end/83623
32173 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32174 check for bswap in mode rather than HImode and use that in expand_unop
32177 Copyright (C) 2018 Free Software Foundation, Inc.
32179 Copying and distribution of this file, with or without modification,
32180 are permitted in any medium without royalty provided the copyright
32181 notice and this notice are preserved.