1 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
3 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4 (*simple_return_indirect_internal<mode>): Rename from
5 simple_return_indirect_internal. Use W mode iterator.
6 (rstorssp): New expander.
7 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
8 (clrssbsy): New expander.
9 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
11 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
13 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
15 (cmse_check_pointed_object): Likewise.
17 2018-06-05 Martin Liska <mliska@suse.cz>
20 * doc/invoke.texi: Document how -fprofile-dir format
23 2018-06-05 Richard Biener <rguenther@suse.de>
25 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
26 removal pretend DOM info isn't available so we do not update
27 it and only remove edges, not dominated blocks. Actually free
28 DOM info in case we removed something. Remove unreachable blocks.
29 (mfb_keep_latches): Work with either DOM info or marked backedges.
30 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
31 first. Mark backedges if DOM info isn't available.
32 (Re-)compute DOM info after cleanup_control_flow_pre.
34 2018-06-05 Richard Biener <rguenther@suse.de>
36 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
37 (locus_discrim_hasher::hash): Adjust.
38 (locus_discrim_hasher::equal): Likewise.
39 (next_discriminator_for_locus): Work on line directly.
40 (same_line_p): Pass in expanded locus1 as well.
41 (assign_discriminators): Avoid redundant location expansions.
43 2018-06-05 Richard Biener <rguenther@suse.de>
45 PR tree-optimization/86046
46 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
47 if required after clearing TREE_ADDRESSABLE.
49 2018-06-05 Richard Biener <rguenther@suse.de>
51 PR tree-optimization/86047
52 * tree-ssa-loop.c (for_each_index): Glob handling of all
53 decls and constants and really handle all of them.
55 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
58 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
59 qualifier_void_pointer and qualifier_const_void_pointer.
60 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
61 (arm_init_builtins): Handle the above.
62 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
63 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
64 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
67 2018-06-05 Martin Liska <mliska@suse.cz>
69 * auto-profile.c (read_autofdo_file): Do not use
70 gcov_ctr_summary struct.
71 (afdo_callsite_hot_enough_for_early_inline): Likewise.
72 * coverage.c (struct counts_entry): Likewise.
73 (read_counts_file): Read just single summary entry.
74 (get_coverage_counts): Use gcov_summary struct.
75 * coverage.h (get_coverage_counts): Likewise.
76 * gcov-dump.c (dump_working_sets): Likewise.
77 (tag_summary): Dump just single summary.
78 * gcov-io.c (gcov_write_summary): Write just histogram
80 (gcov_read_summary): Read just single summary.
81 (compute_working_sets): Use gcov_summary struct.
82 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
83 of GCOV_COUNTERS_SUMMABLE.
84 (GCOV_COUNTERS_SUMMABLE): Remove.
85 (GCOV_FIRST_VALUE_COUNTER): Replace with
86 GCOV_COUNTER_V_INTERVAL.
87 (struct gcov_ctr_summary): Remove.
88 (struct gcov_summary): Directly use fields of former
90 (compute_working_sets): Use gcov_summary struct.
91 * gcov.c (read_count_file): Do not use ctrs fields.
92 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
94 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
96 * profile.c: Likewise.
97 * profile.h: Likewise.
99 2018-06-05 Martin Liska <mliska@suse.cz>
101 PR gcov-profile/84846
102 * gcov.c (output_lines): Print working directory only
103 in intermediate format.
105 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
107 * config/s390/s390-builtin-types.def: Add void function type.
108 * config/s390/s390-builtins.def: Use the function type for the
111 2018-06-04 Jim Wilson <jimw@sifive.com>
113 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
115 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
116 and maybe_eh_return. Change regno to unsigned int. Use new args to
117 handle EH_RETURN_DATA_REGNO registers properly.
118 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
119 (riscv_expand_epilogue): Update comment. Change argument name and
120 type. Update code to use new name and type. Pass new args to
121 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
123 * config/riscv/riscv.md (NORMAL_RETURN): New.
124 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
125 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
126 (eh_return): Call gen_eh_return_internal and emit barrier.
127 (eh_return_internal): Call riscv_expand_epilogue.
129 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
131 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
132 bit_insertion field and declare can_be_merged_into method.
133 (merged_store_group::can_be_merged_into): New method.
134 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
135 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
136 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
138 2018-06-04 Richard Biener <rguenther@suse.de>
140 PR tree-optimization/85955
141 * builtins.c (fold_builtin_sincos): Convert pointers to
142 destination to appropriate type before dereferencing.
144 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
146 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
148 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
150 * expr.c (expand_expr_real_1): Force the operand into memory if
151 its TYPE_MODE is BLKmode and if there is no integer mode for
152 the number of bits being extracted.
154 2018-06-04 Jakub Jelinek <jakub@redhat.com>
158 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
159 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
161 2018-06-04 Richard Biener <rguenther@suse.de>
163 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
164 (cleanup_tree_cfg_noloop): ... single caller. Do
165 start_recording_case_labels later.
167 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
169 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
170 to _IMMINTRIN_H_INCLUDED.
171 * config/i386/pconfigintrin.h: Ditto.
172 * config/i386/waitpkgintrin.h: Ditto.
173 * config/i386/immintrin.h: Add includes for sgxintrin.h,
174 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
175 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
176 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
177 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
178 waitpkgintrin.h and cldemoteintrin.h.
180 2018-06-04 Richard Biener <rguenther@suse.de>
182 PR tree-optimization/86038
183 * tracer.c (find_best_successor): Check probability for
184 being initialized, bail out if not.
186 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
189 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
190 of bits to ignore when comparing architectures.
192 2018-06-04 Jakub Jelinek <jakub@redhat.com>
194 PR tree-optimization/69615
195 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
196 maximum or minimum of the type, try to merge it also as if
197 range1 is + [-, x - 1] or + [x + 1, -].
200 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
202 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
204 PR tree-optimization/86034
205 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
206 the unsigned bitfield type in a bit insertion sequence if it does not
207 have a larger precision than the bitfield size.
208 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
210 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
212 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
214 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
216 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
217 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
218 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
219 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
221 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
223 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
224 Disable -fdelete-null-pointer-checks for ELF toolchain.
226 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
227 Kito Cheng <kito.cheng@gmail.com>
229 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
230 (nds32le-*-*, nds32be-*-*): Integrate checking process.
231 (nds32*-*-*): Add glibc and uclibc conditions.
232 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
233 (TARGET_EXCEPT_UNWIND_INFO): Define.
234 * config/nds32/elf.h: New file.
235 * config/nds32/linux.h: New file.
236 * config/nds32/nds32-elf.opt: New file.
237 * config/nds32/nds32-linux.opt: New file.
238 * config/nds32/nds32-fp-as-gp.c
239 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
240 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
242 (nds32_asm_file_end): Ditto.
243 (nds32_print_operand): Ditto.
244 (nds32_insert_attributes): Ditto.
245 (nds32_init_libfuncs): New function.
246 (TARGET_HAVE_TLS): Define.
247 (TARGET_INIT_LIBFUNCS): Define.
248 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
250 (TARGET_ELF): Apply different mcmodel setting.
251 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
252 been migrated into elf.h and linux.h files.
253 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
254 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
255 (mcmodel): The content has been migrated into nds32-elf.opt and
256 nds32-linux.opt files.
257 * config/nds32/t-elf: New file.
258 * config/nds32/t-linux: New file.
260 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
261 Shiva Chen <shiva0217@gmail.com>
263 * config/nds32/constants.md (unspec_volatile_element): Add
264 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
265 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
267 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
268 (make_pass_nds32_fp_as_gp): Declare.
269 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
271 (nds32_asm_function_end_prologue): Remove unused asm output.
272 (nds32_asm_function_begin_epilogue): Remove unused asm output.
273 (nds32_asm_file_start): Output necessary fp_as_gp information.
274 (nds32_option_override): Adjust register usage.
275 (nds32_expand_prologue): Consider fp_as_gp situation.
276 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
277 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
280 (simple_return): Ditto.
281 (omit_fp_begin): Output special directive for fp_as_gp.
282 (omit_fp_end): Output special directive for fp_as_gp.
283 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
284 mforbid-fp-as-gp): New options.
286 2018-06-01 Mark Wielaard <mark@klomp.org>
288 * dwarf2out.c (dwarf2out_finish): Remove generation of
291 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
293 * gimple-ssa-store-merging.c: Include gimple-fold.h.
294 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
295 (struct merged_store_group): Add bit_insertion field.
296 (dump_char_array): Use standard hexadecimal format.
297 (merged_store_group::merged_store_group): Set bit_insertion to false.
298 (merged_store_group::apply_stores): Use optimal buffer size. Deal
299 with BIT_INSERT_EXPR stores. Move up code updating the mask and
300 also print the mask in the dump file.
301 (pass_store_merging::gate): Minor tweak.
302 (imm_store_chain_info::coalesce_immediate): Fix wrong association
303 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
304 stores with INTEGER_CST stores.
305 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
306 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
307 and use it throughout. Generate bit insertion sequences if need be.
308 (pass_store_merging::process_store): Remove redundant condition.
309 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
311 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
313 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
314 the 128-bit floating point types. Fix function comment.
316 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
318 * config/aarch64/aarch64-simd.md
319 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
321 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
324 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
326 PR tree-optimization/85989
327 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
329 (backprop::intersect_uses): Check it when deciding whether this
330 is a backedge reference.
331 (backprop::process_block): Add each phi to m_visited_phis
332 after visiting it, then clear it at the end.
334 2018-06-01 Richard Biener <rguenther@suse.de>
336 * tree-vectorizer.h (vect_dr_stmt): New function.
337 (vect_get_load_cost): Adjust.
338 (vect_get_store_cost): Likewise.
339 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
340 Use vect_dr_stmt instead of DR_SMTT.
341 (vect_record_base_alignments): Likewise.
342 (vect_calculate_target_alignment): Likewise.
343 (vect_compute_data_ref_alignment): Likewise and make static.
344 (vect_update_misalignment_for_peel): Likewise.
345 (vect_verify_datarefs_alignment): Likewise.
346 (vector_alignment_reachable_p): Likewise.
347 (vect_get_data_access_cost): Likewise. Pass down
348 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
349 (vect_get_peeling_costs_all_drs): Likewise.
350 (vect_peeling_hash_get_lowest_cost): Likewise.
351 (vect_enhance_data_refs_alignment): Likewise.
352 (vect_find_same_alignment_drs): Likewise.
353 (vect_analyze_data_refs_alignment): Likewise.
354 (vect_analyze_group_access_1): Likewise.
355 (vect_analyze_group_access): Likewise.
356 (vect_analyze_data_ref_access): Likewise.
357 (vect_analyze_data_ref_accesses): Likewise.
358 (vect_vfa_segment_size): Likewise.
359 (vect_small_gap_p): Likewise.
360 (vectorizable_with_step_bound_p): Likewise.
361 (vect_prune_runtime_alias_test_list): Likewise.
362 (vect_analyze_data_refs): Likewise.
363 (vect_supportable_dr_alignment): Likewise.
364 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
365 (vect_gen_prolog_loop_niters): Likewise.
366 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
367 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
369 (vect_recog_mask_conversion_pattern): Likewise.
370 (vect_try_gather_scatter_pattern): Likewise.
371 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
372 to vect_get_store_cost.
373 (vect_get_store_cost): Get stmt_info instead of DR.
374 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
375 (vect_get_load_cost): Get stmt_info instead of DR.
377 2018-06-01 Richard Biener <rguenther@suse.de>
380 * gimple-fold.c (var_decl_component_p): Also allow offsetted
381 vars wrapped in MEM_REFs.
383 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
385 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
386 Fix subreg tests so that we only return a choice between
387 GENERAL_REGS and FP_REGS if the original classes included both.
389 2018-06-01 Richard Biener <rguenther@suse.de>
392 * tree-ssa-structalias.c (get_function_part_constraint):
393 Handle NULL fi->decl.
394 (find_func_aliases_for_call): Properly handle indirect
396 (find_func_clobbers): Likewise.
397 (ipa_pta_execute): Likewise.
398 (create_variable_info_for): For functions that are ifunc_resolver
399 resolve to a varinfo that contains the result of the resolver call.
400 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
403 2018-05-31 Michael Collison <michael.collison@arm.com>
405 * config/aarch64/aarch64.md:
406 (*fix_to_zero_extenddfdi2): New pattern.
407 * gcc.target/aarch64/fix_extend1.c: New testcase.
409 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
413 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
414 and BUILT_IN_STRNCMP_EQ.
415 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
417 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
418 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
419 (gimple_fold_builtin): Likewise.
420 * tree-ssa-strlen.c (compute_string_length): New function.
421 (determine_min_obsize): New function.
422 (handle_builtin_string_cmp): New function to handle calls to
423 string compare functions.
424 (strlen_optimize_stmt): Add handling to builtin string compare
426 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
427 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
428 * tree.c (build_common_builtin_nodes): Add new defines of
429 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
431 2018-05-31 Jakub Jelinek <jakub@redhat.com>
434 * bb-reorder.c (pass_partition_blocks::gate): Return false for
435 functions with naked attribute.
437 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
439 * config/i386/sse.md (avx_vec_concat<mode>):
440 Substitute concat_tg_mode mode attribute with xtg_mode.
441 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
442 (concat_tg_mode): Remove mode attribute.
444 2018-05-31 Martin Sebor <msebor@redhat.com>
447 * calls.c (alloc_max_size): Correct a logic error/typo.
448 Treat excessive arguments as infinite. Warn for invalid arguments.
449 * doc/invoke.texi (-Walloc-size-larger-than): Update.
451 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
454 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
455 and movx for Haswell.
457 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
458 Cesar Philippidis <cesar@codesourcery.com>
461 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
462 when emitting error on private/firstprivate reductions.
463 * omp-low.c (lower_omp_target): Avoid reference-type processing
464 on pointers for firstprivate clause.
466 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
468 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
471 * config/aarch64/aarch64-simd.md
472 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
473 (aarch64_ld1_x3_<mode>): Likewise
474 (aarch64_st1x2<VALLDIF:mode>): Likewise
475 (aarch64_st1_x2_<mode>): Likewise
476 (aarch64_st1x3<VALLDIF:mode>): Likewise
477 (aarch64_st1_x3_<mode>): Likewise
478 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
479 (vld1_s8_x3): Likewise.
480 (vld1_u16_x3): Likewise.
481 (vld1_s16_x3): Likewise.
482 (vld1_u32_x3): Likewise.
483 (vld1_s32_x3): Likewise.
484 (vld1_u64_x3): Likewise.
485 (vld1_s64_x3): Likewise.
486 (vld1_f16_x3): Likewise.
487 (vld1_f32_x3): Likewise.
488 (vld1_f64_x3): Likewise.
489 (vld1_p8_x3): Likewise.
490 (vld1_p16_x3): Likewise.
491 (vld1_p64_x3): Likewise.
492 (vld1q_u8_x3): Likewise.
493 (vld1q_s8_x3): Likewise.
494 (vld1q_u16_x3): Likewise.
495 (vld1q_s16_x3): Likewise.
496 (vld1q_u32_x3): Likewise.
497 (vld1q_s32_x3): Likewise.
498 (vld1q_u64_x3): Likewise.
499 (vld1q_s64_x3): Likewise.
500 (vld1q_f16_x3): Likewise.
501 (vld1q_f32_x3): Likewise.
502 (vld1q_f64_x3): Likewise.
503 (vld1q_p8_x3): Likewise.
504 (vld1q_p16_x3): Likewise.
505 (vld1q_p64_x3): Likewise.
506 (vst1_s64_x2): Likewise.
507 (vst1_u64_x2): Likewise.
508 (vst1_f64_x2): Likewise.
509 (vst1_s8_x2): Likewise.
510 (vst1_p8_x2): Likewise.
511 (vst1_s16_x2): Likewise.
512 (vst1_p16_x2): Likewise.
513 (vst1_s32_x2): Likewise.
514 (vst1_u8_x2): Likewise.
515 (vst1_u16_x2): Likewise.
516 (vst1_u32_x2): Likewise.
517 (vst1_f16_x2): Likewise.
518 (vst1_f32_x2): Likewise.
519 (vst1_p64_x2): Likewise.
520 (vst1q_s8_x2): Likewise.
521 (vst1q_p8_x2): Likewise.
522 (vst1q_s16_x2): Likewise.
523 (vst1q_p16_x2): Likewise.
524 (vst1q_s32_x2): Likewise.
525 (vst1q_s64_x2): Likewise.
526 (vst1q_u8_x2): Likewise.
527 (vst1q_u16_x2): Likewise.
528 (vst1q_u32_x2): Likewise.
529 (vst1q_u64_x2): Likewise.
530 (vst1q_f16_x2): Likewise.
531 (vst1q_f32_x2): Likewise.
532 (vst1q_f64_x2): Likewise.
533 (vst1q_p64_x2): Likewise.
534 (vst1_s64_x3): Likewise.
535 (vst1_u64_x3): Likewise.
536 (vst1_f64_x3): Likewise.
537 (vst1_s8_x3): Likewise.
538 (vst1_p8_x3): Likewise.
539 (vst1_s16_x3): Likewise.
540 (vst1_p16_x3): Likewise.
541 (vst1_s32_x3): Likewise.
542 (vst1_u8_x3): Likewise.
543 (vst1_u16_x3): Likewise.
544 (vst1_u32_x3): Likewise.
545 (vst1_f16_x3): Likewise.
546 (vst1_f32_x3): Likewise.
547 (vst1_p64_x3): Likewise.
548 (vst1q_s8_x3): Likewise.
549 (vst1q_p8_x3): Likewise.
550 (vst1q_s16_x3): Likewise.
551 (vst1q_p16_x3): Likewise.
552 (vst1q_s32_x3): Likewise.
553 (vst1q_s64_x3): Likewise.
554 (vst1q_u8_x3): Likewise.
555 (vst1q_u16_x3): Likewise.
556 (vst1q_u32_x3): Likewise.
557 (vst1q_u64_x3): Likewise.
558 (vst1q_f16_x3): Likewise.
559 (vst1q_f32_x3): Likewise.
560 (vst1q_f64_x3): Likewise.
561 (vst1q_p64_x3): Likewise.
563 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
565 * config/msp430/msp430.c (msp430_output_labelref): Prepend
566 user_label_prefix to name.
568 * tree-core.h: Update comment about the format of NAME string
569 passed to handler in attribute_spec.
571 * config/msp430/msp430.md: Remove erroneous subreg expression from
572 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
573 zero_extend{q,h}isi2.
575 2018-05-30 Borislav Petkov <bp@suse.de>
577 * doc/extend.texi: Document some architecture specific
578 constraints and sort entries.
580 2018-05-30 Martin Sebor <msebor@redhat.com>
583 * builtins.c (expand_builtin_stpcpy_1): New function.
584 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
585 only if the former succeeds.
587 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
589 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
592 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
594 * doc/invoke.texi (-flinker-output): Document
596 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
598 * passes.c (ipa_write_summaries): Only modify statements if body
600 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
601 incrementally linking.
602 (ipa_passes): Likewise.
603 * lto-cgraph.c (lto_output_node): When incrementally linking do not
604 pass down resolution info.
605 * common.opt (flag_incremental_link): Update info.
606 * gcc.c (plugin specs): Turn flinker-output=* to
607 -plugin-opt=-linker-output-known
608 * toplev.c (compile_file): Also cut compilation when doing incremental
610 * flag-types. (enum lto_partition_model): Add
611 LTO_LINKER_OUTPUT_NOLTOREL.
612 (invoke.texi): Add -flinker-output docs.
613 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
614 link same way as WPA; do not stream in dead initializers.
616 * dwarf2out.c (dwarf2out_die_ref_for_decl,
617 darf2out_register_external_decl): Support incremental link.
619 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
621 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
623 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
625 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
626 it down to simple_object_copy_lto_debug_sections.
627 (run_gcc): Determine incremental LTO link time and configure
628 lto1 into non-wpa mode, disable renaming of debug sections.
630 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
632 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
633 descriptions of various incorrectly documented functions.
635 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
638 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
639 address check not strict.
641 2018-05-30 Richard Biener <rguenther@suse.de>
643 PR tree-optimization/85964
644 * tracer.c (better_p): Drop initialized count check, we only
645 call the function with initialized counts now.
646 (find_best_successor): Do find a best edge if one
647 has uninitialized count.
648 (find_best_predecessor): Likewise. Do BB frequency check only
649 if count is initialized.
651 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
653 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
654 (aarch64_ldrstr_offset_compare): New.
655 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
656 load/store orderings.
657 (aarch64_gen_adjusted_ldpstp): Likewise.
659 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
661 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
662 Check for subset of GENERAL_REGS and FP_REGS.
663 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
666 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
668 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
669 and wi::to_poly_offset. Add the current offset and then check
670 whether the sum fits, rather than using an unchecked addition of
671 a checked term. Check for a shwi rather than a uhwi.
672 * expr.c (get_bit_range): Use tree_to_poly_uint64.
673 (store_constructor): Use poly_int_tree_p.
674 (expand_expr_real_1): Likewise.
675 * function.c (assign_temp): Likewise.
676 * fold-const.c (const_binop): Use poly_int_tree_p and
678 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
680 * ipa-icf-gimple.c (func_checker::compare_operand): Use
681 to_poly_offset for MEM offsets.
682 * ipa-icf.c (sem_variable::equals): Likewise.
683 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
684 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
685 wi::to_poly_offset for BIT_FIELD_REF offsets.
686 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
688 * var-tracking.c (emit_note_insn_var_location): Use
691 2018-05-29 Jim Wilson <jimw@sifive.com>
693 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
695 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
698 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
699 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
701 (sse4_1_round<mode>2): Use nonimmediate_operand
702 for operand 1 predicate.
704 2018-05-29 Martin Sebor <msebor@redhat.com>
705 Richard Biener <rguenther@suse.de>
708 * calls.c (get_size_range): Call determine_value_range instead
710 * tree-vrp.h (determine_value_range): Declared new function.
711 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
713 2018-05-29 Richard Biener <rguenther@suse.de>
715 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
716 sure to use non-pattern stmts for get_earlier_stmt arguments.
717 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
718 called on pattern stmts.
719 (get_later_stmt): Likewise.
721 2018-05-29 Martin Liska <mliska@suse.cz>
723 PR gcov-profile/85759
724 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
727 2018-05-29 Jakub Jelinek <jakub@redhat.com>
729 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
731 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
735 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
736 VEC_PACK_FLOAT_EXPR): New tree codes.
737 * tree-pretty-print.c (op_code_prio): Handle
738 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
739 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
740 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
741 * tree-inline.c (estimate_operator_cost): Likewise.
742 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
743 * fold-const.c (const_binop): Likewise.
744 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
745 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
746 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
747 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
748 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
749 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
750 * expr.c (expand_expr_real_2): Likewise.
751 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
752 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
753 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
755 * optabs.c (expand_widen_pattern_expr): For
756 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
757 sign from result type rather than operand's type.
758 (expand_binop_directly): For vec_packu_float_optab and
759 vec_packs_float_optab allow result type to be different from operand's
761 * optabs-tree.c (optab_for_tree_code): Handle
762 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
763 VEC_PACK_FLOAT_EXPR. Formatting fixes.
764 * tree-vect-generic.c (expand_vector_operations_1): Handle
765 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
767 * tree-vect-stmts.c (supportable_widening_operation): Handle
769 (supportable_narrowing_operation): Handle FLOAT_EXPR.
770 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
771 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
772 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
773 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
775 (vec_pack<floatprefix>_float_<mode>): New expander.
776 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
778 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
779 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
780 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
781 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
782 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
784 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
785 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
786 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
787 VEC_PACK_FLOAT_EXPR): Document.
789 2018-05-29 Richard Biener <rguenther@suse.de>
791 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
793 (stmt_vec_info_vec): Make pointer.
794 (init_stmt_vec_info_vec): Remove.
795 (free_stmt_vec_info_vec): Likewise.
796 (set_stmt_vec_info_vec): New function.
797 (free_stmt_vec_infos): Likewise.
798 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
799 (set_vinfo_for_stmt): Likewise.
800 (get_earlier_stmt): Likewise.
801 (get_later_stmt): Likewise.
802 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
803 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
804 (vec_info::~vec_info): Free stmt_vec_infos.
805 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
806 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
807 (pass_slp_vectorize::execute): Likewise.
808 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
809 (free_stmt_vec_info_vec): Likewise.
810 (set_stmt_vec_info_vec): New function.
811 (free_stmt_vec_infos): Likewise.
812 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
813 the global stmt_vec_info_vec.
814 * tree-parloops.c (gather_scalar_reductions): Use
815 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
818 2018-05-29 Richard Biener <rguenther@suse.de>
820 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
822 2018-05-29 Martin Liska <mliska@suse.cz>
823 David Malcolm <dmalcolm@redhat.com>
825 * vec.c (test_reverse): New.
826 (vec_c_tests): Add new test.
827 * vec.h (vl_ptr>::reverse): New function.
829 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
831 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
833 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
836 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
838 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
840 2018-05-28 Richard Biener <rguenther@suse.de>
842 PR tree-optimization/85933
843 * tree-vect-data-refs.c (vect_record_base_alignments): Only
844 look at stmts marked as vectorizable.
846 2018-05-28 Richard Biener <rguenther@suse.de>
848 PR tree-optimization/85934
849 * tree-vect-generic.c (expand_vector_operations_1): Hoist
850 vector boolean check before scalar optimization.
852 2018-05-28 Jakub Jelinek <jakub@redhat.com>
854 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
857 2018-05-28 Mark Wielaard <mark@klomp.org>
859 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
860 if it is an expression containing a minus sign.
862 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
864 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
866 2018-05-27 Paul Koning <ni1d@arrl.net>
868 * config/pdp11/pdp11.md (truncsihi2): Remove.
870 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
871 Chung-Ju Wu <jasonwucj@gmail.com>
873 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
875 (unaligned_store_dw): Ditto.
876 * config/nds32/nds32-memory-manipulation.c
877 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
878 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
879 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
880 (emit_setmem_word_loop): Rename to ...
881 (emit_setmem_doubleword_loop): ... this.
882 (nds32_gen_dup_4_byte_to_word_value): New function.
883 (nds32_gen_dup_8_byte_to_double_word_value): New function.
884 (nds32_expand_setmem_loop): Refine implementation.
885 (nds32_expand_setmem_loop_v3m): Ditto.
886 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
889 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
891 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
893 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
895 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
896 (nds32_init_machine_status): Initialize machine->attr_naked_p and
897 machine->attr_no_prologue_p.
898 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
899 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
900 (nds32_expand_epilogue): Consider attr_naked_p.
901 (nds32_expand_epilogue_v3pop): Likewise.
902 (nds32_can_use_return_insn): Likewise.
903 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
904 attr_no_prologue_p fields.
905 * config/nds32/nds32.opt (mret-in-naked-func): New option.
907 2018-05-27 Jakub Jelinek <jakub@redhat.com>
910 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
913 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
915 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
917 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
919 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
921 (*<floatsuffix>floatv2div2sf2): Rename to ...
922 (*float<floatunssuffix>v2div2sf2): ... this.
923 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
924 (float<floatunssuffix>v2div2sf2_mask): ... this.
925 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
926 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
927 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
929 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
931 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
933 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
935 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
937 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
939 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
940 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
941 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
942 gen_ufix_truncv8dfv8si2.
943 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
944 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
945 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
946 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
947 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
948 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
949 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
950 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
952 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
956 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
958 2018-05-25 Jim Wilson <jimw@sifive.com>
960 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
961 * config/riscv/riscv.c (struct machine_function): Add
962 interrupt_handler_p and attribute_checked_p fields.
963 (riscv_attribute_table): Add interrupt.
964 (riscv_interrupt_type_p): New.
965 (riscv_save_reg_p): Save extra regs for interrupt handler.
966 (riscv_use_save_libcall): Return false for interrupt handler.
967 (riscv_first_stack_step): Add forward declaration.
968 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
969 for interrupt handler with large frame. Use it for saved reg list.
970 (riscv_expand_prologue): Move flag_stack_usage_info support to
971 eliminate duplication.
972 (riscv_expand_epilogue): Generate mret for interrupt handler.
973 (riscv_epilogue_uses): New.
974 (riscv_can_use_return_insn): Return false for interrupt handler.
975 (riscv_function_ok_for_sibcall): Likewise.
976 (riscv_set_current_function): Add interrupt handler support.
977 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
978 * config/riscv/riscv.md (UNSPECV_MRET): New.
980 (riscv_frflags, riscv_fsflags): Use tab after opcode.
982 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
984 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
986 PR tree-optimization/85712
987 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
988 this candidate has already been replaced in-situ by a copy.
990 2018-05-25 Jason Merrill <jason@redhat.com>
992 PR c++/80485 - inline function non-zero address.
993 * symtab.c (nonzero_address): Check DECL_COMDAT.
995 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
998 * config/alpha/alpha.md (ashlsi3): New insn pattern.
999 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1000 extension of SImode operation. Use const123_operand predicate.
1001 (*saddsi_1): Remove.
1002 (*saddl_se_1): Ditto.
1004 (*ssubl_se_1): Ditto.
1005 * config/alpha/predicates.md (const123_operand): New predicate.
1006 * config/alpha/constraints.md (P): Use IN_RANGE.
1008 2018-05-25 Richard Biener <rguenther@suse.de>
1010 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1012 (ref_maybe_used_by_stmt_p): Likewise.
1013 (stmt_may_clobber_ref_p): Likewise.
1014 (stmt_may_clobber_ref_p_1): Likewise.
1015 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1017 (ref_maybe_used_by_stmt_p): Likewise.
1018 (stmt_may_clobber_ref_p): Likewise.
1019 (stmt_may_clobber_ref_p_1): Likewise.
1020 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1021 the alias oracle to disambiguate DRs with stmts DR analysis
1023 (vect_analyze_data_refs): Do not give up on not analyzable
1024 DRs for BB vectorization. Remove code truncating the dataref
1027 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1030 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1031 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1032 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1034 2018-05-25 Richard Biener <rguenther@suse.de>
1036 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1037 function, combining stmt data ref gathering and fatal analysis
1039 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1040 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1041 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1042 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1043 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1045 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1047 PR tree-optimization/85720
1048 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1049 SCC if all partitions are builtins.
1050 (version_loop_by_alias_check): New parameter. Generate cancelable
1051 runtime alias check if all partitions are builtins.
1052 (distribute_loop): Update call to above function.
1054 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1056 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1057 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1058 (parm_default_def_partition_arg): Ditto.
1059 (set_parm_default_def_partition): Ditto.
1060 (get_parm_default_def_partitions): Ditto and make it static.
1061 (get_undefined_value_partitions): Ditto and make it static.
1062 (remove_ssa_form): Refactor call to init_var_map here.
1063 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1064 computation for loop region.
1065 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1066 (register_default_def): Delete.
1067 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1068 (parm_default_def_partition_arg): Ditto.
1069 (set_parm_default_def_partition): Ditto.
1070 (get_parm_default_def_partitions): Ditto and make it static.
1071 (get_undefined_value_partitions): Ditto and make it static.
1072 (coalesce_with_default, coalesce_with_default): Update comment.
1073 (create_coalesce_list_for_region): New func factored out from
1074 create_outofssa_var_map.
1075 (populate_coalesce_list_for_outofssa): New func factored out from
1076 create_outofssa_var_map and coalesce_ssa_name.
1077 (create_outofssa_var_map): Delete.
1078 (coalesce_ssa_name): Refactor to support live range computation.
1079 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1080 (get_parm_default_def_partitions): Delete.
1081 (get_undefined_value_partitions): Ditto.
1082 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1083 computation for loop region.
1084 (new_tree_live_info, loe_visit_block): Ditto.
1085 (live_worklist, set_var_live_on_entry): Ditto.
1086 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1087 * tree-ssa-live.h (struct _var_map): New fields.
1088 (init_var_map): Change decl.
1089 (region_contains_p): New.
1091 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1093 * tree-ssa-live.h (live_merge_and_clear): Delete.
1095 2018-05-25 Richard Biener <rguenther@suse.de>
1098 * tree-dump.c (dequeue_and_dump): Remove access to removed
1099 operand 2 of a SWITCH_EXPR.
1101 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1103 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1104 multiplication and division.
1105 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1106 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1107 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1108 (cond_udiv_optab, cond_umod_optab): New optabs.
1109 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1110 (IFN_COND_RDIV): New internal functions.
1111 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1112 TRUNC_MOD_EXPR and RDIV_EXPR.
1113 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1114 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1116 (SVE_INT_BINARY): Include mult.
1117 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1118 (optab, sve_int_op): Handle mult.
1119 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1121 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1122 for SVE_INT_BINARY_SD.
1124 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1126 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1127 (optab, sve_int_op): Handle div and udiv.
1128 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1129 for SVE_INT_BINARY_SD.
1130 (*<optab><mode>3): New insn for the same.
1132 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1134 * tree-vect-patterns.c: Include predict.h.
1135 (vect_recog_divmod_pattern): Restrict check for division support
1136 to when optimizing for size.
1138 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1140 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1141 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1142 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1143 (gimple_match_op::set_op): Likewise.
1144 (gimple_resimplify4): Declare.
1145 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1146 (expr::gen_transform): Likewise.
1147 (decision_tree::gen): Generate a simplification routine for 4 operands.
1148 * gimple-match-head.c (gimple_simplify): Add an overload for
1149 4 operands. In the top-level function, handle up to 4 call
1150 arguments and call gimple_resimplify4.
1151 (gimple_resimplify4): New function.
1152 (build_call_internal): Pass a fourth operand.
1153 (maybe_push_to_seq): Likewise.
1154 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1155 Fold VEC_COND_EXPRs of an operation and a default value into
1156 an IFN_COND_* function if possible.
1157 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1159 (SVE_COND_FP_BINARY): Include them.
1160 (optab, sve_fp_op): Handle them.
1161 (SVE_INT_BINARY_REV): New code iterator.
1162 (SVE_COND_FP_BINARY_REV): New int iterator.
1163 (commutative): New int attribute.
1164 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1166 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1168 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1169 (*cond_<optab><mode>): New patterns for reversed operands.
1171 2018-05-25 Richard Biener <rguenther@suse.de>
1173 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1174 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1175 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1176 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1177 * tree-vect-data-refs.c (everywhere): Adjust users.
1178 * tree-vect-loop.c (everywhere): Likewise.
1179 * tree-vect-slp.c (everywhere): Likewise.
1180 * tree-vect-stmts.c (everywhere): Likewise.
1181 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1183 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1185 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1187 (gcc_cv_as_section_exclude): ... this.
1188 Try Solaris as #exclude syntax.
1189 * configure: Regenerate.
1190 * config.in: Regenerate.
1191 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1193 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1194 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1196 * varasm.c (default_elf_asm_named_section): Don't check if
1197 HAVE_GAS_SECTION_EXCLUDE is defined.
1199 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1201 * doc/md.texi: Update the documentation of the cond_* optabs
1202 to mention the new final operand. Fix GET_MODE_NUNITS call.
1203 Describe the scalar case too.
1204 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1205 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1207 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1208 (get_conditional_internal_fn): Update comment.
1209 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1210 accumulator value as a final argument to conditional functions.
1211 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1212 a define_expand and add an "else" operand. Assert for now that
1213 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1214 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1215 (*cond_<optab><mode>): New patterns.
1216 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1217 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1218 (UNSPEC_COND_EOR): Delete.
1219 (optab): Remove associated mappings.
1220 (SVE_INT_BINARY): New code iterator.
1221 (sve_int_op): Remove int attribute and add "minus" to the code
1223 (SVE_COND_INT_OP): Delete.
1224 (SVE_COND_FP_OP): Rename to...
1225 (SVE_COND_FP_BINARY): ...this.
1227 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1229 * optabs.c (can_reuse_operands_p): New function.
1230 (maybe_legitimize_operands): Try to reuse the results for
1233 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1235 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1236 Add {q} suffix to insn mnemonic.
1238 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1240 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1241 (msp430_warn_func_return): New.
1243 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1245 * fold-const.c (tree_nonzero_bits): New function.
1246 * fold-const.h (tree_nonzero_bits): Likewise.
1247 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1248 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1250 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1254 * varasm.c (assemble_alias): Check ifunc_resolver only on
1257 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1260 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1261 when memory input operand is handled.
1263 2018-05-24 Luis Machado <luis.machado@linaro.org>
1265 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1267 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1269 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1271 * match.pd: Delay FMA folds until after vectorization.
1273 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1276 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1277 address check not strict.
1279 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1281 * gimple-match.h (gimple_match_op): New class.
1282 (mprts_hook): Replace parameters with a gimple_match_op *.
1283 (maybe_build_generic_op): Likewise.
1284 (gimple_simplified_result_is_gimple_val): Replace parameters with
1285 a const gimple_match_op *.
1286 (gimple_simplify): Replace code_helper * and tree * parameters with
1287 a gimple_match_op * parameter.
1288 (gimple_resimplify1): Replace code_helper *, tree and tree *
1289 parameters with a gimple_match_op * parameter.
1290 (gimple_resimplify2): Likewise.
1291 (gimple_resimplify3): Likewise.
1292 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1293 parameters with a gimple_match_op * parameter.
1294 * gimple-match-head.c (gimple_simplify): Change prototypes of
1295 auto-generated functions to take a gimple_match_op * instead of
1296 separate code_helper * and tree * parameters. Make the same
1297 change in the top-level overload and update calls to the
1298 gimple_resimplify routines. Update calls to the auto-generated
1299 functions and to maybe_push_res_to_seq in the publicly-facing
1300 operation-specific gimple_simplify overloads.
1301 (gimple_match_op::MAX_NUM_OPS): Define.
1302 (gimple_resimplify1): Replace rcode and ops with a single res_op
1303 parameter. Update call to gimple_simplify.
1304 (gimple_resimplify2): Likewise.
1305 (gimple_resimplify3): Likewise.
1306 (mprts_hook): Replace parameters with a gimple_match_op *.
1307 (maybe_build_generic_op): Likewise.
1308 (build_call_internal): Replace type, nargs and ops with
1309 a gimple_match_op *.
1310 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1311 with a single gimple_match_op *. Update calls to mprts_hook,
1312 build_call_internal and gimple_simplified_result_is_gimple_val.
1313 Factor out code that is common to the tree_code and combined_fn cases.
1314 * genmatch.c (expr::gen_transform): Replace tem_code and
1315 tem_ops with a gimple_match_op called tem_op. Update calls
1316 to the gimple_resimplify functions and maybe_push_res_to_seq.
1317 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1318 res_ops. Update call to the gimple_resimplify functions.
1319 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1320 (decision_tree::gen): Make the functions take a gimple_match_op *
1321 called res_op instead of separate res_code and res_ops parameters.
1322 Update call accordingly.
1323 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1324 and ops with a single res_op parameter. Update calls to
1325 maybe_build_generic_op and maybe_push_res_to_seq.
1326 (fold_stmt_1): Update calls to gimple_simplify and
1327 replace_stmt_with_simplification.
1328 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1329 and gimple_simplified_result_is_gimple_val.
1330 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1332 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1333 with a gimple_match_op *.
1334 (vn_nary_build_or_lookup): Likewise. Update call to
1335 vn_nary_build_or_lookup_1.
1336 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1337 gimple_match_op *. Update calls to the gimple_resimplify routines
1338 and to gimple_simplified_result_is_gimple_val.
1339 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1340 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1341 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1342 (visit_nary_op): Likewise.
1343 (visit_reference_op_load): Likewise.
1345 2018-05-23 Luis Machado <luis.machado@linaro.org>
1347 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1348 modifier for printing the step amount.
1350 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1353 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1356 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1358 * doc/sourcebuild.texi (Endianness): New subsubsection.
1360 2018-05-23 Luis Machado <luis.machado@linaro.org>
1362 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1363 <prefetch_dynamic_strides>: New const bool field.
1364 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1365 prefetch_dynamic_strides.
1366 (exynosm1_prefetch_tune): Likewise.
1367 (thunderxt88_prefetch_tune): Likewise.
1368 (thunderx_prefetch_tune): Likewise.
1369 (thunderx2t99_prefetch_tune): Likewise.
1370 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1372 (aarch64_override_options_internal): Update to set
1373 PARAM_PREFETCH_DYNAMIC_STRIDES.
1374 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1375 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1376 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1377 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1378 prefetch-dynamic-strides setting.
1380 2018-05-23 Luis Machado <luis.machado@linaro.org>
1382 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1383 <minimum_stride>: New const int field.
1384 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1385 minimum_stride field defaulting to -1.
1386 (exynosm1_prefetch_tune): Likewise.
1387 (thunderxt88_prefetch_tune): Likewise.
1388 (thunderx_prefetch_tune): Likewise.
1389 (thunderx2t99_prefetch_tune): Likewise.
1390 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1391 <default_opt_level>: Set to 3.
1392 (aarch64_override_options_internal): Update to set
1393 PARAM_PREFETCH_MINIMUM_STRIDE.
1394 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1395 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1396 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1397 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1398 stride is constant and is below the minimum stride threshold.
1400 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1402 * config/arm/arm-cpus.in (mode26): Delete.
1403 (armv4): Delete mode26 reference.
1404 * config/arm/arm.c (arm_configure_build_target): Delete use of
1407 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1409 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1411 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1412 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1414 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1415 (floatunsdidf2): Ditto.
1417 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1419 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1420 (fixuns_trunc<mode>si2_avx512f): Ditto.
1421 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1422 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1423 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1425 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1427 PR rtl-optimization/79985
1428 * df-scan.c (df_insn_refs_collect): Remove special case for
1429 global registers and asm statements.
1431 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1433 * extend.texi (Global Register Variables): Rewrite the bullet list.
1434 Note that the register is available for allocation. Note that access
1435 via inline asm must use constraints. Add note about async-signal
1436 handlers. Remove paragraph about automagic register selection.
1438 2018-05-23 Richard Biener <rguenther@suse.de>
1440 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1441 of fixed offset from memset VN.
1443 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1445 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1447 (alloc_cand_and_find_basis): Initialize first_interp field.
1448 (slsr_process_mul): Modify first_interp field.
1449 (slsr_process_add): Likewise.
1450 (slsr_process_cast): Modify first_interp field for each new
1452 (slsr_process_copy): Likewise.
1453 (dump_candidate): Dump first_interp field.
1454 (replace_mult_candidate): Process all interpretations, not just
1456 (replace_rhs_if_not_dup): Likewise.
1457 (replace_one_candidate): Likewise.
1459 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1461 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1463 (aarch64_needs_frame_chain): New function.
1464 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1466 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1469 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1470 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1471 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1472 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1473 as true for strict-align.
1474 (aarch64_can_inline_p): Perform checks even when callee has no
1475 attributes to check for strict alignment.
1476 * doc/extend.texi (AArch64 Function Attributes): Document
1478 * doc/invoke.texi: (AArch64 Options): Likewise.
1480 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1482 PR tree-optimization/85853
1483 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1484 the handling of the root of the node to...
1485 (vect_slp_analyze_node_operations_1): ...this new function,
1486 and run the whole thing with the child nodes' def types
1487 set according to their SLP node's def type.
1489 2018-05-23 Richard Biener <rguenther@suse.de>
1492 * tree-data-ref.c (create_runtime_alias_checks): Defer
1493 and ignore overflow warnings.
1495 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1497 PR tree-optimization/85822
1498 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1501 2018-05-23 Richard Biener <rguenther@suse.de>
1503 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1504 memset constants via native_interpret_expr.
1506 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1509 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1511 (cgraph_node::create_alias): Likewise.
1512 (cgraph_node::get_availability): Check ifunc_resolver instead
1513 of looking up ifunc attribute.
1514 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1515 * varasm.c (do_assemble_alias): Likewise.
1516 (assemble_alias): Likewise.
1517 (default_binds_local_p_3): Likewise.
1518 * cgraph.h (cgraph_node): Add ifunc_resolver.
1519 (cgraph_node::only_called_directly_or_aliased_p): Return false
1521 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1523 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1524 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1525 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1526 instead of looking up ifunc attribute.
1528 2018-05-22 Luis Machado <luis.machado@linaro.org>
1530 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1532 2018-05-22 Martin Sebor <msebor@redhat.com>
1535 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1536 only when expasion succeeds.
1537 (expand_builtin_strcmp): Same.
1538 (expand_builtin_strncmp): Same.
1540 2018-05-22 Martin Sebor <msebor@redhat.com>
1542 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1544 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1545 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1547 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1548 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1550 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1551 Add check for legitimate_address.
1552 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1553 (aarch64_swap_ldrstr_operands): New.
1554 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1557 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1558 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1560 * config/aarch64/aarch64.md: New patterns to generate stp
1562 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1563 single words and double words.
1564 (load_pair_sw, load_pair_dw): Likewise.
1565 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1567 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1569 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1570 for different mode ldpstp and add peephole for merged zero stores.
1572 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1574 (aarch64_gen_store_pair): Rename calls to match new patterns.
1575 (aarch64_gen_load_pair): Rename calls to match new patterns.
1576 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1577 (load_pair<DREG:mode><DREG2:mode>): ... This.
1578 (store_pair<mode>): Rename to...
1579 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1580 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1582 (V_INT_EQUIV): Handle SImode.
1583 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1586 2018-05-22 Martin Sebor <msebor@redhat.com>
1589 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1590 or ajust the presumed bound on an operation to avoid unnecessary
1593 2018-05-22 Martin Sebor <msebor@redhat.com>
1595 PR tree-optimization/85826
1596 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1597 assuming that a DECL necesarily has a constant size.
1599 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1602 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1604 2018-05-22 Richard Biener <rguenther@suse.de>
1606 PR tree-optimization/85834
1607 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1608 non-constant and non-zero memset arguments.
1610 2018-05-22 Martin Liska <mliska@suse.cz>
1613 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1615 2018-05-22 Richard Biener <rguenther@suse.de>
1617 PR tree-optimization/85863
1618 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1619 comparisons when vectype is specified.
1620 (vectorizable_condition): Do not specify vectype for
1621 vect_is_simple_cond when SLP vectorizing.
1623 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1626 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1627 define __ibm128 as long double.
1628 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1629 as a distinct type when IEEE 128-bit support is enabled.
1630 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1631 128-bit types to use the correct functions.
1632 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1633 convert between 128-bit floating point types that have different
1634 modes but the same representation, instead of using gen_lowpart to
1636 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1638 (IFKF_reg): New attributes to give the register constraints for
1640 (extend<mode>tf2_internal): New insns to mark an explicit
1641 conversion between 128-bit floating point types that have a
1642 different mode but share the same representation.
1644 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1646 PR tree-optimization/85814
1647 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1648 a null return from get_strinfo when unsharing the next
1649 strinfo in the chain.
1651 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1654 * varasm.c (weak_finish): Clean up weak_decls.
1656 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1658 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1659 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1660 UNSPEC_UADALP values.
1661 * config/aarch64/iterators.md (ABAL): New int iterator.
1664 (sur): Add mappings for the above.
1665 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1667 (aarch64_<sur>abal<mode>_4): Likewise.
1668 (aarch64_<sur>adalp<mode>_3): Likewise.
1669 (<sur>sadv16qi): New define_expand.
1671 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1673 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1674 (*movdf_internal): Ditto.
1675 (*rcpsf2_sse): Ditto.
1676 (*rsqrtsf2_sse): Ditto.
1677 (*sqrt<mode>2_sse): Ditto.
1679 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1681 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1683 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1684 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1685 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1686 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1688 * config/aarch64/arm_neon.h: Likewise.
1689 * config/aarch64/iterators.md (VQ_I): New.
1691 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1693 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1695 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1698 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1700 * config/nds32/constraints.md (S): New constraint.
1701 * config/nds32/nds32.md (call_internal): Use constraint S.
1702 (call_value_internal): Likewise.
1703 (sibcall_internal): Likewise.
1704 (sibcall_value_internal): Likewise.
1706 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1707 Chung-Ju Wu <jasonwucj@gmail.com>
1709 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1712 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1713 Chung-Ju Wu <jasonwucj@gmail.com>
1715 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1716 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1717 (nds32_rtx_costs_impl): Simplify.
1718 (nds32_address_cost_impl): Simplify.
1719 (nds32_init_rtx_costs): New function.
1720 (nds32_rtx_costs_speed_prefer): Likewise.
1721 (nds32_rtx_costs_size_prefer): Likewise.
1722 (nds32_address_cost_speed_prefer): Likewise.
1723 (nds32_address_cost_speed_fwprop): Likewise.
1724 (nds32_address_cost_size_prefer): Likewise.
1725 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1726 * config/nds32/nds32.c (nds32_option_override): Use
1727 nds32_init_rtx_costs function.
1729 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1731 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1732 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1733 (TARGET_PIPELINE_N8): Likewise.
1734 (TARGET_PIPELINE_N10): Likewise.
1735 (TARGET_PIPELINE_N13): Likewise.
1736 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1738 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1740 * config/nds32/nds32-fpu.md: Update copyright year.
1742 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1744 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1746 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1748 * config/nds32/nds32.c
1749 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1750 * config/nds32/nds32.opt (minline-asm-r15): New option.
1752 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1754 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1756 * config/nds32/nds32.md (abssi2): New pattern.
1758 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1760 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1761 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1762 Merge insn pattern from sse_cvtsi2ss<round_name> and
1763 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1764 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1765 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1766 using SWI48 mode iterator.
1767 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1768 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1769 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1770 pattern from sse_cvttss2si<round_saeonly_name>
1771 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1772 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1773 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1774 using SWI48 mode iterator.
1775 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1776 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1777 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1778 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1779 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1780 using SWI48 mode iterator.
1781 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1782 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1783 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1784 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1785 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1786 SWI48 mode iterator.
1787 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1788 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1789 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1790 pattern from sse_cvttsd2si<round_saeonly_name>
1791 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1793 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1795 * config/nds32/nds32-md-auxiliary.c
1796 (nds32_valid_smw_lwm_base_p): Refine.
1797 (nds32_output_smw_single_word): Refine.
1798 (nds32_output_smw_double_word): New.
1799 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1801 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1803 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1804 (nds32_output_stack_pop): Refine.
1805 (nds32_expand_unaligned_load): Refine.
1806 (nds32_expand_unaligned_store): Refine.
1808 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1809 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config/nds32/constants.md: Add TP_REGNUM constant.
1812 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1813 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1815 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1816 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1817 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1818 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1819 and PIC code generation.
1820 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1822 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1824 * config/nds32/nds32.md: Support TLS and PIC.
1825 * config/nds32/nds32.c: Support TLS and PIC.
1826 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1827 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1830 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1832 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1833 mode with E_ prefix.
1835 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1836 Chung-Ju Wu <jasonwucj@gmail.com>
1838 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1839 * config/nds32/nds32-md-auxiliary.c
1840 (symbolic_reference_mentioned_p): New.
1841 (nds32_legitimize_ict_address): New.
1842 (nds32_expand_ict_move): New.
1843 (nds32_indirect_call_referenced_p): New.
1844 (nds32_symbol_binds_local_p): Delete.
1845 (nds32_long_call_p): Modify.
1846 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1847 * config/nds32/nds32-protos.h
1848 (symbolic_reference_mentioned_p): Declare.
1849 (nds32_legitimize_ict_address): Declare.
1850 (nds32_expand_ict_move): Declare.
1851 (nds32_indirect_call_referenced_p): Declare.
1852 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1853 (nds32_relax_group): Use nds32_ict_const_p as condition.
1854 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1855 (nds32_asm_file_start): Output ict_model directive in asm code.
1856 (nds32_legitimate_address_p): Consider indirect call.
1857 (nds32_print_operand): Consider indirect call.
1858 (nds32_print_operand_address): Consider indirect call.
1859 (nds32_insert_attributes): Handle "indirect_call" attribute.
1860 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1861 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1862 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1863 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1864 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1865 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1866 (TARGET_ICT_MODEL_SMALL): Define.
1867 (TARGET_ICT_MODEL_LARGE): Define.
1868 * config/nds32/nds32.md (movsi): Consider ict model.
1869 (call, call_value): Consider ict model.
1870 (sibcall, sibcall_value): Consider ict model.
1871 * config/nds32/nds32.opt (mict-model): New option.
1872 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1875 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1876 Monk Chiang <sh.chiang04@gmail.com>
1877 Jim Wilson <jimw@sifive.com>
1879 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1880 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1881 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1882 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1883 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1884 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1885 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1886 compute save_libcall_adjustment properly.
1887 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1888 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1889 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1890 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1891 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1892 (ABI_SPEC): Handle mabi=ilp32e.
1893 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1894 (RVE): Add RVE mask.
1895 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1896 <-march>: Add rv32e as an example.
1898 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1901 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1902 (intra_create_variable_infos): Handle C++ constructors.
1904 2018-05-18 Martin Liska <mliska@suse.cz>
1906 * passes.def: Remove a redundant pass.
1908 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1911 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1913 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1915 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1917 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1919 (armv2, armv2a, armv3, armv3m): Delete architectures.
1920 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1921 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1922 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1924 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1925 (*mulsidi3adddi): Likewise.
1926 (mulsidi3): Likewise.
1927 (*mulsidi3_nov6): Likewise.
1928 (umulsidi3): Likewise.
1929 (umulsidi3_nov6): Likewise.
1930 (umaddsidi4): Likewise.
1931 (*umulsidi3adddi): Likewise.
1932 (smulsi3_highpart): Likewise.
1933 (*smulsi3_highpart_nov6): Likewise.
1934 (umulsi3_highpart): Likewise.
1935 (*umulsi3_highpart_nov6): Likewise.
1936 * config/arm/arm.h (arm_arch3m): Delete.
1937 * config/arm/arm.c (arm_arch3m): Delete.
1938 (arm_option_override_internal): Update armv3-related comment.
1939 (arm_configure_build_target): Delete use of isa_bit_mode32.
1940 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1941 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1942 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1944 (mulusa3): Likewise.
1945 * config/arm/arm-protos.h (arm_arch3m): Delete.
1946 * config/arm/arm-tables.opt: Regenerate.
1947 * config/arm/arm-tune.md: Likewise.
1948 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1949 deleted architectures.
1951 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1953 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1954 (armv5t, armv5te): New features.
1955 (ARMv5, ARMv5e): Delete fgroups.
1956 (ARMv5t, ARMv5te): Adjust for above changes.
1958 (armv5, armv5e): Delete arches.
1959 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1961 (*call_reg_arm): Likewise.
1962 (*call_value_reg_armv5): Likewise.
1963 (*call_value_reg_arm): Likewise.
1964 (*call_symbol): Likewise.
1965 (*call_value_symbol): Likewise.
1966 (*sibcall_insn): Likewise.
1967 (*sibcall_value_insn): Likewise.
1969 (prefetch): Likewise.
1970 (define_split and define_peephole2 dependent on arm_arch5):
1972 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1974 (TARGET_ARM_QBIT): Likewise.
1975 (TARGET_DSP_MULTIPLY): Likewise.
1976 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1977 (arm_arch5, arm_arch5e): Delete.
1978 (arm_arch5t, arm_arch5te): Declare.
1979 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1980 (arm_arch5t): Declare.
1981 (arm_option_reconfigure_globals): Update for the above.
1982 (arm_options_perform_arch_sanity_checks): Update comment, replace
1983 use of arm_arch5 with arm_arch5t.
1984 (use_return_insn): Likewise.
1985 (arm_emit_call_insn): Likewise.
1986 (output_return_instruction): Likewise.
1987 (arm_final_prescan_insn): Likewise.
1988 (arm_coproc_builtin_available): Likewise.
1989 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1990 arm_arch5e with arm_arch5t and arm_arch5te.
1991 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1992 (arm_arch5t, arm_arch5te): Declare.
1993 * config/arm/arm-tables.opt: Regenerate.
1994 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1995 * config/arm/t-multilib: Likewise.
1996 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1997 instead of arm_arch5.
1998 (*call_reg_thumb1): Likewise.
1999 (*call_value_reg_thumb1_v5): Likewise.
2000 (*call_value_reg_thumb1): Likewise.
2001 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2003 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2005 2018-05-18 Martin Liska <mliska@suse.cz>
2007 PR gcov-profile/84846
2008 * doc/gcov.texi: Document -t option of gcov tool.
2010 2018-05-18 Martin Liska <mliska@suse.cz>
2012 PR gcov-profile/84846
2013 * gcov.c (print_usage): Add new -t option.
2014 (process_args): Handle the option.
2015 (generate_results): Use stdout as output when requested by
2018 2018-05-18 Martin Liska <mliska@suse.cz>
2020 PR gcov-profile/84846
2021 * coverage.c (coverage_init): Write PWD to .gcno file.
2022 * doc/gcov.texi: Document how working directory is printed.
2023 * gcov-dump.c (dump_gcov_file): Print PWD.
2024 * gcov.c (output_intermediate_file): Likewise.
2025 (read_graph_file): Read PWD string.
2026 (output_lines): Print PWD.
2028 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2031 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2032 for retval and return false if all args to phi are zero.
2034 2018-05-18 Richard Biener <rguenther@suse.de>
2036 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2038 (evrp_dom_walker::before_dom_children): Call it.
2040 2018-05-18 Richard Biener <rguenther@suse.de>
2042 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2043 results when processing array refs with variable index.
2045 2018-05-18 Toon Moene <toon@moene.org>
2047 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2048 directly after that of -floop-interchange. Indicate that both
2049 options are enabled by default when specifying -O3.
2051 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2053 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2054 iterator. Delete separate integer-mode vec_set<mode> expander.
2055 (aarch64_simd_vec_setv2di): Delete.
2056 (vec_setv2di): Delete.
2057 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2058 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2059 the "w, r" alternative.
2061 2018-05-18 Martin Liska <mliska@suse.cz>
2063 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2064 * tree-pass.h (make_pass_lower_switch_O0): New function.
2065 * tree-switch-conversion.c (node_has_low_bound): Remove.
2066 (node_has_high_bound): Likewise.
2067 (node_is_bounded): Likewise.
2068 (class pass_lower_switch): Make it a template type and create
2070 (pass_lower_switch::execute): Add template argument.
2071 (make_pass_lower_switch): New function.
2072 (make_pass_lower_switch_O0): New function.
2073 (do_jump_if_equal): Remove.
2074 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2075 all the hard work to tree optimization passes.
2077 2018-05-18 Martin Liska <mliska@suse.cz>
2079 * dbgcnt.c (limit_low): Renamed from limit.
2080 (limit_high): New variable.
2081 (dbg_cnt_is_enabled): Check for upper limit.
2082 (dbg_cnt): Adjust dumping.
2083 (dbg_cnt_set_limit_by_index): Add new argument for high
2085 (dbg_cnt_set_limit_by_name): Likewise.
2086 (dbg_cnt_process_single_pair): Parse new format.
2087 (dbg_cnt_process_opt): Use strtok.
2088 (dbg_cnt_list_all_counters): Remove 'value' and add
2090 * doc/invoke.texi: Document changes.
2092 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2094 * doc/sourcebuild.texi (scalar_all_fma): Document.
2095 * tree.def (FMA_EXPR): Delete.
2096 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2097 * internal-fn.c (ternary_direct): New macro.
2098 (expand_ternary_optab_fn): Likewise.
2099 (direct_ternary_optab_supported_p): Likewise.
2100 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2101 * builtins.c (fold_builtin_fma): Delete.
2102 (fold_builtin_3): Don't call it.
2103 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2104 * expr.c (expand_expr_real_2): Likewise.
2105 * fold-const.c (operand_equal_p): Likewise.
2106 (fold_ternary_loc): Likewise.
2107 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2108 * gimple.c (DEFTREECODE): Likewise.
2109 * gimplify.c (gimplify_expr): Likewise.
2110 * optabs-tree.c (optab_for_tree_code): Likewise.
2111 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2112 * tree-eh.c (operation_could_trap_p): Likewise.
2113 (stmt_could_throw_1_p): Likewise.
2114 * tree-inline.c (estimate_operator_cost): Likewise.
2115 * tree-pretty-print.c (dump_generic_node): Likewise.
2116 (op_code_prio): Likewise.
2117 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2118 * tree-ssa-operands.c (get_expr_operands): Likewise.
2119 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2120 * fold-const-call.h (fold_fma): Delete.
2121 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2122 CFN_FNMA and CFN_FNMS.
2124 * genmatch.c (combined_fn): New enum.
2125 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2126 (commutative_op): New function.
2127 (commutate): Use it. Handle more than 2 operands.
2128 (dt_operand::gen_gimple_expr): Use commutative_op.
2129 (parser::parse_expr): Allow :c to be used with non-binary
2130 operators if the commutative operand is known.
2131 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2132 CFN_FMS, CFN_FNMA and CFN_FNMS.
2133 (backprop::process_assign_use): Remove FMA_EXPR handling.
2134 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2135 (gen_hsa_fma): New function.
2136 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2137 IFN_FNMA and IFN_FNMS.
2138 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2139 * gimple-fold.h (follow_all_ssa_edges): Declare.
2140 * gimple-fold.c (follow_all_ssa_edges): New function.
2141 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2142 gimple_build interface and use follow_all_ssa_edges to fold the result.
2143 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2144 instead of checking for optabs directly.
2145 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2146 rather than FMA_EXPRs.
2147 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2148 call to IFN_FMA instead of an FMA_EXPR.
2150 2018-05-17 Jim Wilson <jimw@sifive.com>
2152 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2153 sign extended promoted subreg, and the range does not have the sign bit
2154 set, then do a sign extend.
2156 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2157 test, check for sign extended subreg and/or constant operands, and
2158 do a sign extend in that case.
2160 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2162 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2163 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2165 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2166 Change logics_shift_reg to logics_shift_imm.
2167 (thunderx2t99_fp_loadpair_basic): Delete.
2168 (thunderx2t99_fp_storepair_basic): Delete.
2169 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2170 (thunderx2t99_asimd_polynomial): Delete.
2171 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2172 and neon_fp_mul_d_scalar_q.
2173 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2174 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2175 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2176 (thunderx2t99_asimd_lut): Add missing tbl types.
2177 (thunderx2t99_asimd_ext): Delete.
2178 (thunderx2t99_asimd_load1_1_mult): Delete.
2179 (thunderx2t99_asimd_load1_2_mult): Delete.
2180 (thunderx2t99_asimd_load1_ldp): New.
2181 (thunderx2t99_asimd_load1): New.
2182 (thunderx2t99_asimd_load2): Add missing *load2* types.
2183 (thunderx2t99_asimd_load3): New.
2184 (thunderx2t99_asimd_load4): New.
2185 (thunderx2t99_asimd_store1_1_mult): Delete.
2186 (thunderx2t99_asimd_store1_2_mult): Delete.
2187 (thunderx2t99_asimd_store2_mult): Delete.
2188 (thunderx2t99_asimd_store2_onelane): Delete.
2189 (thunderx2t99_asimd_store_stp): New.
2190 (thunderx2t99_asimd_store1): New.
2191 (thunderx2t99_asimd_store2): New.
2192 (thunderx2t99_asimd_store3): New.
2193 (thunderx2t99_asimd_store4): New.
2195 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2197 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2198 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2200 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2201 Segher Boessenkool <segher@kernel.crashing.org>
2204 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2207 2018-05-17 Richard Biener <rguenther@suse.de>
2209 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2210 for pruning loop and prune defs feeding only already visited PHIs.
2212 2018-05-17 Richard Biener <rguenther@suse.de>
2214 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2216 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2217 Richard Biener <rguenther@suse.de>
2219 PR tree-optimization/85793
2220 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2221 for VMAT_ELEMENTWISE.
2223 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2225 * internal-fn.h (lookup_internal_fn): Declare
2226 * internal-fn.c (lookup_internal_fn): New function.
2227 * gimple.c (gimple_build_call_from_tree): Handle calls to
2229 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2230 internal function names.
2231 * tree-pretty-print.c (dump_generic_node): Likewise.
2232 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2234 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2236 * gimple-fold.h (gimple_build): Make the function forms take
2237 combined_fn rather than built_in_function.
2238 (gimple_simplify): Likewise.
2239 * gimple-match-head.c (gimple_simplify): Likewise.
2240 * gimple-fold.c (gimple_build): Likewise.
2241 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2242 rather than gimple_build_call_internal.
2243 (get_initial_defs_for_reduction): Likewise.
2244 (vect_create_epilog_for_reduction): Likewise.
2245 (vectorizable_live_operation): Likewise.
2247 2018-05-17 Martin Liska <mliska@suse.cz>
2249 * gimple-ssa-sprintf.c (format_directive): Do not use
2250 space in between 'G_' and '('.
2252 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2255 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2256 even if the mask is not all ones.
2259 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2261 (ix86_gimple_fold_builtin): Likewise.
2264 * config/i386/i386.c: Include tree-vector-builder.h.
2265 (ix86_vector_shift_count): New function.
2266 (ix86_fold_builtin): Fold shift builtins by scalar count.
2267 (ix86_gimple_fold_builtin): Likewise.
2269 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2270 _mm512_setzero): New intrinsics.
2272 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2273 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2275 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2276 code generation for cases where splatting a value is not useful.
2277 * simplify-rtx.c (simplify_ternary_operation): Simplify
2278 vec_merge across a vec_duplicate and a paradoxical subreg forming
2279 a vector mode to a vec_concat.
2281 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2283 * config.gcc: Support "goldmont-plus".
2284 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2286 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2287 PROCESSOR_GOLDMONT_PLUS.
2288 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2289 (processor_target_table): Add "goldmont-plus".
2290 (PTA_GOLDMONT_PLUS): Define.
2291 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2292 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2293 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2294 (fold_builtin_cpu): Add "goldmont-plus".
2295 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2296 (ix86_option_override_internal): Add "goldmont-plus".
2297 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2298 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2299 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2300 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2302 2018-05-17 Richard Biener <rguenther@suse.de>
2304 PR tree-optimization/85757
2305 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2306 remove defs that only feed that PHI from further processing.
2308 2018-05-16 Jim Wilson <jimw@sifive.com>
2310 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2312 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2314 2018-05-16 Mark Wielaard <mark@klomp.org>
2316 * dwarf2out.c (count_index_strings): New function.
2317 (output_indirect_strings): Call count_index_strings and generate
2318 header for dwarf_version >= 5.
2320 2018-05-16 Mark Wielaard <mark@klomp.org>
2322 * dwarf2out.c (dwarf_FORM): New function.
2323 (set_indirect_string): Use dwarf_FORM.
2324 (reset_indirect_string): Likewise.
2325 (size_of_die): Likewise.
2326 (value_format): Likewise.
2327 (output_die): Likewise.
2328 (add_skeleton_AT_string): Likewise.
2329 (output_macinfo_op): Likewise.
2330 (index_string): Likewise.
2331 (output_index_string_offset): Likewise.
2332 (output_index_string): Likewise.
2333 (count_index_strings): Likewise.
2335 2018-05-16 Carl Love <cel@us.ibm.com>
2337 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2338 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2340 2018-05-16 Martin Jambor <mjambor@suse.cz>
2342 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2343 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2345 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2347 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2348 (fnma<mode>4): Likewise.
2349 (fms<mode>4): Likewise.
2350 (fnms<mode>4): Likewise.
2351 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2352 (aarch64_fnma<mode>4): Likewise.
2353 (aarch64_fms<mode>4): Likewise.
2354 (aarch64_fnms<mode>4): Likewise.
2355 (aarch64_fnmadd<mode>4): Likewise.
2357 2018-05-16 Jason Merrill <jason@redhat.com>
2359 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2361 2018-05-16 Richard Biener <rguenther@suse.de>
2363 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2364 (dump_stmt_cost): Declare.
2365 (add_stmt_cost): Dump cost we add.
2366 (add_stmt_costs): New function.
2367 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2369 (vect_analyze_stmt): Adjust prototype.
2370 (vectorizable_condition): Likewise.
2371 (vectorizable_live_operation): Likewise.
2372 (vectorizable_reduction): Likewise.
2373 (vectorizable_induction): Likewise.
2374 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2375 cost vector to pass to vectorizable_ and record afterwards.
2376 (vect_model_reduction_cost): Take cost vector argument and adjust.
2377 (vect_model_induction_cost): Likewise.
2378 (vectorizable_reduction): Likewise.
2379 (vectorizable_induction): Likewise.
2380 (vectorizable_live_operation): Likewise.
2381 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2382 SLP_TREE_NUMBER_OF_VEC_STMTS.
2383 (vect_analyze_slp_cost_1): Remove.
2384 (vect_analyze_slp_cost): Likewise.
2385 (vect_slp_analyze_node_operations): Take visited args and
2386 a target cost vector. Avoid processing already visited stmt sets.
2387 (vect_slp_analyze_operations): Use a local cost vector to gather
2388 costs and register those of non-discarded instances.
2389 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2390 (vect_schedule_slp_instance): Remove copying of
2391 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2393 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2394 adding cost. Record cost entry location.
2395 (vect_prologue_cost_for_slp_op): Function to compute cost of
2396 a constant or invariant generated for SLP vect in the prologue,
2397 split out from vect_analyze_slp_cost_1.
2398 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2399 (vect_model_promotion_demotion_cost): Likewise.
2400 (vect_model_store_cost): Likewise, make static.
2401 (vect_model_load_cost): Likewise.
2402 (vectorizable_bswap): Add cost vector arg and adjust.
2403 (vectorizable_call): Likewise.
2404 (vectorizable_simd_clone_call): Likewise.
2405 (vectorizable_conversion): Likewise.
2406 (vectorizable_assignment): Likewise.
2407 (vectorizable_shift): Likewise.
2408 (vectorizable_operation): Likewise.
2409 (vectorizable_store): Likewise.
2410 (vectorizable_load): Likewise.
2411 (vectorizable_condition): Likewise.
2412 (vectorizable_comparison): Likewise.
2413 (can_vectorize_live_stmts): Likewise.
2414 (vect_analyze_stmt): Likewise.
2415 (vect_transform_stmt): Adjust calls to vectorizable_*.
2416 * tree-vectorizer.c: Include gimple-pretty-print.h.
2417 (dump_stmt_cost): New function.
2419 2018-05-16 Richard Biener <rguenther@suse.de>
2421 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2422 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2423 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2424 (check_name): New callback.
2425 (dse_classify_store): Track cycles via a visited bitmap of PHI
2426 defs and simplify handling of in-loop and across loop dead stores
2427 and properly fail for loop-variant refs. Handle byte-tracking with
2428 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2431 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2433 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2434 (vect_get_mask_type_for_stmt): Likewise.
2435 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2437 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2438 to determine the statement's vector type and the vector type that
2439 should be used for calculating nunits. Deal with cases in which
2440 the type has to be deferred.
2441 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2442 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2443 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2444 (vect_determine_vf_for_stmt): New functions, split out from...
2445 (vect_determine_vectorization_factor): ...here.
2446 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2447 (vect_get_mask_type_for_stmt): New functions, split out from
2448 vect_determine_vectorization_factor.
2450 2018-05-16 Richard Biener <rguenther@suse.de>
2452 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2453 verify the [VEC_]COND_EXPR embedded comparison.
2455 2018-05-15 Martin Sebor <msebor@redhat.com>
2457 PR tree-optimization/85753
2458 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2459 RECORD_TYPE in addition to ARRAY_TYPE.
2461 2018-05-15 Martin Sebor <msebor@redhat.com>
2464 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2466 2018-05-15 Richard Biener <rguenther@suse.de>
2468 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2469 add by_clobber_p one. Change algorithm to collect all defs
2470 representing uses we need to walk and try reducing them to
2471 a single one before failing.
2472 (dse_dom_walker::dse_optimize_stmt): Adjust.
2474 2018-05-13 Mark Wielaard <mark@klomp.org>
2476 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2477 (size_of_loc_descr): Likewise.
2478 (output_loc_operands): Likewise.
2479 (output_loc_operands_raw): Likewise.
2480 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2481 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2482 (hash_loc_operands): Likewise.
2483 (compare_loc_operands): Likewise.
2485 2018-05-14 Mark Wielaard <mark@klomp.org>
2487 * dwarf2out.c (count_index_addrs): New function.
2488 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2490 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2492 PR tree-optimization/83648
2493 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2494 return value as malloc candidate.
2496 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2499 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2500 param as true in call to suggest_attribute.
2502 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2504 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2505 -mreadonly-in-sdata.
2507 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2509 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2511 (aarch64_crypto_aesd_fused): Likewise.
2513 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2515 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2516 (movsi_aarch64): Likewise.
2517 (load_pairsi): Likewise.
2518 (load_pairdi): Likewise.
2519 (store_pairsi): Likewise.
2520 (store_pairdi): Likewise.
2521 (load_pairsf): Likewise.
2522 (load_pairdf): Likewise.
2523 (store_pairsf): Likewise.
2524 (store_pairdf): Likewise.
2525 (zero_extend): Likewise.
2526 (trunc): Swap alternatives.
2527 (fcvt_target): Add '?' to prefer w over r.
2529 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2532 * config/i386/i386.md: Disallow non-commutative arithmetics in
2533 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2534 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2535 in the peephole2 before it.
2537 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2539 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2540 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2541 (ix86_handle_option): Handle -mcldemote.
2542 * config.gcc: New header.
2543 * config/i386/cldemoteintrin.h: New file.
2544 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2545 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2547 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2548 OPTION_MASK_ISA_CLDEMOTE.
2549 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2550 (ix86_valid_target_attribute_inner_p): Ditto.
2551 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2552 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2553 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2554 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2555 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2557 * config/i386/i386.opt: Add -mcldemote.
2558 * config/i386/x86intrin.h: New header.
2559 * doc/invoke.texi: Add -mcldemote.
2561 2018-05-14 Richard Biener <rguenther@suse.de>
2563 * doc/match-and-simplify.texi: Adjust :s documentation.
2565 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2567 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2568 intended memcpy size.
2569 (REORDER_45): Likewise.
2571 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2573 * sort.cc: New file.
2574 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2575 * vec.c (qsort_chk): Use gcc_qsort.
2576 * Makefile.in (OBJS-libcommon): Add sort.o.
2577 (build/sort.o): New target. Use it...
2578 (BUILD_RTL): ... here, and...
2579 (build/gencfn-macros): ... here, and...
2580 (build/genmatch): ... here.
2582 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2583 Chung-Ju Wu <jasonwucj@gmail.com>
2585 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2586 * config/nds32/nds32-graywolf.md: New file.
2587 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2588 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2590 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2591 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2592 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2593 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2594 * config/nds32/pipelines.md: Include n15 settings.
2596 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2597 Chung-Ju Wu <jasonwucj@gmail.com>
2599 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2600 * config/nds32/nds32-n13.md: New file.
2601 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2602 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2604 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2605 * config/nds32/nds32.md (pipeline_model): Add n13.
2606 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2607 * config/nds32/pipelines.md: Include n13 settings.
2609 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2610 Chung-Ju Wu <jasonwucj@gmail.com>
2612 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2613 * config/nds32/nds32-n10.md: New file.
2614 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2615 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2617 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2618 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2619 * config/nds32/nds32.md (pipeline_model): Add n10.
2620 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2621 * config/nds32/pipelines.md: Include n10 settings.
2623 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2624 Kito Cheng <kito.cheng@gmail.com>
2625 Chung-Ju Wu <jasonwucj@gmail.com>
2627 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2628 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2629 Add enum values for DSP extension instructions.
2630 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2632 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2633 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2635 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2636 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2637 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2638 * config/nds32/nds32-intrinsic.md: Likewise.
2639 * config/nds32/nds32_intrinsic.h: Likewise.
2640 * config/nds32/nds32-md-auxiliary.c: Likewise.
2641 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2642 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2643 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2644 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2645 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2646 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2647 TYPE_DMAC in switch statement.
2648 * config/nds32/nds32.c: New checking and implementation for DSP
2649 extension instructions.
2650 * config/nds32/nds32.h: Likewise.
2651 * config/nds32/nds32.md: Likewise.
2652 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2653 * config/nds32/predicates.md: Implement new predicates for DSP
2656 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2658 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2659 Reformat alternatives and attributes so it is easier to identify
2660 which constraints/attributes go with which instruction.
2661 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2662 (mov<mode>_softfloat32, FMOVE64): Likewise.
2663 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2664 (mov<mode>_softfloat64, FMOVE64): Likewise.
2666 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2668 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2670 (Basic PowerPC Built-in Functions): The new name of the
2671 subsection previously known as "PowerPC Built-in Functions".
2672 (Basic PowerPC Built-in Functions Available on all Configurations):
2674 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2675 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2676 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2677 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2679 2018-05-11 Martin Jambor <mjambor@suse.cz>
2682 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2685 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2688 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2690 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2692 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2693 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2694 (ix86_handle_option): Handle -mwaitpkg.
2695 * config.gcc: New header.
2696 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2697 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2698 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2700 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2701 OPTION_MASK_ISA_WAITPKG.
2702 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2703 (ix86_option_override_internal): Add PTA_WAITPKG.
2704 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2705 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2706 IX86_BUILTIN_TPAUSE.
2707 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2708 __builtin_ia32_umwait and __builtin_ia32_tpause.
2709 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2710 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2711 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2712 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2713 UNSPECV_TPAUSE): New.
2714 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2715 * config/i386/i386.opt: Add -mwaitpkg.
2716 * config/i386/waitpkgintrin.h: New file.
2717 * config/i386/x86intrin.h: New header.
2718 * doc/invoke.texi: Add -mwaitpkg.
2720 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2723 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2725 (cortex-m0): Use armv6s-m isa.
2726 (cortex-m0plus): Likewise.
2727 (cortex-m1): Likewise.
2728 (cortex-m0.small-multiply): Likewise.
2729 (cortex-m0plus.small-multiply): Likewise.
2730 (cortex-m1.small-multiply): Likewise.
2732 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2733 Jakub Jelinek <jakub@redhat.com>
2735 PR tree-optimization/85692
2736 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2737 source permute as well.
2739 2018-05-11 Martin Liska <mliska@suse.cz>
2742 * doc/extend.texi: Document LLVM style format for no_sanitize
2745 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2747 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2748 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2749 (mode_supports_vsx_dform_quad): Likewise.
2750 (mode_supports_vmx_dform): Move these functions to be next to the
2751 other mode_supports functions.
2752 (mode_supports_dq_form): Likewise.
2753 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2754 mode_supports_dq_form.
2755 (reg_offset_addressing_ok_p): Likewise.
2756 (offsettable_ok_by_alignment): Likewise.
2757 (rs6000_legitimate_offset_address_p): Likewise.
2758 (legitimate_lo_sum_address_p): Likewise.
2759 (rs6000_legitimize_address): Likewise.
2760 (rs6000_legitimize_reload_address): Likewise.
2761 (rs6000_secondary_reload_inner): Likewise.
2762 (rs6000_preferred_reload_class): Likewise.
2763 (rs6000_output_move_128bit): Likewise.
2765 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2767 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2768 Generate SImode target register for null target.
2769 <case IX86_BUILTIN_XGETBV>: Ditto.
2770 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2771 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2773 2018-05-10 Carl Love <cel@us.ibm.com>
2775 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2776 dcbtt and dcbtstt if operands[2] is 0.
2778 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2781 * config/i386/sse.md (usadv64qi): New expander.
2783 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2785 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2786 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2787 -maltivec=be support.
2788 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2789 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2790 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2791 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2792 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2793 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2794 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2795 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2796 altivec_vsumsws): Adjust.
2797 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2798 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2799 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2801 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2802 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2803 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2804 (altivec_lve<VI_char>x): Delete expand.
2805 (*altivec_lve<VI_char>x_internal): Rename to...
2806 (altivec_lve<VI_char>x): ... this.
2807 (altivec_lvxl_<mode>): Delete expand.
2808 (*altivec_lvxl_<mode>_internal): Rename to ...
2809 (altivec_lvxl_<mode>): ... this.
2810 (altivec_stvxl_<mode>): Delete expand.
2811 (*altivec_stvxl_<mode>_internal): Rename to ...
2812 (altivec_stvxl_<mode>): ... this.
2813 (altivec_stve<VI_char>x): Delete expand.
2814 (*altivec_stve<VI_char>x_internal): Rename to ...
2815 (altivec_stve<VI_char>x): ... this.
2816 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2817 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2818 reduc_plus_scal_<mode>): Adjust.
2819 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2821 (rs6000_cpu_cpp_builtins): Adjust.
2822 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2823 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2824 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2825 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2826 -maltivec=be support.
2827 (rs6000_split_vec_extract_var): Adjust.
2828 (rs6000_split_v4si_init): Adjust.
2829 (swap_selector_for_mode): Delete.
2830 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2831 altivec_expand_stvex_be): Delete.
2832 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2833 -maltivec=be support.
2834 (rs6000_gimple_fold_builtin): Ditto.
2835 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2837 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2838 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2839 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2840 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2841 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2842 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2843 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2844 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2845 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2846 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2847 anonymous split): Adjust.
2848 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2849 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2851 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2853 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2854 when --with-gxx-include-dir is also specified.
2855 * configure: Regenerate.
2857 2018-05-09 Jim Wilson <jimw@sifive.com>
2860 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2861 * config/riscv/t-withmultilib: New.
2862 * config/riscv/withmultilib.h: New.
2863 * doc/install.texi: Document RISC-V --with-multilib-list support.
2865 2018-05-09 Richard Biener <rguenther@suse.de>
2867 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2869 (vect_bb_vectorization_profitable_p): Adjust. Compute
2870 actual scalar cost using the cost vector and the add_stmt_cost
2873 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2875 PR rtl-optimization/85645
2876 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2877 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2879 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2881 PR rtl-optimization/85645
2882 * shrink-wrap.c (spread_components): Return a boolean saying if
2883 anything was changed.
2884 (try_shrink_wrapping_separate): Iterate spread_components until
2885 nothing changes anymore.
2887 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2889 PR rtl-optimization/85645
2890 * regrename.c (build_def_use): Also kill the chains that include the
2891 destination of a REG_CFA_REGISTER note.
2893 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2895 PR rtl-optimization/85645
2896 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2897 insn that has a REG_CFA_REGISTER note.
2899 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2901 * cfgexpand.c (expand_clobber): New function.
2902 (expand_gimple_stmt_1): Use it.
2903 * tree-vect-stmts.c (vect_clobber_variable): New function,
2905 (vectorizable_simd_clone_call): ...here.
2906 (vectorizable_store): Emit a clobber either side of an
2907 IFN_STORE_LANES sequence.
2908 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2910 2018-05-09 Tom de Vries <tom@codesourcery.com>
2913 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2914 (define_insn "trap_if_false"): Add exit after trap.
2916 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2918 PR rtl-optimization/85638
2919 * bb-reorder.c: Include common/common-target.h.
2920 (create_forwarder_block): New function extracted from...
2921 (fix_up_crossing_landing_pad): ...here. Rename into...
2922 (dw2_fix_up_crossing_landing_pad): ...this.
2923 (sjlj_fix_up_crossing_landing_pad): New function.
2924 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2925 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2926 from both partitions and exit the loop after one iteration.
2928 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2931 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2933 (Basic PowerPC Built-in Functions): The new name of the
2934 subsection previously known as "PowerPC Built-in Functions".
2935 (Basic PowerPC Built-in Functions Available on all Configurations):
2937 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2939 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2941 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2943 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2946 2018-05-08 Jim Wilson <jimw@sifive.com>
2948 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2949 (LD_EMUL_SUFFIX): New.
2950 (LINK_SPEC): Use it.
2952 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2954 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2956 (Basic PowerPC Built-in Functions): The new name of the
2957 subsection previously known as "PowerPC Built-in Functions".
2958 (Basic PowerPC Built-in Functions Available on all Configurations):
2960 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2962 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2964 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2966 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2969 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2972 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2973 after cmpelim optimization.
2975 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2977 * config.gcc: Support "goldmont".
2978 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2979 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2981 * config/i386/i386.c (m_GOLDMONT): Define.
2982 (processor_target_table): Add "goldmont".
2983 (PTA_GOLDMONT): Define.
2984 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2985 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2986 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2987 (fold_builtin_cpu): Add "goldmont".
2988 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2989 (ix86_option_override_internal): Add "goldmont".
2990 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2991 (processor_type): Add PROCESSOR_GOLDMONT.
2992 * config/i386/i386.md: Add CPU "glm".
2993 * config/i386/glm.md: New file.
2994 * config/i386/x86-tune.def: Add m_GOLDMONT.
2995 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2997 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3000 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3002 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3003 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3004 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3007 * config/i386/i386.c (ix86_fold_builtin): Handle
3008 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3011 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3012 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3014 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3017 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3018 (check_arch): Likewise.
3019 (check_fpu): Return the result rather than printing it.
3020 (end arch): Fix operator precedence.
3021 (end cpu): Likewise.
3022 (END): Print the result from check_fpu.
3024 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3025 Alan Hayward <alan.hayward@arm.com>
3026 David Sherwood <david.sherwood@arm.com>
3028 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3029 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3030 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3031 (*fcmuo<mode>_and): New patterns.
3033 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3035 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3036 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3037 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3038 (cmp_op, sve_imm_con): New code attributes.
3039 (SVE_COND_INT_CMP, imm_con): Delete.
3040 (cmp_op): Remove above unspecs from int attribute.
3041 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3043 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3044 comparison-specific unspecs.
3045 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3046 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3047 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3048 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3049 (*vec_fcm<cmp_op><mode>): Rename to...
3050 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3051 (*vec_fcmuo<mode>): Rename to...
3052 (*fcmuo<mode>): ...this and adjust likewise.
3053 (*pred_fcm<cmp_op><mode>): New pattern.
3054 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3055 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3057 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3059 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3060 (aarch64_emit_sve_predicated_cond): New function.
3061 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3062 (aarch64_emit_unspec_cond_or): Replace with...
3063 (aarch64_emit_sve_or_conds): ...this new function. Use
3064 aarch64_emit_sve_ptrue_op for the individual comparisons and
3065 aarch64_emit_binop to OR them together.
3066 (aarch64_emit_inverted_unspec_cond): Replace with...
3067 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3068 aarch64_emit_sve_ptrue_op for the comparison and
3069 aarch64_emit_unop to invert the result.
3070 (aarch64_expand_sve_vec_cmp_float): Update after the above
3071 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3073 2018-05-07 Nathan Sidwell <nathan@acm.org>
3075 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3076 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3077 (Backwards Compatibility): Likewise.
3079 2018-05-07 Luis Machado <luis.machado@linaro.org>
3083 2018-05-07 Luis Machado <luis.machado@linaro.org>
3085 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3086 <prefetch_dynamic_strides>: New const bool field.
3087 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3088 prefetch_dynamic_strides.
3089 (exynosm1_prefetch_tune): Likewise.
3090 (thunderxt88_prefetch_tune): Likewise.
3091 (thunderx_prefetch_tune): Likewise.
3092 (thunderx2t99_prefetch_tune): Likewise.
3093 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3095 (aarch64_override_options_internal): Update to set
3096 PARAM_PREFETCH_DYNAMIC_STRIDES.
3097 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3098 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3099 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3100 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3101 prefetch-dynamic-strides setting.
3103 2018-05-07 Luis Machado <luis.machado@linaro.org>
3105 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3106 <minimum_stride>: New const int field.
3107 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3108 minimum_stride field.
3109 (exynosm1_prefetch_tune): Likewise.
3110 (thunderxt88_prefetch_tune): Likewise.
3111 (thunderx_prefetch_tune): Likewise.
3112 (thunderx2t99_prefetch_tune): Likewise.
3113 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3114 (aarch64_override_options_internal): Update to set
3115 PARAM_PREFETCH_MINIMUM_STRIDE.
3116 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3117 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3118 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3119 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3120 stride is constant and is below the minimum stride threshold.
3122 2018-05-07 Luis Machado <luis.machado@linaro.org>
3124 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3127 2018-05-07 Luis Machado <luis.machado@linaro.org>
3129 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3130 <prefetch_dynamic_strides>: New const bool field.
3131 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3132 prefetch_dynamic_strides.
3133 (exynosm1_prefetch_tune): Likewise.
3134 (thunderxt88_prefetch_tune): Likewise.
3135 (thunderx_prefetch_tune): Likewise.
3136 (thunderx2t99_prefetch_tune): Likewise.
3137 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3139 (aarch64_override_options_internal): Update to set
3140 PARAM_PREFETCH_DYNAMIC_STRIDES.
3141 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3142 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3143 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3144 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3145 prefetch-dynamic-strides setting.
3147 2018-05-07 Luis Machado <luis.machado@linaro.org>
3149 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3150 <minimum_stride>: New const int field.
3151 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3152 minimum_stride field.
3153 (exynosm1_prefetch_tune): Likewise.
3154 (thunderxt88_prefetch_tune): Likewise.
3155 (thunderx_prefetch_tune): Likewise.
3156 (thunderx2t99_prefetch_tune): Likewise.
3157 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3158 (aarch64_override_options_internal): Update to set
3159 PARAM_PREFETCH_MINIMUM_STRIDE.
3160 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3161 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3162 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3163 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3164 stride is constant and is below the minimum stride threshold.
3166 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3169 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3170 the type is addressable. Don't force op into register if it has
3173 2018-05-05 Roland McGrath <mcgrathr@google.com>
3176 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3177 any section for which we don't know a specific type it should have,
3178 regardless of name. Previously this was done only for the exact
3179 names ".init_array", ".fini_array", and ".preinit_array".
3180 (default_elf_asm_named_section): Add comment about
3181 relationship with default_section_type_flags and SECTION_NOTYPE.
3182 (get_section): Don't consider it a type conflict if one side has
3183 SECTION_NOTYPE and the other doesn't, as long as neither has the
3184 SECTION_BSS et al used in the default_section_type_flags logic.
3186 2018-05-05 Tom de Vries <tom@codesourcery.com>
3189 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3190 (workaround_barsyncs): New function.
3191 (nvptx_reorg): Use workaround_barsyncs.
3192 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3193 (define_expand "nvptx_membar_cta"): New define_expand.
3194 (define_insn "*nvptx_membar_cta"): New insn.
3196 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3198 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3199 To improve optimization opportunities.
3200 * builtin-types.def: The new needed builtin types for the above.
3202 2018-05-04 Richard Biener <rguenther@suse.de>
3204 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3205 * gimple-ssa-store-merging.c
3206 (imm_store_chain_info::output_merged_store): Remove redundant create,
3207 release split_store vector contents on failure.
3208 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3209 scalar stmt vector on cache hit.
3211 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3213 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3215 * config.gcc (powerpc-xilinx-eabi*): Remove.
3216 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3218 (fusion_addis_mem_combo_load): Ditto.
3219 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3221 (rs6000_cpu_cpp_builtins): Ditto.
3222 * config/rs6000/rs6000-linux.c
3223 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3224 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3225 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3227 (rs6000_setup_reg_addr_masks): Ditto.
3228 (rs6000_init_hard_regno_mode_ok): Ditto.
3229 (rs6000_option_override_internal): Ditto.
3230 (legitimate_lo_sum_address_p): Ditto.
3231 (rs6000_legitimize_address): Ditto.
3232 (rs6000_legitimize_reload_address): Ditto.
3233 (rs6000_legitimate_address_p): Ditto.
3234 (abi_v4_pass_in_fpr): Ditto.
3235 (setup_incoming_varargs): Ditto.
3236 (rs6000_gimplify_va_arg): Ditto.
3237 (rs6000_split_multireg_move): Ditto.
3238 (rs6000_savres_strategy): Ditto.
3239 (rs6000_emit_prologue_components): Ditto.
3240 (rs6000_emit_epilogue_components): Ditto.
3241 (rs6000_emit_prologue): Ditto.
3242 (rs6000_emit_epilogue): Ditto.
3243 (rs6000_elf_file_end): Ditto.
3244 (rs6000_function_value): Ditto.
3245 (rs6000_libcall_value): Ditto.
3246 * config/rs6000/rs6000.h: Ditto.
3247 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3248 (TARGET_MINMAX): ... this. New.
3249 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3250 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3251 (*movsi_internal1_single): Delete.
3252 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3253 mfpu=, mxilinx-fpu): Delete.
3254 * config/rs6000/singlefp.h: Delete.
3255 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3256 * config/rs6000/t-rs6000: Ditto.
3257 * config/rs6000/t-xilinx: Delete.
3258 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3259 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3260 (VStype_simple): Delete.
3261 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3262 * config/rs6000/xfpu.h: Delete.
3263 * config/rs6000/xfpu.md: Delete.
3264 * config/rs6000/xilinx.h: Delete.
3265 * config/rs6000/xilinx.opt: Delete.
3266 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3267 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3269 2018-05-04 Tom de Vries <tom@codesourcery.com>
3272 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3275 2018-05-04 Richard Biener <rguenther@suse.de>
3278 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3279 (expand_complex_div_wide): Likewise.
3280 (expand_complex_operations_1): Likewise.
3281 (expand_complex_libcall): Preserve EH info of the original stmt.
3282 (tree_lower_complex): Handle removed blocks.
3283 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3284 on complex multiplication and division libcall builtins.
3286 2018-05-04 Richard Biener <rguenther@suse.de>
3289 * fold-const.c (negate_expr_p): Restrict negation of operand
3290 zero of a division to when we know that can happen without
3292 (fold_negate_expr_1): Likewise.
3294 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3297 * real.h (real_nextafter): Declare.
3298 * real.c (real_nextafter): New function.
3299 * fold-const-call.c (fold_const_nextafter): New function.
3300 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3301 CASE_CFN_NEXTTOWARD.
3302 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3303 even when arg1_mode is different from arg0_mode.
3305 2018-05-03 Nathan Sidwell <nathan@acm.org>
3307 * doc/extend.texi (Deprecated Features): Remove
3309 (Backwards Compatibility): Likewise.
3310 * doc/invoke.texi (C++ Language Options): Likewise.
3311 (C++ Dialect Options): Likewise.
3313 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3316 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3317 _mm512_mask_mullox_epi64): New intrinsics.
3319 2018-05-03 Tom de Vries <tom@codesourcery.com>
3322 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3323 dump files): Add offload-tree.
3325 2018-05-03 Richard Biener <rguenther@suse.de>
3327 PR tree-optimization/85615
3328 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3329 to loops not nested in BBs loop father to avoid creating multi-entry
3332 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3334 PR tree-optimization/70291
3335 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3336 arguments. Change return type to tree. Emit libcall as a new
3337 statement rather than replacing existing one when inplace_p is true.
3338 (expand_complex_multiplication_components): New function.
3339 (expand_complex_multiplication): Expand floating-point complex
3340 multiplication using the above.
3341 (expand_complex_division): Rename inner_type parameter to type.
3342 Update expand_complex_libcall call-site.
3343 (expand_complex_operations_1): Update expand_complex_multiplication
3344 and expand_complex_division call-sites.
3346 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3349 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3350 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3351 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3352 the highest significant bit of the shift count mask is clear. In
3353 check whether and[sq]i3 is needed verify that all significant bits
3354 of the shift count other than the highest are set.
3356 2018-05-02 Tom de Vries <tom@codesourcery.com>
3359 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3360 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3361 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3362 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3363 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3364 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3365 __builtin_goacc_parlevel_size.
3367 2018-05-02 Richard Biener <rguenther@suse.de>
3369 PR tree-optimization/85597
3370 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3371 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3374 2018-05-02 Tom de Vries <tom@codesourcery.com>
3377 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3378 dump files): Add ltrans-tree.
3380 2018-05-02 Tom de Vries <tom@codesourcery.com>
3383 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3384 dump files): Add wpa-ipa.
3386 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3388 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3389 powerpc*-*-linux*paired* target.
3390 * config/rs6000/750cl.h: Delete.
3391 * config/rs6000/paired.h: Delete.
3392 * config/rs6000/paired.md: Delete.
3393 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3395 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3396 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3397 comment. Remove paired float support.
3398 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3399 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3401 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3402 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3404 * config/rs6000/rs6000.c: Remove paired float support.
3405 (paired_expand_vector_init, paired_expand_vector_move,
3406 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3407 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3408 paired_expand_builtin, paired_expand_predicate_builtin,
3409 paired_init_builtins): Delete.
3410 * config/rs6000/rs6000.h: Remove paired float support.
3411 * config/rs6000/rs6000.md: Remove paired float support.
3412 (move_from_CR_ov_bit): Delete.
3413 * config/rs6000/rs6000.opt (mpaired): Delete.
3414 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3415 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3417 2018-05-02 Richard Biener <rguenther@suse.de>
3420 * gimplify.c (gimplify_save_expr): When in SSA form allow
3421 SAVE_EXPRs to compute to SSA vars.
3423 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3426 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3427 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3428 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3429 clobber operands[2], instead use a new pseudo. Formatting fixes.
3431 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3433 PR tree-optimization/85586
3434 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3435 exit early for statements in the same group if the accesses are
3438 2018-05-02 Tom de Vries <tom@codesourcery.com>
3441 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3444 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3446 PR tree-optimization/85143
3447 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3449 2018-05-01 Tom de Vries <tom@codesourcery.com>
3452 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3453 not found" error message.
3455 2018-05-01 Tom de Vries <tom@codesourcery.com>
3458 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3459 * vec.c (test_ordered_remove_if): New function.
3460 (vec_c_tests): Call test_ordered_remove_if.
3461 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3462 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3463 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3464 VEC_ORDERED_REMOVE_IF.
3466 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3468 PR tree-optimization/82665
3469 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3470 pointer subtraction where arguments come from a memchr call.
3472 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3474 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3475 --push-state --as-needed and --pop-state instead of --as-needed and
3476 --no-as-needed if ld supports it.
3477 * configure: Regenerated.
3480 * doc/install.texi2html: Replace _002d with - and _002a with * in
3481 generated html files using sed.
3483 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3486 * gcc-rich-location.c (blank_line_before_p): New function.
3487 (use_new_line): New function.
3488 (gcc_rich_location::add_fixit_insert_formatted): New function.
3489 * gcc-rich-location.h
3490 (gcc_rich_location::add_fixit_insert_formatted): New function.
3492 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3494 * selftest.c (assert_streq): Rename "expected" and "actual" to
3495 "val1" and "val2". Extend NULL-handling to cover both inputs
3496 symmetrically, while still requiring both to be non-NULL for a pass.
3497 * selftest.h (assert_streq): Rename "expected" and "actual" to
3499 (ASSERT_EQ): Likewise.
3500 (ASSERT_EQ_AT): Likewise.
3501 (ASSERT_KNOWN_EQ): Likewise.
3502 (ASSERT_KNOWN_EQ_AT): Likewise.
3503 (ASSERT_NE): Likewise.
3504 (ASSERT_MAYBE_NE): Likewise.
3505 (ASSERT_MAYBE_NE_AT): Likewise.
3506 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3507 the assertion to pass.
3508 (ASSERT_STREQ_AT): Likewise.
3510 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3512 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3513 interaction with -pie.
3515 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3517 * selftest.h: Fix alphabetization of per-source-file selftest
3520 2018-04-30 Jason Merrill <jason@redhat.com>
3522 PR c++/61982 - dead stores to destroyed objects.
3523 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3526 2018-04-30 Jason Merrill <jason@redhat.com>
3528 * tree.c (build_clobber): New.
3529 * tree.h: Declare it.
3530 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3532 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3534 * diagnostic-show-locus.c (layout::layout): Update for
3535 location_get_source_line returning a char_span.
3536 (struct char_span): Move to input.h.
3537 (struct correction): Update for fields in char_span becoming
3539 (struct source_line): Update for location_get_source_line
3540 returning a char_span.
3541 (layout::print_line): Likewise.
3542 * edit-context.c (edited_file::print_content): Likewise.
3543 (edited_file::print_diff_hunk): Likewise.
3544 (edited_file::print_run_of_changed_lines): Likewise.
3545 (edited_file::get_num_lines): Likewise.
3546 (edited_line::edited_line): Likewise.
3547 * final.c (asm_show_source): Likewise.
3548 * input.c (location_get_source_line): Convert return type
3549 from const char * to char_span, losing the final "line_len"
3551 (dump_location_info): Update for the above.
3552 (get_substring_ranges_for_loc): Likewise. Use a char_span
3553 when handling the literal within the line.
3554 (test_reading_source_line): Update for location_get_source_line
3555 returning a char_span.
3556 * input.h (class char_span): Move here from
3557 diagnostic-show-locus.c, converting from a struct to a class.
3558 Make data members private.
3559 (char_span::operator bool): New.
3560 (char_span::length): New.
3561 (char_span::get_buffer): New.
3562 (char_span::operator[]): New.
3563 (char_span::subspan): Make const.
3564 (char_span::xstrdup): New.
3565 (location_get_source_line): Convert return type from const char *
3566 to char_span, losing the final "line_size" param.
3568 2018-04-30 Jan Hubicka <jh@suse.cz>
3570 * lto-wrapper.c (ltrans_priorities): New static var.
3571 (cmp_priority): New.
3572 (run_gcc): Read priorities and if doing parallel build order
3573 the Makefile by them.
3575 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3577 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3579 2018-04-30 Richard Biener <rguenther@suse.de>
3581 * tree-cfg.c (verify_address): Remove base argument, add
3582 flag whether to check TREE_ADDRESSABLE and do that.
3583 (verify_expr): Remove.
3584 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3585 (verify_gimple_assign_single): Likewise.
3586 (verify_gimple_switch): Likewise.
3587 (verify_expr_location_1): Dereference tp once. Add (disabled)
3588 piece from verify_expr.
3589 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3591 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3593 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3595 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3597 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3598 (small_data_pattern): Likewise.
3599 (arc_rewrite_small_data): Likewise.
3600 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3601 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3602 (get_symbol_alignment): New function.
3603 (legitimate_small_data_address_p): Likewise.
3604 (legitimate_scaled_address): Update, call
3605 legitimate_small_data_address_p.
3606 (output_sdata): New static variable.
3607 (arc_print_operand): Update how we handle small data operands.
3608 (arc_print_operand_address): Likewise.
3609 (arc_legitimate_address_p): Update, use
3610 legitimate_small_data_address_p.
3611 (arc_rewrite_small_data_p): Remove.
3612 (arc_rewrite_small_data_1): Likewise.
3613 (arc_rewrite_small_data): Likewise.
3614 (small_data_pattern): Likewise.
3615 (compact_sda_memory_operand): Update to use
3616 legitimate_small_data_address_p and get_symbol_alignment.
3617 (prepare_move_operands): Don't rewite sdata pattern.
3618 (prepare_extend_operands): Remove.
3619 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3621 (zero_extendqisi2): Likewise.
3622 (zero_extendhisi2): Likewise.
3623 (extendqihi2): Likewise.
3624 (extendqisi2): Likewise.
3625 (extendhisi2): Likewise.
3629 * config/arc/constraints.md (Usd): Change it to memory constraint.
3631 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3633 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3634 as source of std instructions.
3635 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3636 allow 6-bit constants as source for store instructions.
3637 (movdi_insn): Update instruction pattern to allow 6-bit constants
3638 as source for store instructions.
3640 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3642 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3644 2018-04-30 Nathan Sidwell <nathan@acm.org>
3645 Sandra Loosemore <sandra@codesourcery.com>
3647 * dumpfile.c (dump_open): Allow '-' for stdout.
3648 * doc/invoke.texi (Developer Options): Document dump filename
3649 determination early. Document stdin/stdout selection.
3651 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3653 Microblaze Target: PIC data text relative
3655 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3656 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3658 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3659 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3660 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3661 New addressing mode for data-text relative position indepenedent code.
3662 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3663 'ADDRESS_SYMBOLIC_TXT_REL'.
3664 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3665 (microblaze_legitimate_pic_operand): Exclude function calls from
3666 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3667 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3669 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3670 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3671 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3672 for 'address + offset'.
3673 (microblaze_expand_prologue): Add new function prologue call for
3675 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3676 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3677 table in case of TARGET_PIC_DATA_TEXT_REL.
3678 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3679 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3680 Add new macros 'UNSPEC_TEXT',
3681 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3682 + exclude function calls from 'UNSPEC_PLT' in case of data text
3684 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3685 new target hook for generating address diff vector tables in case of
3687 * doc/tm.texi : Regenerate.
3688 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3689 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3690 of addr diff vector generation.
3691 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3692 target hook definition.
3693 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3694 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3695 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3696 option for data text relative.
3698 2018-04-30 Richard Biener <rguenther@suse.de>
3700 * tree-chrec.h (evolution_function_is_constant_p): Remove
3702 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3704 2018-04-30 Richard Biener <rguenther@suse.de>
3707 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3709 2018-04-30 Richard Biener <rguenther@suse.de>
3711 PR tree-optimization/28364
3712 PR tree-optimization/85275
3713 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3714 copying first exit test.
3716 2018-04-28 Mark Wielaard <mark@klomp.org>
3718 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3720 (dwarf_AT): Handle DW_AT_addr_base.
3721 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3723 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3726 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3727 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3728 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3729 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3731 2018-04-28 Richard Biener <rguenther@suse.de>
3733 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3734 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3735 to reflect use. Only add interesting stmts.
3737 2018-04-27 Martin Jambor <mjambor@suse.cz>
3740 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3741 the jump function allows for passing through aggregate values.
3743 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3745 * input.h (in_system_header_at): Convert from macro to inline
3747 (from_macro_expansion_at): Likewise.
3748 (from_macro_definition_at): Likewise.
3750 2018-04-27 Jeff Law <law@redhat.com>
3752 * config.gcc: Mark tile* targets as deprecated/obsolete.
3754 2018-04-27 Richard Biener <rguenther@suse.de>
3756 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3759 2018-04-27 Richard Biener <rguenther@suse.de>
3761 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3763 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3765 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3766 with Yd constraint. Set "preferred_for_speed" attribute from
3767 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3769 (*movdi_internal): Ditto.
3770 (movti_interunit splitters): Remove
3771 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3772 (movdi_interunit splitters): Ditto.
3773 * config/i386/constraints.md (Ye): Remove.
3774 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3776 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3779 * config/aarch64/constraints.md (Usg): Limit to 31.
3782 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3784 PR tree-optimization/85529
3785 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3786 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3787 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3788 zero extension or masking of the MSB bit.
3789 (optimize_range_tests): Add FIRST_BB argument, pass it through
3790 to optimize_range_tests_var_bound.
3791 (maybe_optimize_range_tests, reassociate_bb): Adjust
3792 optimize_range_tests callers.
3794 2018-04-26 Richard Biener <rguenther@suse.de>
3795 Jakub Jelinek <jakub@redhat.com>
3797 * cgraph.h (symbol_table): Just declare debug method here.
3798 * symtab.c (symbol_table::debug): Define.
3800 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3802 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3804 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3806 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3807 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3808 (*movdi_internal): Substitute Yi and Yj constraint with x
3809 and Ym and Yn constraint with y constraint. Update "isa"
3810 attribute and set "preferred_for_speed" attribute from
3811 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3812 (*movsi_internal): Ditto.
3813 (*movdf_internal): Ditto.
3814 (*movsf_internal): Ditto.
3815 (*zero_extendsidi2): Ditto.
3816 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3817 (sse2_loadld): Ditto.
3818 (*vec_extract<ssevecmodelower>_0): Ditto.
3819 (*vec_extractv4si_0_zext_sse4): Ditto.
3820 (vec_concatv2di): Ditto.
3821 (*vec_dup<mode>): Ditto.
3822 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3823 * config/i386/constraints.md (Yi): Remove.
3828 2018-04-26 Nathan Sidwell <nathan@acm.org>
3830 * dumpfile.c (dump_open): New.
3831 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3832 (dump_finish): Detect stdio/stderr by value not name.
3834 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3836 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3838 2018-04-26 Tom de Vries <tom@codesourcery.com>
3841 * config/nvptx/nvptx.c (verify_neutering_jumps)
3842 (verify_neutering_labels): New function
3843 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3845 2018-04-26 Tom de Vries <tom@codesourcery.com>
3848 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3849 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3852 2018-04-26 Richard Biener <rguenther@suse.de>
3853 Tom de Vries <tom@codesourcery.com>
3856 * lto-streamer-out.c (output_function): Fixup loops if required to match
3857 discovery done in the reader.
3859 2018-04-26 Richard Biener <rguenther@suse.de>
3861 PR tree-optimization/85116
3862 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3863 have a loop exit from the single latch predecessor. Remove
3864 case of header with just condition.
3865 (ch_base::copy_headers): Exclude infinite loops from any
3867 (pass_ch::execute): Record exits.
3869 2018-04-26 Richard Biener <rguenther@suse.de>
3871 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3872 prologue cost vector and pass it to vect_get_load_cost.
3873 (vect_get_peeling_costs_all_drs): Likewise.
3874 (vect_peeling_hash_get_lowest_cost): Likewise.
3875 (vect_enhance_data_refs_alignment): Likewise.
3877 2018-04-26 Richard Biener <rguenther@suse.de>
3880 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3881 checking of integer<->pointer conversions.
3882 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3883 sign-/zero-extending pointer types.
3884 (expand_omp_for_static_chunk): Likewise.
3886 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3887 Jean Lee <xiaoyur347@gmail.com>
3889 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3890 (TARGET_ASAN_SHADOW_OFFSET): Define.
3891 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3892 true for -fsanitize=address.
3894 2018-04-25 Mark Wielaard <mark@klomp.org>
3896 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3899 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3901 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3902 than "alu", remove explicit "memory" and "imm_disp" attributes.
3903 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3906 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3907 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3908 gen_lowpart_no_emit.
3910 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3913 * config/i386/i386.c (ix86_expand_builtin): Change memory
3914 operand to XI, extend p0 to Pmode.
3915 * config/i386/i386.md: Change unspec volatile and operand
3916 1 mode to XI, change operand 0 mode to P.
3918 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3920 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3921 GET_MODE_MASK before any checking.
3922 (nds32_can_use_bset_p): Likewise.
3923 (nds32_can_use_btgl_p): Likewise.
3925 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3927 * config/nds32/nds32-doubleword.md: New define_split pattern for
3928 illegal register number.
3930 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3934 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3936 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3938 2018-04-25 Richard Biener <rguenther@suse.de>
3940 * lto-streamer.h (LTO_major_version): Bump to 8.
3942 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3944 * BASE-VER: Set to 9.0.0.
3946 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3948 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3949 in __abskf2 and __powikf2.
3951 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3954 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3955 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3956 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3957 Use the above on operand 2. Reindent.
3958 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3960 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3963 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3965 * config/i386/i386.opt (mcet): Removed.
3966 * doc/install.texi: Remove -mcet documentation.
3967 * doc/invoke.texi: Likewise.
3969 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3972 * doc/install.texi: Remove -mcet from bootstrap-cet.
3974 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3977 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3978 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3982 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3983 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3984 containing a CONST_VECTOR.
3986 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3988 * doc/install.texi: Update newlib dependency for nvptx.
3990 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3993 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3994 instead of INTVAL when shifting x left.
3996 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3998 PR tree-optimization/85478
3999 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4000 vect_grouped_store_supported for single element vectors.
4002 2018-04-24 Richard Biener <rguenther@suse.de>
4005 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4006 load cost increase to the case of non-constant step.
4008 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4011 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4012 destination if any_malformed_asm.
4014 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4017 * expr.c (store_field): In the bitfield case, if the value comes from
4018 a function call and is returned in registers by means of a PARALLEL,
4019 do not change the mode of the temporary unless BLKmode and VOIDmode.
4021 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4023 PR rtl-optimization/85423
4024 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4025 dependencies to debug insns when the previous insn is non-debug.
4027 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4029 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4030 enums into a single definition.
4031 (fls): Fix predicates and printing.
4034 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4036 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4037 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4038 and short u6 immediate.
4039 (check_if_valid_sleep_operand): Remove.
4040 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4042 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4044 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4045 flag_always_save_lp condition.
4046 * config/nds32/nds32.opt (malways-save-lp): New option.
4048 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4050 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4051 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4052 * config/nds32/nds32.h
4053 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4054 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4056 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4058 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4059 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4061 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4062 Chung-Ju Wu <jasonwucj@gmail.com>
4064 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4066 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4067 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4069 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4071 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4073 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4075 * config/nds32/nds32-protos.h (nds32_data_alignment,
4076 nds32_local_alignment): Declare.
4077 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4078 nds32_local_alignment): New functions.
4079 (TARGET_CONSTANT_ALIGNMENT): Define.
4080 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4082 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4084 * config/nds32/nds32.c
4085 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4086 (TARGET_MODES_TIEABLE_P): Likewise.
4088 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4090 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4093 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4094 Chung-Ju Wu <jasonwucj@gmail.com>
4096 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4097 for unaligned access.
4098 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4100 * config/nds32/nds32-intrinsic.md: Likewise.
4101 * config/nds32/nds32_intrinsic.h: Likewise.
4102 * config/nds32/nds32.h (nds32_builtins): Likewise.
4103 * config/nds32/nds32.opt (munaligned-access): New option.
4104 * config/nds32/nds32.c (nds32_asm_file_start): Display
4105 flag_unaligned_access status.
4107 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4109 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4110 -mno-relax is present.
4111 * config/riscv/linux.h (LINK_SPEC): Ditto.
4113 2018-04-20 Martin Sebor <msebor@redhat.com>
4116 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4118 (gimple_fold_builtin_stxcpy_chk): Same.
4119 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4121 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4124 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4125 __powikf2 when long double is IEEE 128-bit.
4127 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4129 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4130 step to make sure stack always aligned.
4132 2018-04-20 Carl Love <cel@us.ibm.com>
4135 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4136 size check for arg0.
4138 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4139 Tom de Vries <tom@codesourcery.com>
4142 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4143 Emit insns for calls too.
4144 (nvptx_find_par): Always look for worker-level predecessor insn.
4145 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4147 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4148 (nvptx_process_pars): Propagate frames for calls.
4150 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4153 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4155 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4156 (ix86_handle_option): Don't handle OPT_mibt.
4157 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4159 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4161 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4162 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4163 (ix86_target_macros): Define __CET__ with flag_cf_protection
4164 for -fcf-protection.
4165 * config/i386/i386.c (isa2_opts): Remove -mibt.
4166 * config/i386/i386.h (TARGET_IBT): Removed.
4167 (TARGET_IBT_P): Likewise.
4168 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4169 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4170 * config/i386/i386.opt (mcet): Update help message.
4173 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4174 -mcet as an alias for -mshstk.
4176 2018-04-20 Richard Biener <rguenther@suse.de>
4179 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4180 complexity by forcing a single use of the multiply operand.
4182 2018-04-20 Martin Jambor <mjambor@suse.cz>
4185 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4186 recursion dependency to only apply to non-clones.
4188 2018-04-20 Martin Jambor <mjambor@suse.cz>
4191 * ipa-cp.c (create_specialized_node): Check that clones of
4192 self-recursive edges exist during IPA-CP.
4194 2018-04-19 Toon Moene <toon@moene.org>
4196 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4199 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4201 PR tree-optimization/85467
4202 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4203 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4204 VECTOR_CST element to type.
4206 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4209 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4210 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4211 (builtin_longjmp): Likewise.
4212 (save_stack_nonlocal): New pattern.
4213 (restore_stack_nonlocal): Likewise.
4215 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4218 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4219 Replace ASM_OUTPUT_LABEL with fprintf.
4221 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4224 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4225 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4226 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4227 define __IBT__ and __SHSTK__ for -fcf-protection.
4228 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4230 (ix86_trampoline_init): Likewise.
4231 (x86_output_mi_thunk): Likewise.
4232 (ix86_notrack_prefixed_insn_p): Likewise.
4233 (ix86_option_override_internal): Don't disallow -fcf-protection.
4234 * config/i386/i386.md (rdssp<mode>): Also enable for
4236 (incssp<mode>): Likewise.
4237 (nop_endbr): Likewise.
4238 * config/i386/i386.opt (mcet): Change help message to built-in
4242 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4243 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4244 enable CET built-in functions.
4246 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4248 * common/config/i386/i386-common.c
4249 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4250 OPTION_MASK_ISA_MOVDIRI_UNSET,
4251 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4252 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4253 * config.gcc (movdirintrin.h): New header.
4254 * config/i386/cpuid.h (bit_MOVDIRI,
4255 bit_MOVDIR64B): New bits.
4256 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4258 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4259 (VOID, PVOID, PCVOID)): New function types.
4260 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4261 __builtin_ia32_directstoreu_u64,
4262 __builtin_ia32_movdir64b): New builtins.
4263 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4264 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4266 (ix86_valid_target_attribute_inner_p): Ditto.
4267 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4268 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4269 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4270 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4271 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4272 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4273 (movdiri<mode>, movdir64b_<mode>): New.
4274 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4275 * config/i386/immintrin.h: Include movdirintrin.h.
4276 * config/i386/movdirintrin.h: New file.
4277 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4279 2018-04-19 Richard Biener <rguenther@suse.de>
4282 * cfg.c (clear_bb_flags): When loop state says we have
4283 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4285 2018-04-19 Richard Biener <rguenther@suse.de>
4287 PR tree-optimization/84737
4288 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4289 copying restrict info.
4290 (vect_setup_realignment): Use it.
4291 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4292 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4293 the first DR to all generated stores.
4294 (vectorizable_load): Likewise for loads.
4296 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4298 PR tree-optimization/85446
4299 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4300 the integral and pointer types to have the same precision.
4302 * doc/install.texi: Document --disable-cet being the default and
4305 2018-04-18 Martin Liska <mliska@suse.cz>
4307 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4310 2018-04-18 Martin Liska <mliska@suse.cz>
4313 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4316 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4317 arguments if they are comparable.
4319 2018-04-18 Martin Liska <mliska@suse.cz>
4322 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4325 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4328 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4331 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4332 ENDBR after calling __morestack.
4334 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4337 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4338 by using gcc_base_ver to generate a gcc_driver_version, and use
4339 it when generating GCC_DRIVER_NAME.
4340 * configure: Regenerate.
4342 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4345 * config.gcc: Obsolete powerpc*-*-*spe*.
4347 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4350 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4351 (stabstr_D): Change type of unum from unsigned int to
4352 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4355 2018-04-17 Jim Wilson <jimw@sifive.com>
4358 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4359 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4360 Set arg_pointer_offset after using pretend_args_size.
4362 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4364 PR rtl-optimization/85431
4365 * dse.c (record_store): Ignore zero width stores.
4368 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4369 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4370 __builtin_stack_restore rather than after it.
4371 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4372 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4373 argument instead of virtual_dynamic_stack_rtx.
4375 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4377 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4379 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4380 Add note to error message to explain internal mapping of overloaded
4381 built-in function name to non-overloaded built-in function name.
4382 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4385 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4388 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4389 where the inputs overlap with the output.
4391 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4393 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4394 (=v, v) alternative and explicit "memory" attribute.
4395 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4396 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4398 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4399 "sselog1" type instead of "sselog".
4400 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4401 "sselog". Remove explicit "memory" attribute.
4402 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4403 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4405 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4406 "sselog1" type instead of "sselog". Remove explicit "memory"
4408 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4409 use "sselog1" type instead of "sselog". Remove explicit "memory"
4411 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4412 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4414 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4415 "sselog1" type instead of "sselog". Remove explicit "memory"
4417 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4418 use "sselog1" type instead of "sselog". Remove explicit "memory"
4422 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4425 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4428 2018-04-17 Martin Jambor <mjambor@suse.cz>
4431 * ipa-cp.c (create_specialized_node): Call
4432 expand_all_artificial_thunks if necessary.
4434 2018-04-17 Martin Liska <mliska@suse.cz>
4437 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4438 in message, remote space in between '_G' and '('.
4440 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4443 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4444 avx512f_vmcmp<mode>3<round_saeonly_name>,
4445 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4446 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4447 avx512f_rndscale<mode><round_saeonly_name>,
4448 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4449 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4450 Use %<iptr>2 instead of %2 for -masm=intel.
4451 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4452 avx512f_vcvttss2usi<round_saeonly_name>,
4453 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4455 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4456 avx512f_vcvttsd2usi<round_saeonly_name>,
4457 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4458 Use %q1 instead of %1 for -masm=intel.
4459 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4460 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4461 of %3 for -masm=intel.
4462 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4464 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4466 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4468 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4469 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4471 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4473 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4474 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4475 %g1 and one with %0 and %1.
4476 (avx512er_vmrcp28<mode><round_saeonly_name>,
4477 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4479 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4480 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4481 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4482 of %0 and %{%4%} for -masm=intel.
4483 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4484 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4485 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4486 order of %0 and %{%5%}%{z%} for -masm=intel.
4488 2018-04-17 Jan Hubicka <jh@suse.cz>
4491 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4493 2018-04-17 Martin Liska <mliska@suse.cz>
4496 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4497 for target_clone error message. Make default implementation
4498 clone to be a local declaration.
4499 (separate_attrs): Add new argument and check for an empty
4501 (expand_target_clones): Handle it.
4502 (ipa_target_clone): Make redirection just for target_clones
4505 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4506 Tom de Vries <tom@codesourcery.com>
4509 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4510 tiled basic blocks without omp continue statements.
4512 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4515 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4516 vec_extract expression as having side effects to make sure it gets
4519 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4522 * config/i386/i386.c (get_builtin_code_for_version): Check
4525 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4528 * gcc/config.gcc: Support "skylake".
4529 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4531 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4532 (processor_target_table): Add "skylake".
4533 (ix86_option_override_internal): Add "skylake".
4534 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4535 PROCESSOR_CANNONLAKE.
4536 (get_builtin_code_for_version): Fix priority for
4537 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4538 PROCESSOR_SKYLAKE-AVX512.
4539 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4540 (processor_type): Add PROCESSOR_SKYLAKE.
4542 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4543 Jason Merrill <jason@redhat.com>
4546 * convert.c (convert_to_integer_1): Use direct recursion for
4547 enumeral types and types with a precision less than the number
4548 of bits in their mode.
4550 2018-04-16 Julia Koval <julia.koval@intel.com>
4553 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4554 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4556 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4559 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4560 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4561 and -mno-direct-move.
4563 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4566 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4567 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4568 Ensure negative shifts result in {0}.
4570 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4572 PR rtl-optimization/79916
4573 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4574 regs (if any) to define how to gnerate SD moves when LRA is in
4577 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4579 PR rtl-optimization/85393
4580 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4581 * except.c (expand_dw2_landing_pad_for_region): Make static.
4582 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4583 a label and unconditional jump to old_bb, rather than
4584 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4587 PR rtl-optimization/85376
4588 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4589 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4590 instead of a specific value.
4592 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4593 Bin Cheng <bin.cheng@arm.com>
4595 PR tree-optimization/82965
4596 PR tree-optimization/83991
4597 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4598 by_profile_only parameter.
4599 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4600 information if the loop was predicted to iterate too many times.
4601 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4603 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4606 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4609 2018-04-13 Martin Liska <mliska@suse.cz>
4610 Jakub Jelinek <jakub@redhat.com>
4613 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4614 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4615 * builtins.c (expand_builtin_memory_copy_args): Use
4616 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4617 handle dest_addr == pc_rtx.
4619 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4622 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4623 asked to not generate direct moves.
4624 (fix_trunc<mode>si2_stfiwx): Similar.
4625 (fix_trunc<mode>si2_internal): Similar.
4627 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4630 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4631 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4632 lookup if dest in some wider mode is known to be const0_rtx and
4633 if so, record permanent equivalence for it to be ZERO_EXTEND of
4634 the narrower mode destination.
4636 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4638 * lto-streamer-out.c (output_function): Revert 259346.
4639 * omp-expand.c (expand_oacc_for): Likewise.
4641 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4643 PR rtl-optimization/85354
4644 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4645 * sel-sched.c (sel_global_init): ... here.
4647 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4650 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4651 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4652 mode for PE-COFF targets.
4653 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4654 (i386_pe_asm_lto_end): Likewise.
4655 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4656 (TARGET_ASM_LTO_END): Likewise.
4657 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4658 (i386_pe_asm_lto_start): New function.
4659 (i386_pe_asm_lto_end): Likewise.
4661 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4662 Richard Biener <rguenther@suse.de>
4665 * lto-streamer-out.c (output_function): Fix CFG loop state before
4667 * omp-expand.c (expand_oacc_for): Handle calls to internal
4668 functions like regular functions.
4670 2018-04-12 Richard Biener <rguenther@suse.de>
4673 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4674 for the early LTO debug to properly generate references to it
4675 during DIE emission. Do not re-use that for the skeleton for
4677 (dwarf2out_early_finish): Likewise.
4679 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4682 * config/i386/sse.md
4683 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4684 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4685 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4686 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4687 and output is a reg, avoid creating invalid lowpart subreg, but
4688 instead split into a 512-bit move. Don't split if not AVX512VL,
4689 input is xmm16+ reg and output is a mem.
4690 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4691 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4692 xmm16+ reg and output is a mem.
4694 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4696 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4697 also for flag_dwarf2_cfi_asm.
4699 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4701 PR rtl-optimization/85342
4702 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4703 a bool scalar var inside of the loop instead. Don't try to update
4704 recog_data.operand after failed apply_change_group.
4706 2018-04-12 Tom de Vries <tom@codesourcery.com>
4709 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4710 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4711 array with flexible array member as array without given dimension.
4712 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4713 argument for undefined param to true.
4715 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4718 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4719 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4720 from PowerPC section.
4721 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4722 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4723 help text that is too long.
4724 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4725 help text that is too long.
4726 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4727 help text that is too long.
4729 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4731 * config/alpha/alpha.md (stack_probe_internal): Rename
4732 from "probe_stack". Update all callers.
4734 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4736 PR rtl-optimization/84566
4737 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4738 sched_macro_fuse_insns.
4740 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4743 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4744 (compute_block_dependences): ... from here.
4746 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4748 PR tree-optimization/85331
4749 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4750 from int to HOST_WIDE_INT.
4752 2018-04-11 Martin Jambor <mjambor@suse.cz>
4755 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4756 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4757 not the same as the source val.
4758 (cgraph_edge_brings_value_p): New parameter.
4759 (gather_edges_for_value): Pass destination value to
4760 cgraph_edge_brings_value_p.
4761 (perhaps_add_new_callers): Likewise.
4762 (get_info_about_necessary_edges): Likewise and exclude values brought
4763 only by self-recursive edges.
4764 (create_specialized_node): Redirect only clones of self-calling edges.
4765 (+self_recursive_pass_through_p): New function.
4766 (find_more_scalar_values_for_callers_subset): Use it.
4767 (find_aggregate_values_for_callers_subset): Likewise.
4768 (known_aggs_to_agg_replacement_list): Removed.
4769 (decide_whether_version_node): Re-calculate known constants for all
4770 remaining context clones.
4772 2018-04-11 Richard Biener <rguenther@suse.de>
4775 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4776 from early DWARF output.
4777 (dwarf2out_early_finish): Output line info unconditionally into
4778 early DWARF and add reference to it.
4780 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4783 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4784 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4785 other than V2DFmode using iptr mode attribute.
4786 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4788 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4790 PR rtl-optimization/84659
4791 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4793 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4796 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4798 (output_loc_list): Pass address of a dummy size variable even in the
4799 locview handling loop.
4800 (index_location_lists): Add comment on why skip_loc_list_entry can't
4803 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4806 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4809 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4812 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4813 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4814 and -mstring-compare-inline-limit.
4816 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4819 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4820 for stack clash protection in a register whenever we need it to be in
4823 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4825 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4826 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4828 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4831 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4833 (mlong-double-): Ditto.
4834 * config/rs6000/sysv4.opt (msdata=): Ditto.
4835 (mtls-size=): Ditto.
4837 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4839 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4840 erroneous entries for
4841 "vector int vec_ldl (int, long int *)", and
4842 "vector unsigned int vec_ldl (int, unsigned long int *)".
4843 Add comments and entries for
4844 "vector bool char vec_ldl (int, bool char *)",
4845 "vector bool short vec_ldl (int, bool short *)",
4846 "vector bool int vec_ldl (int, bool int *)",
4847 "vector bool long long vec_ldl (int, bool long long *)",
4848 "vector pixel vec_ldl (int, pixel *)",
4849 "vector long long vec_ldl (int, long long *)",
4850 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4851 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4852 type tree bool_long_long_type_node and correct definition of
4853 bool_V2DI_type_node to make reference to this new type tree.
4854 (rs6000_mangle_type): Replace erroneous reference to
4855 bool_long_type_node with bool_long_long_type_node.
4856 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4857 comments to emphasize sign distinctions for char and int types and
4858 replace RS6000_BTI_bool_long constant with
4859 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4860 use of RS6000_BTI_pixel.
4861 (bool_long_type_node): Remove this macro definition.
4862 (bool_long_long_type_node): New macro definition
4864 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4866 PR rtl-optimization/85300
4867 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4868 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4869 simplify_unary_operation fails.
4871 2018-04-10 Martin Liska <mliska@suse.cz>
4873 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4874 cgraph_edge and ipa_ref.
4876 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4880 * config/i386/sse.md
4881 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4882 computation of the VEC_MERGE selector from mask.
4883 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4884 Fix decoding of the VEC_MERGE selector into mask.
4886 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4888 PR tree-optimization/85286
4889 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4891 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4893 * final.c (final_1): Set insn_last_address as well as
4894 insn_current_address.
4896 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4899 * explow.c (emit_stack_probe): Call validize_mem on memory location
4900 before passing it to gen_probe_stack. Create address operand and
4901 legitimize it for the probe_stack_address case.
4903 2018-04-09 Jan Hubicka <jh@suse.cz>
4906 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4907 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4908 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4910 (free_lang_data): Rebuild type inheritance graph.
4912 2018-04-09 Martin Sebor <msebor@redhat.com>
4914 * invoke.texi (-finline-small-functions): Mention other optimization
4916 (-findirect-inlining, -fpartial-inlining): Same.
4917 (-finline-functions-called-once): Same.
4918 (-freorder-blocks-and-partition): Same.
4920 2018-04-09 Jan Hubicka <jh@suse.cz>
4923 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4924 jumps; choose last target that matches the criteria (i.e.
4925 no partition changes for non-crossing jumps).
4926 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4927 support for redirecting crossing jumps to non-crossing.
4929 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4931 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4932 also for naked functions.
4934 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4936 * config/arc/arc.md (add_shift): New pattern.
4937 (add_shift2): Likewise.
4938 (sub_shift): Likewise.
4939 (sub_shift_cmp0_noout): Likewise.
4940 (compare_si_ashiftsi): Likewise.
4941 (xbfu_cmp0_noout): New combine pattern.
4942 (xbfu_cmp0"): Likewise.
4943 (movsi_set_cc_insn): Place the predicable variant first.
4944 (commutative_binary_cmp0_noout): Remove clobber.
4945 (commutative_binary_cmp0): New pattern.
4946 (noncommutative_binary_cmp0): Likewise.
4947 (noncommutative_binary_cmp0_noout): Likewise.
4948 (noncommutative_binary_comparison_result_used): Removed.
4949 (rsub_cmp0): New pattern.
4950 (rsub_cmp0_noout): Likewise.
4951 (extzvsi): Changed, keep only meaningful variants.
4952 (SQH, SEZ): New iterators.
4953 (SQH_postfix): New mode attribute.
4954 (SEZ_prefix): New code attribute.
4955 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4956 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4957 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4959 (noncommutative_operator): Check the availability of barrel
4962 2018-04-09 Richard Biener <rguenther@suse.de>
4964 PR tree-optimization/85284
4965 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4966 Only use the niter constraining form of simple_iv when the exit
4969 2018-04-09 Tom de Vries <tom@codesourcery.com>
4972 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4973 (define_expand "*memory_barrier"): New define_expand.
4974 (define_insn "memory_barrier"): New insn.
4976 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4978 PR rtl-optimization/80463
4979 PR rtl-optimization/83972
4980 PR rtl-optimization/83480
4982 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4983 correct producer for the insn.
4984 (tidy_control_flow): Fixup seqnos in case of debug insns.
4986 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4988 PR rtl-optimization/83913
4990 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4991 different sched-times when merging exprs.
4993 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4995 PR rtl-optimization/83962
4997 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4998 tidy_fallthru_edge and tidy_control_flow.
5000 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5002 PR rtl-optimization/83530
5004 * sel-sched.c (force_next_insn): New global variable.
5005 (remove_insn_for_debug): When force_next_insn is true, also leave only
5006 next insn in the ready list.
5007 (sel_sched_region): When the region wasn't scheduled, make another pass
5008 over it with force_next_insn set to 1.
5010 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5012 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5014 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5015 for interrupt control.
5016 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5017 functions for interrupt control.
5018 * config/nds32/nds32-intrinsic.md: Likewise.
5019 * config/nds32/nds32_intrinsic.h: Likewise.
5020 * config/nds32/nds32.h (nds32_builtins): Likewise.
5022 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5024 * config/nds32/nds32.c (nds32_init_machine_status,
5025 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5026 strict_aligned_p field.
5027 (nds32_expand_to_rtl_hook): New function.
5028 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5029 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5031 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5032 Chung-Ju Wu <jasonwucj@gmail.com>
5034 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5035 * config/nds32/nds32-n7.md: New file.
5036 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5037 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5039 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5040 * config/nds32/nds32.md (pipeline_model): Add n7.
5041 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5042 * config/nds32/pipelines.md: Include n7 settings.
5044 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5045 Chung-Ju Wu <jasonwucj@gmail.com>
5047 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5048 * config/nds32/nds32-e8.md: New file.
5049 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5050 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5052 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5053 * config/nds32/nds32.md (pipeline_model): Add e8.
5054 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5055 * config/nds32/pipelines.md: Include e8 settings.
5057 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5058 Chung-Ju Wu <jasonwucj@gmail.com>
5060 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5061 * config/nds32/nds32-n8.md: New file.
5062 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5063 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5065 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5066 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5067 * config/nds32/nds32.md (pipeline_model): Add n8.
5068 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5069 * config/nds32/pipelines.md: Include n8 settings.
5071 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5072 Chung-Ju Wu <jasonwucj@gmail.com>
5074 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5075 * config/nds32/nds32-n9-2r1w.md: New file.
5076 * config/nds32/nds32-n9-3r2w.md: New file.
5077 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5078 nds32_register_ports): New or modify for cpu n9.
5079 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5081 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5082 * config/nds32/nds32-utils.c: New file.
5083 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5084 TARGET_MUL_SLOW): Define.
5085 * config/nds32/nds32.md (pipeline_model): New attribute.
5086 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5087 New options that support cpu n9.
5088 * config/nds32/pipelines.md: Include n9 settings.
5089 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5091 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5093 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5094 information if necessary.
5095 (output_cond_branch_compare_zero): Likewise.
5096 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5097 (nds32_target_alignment): Refine for alignment.
5098 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5099 (FUNCTION_BOUNDARY): Modify.
5100 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5102 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5104 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5106 * config/nds32/constants.md (unspec_volatile_element): Add values for
5107 TLB operation and data prefetch.
5108 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5109 functions for TLB operation and data prefetch.
5110 * config/nds32/nds32-intrinsic.md: Likewise.
5111 * config/nds32/nds32_intrinsic.h: Likewise.
5112 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5113 (nds32_print_operand): Likewise.
5114 * config/nds32/nds32.h (nds32_builtins): Likewise.
5116 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5117 Andrew Pinski <pinsika@gcc.gnu.org>
5120 * match.pd: Use constant_boolean_node of correct type instead of
5121 boolean_true_node or boolean_false_node for simplifying
5122 pointer comparisons to zero.
5124 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5126 PR tree-optimization/80021
5127 * tree.c (verify_type_variant): Make error call in verify_variant_match
5128 translatable and remove final full stop.
5130 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5132 * config/nds32/constants.md (unspec_volatile_element): Add
5133 UNSPEC_VOLATILE_EH_RETURN.
5134 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5135 nds32_output_stack_pop): Support dwarf exception handling process.
5136 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5137 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5138 exception handling process.
5139 (nds32_compute_stack_frame): Likewise.
5140 (nds32_return_addr_rtx): Likewise.
5141 (nds32_initial_elimination_offset): Likewise.
5142 (nds32_expand_prologue): Likewise.
5143 (nds32_expand_epilogue): Likewise.
5144 (nds32_dynamic_chain_address): New function.
5145 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5147 (DYNAMIC_CHAIN_ADDRESS): Define.
5148 (EH_RETURN_DATA_REGNO): Define.
5149 (EH_RETURN_STACKADJ_RTX): Define.
5150 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5151 patterns for dwarf exception handling.
5153 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5155 * config/nds32/nds32.h: Clean up obsolete macros.
5157 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5159 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5160 Add enum values for particular instructions.
5161 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5162 particular intrinsic functions.
5163 * config/nds32/nds32-intrinsic.md: Likewise.
5164 * config/nds32/nds32_intrinsic.h: Likewise.
5165 * config/nds32/nds32.h (nds32_builtins): Likewise.
5166 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5167 (btst, ave): New patterns for particular instructions.
5169 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5171 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5172 Add enum values for atomic load/store and memory sync.
5173 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5175 * config/nds32/nds32-intrinsic.md: Likewise.
5176 * config/nds32/nds32_intrinsic.h: Likewise.
5177 * config/nds32/nds32.h (nds32_builtins): Likewise.
5179 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5181 PR tree-optimization/85257
5182 * fold-const.c (native_encode_vector): If not all elts could fit
5183 and off is -1, return 0 rather than offset.
5184 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5185 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5186 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5187 adjust buffer in native_interpret_expr call.
5189 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5191 * config/nds32/constants.md (unspec_volatile_element): Add cache
5192 control enum values.
5193 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5194 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5195 * config/nds32/nds32.c (nds32_cctl_names): New.
5196 (nds32_print_operand): Handle cache control register names.
5197 * config/nds32/nds32.h (nds32_builtins): New enum values.
5198 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5200 * config/nds32/nds32.md (type): Add mmu.
5201 * config/nds32/pipelines.md (simple_insn): Add mmu.
5203 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5205 * config/nds32/nds32.md (type): Remove call.
5206 * config/nds32/pipelines.md (simple_insn): Likewise.
5208 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5210 * config/nds32/constants.md (unspec_volatile_element): Add
5211 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5212 UNSPEC_VOLATILE_FMFCFG.
5213 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5214 description for fmfcfg and fmfcsr.
5215 (bdesc_1arg): Add fmtcsr.
5216 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5217 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5218 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5219 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5220 unspec_fmfcfg): New patterns.
5221 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5222 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5223 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5224 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5225 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5226 __nds32__fmfcfg): Define.
5228 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5230 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5231 intrinsic register names.
5232 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5233 intrinsic register enum values and macros.
5235 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5237 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5238 for load/store addressing form.
5239 (nds32_print_operand_address): Likewise.
5241 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5244 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5245 based on LABEL_REF. Remove useless assertion.
5246 (pic_address_needs_scratch): Fix formatting.
5247 (sparc_legitimize_pic_address): Minor tweaks.
5248 (sparc_delegitimize_address): Adjust assertion accordingly.
5249 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5250 into symbolic_operand.
5251 (movsi_high_pic_label_ref): Likewise.
5252 (movsi_lo_sum_pic_label_ref): Likewise.
5253 (movdi_pic_label_ref): Likewise.
5254 (movdi_high_pic_label_ref): Likewise.
5255 (movdi_lo_sum_pic_label_ref): Likewise.
5257 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5259 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5260 custom LIB_SPEC setup.
5262 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5263 Kito Cheng <kito.cheng@gmail.com>
5265 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5266 * config/riscv/freebsd.h: New.
5268 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5270 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5271 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5274 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5275 Kito Cheng <kito.cheng@gmail.com>
5277 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5278 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5279 * config/nds32/nds32-protos.h (nds32_output_call,
5280 nds32_output_return): Declare.
5281 * config/nds32/nds32.md: Refine all the call and return patterns.
5283 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5286 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5287 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5289 PR rtl-optimization/84872
5290 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5291 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5294 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5296 * expr.c (copy_blkmode_to_reg): Revert 254862.
5297 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5299 2018-04-06 Richard Biener <rguenther@suse.de>
5302 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5303 after seeing a component reference with an adjacent field. Treat
5304 refs to arrays at struct end of external decls similar to
5305 refs to unconstrained commons.
5307 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5310 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5311 look through SAVE_EXPRs with non-side-effects argument. Adjust
5313 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5316 2018-04-06 Richard Biener <rguenther@suse.de>
5319 * alias.c (find_base_term): New wrapper around find_base_term
5320 unwinding CSELIB_VAL_PTR changes.
5321 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5324 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5326 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5328 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5329 constant definitions.
5330 ("nop"): lr 0,0 -> nopr r0
5331 ("nop_lr0", "nop_lr1"): New insn definitions.
5333 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5335 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5336 NDS32_V3PUSH_AVAILABLE_P macro.
5338 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5339 Chung-Ju Wu <jasonwucj@gmail.com>
5341 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5342 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5343 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5344 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5345 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5346 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5347 * config/nds32/constraints.md: New constraints and checking for hard
5348 float configuration.
5349 * config/nds32/iterators.md: New mode iterator and attribute for hard
5350 float configuration.
5351 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5353 * config/nds32/nds32-fpu.md: New file.
5354 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5355 deal with hard float code generation.
5356 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5358 (abi_type, float_reg_number): New enum type.
5359 * config/nds32/nds32-predicates.c: New predicates for hard float.
5360 * config/nds32/nds32-protos.h: Declare functions for hard float.
5361 * config/nds32/nds32.c: Implementation for hard float configuration.
5362 * config/nds32/nds32.h: Definitions for hard float configuration.
5363 * config/nds32/nds32.md: Include hard float machine description and
5364 modify patterns for hard float configuration.
5365 * config/nds32/nds32.opt: New options for hard float configuration.
5366 * config/nds32/predicates.md: New predicates for hard float
5369 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5371 * common/config/nds32/nds32-common.c
5372 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5374 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5377 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5378 CONSTRUCTOR_ELT (ctor, ...)->value.
5380 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5383 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5385 2018-04-05 Tom de Vries <tom@codesourcery.com>
5388 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5391 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5392 Kito Cheng <kito.cheng@gmail.com>
5394 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5395 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5396 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5397 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5399 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5400 Kito Cheng <kito.cheng@gmail.com>
5402 * config/nds32/constraints.md (Ufe): New memory constraint.
5403 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5404 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5405 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5407 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5408 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5410 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5412 * config/nds32/nds32.md: Use optimize_size in the condition for
5413 alu-shift instructions.
5415 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5417 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5419 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5421 * config/nds32/nds32.md (negsi2): Refine pattern.
5423 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5424 Chung-Ju Wu <jasonwucj@gmail.com>
5426 * config/nds32/iterators.md (shift_rotate): New code iterator.
5427 (shift): New code attribute.
5428 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5429 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5430 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5431 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5432 bit-wise operations.
5433 (andsi3, *andsi3): Ditto.
5434 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5435 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5436 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5437 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5438 nds32_ior_operand, nds32_xor_operand): New predicates.
5440 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5442 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5443 (addsi3, subsi3): ... this.
5445 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5447 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5449 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5451 * config/nds32/nds32.md: Adjust indention.
5453 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5455 * config/nds32/nds32.md (feature): New attribute.
5457 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5459 * config/nds32/nds32.md (subtype): New attribute.
5461 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5464 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5465 expansion to perform a bitwise AND of the argument followed by a
5466 boolean negation of the result.
5468 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5470 PR rtl-optimization/84878
5471 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5472 the basic block. Assert the use reference is not artificial and that
5473 it has an associated insn.
5475 2018-04-04 Michael Matz <matz@suse.de>
5477 * builtins.c (compute_objsize): Pass correct operand
5478 to array_at_struct_end_p.
5480 2018-04-04 Richard Biener <rguenther@suse.de>
5483 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5484 from contexts for DINFO_LEVEL_TERSE and below.
5486 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5488 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5489 resiter_operand condition.
5490 * config/nds32/nds32.md (*move<mode>): Ditto.
5492 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5493 Monk Chiang <sh.chiang04@gmail.com>
5495 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5497 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5499 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5501 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5502 Kito Cheng <kito.cheng@gmail.com>
5504 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5505 nds32_cond_code_str, output_cond_branch,
5506 output_cond_branch_compare_zero, nds32_expand_cbranch,
5507 nds32_expand_cstore, nds32_expand_movcc,
5508 nds32_output_cbranchsi4_equality_zero,
5509 nds32_output_cbranchsi4_equality_reg,
5510 nds32_output_cbranchsi4_equality_reg_or_const_int,
5511 nds32_output_cbranchsi4_greater_less_zero: New functions.
5512 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5513 nds32_expand_cstore, nds32_expand_movcc,
5514 nds32_output_cbranchsi4_equality_zero,
5515 nds32_output_cbranchsi4_equality_reg,
5516 nds32_output_cbranchsi4_equality_reg_or_const_int,
5517 nds32_output_cbranchsi4_greater_less_zero): Declare.
5518 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5519 nds32_rimm11s_operand): New predicates.
5520 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5521 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5524 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5526 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5527 * config/nds32/nds32.md: Ditto.
5528 * config/nds32/pipelines.md: Ditto.
5530 2018-04-04 Richard Biener <rguenther@suse.de>
5532 PR tree-optimization/85168
5533 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5534 propagating abnormals.
5536 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5538 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5540 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5541 Kito Cheng <kito.cheng@gmail.com>
5543 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5544 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5545 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5546 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5547 * config/nds32/nds32.md (sibcall_internal): New.
5548 (sibcall_register): Remove.
5549 (sibcall_immediate): Remove.
5550 (sibcall_value_internal): New.
5551 (sibcall_value_register): Remove.
5552 (sibcall_value_immediate): Remove.
5553 * config/nds32/predicates.md (nds32_general_register_operand): New.
5554 (nds32_call_address_operand): New.
5556 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5558 PR rtl-optimization/85167
5559 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5560 bb_defs if *split_p, instead preinitialize it to NULL.
5562 PR tree-optimization/85156
5563 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5564 evaluating the argument multiple times.
5566 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5568 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5570 (_mm_cvtpd_ps): Likewise.
5571 (_mm_cvttpd_epi32): Likewise.
5572 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5573 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5574 vector, pixel, and bool following altivec.h include.
5576 2018-04-03 Martin Sebor <msebor@redhat.com>
5578 * doc/extend.texi (Common Function Attributes): Clarify.
5579 (const attribute): Likewise.
5580 (pure attribute): Likewise.
5582 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5585 * config/i386/i386.c (ix86_expand_vector_set): Use
5586 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5588 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5590 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5591 instructions when changing rounding bits to preserve precision bits
5592 in the x87 control word.
5594 2018-04-03 Martin Liska <mliska@suse.cz>
5596 PR tree-optimization/82491
5597 * rtl.h (strip_offset_and_add): Replace += suboffset with
5598 poly_uint64 () + suboffset.
5600 2018-03-29 Martin Liska <mliska@suse.cz>
5601 Martin Jambor <mjambor@suse.cz>
5604 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5605 param_type is not an integral or pointer type.
5607 2018-04-03 Richard Biener <rguenther@suse.de>
5609 * sese.h (recompute_all_dominators): Remove.
5611 2018-04-02 Martin Sebor <msebor@redhat.com>
5613 * doc/invoke.texi (-Wrestrict): Fix typos.
5615 2018-04-02 Jim Wilson <jimw@sifive.com>
5617 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5618 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5619 (<optab>di3, <optab>si3_extend): Likewise.
5620 (<optab>si3_mask, <optab>si3_mask_1): New.
5621 (<optab>di3_mask, <optab>di3_mask_1): New.
5622 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5623 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5624 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5626 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5628 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5631 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5633 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5634 (nds32_canonicalize_comparison): New function.
5636 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5637 Kito Cheng <kito.cheng@gmail.com>
5638 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5640 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5641 * config/nds32/constants.md (unspec_volatile_element): Add
5642 UNSPEC_VOLATILE_RELAX_GROUP.
5643 * config/nds32/nds32-relax-opt.c: New file.
5644 * config/nds32/nds32-predicates.c
5645 (nds32_symbol_load_store_p): New function.
5646 * config/nds32/nds32-protos.h
5647 (nds32_symbol_load_store_p): Declare function.
5648 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5649 * config/nds32/nds32.c
5650 (nds32_register_pass): New function to register pass.
5651 (nds32_register_passes): New function to register passes.
5652 * config/nds32/nds32.md (relax_group): New pattern.
5653 * config/nds32/nds32.opt (mrelax-hint): New option.
5654 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5656 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5658 * config/nds32/t-nds32: Modify files dependency.
5660 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5662 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5663 (PROFILE_HOOK): Define its implementation.
5665 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5667 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5668 type and 32-bit size.
5670 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5673 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5674 (V_128_256): New mode iterator.
5675 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5676 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5677 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5679 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5680 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5682 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5685 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5686 NaN inputs correctly.
5688 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5691 * config/rs6000/vsx.md (??r): New mode attribute.
5692 (*vsx_mov<mode>_64bit): Use it.
5693 (*vsx_mov<mode>_32bit): Likewise.
5695 2018-03-30 Martin Sebor <msebor@redhat.com>
5697 PR tree-optimization/84818
5698 * builtins.c (check_access): Use warning_n.
5700 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5703 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5705 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5708 2018-03-30 Julia Koval <julia.koval@intel.com>
5711 * x86-tune.def (movx, partial_reg_dependency): Enable for
5714 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5717 * lra-constraints.c (process_alt_operands): Move setting
5718 this_alternative_matches below.
5720 2018-03-29 Martin Liska <mliska@suse.cz>
5723 * doc/invoke.texi: Document how LTO works with debug info.
5724 Describe auto-load support of binutils. Mention 'x86-64'
5725 as valid option value of -march option.
5727 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5729 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5732 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5733 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5734 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5737 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5740 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5741 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5742 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5743 for RS6000_BTM_POWERPC64.
5744 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5745 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5746 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5751 2018-03-28 Carl Love <cel@us.ibm.com>
5754 2017-09-27 Carl Love <cel@us.ibm.com>
5756 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5757 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5758 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5761 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5763 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5764 instead of __vector bool.
5765 (_mm_max_pu8): Likewise.
5766 (_mm_min_pi16): Likewise.
5768 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5771 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5772 (DIVWEUO): Likewise.
5774 (DIVDEUO): Likewise.
5775 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5776 DIVWEUO and DIVDEUO.
5777 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5778 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5779 (div_extend): Likewise.
5780 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5782 (__builtin_divweuo): Likewise.
5783 (__builtin_divdeo): Likewise.
5784 (__builtin_divdeuo): Likewise.
5786 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5789 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5790 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5792 PR tree-optimization/82004
5793 * gimple-match-head.c (optimize_pow_to_exp): New function.
5794 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5795 Don't fold to exp if optimize_pow_to_exp is false.
5797 2018-03-28 Martin Liska <mliska@suse.cz>
5800 * calls.c (initialize_argument_information): Fix trailing space.
5801 * common.opt: Fix typo and provide better explanation for
5802 -fsanitize-coverage option.
5803 * config/i386/i386.opt: Fix typo.
5805 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5806 Martin Liska <mliska@suse.cz>
5809 * gimplify.c (asan_poison_variable): Don't do the check for
5810 gimplify_omp_ctxp here.
5811 (gimplify_decl_expr): Do it here.
5812 (gimplify_target_expr): Likewise.
5814 2018-03-28 Martin Liska <mliska@suse.cz>
5817 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5818 chkp_type_bounds_count if MPX is not enabled.
5820 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5822 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5824 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5827 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5828 function to create the function decl for complex long double
5829 multiply and divide for -mabi=ieeelongdouble.
5830 (init_float128_ieee): Call it.
5832 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5835 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5836 -fcf-protection=branch -mibt.
5837 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5839 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5842 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5843 arm_word_relocations.
5845 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5848 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5849 extern array declarations.
5851 2018-03-27 Richard Biener <rguenther@suse.de>
5854 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5855 explicit single_use checks.
5857 2018-03-27 Richard Biener <rguenther@suse.de>
5859 PR tree-optimization/85082
5860 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5863 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5865 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5866 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5867 Turn on fasynchronous-unwind-tables and funwind-tables.
5869 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5872 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5873 (*bmi_blsr_<mode>_ccz): Ditto.
5875 2018-03-26 Tom de Vries <tom@codesourcery.com>
5877 PR tree-optimization/85063
5878 * omp-general.c (offloading_function_p): New function. Factor out
5880 * omp-offload.c (pass_omp_target_link::gate): ... here.
5881 * omp-general.h (offloading_function_p): Declare.
5882 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5883 with attribute omp declare target for offloading functions.
5885 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5887 PR tree-optimization/84005
5888 * tree-data-ref.h (get_base_for_alignment): Declare.
5889 * tree-data-ref.c (get_base_for_alignment_1): New function.
5890 (get_base_for_alignment): Likewise.
5891 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5892 get_base_for_alignment to find a suitable base object, instead
5893 of always using drb->base_address.
5895 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5898 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5899 known size by default.
5901 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5904 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5905 and non BLKmode operands.
5907 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5910 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5911 Clean up attributes.
5913 2018-03-23 Richard Biener <rguenther@suse.de>
5916 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5917 we are going to emit early debug for LTO.
5919 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5922 * function.c (match_asm_constraints_1): Don't optimize if input
5923 doesn't satisfy general_operand predicate for output's mode.
5926 * alias.c (write_dependence_p): Don't require for x_canonicalized
5927 non-VOIDmode if x has VOIDmode.
5930 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5931 just don't try to optimize it rather than assert it never happens.
5933 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5935 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5936 macro expansions for definition of ST_INTERNAL_<mode> and
5937 LD_INTERNAL_<mode> builtins.
5938 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5940 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5942 (altivec_expand_st_builtin): Likewise.
5943 (altivec_expand_builtin): Remove calls to deleted functions.
5944 (rs6000_address_for_altivec): Delete this function.
5945 * config/rs6000/vector.md: Remove expands for
5946 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5948 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5951 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5952 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5953 re-computing once computed.
5954 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5955 (arm_init_machine_status): Initialize
5956 machine->static_chain_stack_bytes.
5958 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5961 * doc/extend.texi: Add four new prototypes for vec_ld.
5962 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5963 definitions for more logical presentation.
5964 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5965 entries for V1TI variants of __builtin_altivec_ld builtin.
5966 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5967 handling of V1TI variant of LVX icode pattern.
5968 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5969 (rs6000_gimple_fold_builtin): Likewise.
5970 (altivec_init_builtins): Add code to define
5971 __builtin_altivec_lvx_v1ti function.
5973 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5976 * function.c (match_asm_constraints_1): Don't do the optimization
5977 if input isn't a REG, SUBREG, MEM or constant.
5979 2018-03-22 Tom de Vries <tom@codesourcery.com>
5981 PR tree-optimization/84956
5982 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5983 bb_has_abnormal_pred.
5985 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5988 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5989 DECL_INITIAL (decl) to decl at the end.
5990 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5993 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5995 * doc/extend.texi (__builtin_tgmath): Document when complex
5996 integer types are treated as _Complex _Float64.
5998 2018-03-21 Tom de Vries <tom@codesourcery.com>
6000 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6002 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6004 PR tree-optimization/84960
6005 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6006 if it is ENTRY block, move them into single succ of ENTRY in that case.
6008 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6010 PR tree-optimization/84811
6011 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6012 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6013 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6014 (known_subrange_p): Update accordingly. Cast each value involved
6015 in the size comparison, rather than casting the result of the
6018 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6020 PR tree-optimization/84982
6021 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6022 by flipping the least significant bit rather than all bits from
6023 bitpos to bitpos + bitsize - 1.
6025 2018-03-21 Nathan Sidwell <nathan@acm.org>
6027 * doc/extend.texi (Deprecated Features): Remove mention of
6028 long-deleted deprecations.
6030 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6033 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6034 * configure: Regenerate.
6036 2018-03-21 Tom de Vries <tom@codesourcery.com>
6038 PR tree-optimization/83126
6039 * tree-parloops.c (num_phis): New function.
6040 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6042 2018-03-21 Nathan Sidwell <nathan@acm.org>
6044 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6045 mention anon-struct/union members and trailing attributes.
6047 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6049 PR tree-optimization/84969
6050 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6051 builtin memset partitions if they set different rhs values.
6053 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6055 PR rtl-optimization/84989
6056 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6057 VEC_DUPLICATE with scalar result mode.
6059 2018-03-21 Martin Liska <mliska@suse.cz>
6062 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6063 not intended return statement.
6065 2018-03-21 Martin Liska <mliska@suse.cz>
6068 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6069 (chkp_find_bound_slots_1): Limit number of iterations.
6071 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6074 * Minor grammar fixes for x86 options.
6076 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6079 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6080 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6082 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6085 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6086 (altivec_lvx_<mode>_1op): Likewise.
6087 (altivec_stvx_<mode>_2op): Likewise.
6088 (altivec_stvx_<mode>_1op): Likewise.
6089 (altivec_lvx_<VM2:mode>): New define_expand.
6090 (altivec_stvx_<VM2:mode>): Likewise.
6091 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6092 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6093 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6094 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6095 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6096 (rs6000_gen_lvx): Likewise.
6097 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6098 (altivec_expand_stv_builtin): Likewise.
6099 (altivec_expand_builtin): Likewise.
6100 * config/rs6000/vector.md: Likewise.
6102 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6105 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6108 2018-03-20 Richard Biener <rguenther@suse.de>
6111 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6112 sign-conversions as zero, fall back to standard scalar_stmt
6115 2018-03-20 Martin Liska <mliska@suse.cz>
6118 * predict.c (rebuild_frequencies): Handle case when we have
6119 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6121 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6124 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6125 flag_section_anchors.
6126 * varasm.c (use_blocks_for_decl_p): Remove hack for
6127 dw2_force_const_mem.
6130 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6132 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6133 be created, use lowpart_subreg of operands[0] rather than operands[0]
6135 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6136 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6137 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6138 and n constraint instead of aarch64_shift_imm_di and Usd.
6139 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6140 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6142 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6145 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6146 to favor GPR over NEON registers.
6147 (<shift>di3_neon): Likewise.
6149 2018-03-20 Tom de Vries <tom@codesourcery.com>
6152 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6153 (nvptx_process_pars): Emit bar.sync asap and alap.
6155 2018-03-20 Tom de Vries <tom@codesourcery.com>
6158 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6159 seen_label if seen_label is already set.
6161 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6164 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6165 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6166 Use 1U instead of 1. Formatting fixes.
6169 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6170 instead of TREE_TYPE (s1) for the return value.
6172 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6174 PR tree-optimization/84946
6175 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6176 bitsize + bitsize in poly_uint64 rather than poly_int64.
6179 * dwarf2asm.c: Include fold-const.c.
6180 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6181 of decl rather than decl itself.
6183 PR rtl-optimization/84643
6184 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6186 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6189 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6190 calling assemble_variable.
6192 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6195 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6196 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6198 2018-03-19 Jim Wilson <jimw@sifive.com>
6201 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6202 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6203 (riscv_first_stack_step): Likewise.
6204 (riscv_option_override): Use STACK_BOUNDARY instead of
6206 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6208 (BIGGEST_ALIGNMENT): Set to 128.
6209 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6210 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6213 2018-03-19 Richard Biener <rguenther@suse.de>
6215 PR tree-optimization/84933
6216 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6217 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6219 2018-03-19 Richard Biener <rguenther@suse.de>
6221 PR tree-optimization/84859
6222 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6223 (cond_if_else_store_replacement): Perform sinking operation on
6224 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6225 Generalize what a BB with a single eligible store is.
6227 2018-03-19 Richard Biener <rguenther@suse.de>
6229 PR tree-optimization/84929
6230 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6231 chrec_is_positive against non-chrec arg.
6233 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6236 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6238 2018-03-18 Martin Liska <mliska@suse.cz>
6240 PR rtl-optimization/84635
6241 * regrename.c (build_def_use): Use matches_mode only when
6244 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6246 PR tree-optimization/84913
6247 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6248 vectorize chains of COND_EXPRs.
6250 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6252 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6254 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6256 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6258 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6260 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6262 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6263 Kito Cheng <kito.cheng@gmail.com>
6265 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6266 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6267 (nds32_adjust_reg_alloc_order): New function.
6268 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6270 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6272 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6273 nds32_print_operand, nds32_print_operand_address): Use
6274 HOST_WIDE_INT_PRINT_DEC instead.
6276 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6278 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6280 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6283 * config/i386/i386.c (initial_ix86_tune_features,
6284 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6286 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6287 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6288 rather than 1u << ix86_tune. Formatting fix.
6289 (ix86_option_override_internal): Change ix86_arch_mask from
6290 unsigned int to unsigned HOST_WIDE_INT, initialize to
6291 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6292 (ix86_function_specific_restore): Likewise.
6294 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6297 * postreload.c (reload_combine_recognize_pattern): Perform
6298 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6299 truncate_int_for_mode the result for the destination's mode.
6302 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6303 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6305 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6308 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6311 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6312 regno_allocno_class_array and sorted_pseudos.
6313 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6314 insns where regno is used.
6316 2018-03-16 Martin Liska <mliska@suse.cz>
6319 * multiple_target.c (create_dispatcher_calls): Redirect
6320 reference in the symbol table.
6322 2018-03-16 Martin Liska <mliska@suse.cz>
6325 * multiple_target.c (create_dispatcher_calls): Redirect also
6328 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6332 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6333 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6334 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6336 2018-03-16 Julia Koval <julia.koval@intel.com>
6338 * doc/invoke.texi (Skylake Server): Add CLWB.
6339 Cannonlake): Remove CLWB.
6341 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6343 PR tree-optimization/84841
6344 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6346 (FLOAT_ONE_CONST_TYPE): Define.
6347 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6348 (sort_by_operand_rank): Put entries with higher constant_type last
6349 rather than first to match comments.
6351 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6353 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6356 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6359 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6360 check and use address of target always.
6362 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6365 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6366 (indirect_thunk_bnd_needed): Likewise.
6367 (indirect_thunks_used): Likewise.
6368 (indirect_thunks_bnd_used): Likewise.
6369 (indirect_return_needed): New.
6370 (indirect_return_bnd_needed): Likewise.
6371 (output_indirect_thunk_function): Add a bool argument for
6373 (output_indirect_thunk_function): Don't generate alias for
6374 function return thunk.
6375 (ix86_code_end): Call output_indirect_thunk_function to generate
6376 function return thunks.
6377 (ix86_output_function_return): Set indirect_return_bnd_needed
6378 and indirect_return_needed instead of indirect_thunk_bnd_needed
6379 and indirect_thunk_needed.
6381 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6383 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6384 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6385 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6387 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6388 Paul Hua <paul.hua.gm@gmail.com>
6391 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6393 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6395 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6396 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6399 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6402 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6403 instead of GET_MODE_SIZE when comparing Units.
6405 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6408 * varasm.c (hash_section): Return an unchangeble hash value
6409 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6410 Return !aarch64_can_use_per_function_literal_pools_p ().
6412 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6415 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6416 rather than address of cmode as last argument to prepare_cmp_insn.
6418 2018-03-15 Julia Koval <julia.koval@intel.com>
6420 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6421 F_AVX512VNNI, F_AVX512BITALG): New.
6423 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6426 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6427 insn for floating-point loads and stores.
6429 2018-03-14 Carl Love <cel@us.ibm.com>
6431 * config/rs6000/rs6000-c.c: Add macro definitions for
6432 ALTIVEC_BUILTIN_VEC_PERMXOR.
6433 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6434 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6435 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6436 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6438 * config/doc/extend.texi: Add prototypes for vec_permxor.
6440 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6443 * diagnostic-show-locus.c (class layout_point): Convert m_line
6444 from int to linenum_type.
6445 (line_span::comparator): Use linenum "compare" function when
6446 comparing line numbers.
6447 (test_line_span): New function.
6448 (layout_range::contains_point): Convert param "row" from int to
6450 (layout_range::intersects_line_p): Likewise.
6451 (layout::will_show_line_p): Likewise.
6452 (layout::print_source_line): Likewise.
6453 (layout::should_print_annotation_line_p): Likewise.
6454 (layout::print_annotation_line): Likewise.
6455 (layout::print_leading_fixits): Likewise.
6456 (layout::annotation_line_showed_range_p): Likewise.
6457 (struct line_corrections): Likewise for field m_row.
6458 (line_corrections::line_corrections): Likewise for param "row".
6459 (layout::print_trailing_fixits): Likewise.
6460 (layout::get_state_at_point): Likewise.
6461 (layout::get_x_bound_for_row): Likewise.
6462 (layout::print_line): Likewise.
6463 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6464 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6465 * input.c (selftest::test_linenum_comparisons): New function.
6466 (selftest::input_c_tests): Call it.
6467 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6468 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6469 * selftest.h (ASSERT_GT): New macro.
6470 (ASSERT_GT_AT): New macro.
6471 (ASSERT_LT): New macro.
6472 (ASSERT_LT_AT): New macro.
6474 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6476 PR rtl-optimization/84780
6477 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6479 2018-03-14 Martin Liska <mliska@suse.cz>
6481 * tree.c (record_node_allocation_statistics): Use
6482 get_stats_node_kind.
6483 (get_stats_node_kind): New function extracted from
6484 record_node_allocation_statistics.
6485 (free_node): Use get_stats_node_kind.
6487 2018-03-14 Richard Biener <rguenther@suse.de>
6489 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6490 that the value-set of ANTIC_IN doesn't grow.
6493 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6495 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6496 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6498 2018-03-14 Julia Koval <julia.koval@intel.com>
6500 * config.gcc (icelake-client, icelake-server): New.
6502 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6503 (initial_ix86_arch_features): Ditto.
6504 (PTA_SKYLAKE): Add SGX.
6505 (PTA_ICELAKE): Remove.
6506 (PTA_ICELAKE_CLIENT): New.
6507 (PTA_ICELAKE_SERVER): New.
6508 (ix86_option_override_internal): Split up icelake on icelake client and
6510 (get_builtin_code_for_version): Ditto.
6511 (fold_builtin_cpu): Ditto.
6512 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6513 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6514 * config/i386/i386.h (processor_type): Ditto.
6515 * doc/invoke.texi: Ditto.
6517 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6520 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6521 INTEGER_CST offset, add it together with bitpos / 8 and
6522 sign extend based on POINTER_SIZE.
6526 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6529 * config/i386/constraints.md (Yc): New register constraint.
6530 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6531 Use Yc constraint for alternative 2 of operand 0. Remove
6532 preferred_for_speed attribute.
6534 2018-03-14 Richard Biener <rguenther@suse.de>
6536 PR tree-optimization/84830
6537 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6538 with the old one to avoid oscillations.
6540 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6543 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6545 (assign_by_spills): Return a flag of reload assignment failure.
6546 Do not process the reload assignment failures. Do not spill other
6547 reload pseudos if they has the same reg class. Update n if
6549 (lra_assign): Add a return arg. Set up from the result of
6550 assign_by_spills call.
6551 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6552 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6553 usage_insns if it is not NULL.
6554 (spill_hard_reg_in_range): New function.
6555 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6556 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6557 function prototypes.
6558 (lra_assign): Change prototype.
6559 * lra.c (lra): Add code to deal with fails by splitting hard reg
6562 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6564 * config/riscv/riscv.opt (mrelax): New option.
6565 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6566 "norelax" when riscv_mrelax is disabled.
6567 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6569 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6572 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6573 reassociation for int modes.
6575 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6577 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6578 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6580 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6581 * config/aarch64/aarch64-sve.md
6582 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6583 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6584 (*extend<mode><Vwide>2): Rename to...
6585 (aarch64_sve_extend<mode><Vwide>2): ...this.
6586 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6587 renaming the old pattern to...
6588 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6590 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6591 define_expand, renaming the old pattern to...
6592 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6593 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6594 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6595 account when deciding which SVE instruction the optab should use.
6596 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6598 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6600 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6601 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6602 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6603 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6604 (tlsdesc_small_<mode>): Turn a define_expand and use
6605 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6606 (tlsdesc_small_advsimd_<mode>): ...this.
6607 (tlsdesc_small_sve_<mode>): New pattern.
6609 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6611 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6612 (UNSPEC_UMUL_HIGHPART): New constants.
6613 (MUL_HIGHPART): New int iteraor.
6614 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6615 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6617 (*<su>mul<mode>3_highpart): New define_insn.
6619 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6622 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6625 2018-03-13 Martin Liska <mliska@suse.cz>
6628 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6630 (sem_item_optimizer::~sem_item_optimizer): Release it.
6631 (sem_item_optimizer::merge_classes): Register variable aliases.
6632 (sem_item_optimizer::fixup_pt_set): New function.
6633 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6634 * ipa-icf.h: Declare new variables and functions.
6636 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6639 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6640 integer_pow2p@2 and test integer_pow2p in condition.
6641 (A < 0 ? C : 0): Similarly for @1.
6644 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6645 characters starting at p contain '\0' character, don't look beyond
6649 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6650 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6653 * reg-stack.c (change_stack): Change update_end var from int to
6654 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6655 also call set_block_for_insn on the newly added insns and rescan.
6658 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6659 on the last operand.
6662 * tree.c (stabilize_reference_1): Return save_expr (e) for
6663 STATEMENT_LIST even if it doesn't have side-effects.
6665 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6667 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6669 2018-03-12 Renlin Li <renlin.li@arm.com>
6671 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6672 aarch64_output_scalar_simd_mov_immediate.
6674 2018-03-12 Martin Sebor <msebor@redhat.com>
6676 PR tree-optimization/83456
6677 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6678 for perfectly overlapping calls to memcpy.
6679 (gimple_fold_builtin_memory_chk): Same.
6680 (gimple_fold_builtin_strcpy): Handle no-warning.
6681 (gimple_fold_builtin_stxcpy_chk): Same.
6682 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6684 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6686 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6687 parameter. Use it for SFmode.
6688 (rs6000_function_arg_advance_1): Adjust.
6689 (rs6000_function_arg): Adjust.
6690 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6692 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6694 PR rtl-optimization/84169
6695 PR rtl-optimization/84780
6696 * combine.c (can_combine_p): Check for a 2-insn combination whether
6697 the destination register is used between the two insns, too.
6699 2018-03-12 Richard Biener <rguenther@suse.de>
6701 PR tree-optimization/84803
6702 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6703 for refs DR analysis didn't process.
6705 2018-03-12 Richard Biener <rguenther@suse.de>
6707 PR tree-optimization/84777
6708 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6709 force-vectorize loops ignore whether we are optimizing for size.
6711 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6713 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6714 (TARGET_MD_ASM_ADJUST): Define.
6716 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6717 Kito Cheng <kito.cheng@gmail.com>
6718 Chung-Ju Wu <jasonwucj@gmail.com>
6720 * config/nds32/nds32.c (nds32_compute_stack_frame,
6721 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6722 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6723 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6724 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6725 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6726 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6727 * config/nds32/nds32.md (prologue, epilogue): Use macro
6728 NDS32_V3PUSH_AVAILABLE_P to do checking.
6730 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6733 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6734 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6735 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6736 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6737 addition of most attributes on !orig_type_die or the attribute not
6738 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6740 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6741 Chung-Ju Wu <jasonwucj@gmail.com>
6743 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6745 * config/nds32/nds32.opt (mvh): New option.
6747 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6748 Chung-Ju Wu <jasonwucj@gmail.com>
6750 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6752 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6753 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6756 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6757 Chung-Ju Wu <jasonwucj@gmail.com>
6759 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6761 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6762 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6764 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6765 Kito Cheng <kito.cheng@gmail.com>
6766 Chung-Ju Wu <jasonwucj@gmail.com>
6768 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6769 UNSPEC_FFMISM and UNSPEC_FLMISM.
6770 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6771 for ffb, ffmism and flmism.
6772 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6773 (unspec_ffmism): Ditto.
6774 (unspec_flmism): Ditto.
6775 (nds32_expand_builtin_impl): Check if string extension is available.
6776 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6777 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6779 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6782 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6785 * lra-assigns.c (assign_by_spills): Return a flag of reload
6786 assignment failure. Do not process the reload assignment
6787 failures. Do not spill other reload pseudos if they has the same
6789 (lra_assign): Add a return arg. Set up from the result of
6790 assign_by_spills call.
6791 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6792 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6793 usage_insns if it is not NULL.
6794 (spill_hard_reg_in_range): New function.
6795 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6796 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6797 function prototypes.
6798 (lra_assign): Change prototype.
6799 * lra.c (lra): Add code to deal with fails by splitting hard reg
6802 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6805 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6807 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6810 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6811 (dw_val_node): Add val_symbolic_view.
6812 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6813 (symview_upper_bound): New.
6814 (new_line_info_table): Initialize symviews_since_reset.
6815 (dwarf2out_source_line): Count symviews_since_reset and set
6816 symview_upper_bound.
6817 (dw_val_equal_p): Handle symview.
6818 (add_AT_symview): New.
6819 (print_dw_val): Handle symview.
6820 (attr_checksum, attr_checksum_ordered): Likewise.
6821 (same_dw_val_p, size_of_die): Likewise.
6822 (value_format, output_die): Likewise.
6823 (add_high_low_attributes): Use add_AT_symview for entry_view.
6824 (dwarf2out_finish): Reset symview_upper_bound, clear
6827 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6830 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6831 Add strict argument and use it.
6832 (rs6000_split_multireg_move): Update for new strict argument.
6833 (mem_operand_gpr): Disallow all non-offsettable addresses.
6834 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6836 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6839 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6840 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6841 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6844 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6845 decl, use remap_type if we want to use the type.
6847 2018-03-09 Martin Sebor <msebor@redhat.com>
6849 PR tree-optimization/84526
6850 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6852 (builtin_access::generic_overlap): Be prepared to handle non-array
6855 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6857 PR rtl-optimization/84682
6858 * lra-constraints.c (process_address_1): Check is_address flag
6859 for address constraints.
6860 (process_alt_operands): Likewise.
6861 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6862 preprocess_constraints.
6863 * recog.h (preprocess_constraints): Add oploc parameter.
6865 * recog.c (preprocess_constraints): Test address_operand for
6866 CT_ADDRESS constraints.
6868 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6871 * lra-assigns.c (assign_by_spills): Return a flag of reload
6872 assignment failure. Do not process the reload assignment
6873 failures. Do not spill other reload pseudos if they has the same
6875 (lra_assign): Add a return arg. Set up from the result of
6876 assign_by_spills call.
6877 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6878 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6879 usage_insns if it is not NULL.
6880 (spill_hard_reg_in_range): New function.
6881 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6882 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6883 function prototypes.
6884 (lra_assign): Change prototype.
6885 * lra.c (lra): Add code to deal with fails by splitting hard reg
6888 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6891 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6892 Accept complain bool parameter. Only emit errors if it is true.
6893 (arm_parse_cpu_option_name): Likewise.
6894 (arm_target_thumb_only): Adjust callers of the above.
6895 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6896 prototype to take a default true bool parameter.
6897 (arm_parse_arch_option_name): Likewise.
6899 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6900 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6904 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6905 * configure: Regenerate.
6906 * configure.ac ("linker --version-script option"): New.
6907 ("linker soname option"): New.
6909 2018-03-09 Richard Biener <rguenther@suse.de>
6911 PR tree-optimization/84775
6912 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6913 immediate uses of predicate stmts and mark them modified.
6916 PR tree-optimization/84178
6917 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6919 (version_loop_for_if_conversion): Delay update_ssa call.
6920 (tree_if_conversion): Delay update_ssa until after predicate
6923 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6926 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6927 when the function accesses prior frames.
6929 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6932 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6933 gen_llsym, otherwise call maybe_gen_llsym.
6936 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6937 has ',' character inside of it.
6939 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6942 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6943 as clobbering CC_REGNUM.
6945 2018-03-08 Richard Biener <rguenther@suse.de>
6948 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6949 (follow_copies_to_constant): Do not follow SSA names registered
6952 2018-03-08 Richard Biener <rguenther@suse.de>
6954 PR tree-optimization/84178
6955 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6957 (version_loop_for_if_conversion): Delay update_ssa call.
6958 (tree_if_conversion): Delay update_ssa until after predicate
6961 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6963 PR tree-optimization/84178
6964 * tree-if-conv.c (release_bb_predicate): Remove the
6965 the assertion that the stmts have NULL use_ops.
6966 Discard the statements, asserting that they haven't
6967 yet been added to a BB.
6969 2018-03-08 Richard Biener <rguenther@suse.de>
6971 PR tree-optimization/84746
6972 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6973 (phi_translate): Pass in destination ANTIC_OUT set.
6974 (phi_translate_1): Likewise. For a simplified result lookup
6975 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6976 (phi_translate_set): Adjust.
6977 (do_pre_regular_insertion): Likewise.
6978 (do_pre_partial_partial_insertion): Likewise.
6980 2018-03-08 Martin Liska <mliska@suse.cz>
6982 PR gcov-profile/84735
6983 * doc/gcov.texi: Document usage of profile files.
6984 * gcov-io.h: Document changes in the format.
6986 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6990 * dwarf2out.c (struct dw_line_info_table): Update comments for
6992 (FORCE_RESET_NEXT_VIEW): New.
6993 (FORCE_RESETTING_VIEW_P): New.
6994 (RESETTING_VIEW_P): Check for -1 too.
6995 (ZERO_VIEW_P): Likewise.
6996 (new_line_info_table): Force-reset next view.
6997 (dwarf2out_begin_function): Likewise.
6998 (dwarf2out_source_line): Simplify zero_view_p initialization.
6999 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7000 view directly. Omit view when omitting .loc at line 0.
7002 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7004 PR tree-optimization/84740
7005 * tree-switch-conversion.c (process_switch): Call build_constructors
7006 only if info.phi_count is non-zero.
7008 PR tree-optimization/84739
7009 * tree-tailcall.c (find_tail_calls): Check call arguments against
7010 DECL_ARGUMENTS (current_function_decl) rather than
7011 DECL_ARGUMENTS (func) when checking for tail recursion.
7013 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7015 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7016 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7017 Volker Reichelt's entry and add entries for people that perform
7018 GCC fuzzy testing and report numerous bugs.
7020 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7023 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7024 readonly data in sdata, if that is disabled.
7025 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7026 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7027 -mreadonly-in-sdata option.
7029 2018-03-07 Martin Sebor <msebor@redhat.com>
7031 PR tree-optimization/84468
7032 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7033 basic block when looking for nul assignment.
7035 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7038 * except.h (output_function_exception_table): Adjust prototype.
7039 * except.c (output_function_exception_table): Remove FNNAME parameter
7040 and add SECTION parameter. Ouput one part of the table at a time.
7041 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7042 the first part of the exception table and emit unwind directives.
7043 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7044 (i386_pe_seh_cold_init): Likewise.
7045 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7046 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7047 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7048 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7049 * config/i386/winnt.c: Include except.h.
7050 (struct seh_frame_state): Add reg_offset, after_prologue and
7051 in_cold_section fields.
7052 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7053 (i386_pe_seh_cold_init): New function.
7054 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7055 to seh->in_cold_section.
7056 (seh_emit_push): Record the offset of the push.
7057 (seh_emit_save): Record the offet of the save.
7058 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7059 Test seh->after_prologue to disregard the epilogue.
7060 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7061 (i386_pe_end_cold_function): New function.
7063 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7066 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7067 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7070 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7072 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7073 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7076 * multiple_target.c: Include tree-inline.h and intl.h.
7077 (expand_target_clones): Diagnose and fail if node->definition and
7078 !tree_versionable_function_p (node->decl).
7080 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7082 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7084 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7085 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7086 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7088 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7091 * combine.c (try_combine): Use reg_or_subregno instead of handling
7092 just paradoxical SUBREGs and REGs.
7094 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7096 * config/arc/arc.c (arc_finalize_pic): Remove function.
7097 (arc_must_save_register): We use single base PIC register, remove
7098 checks to save/restore the PIC register.
7099 (arc_expand_prologue): Likewise.
7100 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7102 (arc_verify_short): Likewise.
7103 (arc_attr_type): Likewise.
7104 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7105 (walk_stores): Likewise.
7106 (arc_address_cost): Make it static.
7107 (arc_verify_short): Likewise.
7108 (branch_dest): Likewise.
7109 (arc_attr_type): Likewise.
7110 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7111 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7112 (arc_final_prescan_insn): Remove inserting the nops due to
7113 hardware hazards. It is done in reorg step.
7114 (insn_length_variant_t): Remove.
7115 (insn_length_parameters_t): Likewise.
7116 (arc_insn_length_parameters): Likewise.
7117 (arc_get_insn_variants): Likewise.
7118 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7120 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7123 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7126 PR tree-optimization/84687
7127 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7129 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7131 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7133 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7134 Rename to ppc_speculation_barrier.
7135 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7136 __builtin_ppc_speculation_barrier.
7138 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7141 * combine.c (combine_simplify_rtx): Don't try to simplify if
7142 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7145 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7147 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7148 to 32 bytes when compiling for POWER9.
7150 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7153 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7154 regparm >= 3 with no arg reg available also for calls with
7155 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7158 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7160 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7162 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7165 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7167 2018-03-05 Richard Biener <rguenther@suse.de>
7169 PR tree-optimization/84486
7170 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7171 When inserting a __builtin_assume_aligned call set the LHS
7172 SSA name alignment info accordingly.
7174 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7176 PR tree-optimization/84114
7177 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7178 Avoid reassociation of FLOAT_MODE addition.
7180 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7182 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7183 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7184 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7185 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7186 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7187 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7188 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7190 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7191 __builtin_ia32_wbinvd): New builtins.
7192 (SPECIAL_ARGS2): New.
7193 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7194 (SPECIAL_ARGS2): New.
7195 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7196 (ix86_valid_target_attribute_inner_p): Ditto.
7197 (ix86_init_mmx_sse_builtins): Add special_args2.
7198 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7199 TARGET_WBNOINVD_P): New.
7200 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7201 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7202 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7203 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7204 * config/i386/pconfigintrin.h: New file.
7205 * config/i386/wbnoinvdintrin.h: Ditto.
7206 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7208 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7210 2018-03-05 Richard Biener <rguenther@suse.de>
7212 PR tree-optimization/84670
7213 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7215 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7216 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7217 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7218 doesn't grow if all successors (recursively) were visited at least
7221 2018-03-05 Richard Biener <rguenther@suse.de>
7223 PR tree-optimization/84650
7224 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7225 if executed in the loop pipeline.
7227 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7229 * doc/configfiles.texi (Configuration Files): Move info about
7230 conditionalizing $target-protos.h to...
7231 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7232 differs from $target-protos.h.
7234 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7235 Chung-Ju Wu <jasonwucj@gmail.com>
7237 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7238 * config/nds32/nds32-multiple.md (setmemsi): Define.
7239 * config/nds32/nds32-memory-manipulation.c
7240 (nds32_gen_dup_4_byte_to_word_value): New.
7241 (emit_setmem_word_loop): New.
7242 (emit_setmem_byte_loop): New.
7243 (nds32_expand_setmem_loop): New.
7244 (nds32_expand_setmem_loop_v3m): New.
7245 (nds32_expand_setmem_unroll): New.
7246 (nds32_expand_setmem): New.
7248 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7249 Chung-Ju Wu <jasonwucj@gmail.com>
7251 * config/nds32/nds32-memory-manipulation.c
7252 (nds32_emit_load_store): New.
7253 (nds32_emit_post_inc_load_store): New.
7254 (nds32_emit_mem_move): New.
7255 (nds32_emit_mem_move_block): New.
7256 (nds32_expand_movmemsi_loop_unknown_size): New.
7257 (nds32_expand_movmemsi_loop_known_size): New.
7258 (nds32_expand_movmemsi_loop): New.
7259 (nds32_expand_movmemsi_unroll): New.
7260 (nds32_expand_movmemqi): Rename ...
7261 (nds32_expand_movmemsi): ... to this.
7262 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7263 (movmemsi): ... to this.
7264 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7265 (nds32_expand_movmemsi): ... to this.
7267 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7268 Monk Chiang <sh.chiang04@gmail.com>
7269 Chung-Ju Wu <jasonwucj@gmail.com>
7271 * config/nds32/nds32-protos.h
7272 (nds32_expand_load_multiple): New arguments.
7273 (nds32_expand_store_multiple): Ditto.
7274 (nds32_valid_multiple_load_store): Rename ...
7275 (nds32_valid_multiple_load_store_p): ... to this.
7276 * config/nds32/nds32-memory-manipulation.c
7277 (nds32_expand_load_multiple): Refine implementation.
7278 (nds32_expand_store_multiple): Ditto.
7279 * config/nds32/nds32-multiple.md
7280 (load_multiple): Update nds32_expand_load_multiple interface.
7281 (store_multiple): Update nds32_expand_store_multiple interface.
7282 * config/nds32/nds32-predicates.c
7283 (nds32_valid_multiple_load_store): Rename ...
7284 (nds32_valid_multiple_load_store_p): ... to this and refine
7286 * config/nds32/predicates.md
7287 (nds32_load_multiple_and_update_address_operation): New predicate.
7288 (nds32_store_multiple_and_update_address_operation): New predicate.
7290 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7291 Chung-Ju Wu <jasonwucj@gmail.com>
7293 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7294 (combo): New attribute.
7295 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7297 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7299 * config/nds32/nds32.opt: Change -mcmodel= default value.
7301 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7302 Monk Chiang <sh.chiang04@gmail.com>
7303 Chung-Ju Wu <jasonwucj@gmail.com>
7305 * config/nds32/constants.md (unspec_element): New enum.
7306 * config/nds32/constraints.md (Umw): New constraint.
7307 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7308 * config/nds32/nds32-intrinsic.md: Likewise.
7309 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7310 (nds32_valid_smw_lwm_base_p): New.
7311 (nds32_output_smw_single_word): New.
7312 (nds32_output_lmw_single_word): New.
7313 (nds32_expand_unaligned_load): New.
7314 (nds32_expand_unaligned_store): New.
7315 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7316 (nds32_output_smw_single_word): Declare.
7317 (nds32_output_lmw_single_word): Declare.
7318 (nds32_expand_unaligned_load): Declare.
7319 (nds32_expand_unaligned_store): Declare.
7320 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7321 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7322 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7323 NDS32_BUILTIN_UASTORE_DW.
7324 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7327 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7328 Kito Cheng <kito.cheng@gmail.com>
7329 Chung-Ju Wu <jasonwucj@gmail.com>
7331 * config/nds32/nds32-intrinsic.c
7332 (nds32_expand_builtin_null_ftype_reg): Delete.
7333 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7334 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7335 (nds32_read_argument): New.
7336 (nds32_legitimize_target): Ditto.
7337 (nds32_legitimize_argument): Ditto.
7338 (nds32_check_constant_argument): Ditto.
7339 (nds32_expand_unop_builtin): Ditto.
7340 (nds32_expand_unopimm_builtin): Ditto.
7341 (nds32_expand_binop_builtin): Ditto.
7342 (nds32_builtin_decl_impl): Ditto.
7343 (builtin_description): Ditto.
7344 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7345 (nds32_init_builtins_impl): Ditto.
7346 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7347 (nds32_builtin_decl): New.
7348 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7349 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7351 2018-03-02 Jeff Law <law@redhat.com>
7353 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7354 (redundant_insn, fill_simple_delay_slots): Likewise.
7355 (fill_slots_from_thread): Likewise.
7356 * resource.c (mark_referenced_resources): Likewise.
7357 (mark_set_resources, find_dead_or_set_registers): Likewise.
7359 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7361 * substring-locations.h (format_warning_va): Formatting fix for
7363 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7365 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7366 * substring-locations.c: Include intl.h.
7367 (format_warning_va): Turned into small wrapper around
7368 format_warning_n_va, renamed to ...
7369 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7370 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7372 (format_warning_at_substring_n): New function.
7373 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7374 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7375 format_warning_at_substring with just a shorter name instead of
7376 const function pointer.
7377 (fmtwarn_n): New function.
7378 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7379 appropriate, get rid of all the fmtstr temporaries, move conditionals
7380 with G_() wrapped string literals directly into fmtwarn arguments,
7381 cast dir.len to (int), formatting fixes.
7383 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7385 * doc/invoke.texi: Remove "Cilk Plus" references.
7387 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7388 Richard Biener <rguenther@suse.de>
7391 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7392 for error or warning attributes if CALL_FROM_THUNK_P is set.
7395 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7398 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7399 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7402 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7403 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7405 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7406 instead of -1U in last predictors element's probability member.
7408 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7411 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7412 arguments if they are comparable.
7414 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7416 PR tree-optimization/84634
7417 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7418 masks and masked_loop_p with a single loop_masks, making sure it's
7419 null for bb vectorization.
7421 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7423 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7424 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7425 loop->force_vectorize to check whether there is no alias.
7427 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7430 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7432 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7434 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7435 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7436 instead of a loop around prev_real_insn.
7437 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7441 * config/i386/i386.c (ix86_print_operand): Use conditional
7442 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7445 2018-03-02 Richard Biener <rguenther@suse.de>
7447 PR tree-optimization/84427
7448 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7449 (bitmap_set_subtract_values): Rewrite to handle multiple
7452 (prune_clobbered_mems): Likewise.
7453 (phi_translate): Take edge instead of pred/phiblock.
7454 (phi_translate_1): Likewise.
7455 (phi_translate_set): Likewise. Insert all translated
7456 exprs for a value into the set, keeping possibly multiple
7457 expressions per value.
7458 (compute_antic_aux): Adjust for phi_translate changes.
7459 When intersecting union the expressions and prune those
7460 not in the final value set, keeping possibly multiple
7461 expressions per value. Do not use value-insertion
7462 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7463 all expressions. Add verification that the value-sets
7464 only shrink during iteration.
7465 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7466 (do_pre_regular_insertion): Likewise.
7467 (do_pre_partial_partial_insertion): Likewise.
7469 2018-03-02 Richard Biener <rguenther@suse.de>
7472 * config/darwin.c (saved_debug_info_level): New static global.
7473 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7474 (darwin_asm_lto_end): Restore debug info generation settings.
7476 2018-03-01 Martin Liska <mliska@suse.cz>
7479 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7482 2018-03-01 Richard Biener <rguenther@suse.de>
7485 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7486 type annotation with LTO when debug was disabled at compile-time.
7488 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7490 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7492 (mips_final_postscan_insn): Likewise.
7494 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7496 PR rtl-optimization/84528
7497 * alias.c (init_alias_target): Add commentary.
7498 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7499 a unique base value if the frame pointer is not eliminated
7500 to the stack pointer.
7502 2018-03-01 Tom de Vries <tom@codesourcery.com>
7504 PR rtl-optimization/83327
7505 * lra-int.h (hard_regs_spilled_into): Declare.
7506 * lra.c (hard_regs_spilled_into): Define.
7507 (init_reg_info): Init hard_regs_spilled_into.
7508 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7509 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7510 (process_bb_lives): Handle hard_regs_spilled_into.
7511 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7512 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7514 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7516 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7517 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7518 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7519 * config/rs6000/aix72.h: New file.
7521 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7523 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7524 instead of warning_at with conditional singular and plural messages
7528 * stor-layout.c (update_alignment_for_field): For
7529 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7530 && !DECL_PACKED (field), do the alignment update, just use
7531 only desired_align instead of MAX (type_align, desired_align)
7533 (place_field): Don't do known_align < desired_align handling
7534 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7535 is non-NULL, instead do it after rli->prev_field handling and
7536 only if not within a bitfield word. For DECL_PACKED (field)
7537 use type_align of BITS_PER_UNIT.
7539 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7541 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7542 superfluous parentheses and trailing spaces.
7544 2018-02-28 Richard Biener <rguenther@suse.de>
7546 PR tree-optimization/84584
7547 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7548 SCoPs with fake exit edge.
7550 2018-02-28 Martin Liska <mliska@suse.cz>
7553 * timevar.c (timer::print): Fix format to properly print 100%
7556 2018-02-28 Richard Biener <rguenther@suse.de>
7559 * genmatch.c (capture_info::walk_match): Do not mark
7560 captured expressions without operands as expr_p given
7561 they act more like predicates and should be subject to
7562 "lost tail" side-effect preserving.
7564 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7566 PR rtl-optimization/81611
7567 * auto-inc-dec.c (attempt_change): Move dead note from
7568 mem_insn if it's the next use of regno
7569 (find_address): Take address use of reg holding
7570 non-incremented value. Add parm to limit search to the named
7572 (merge_in_block): Attempt to use a mem insn that is the next
7573 use of the original regno.
7575 2018-02-27 Martin Sebor <msebor@redhat.com>
7578 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7579 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7581 2018-02-27 Martin Sebor <msebor@redhat.com>
7583 PR translation/84207
7584 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7585 n argument to unsigned HOST_WIDE_INT.
7586 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7587 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7588 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7589 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7591 2018-02-27 Richard Biener <rguenther@suse.de>
7593 PR tree-optimization/84512
7594 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7595 Do not use the estimate returned from record_stmt_cost for
7596 the scalar iteration cost but sum properly using add_stmt_cost.
7598 2018-02-27 Richard Biener <rguenther@suse.de>
7600 PR tree-optimization/84466
7601 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7602 Adjust last change to less strictly validate use operands.
7604 2018-02-27 Martin Liska <mliska@suse.cz>
7606 PR gcov-profile/84548
7607 * gcov.c (process_file): Allow partial overlap and consider it
7608 also as group functions.
7609 (output_lines): Properly calculate range of lines for a group.
7611 2018-02-27 Martin Liska <mliska@suse.cz>
7613 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7614 'ggc' suffixes. Change first column width.
7615 (timer::print): Fix formatting of the column.
7617 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7619 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7620 preserve inline entry blocks for the sake of debug inline
7621 entry point markers alone.
7622 (remove_unused_locals): Suggest in comments a better place to
7623 force the preservation of inline entry blocks that are
7624 otherwise unused, but do not preserve them.
7626 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7628 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7630 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7633 * config/i386/constraints.md (Bs): Replace
7634 ix86_indirect_branch_register with
7635 TARGET_INDIRECT_BRANCH_REGISTER.
7637 * config/i386/i386.md (indirect_jump): Likewise.
7638 (tablejump): Likewise.
7639 (*sibcall_memory): Likewise.
7640 (*sibcall_value_memory): Likewise.
7641 Peepholes of indirect call and jump via memory: Likewise.
7642 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7643 (*sibcall_value_GOT_32): Likewise.
7644 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7645 (GOT_memory_operand): Likewise.
7646 (call_insn_operand): Likewise.
7647 (sibcall_insn_operand): Likewise.
7648 (GOT32_symbol_operand): Likewise.
7649 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7651 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7653 PR rtl-optimization/83496
7654 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7655 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7656 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7657 redundant insn, if any.
7658 (relax_delay_slots): Likewise.
7659 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7661 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7663 PR tree-optimization/83965
7664 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7665 that grouped statements are part of a reduction chain. Return
7666 true if the statement is not marked as a reduction itself but
7668 (vect_recog_dot_prod_pattern): Don't check whether the statement
7669 is part of a group here.
7670 (vect_recog_sad_pattern): Likewise.
7671 (vect_recog_widen_sum_pattern): Likewise.
7673 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7676 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7678 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7681 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7683 (ix86_output_indirect_function_return): New prototype.
7684 (ix86_split_simple_return_pop_internal): Likewise.
7685 * config/i386/i386.c (indirect_return_via_cx): New.
7686 (indirect_return_via_cx_bnd): Likewise.
7687 (indirect_thunk_name): Handle return va CX_REG.
7688 (output_indirect_thunk_function): Create alias for
7689 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7690 (ix86_output_indirect_jmp): Remove the bool argument.
7691 (ix86_output_indirect_function_return): New function.
7692 (ix86_split_simple_return_pop_internal): Likewise.
7693 * config/i386/i386.md (*indirect_jump): Don't pass false
7694 to ix86_output_indirect_jmp.
7695 (*tablejump_1): Likewise.
7696 (simple_return_pop_internal): Change it to define_insn_and_split.
7697 Call ix86_split_simple_return_pop_internal to split it for
7699 (simple_return_indirect_internal): Call
7700 ix86_output_indirect_function_return instead of
7701 ix86_output_indirect_jmp.
7703 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7706 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7707 memset and value initialization afterwards.
7709 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7711 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7713 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7716 * common/config/aarch64/aarch64-common.c
7717 (aarch_option_optimization_table[]): Switch
7718 off fomit-frame-pointer
7720 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7721 Chung-Ju Wu <jasonwucj@gmail.com>
7723 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7725 (store_multiple): Ditto.
7727 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7729 * config.gcc: Add --with-cpu support for nds32 target.
7730 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7731 * config/nds32/nds32.opt: Add -mcpu= option.
7733 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7735 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7736 isel=yes): Warn for these deprecated options.
7738 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7740 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7741 ISA_2_5_MASKS_EMBEDDED.
7743 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7745 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7746 p->max as pointers rather than using iterative_hash_expr.
7748 2018-02-23 Carl Love <cel@us.ibm.com>
7750 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7751 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7753 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7754 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7755 P8V_BUILTIN_VEC_VUNSIGNED2.
7757 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7760 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7761 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7763 * lra-constraints.c (curr_insn_transform): Set up
7764 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7766 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7767 * lra-eliminations.c (spill_pseudos): Ditto.
7768 (process_insn_for_elimination): Ditto.
7769 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7770 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7771 LRA_NON_CLOBBERED_ALT.
7773 2018-02-22 Martin Sebor <msebor@redhat.com>
7775 PR tree-optimization/84480
7776 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7777 to maybe_diag_stxncpy_trunc. Call it.
7778 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7779 from gimple_fold_builtin_strcpy. Print inlining stack.
7780 (handle_builtin_stxncpy): Print inlining stack.
7781 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7783 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7786 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7787 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7788 and -fcheck-pointer-bounds are used together.
7789 (indirect_thunk_prefix): New enum.
7790 (indirect_thunk_need_prefix): New function.
7791 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7792 "_nt" instead of "_bnd" for NOTRACK prefix.
7793 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7794 (output_indirect_thunk_function): Likewise.
7796 (ix86_code_end): Update output_indirect_thunk_function calls.
7797 (ix86_output_indirect_branch_via_reg): Replace
7798 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7799 (ix86_output_indirect_branch_via_push): Likewise.
7800 (ix86_output_function_return): Likewise.
7801 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7802 incompatible with -fcf-protection=branch and
7803 -fcheck-pointer-bounds.
7805 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7808 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7809 Change gcc_assert call to output_operand_lossage.
7811 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7813 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7815 2018-02-22 DJ Delorie <dj@redhat.com>
7816 Sebastian Perta <sebastian.perta@renesas.com>
7817 Oleg Endo <olegendo@gcc.gnu.org>
7819 * config/rx/rx.c (rx_rtx_costs): New function.
7820 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7822 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7824 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7826 2018-02-22 Martin Liska <mliska@suse.cz>
7829 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7830 Add "native" as a possible value.
7832 2018-02-22 Martin Liska <mliska@suse.cz>
7835 * config/i386/i386.c (ix86_option_override_internal):
7836 Add "native" as a possible value for -march and -mtune.
7838 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7841 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7842 to all type variants.
7844 PR tree-optimization/84503
7845 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7846 width as info->bitpos + info->bitsize - start.
7847 (merged_store_group::merge_overlapping): Simplify width computation.
7848 (check_no_overlap): New function.
7849 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7850 start + width and last_order of the group, fail if check_no_overlap
7852 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7853 to group if check_no_overlap fails.
7855 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7857 * config/rs6000/altivec.md: Delete contraint arguments to
7858 define_expand, define_split, and define_peephole2, and in
7859 define_insn_and_split if always unused.
7860 * config/rs6000/darwin.md: Ditto.
7861 * config/rs6000/dfp.md: Ditto.
7862 * config/rs6000/rs6000.md: Ditto.
7863 * config/rs6000/sync.md: Ditto.
7864 * config/rs6000/vector.md: Ditto.
7865 * config/rs6000/vsx.md: Ditto.
7867 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7869 * config/rs6000/altivec.md: Write output control strings as braced
7870 blocks instead of double-quoted strings.
7871 * config/rs6000/darwin.md: Ditto.
7872 * config/rs6000/rs6000.md: Ditto.
7873 * config/rs6000/vector.md: Ditto.
7874 * config/rs6000/vsx.md: Ditto.
7876 2018-02-21 Jason Merrill <jason@redhat.com>
7878 PR c++/84314 - ICE with templates and fastcall attribute.
7879 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7881 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7883 * ipa-cp.c (determine_versionability): Fix comment typos.
7885 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7888 * ipa-cp.c (determine_versionability): Do not version functions caling
7891 2018-02-21 Martin Liska <mliska@suse.cz>
7894 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7895 Add "native" as a possible value.
7896 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7897 the macro when native cpu detection is available.
7899 2018-02-21 Martin Liska <mliska@suse.cz>
7902 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7903 Add "native" as a possible value.
7904 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7905 when native cpu detection is available.
7907 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7908 Martin Sebor <msebor@redhat.com>
7910 PR tree-optimization/84478
7911 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7913 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7914 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7915 support which is conservatively correct, for 2 only stay conservative
7916 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7917 argument to the 2 argument get_range_strlen, adjust 6 arg
7918 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7920 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7921 (gimple_fold_builtin_strlen): Pass true as last argument to
7924 2018-02-20 Martin Sebor <msebor@redhat.com>
7927 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7928 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7929 (builtin_memref::builtin_memref): Factor out parts into
7930 set_base_and_offset and call it.
7932 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7935 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7936 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7937 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7938 search at the associated MODE_INT.
7940 2018-02-20 Jeff Law <law@redhat.com>
7943 PR tree-optimization/81592
7945 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7946 for range data rather than using global data.
7947 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7948 range data rather than using global data.
7949 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7950 pass it to children as needed.
7951 (struct directive::fmtresult): Similarly.
7952 (struct directive::set_width): Similarly.
7953 (struct directive::set_precision): Similarly.
7954 (format_integer, format_directive, parse_directive): Similarly.
7955 (format_none): Accept unnamed vr_values parameter.
7956 (format_percent, format_floating, format_character): Similarly.
7957 (format_string, format_plain): Similarly.
7958 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7959 the EVRP range analyzer for range data rather than using global data.
7960 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7961 gimple-ssa-evrp-analyze.h
7962 (class sprintf_dom_walker): Add after_dom_children member function.
7963 Add evrp_range_analyzer member.
7964 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7965 range analyzer as needed.
7966 (sprintf_dom_walker::after_dom_children): New member function.
7967 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7969 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7970 (evrp_range_analyzer::pop_to_marker): Likewise.
7972 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7974 PR tree-optimization/84419
7975 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7976 with the required type if its current type is compatible but
7979 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7982 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7983 after vectorization.
7985 2018-02-20 Martin Liska <mliska@suse.cz>
7988 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7989 possible values if we don't have a hint.
7991 2018-02-20 Martin Liska <mliska@suse.cz>
7995 * final.c (shorten_branches): Build align_tab array with one
7997 * opts.c (finish_options): Add alignment option limit check.
7998 (MAX_CODE_ALIGN): Likewise.
7999 (MAX_CODE_ALIGN_VALUE): Likewise.
8000 * doc/invoke.texi: Document maximum allowed option value for
8001 all -falign-* options.
8003 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8006 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8007 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8008 * var-tracking.c (emit_note_insn_var_location): Remove all references
8009 to NOTE_INSN_CALL_ARG_LOCATION.
8010 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8011 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8012 Use copy_rtx_if_shared.
8013 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8014 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8015 (dwarf2out_var_location): Remove handling of
8016 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8018 * final.c (final_scan_insn): Remove all references to
8019 NOTE_INSN_CALL_ARG_LOCATION.
8020 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8021 before dumping final insns.
8022 * except.c (emit_note_eh_region_end): Remove all references to
8023 NOTE_INSN_CALL_ARG_LOCATION.
8024 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8025 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8026 * config/arc/arc.c (hwloop_optimize): Likewise.
8027 * config/arm/arm.c (create_fix_barrier): Likewise.
8028 * config/s390/s390.c (s390_chunkify_start): Likewise.
8029 * config/sh/sh.c (find_barrier): Likewise.
8030 * config/i386/i386.c (rest_of_insert_endbranch,
8031 ix86_seh_fixup_eh_fallthru): Likewise.
8032 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8033 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8034 * config/frv/frv.c (frv_function_prologue): Likewise.
8035 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8037 (note_outside_basic_block_p): Remove all references to
8038 NOTE_INSN_CALL_ARG_LOCATION.
8039 * gengtype.c (adjust_field_rtx_def): Likewise.
8040 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8042 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8043 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8046 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8049 PR tree-optimization/84452
8050 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8051 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8054 2018-02-19 Martin Liska <mliska@suse.cz>
8057 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8059 2018-02-19 Martin Liska <mliska@suse.cz>
8060 Richard Sandiford <richard.sandiford@linaro.org>
8062 PR tree-optimization/82491
8063 * gimple-fold.c (get_base_constructor): Make earlier bail out
8066 2018-02-19 Carl Love <cel@us.ibm.com>
8068 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8069 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8071 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8072 P8V_BUILTIN_VEC_NEG.
8074 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8076 * config/rl78/rl78.md (movdf): New define expand.
8078 2018-02-19 Martin Liska <mliska@suse.cz>
8081 * doc/invoke.texi: Fix typo.
8082 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8084 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8086 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8087 handle rs6000_single_float and rs6000_double_float specially for
8090 2018-02-16 Jeff Law <law@redhat.com>
8092 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8095 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8098 * ipa-inline.c (inline_small_functions): Fix a typo.
8100 2018-02-16 Nathan Sidwell <nathan@acm.org>
8102 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8104 2018-02-16 Carl Love <cel@us.ibm.com>
8106 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8107 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8108 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8109 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8110 expansion to P8V_BUILTIN_VEC_FLOAT2.
8112 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8114 PR rtl-optimization/70023
8115 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8116 src_regno into account.
8118 2018-02-16 Carl Love <cel@us.ibm.com>
8120 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8121 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8122 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8123 * config/rs6000/rs6000.c: Remove case statements for
8124 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8125 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8126 and P9V_BUILTIN_VEC_VINSERT4B.
8127 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8128 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8129 * config/rs6000/vsx.md:
8130 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8133 2018-02-16 Carl Love <cel@us.ibm.com>
8135 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8137 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8139 * config/rs6000/rs6000-c.c: Add the definitions for
8140 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8141 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8142 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8143 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8144 definition for insert4b and define insn *insert3b_internal.
8145 * doc/extend.texi: Add documentation for vec_extract4b.
8147 2018-02-16 Nathan Sidwell <nathan@acm.org>
8149 * doc/extend.texi (Backwards Compatibility): Mention friend
8150 injection. Note for-scope is deprecated.
8151 * doc/invoke.texi (-ffriend-injection): Deprecate.
8153 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8155 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8156 that moved to I2, also allow destinations that are a paradoxical
8157 subreg (instead of a normal reg).
8159 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8162 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8165 2018-02-16 Richard Biener <rguenther@suse.de>
8167 PR tree-optimization/84037
8168 PR tree-optimization/84016
8170 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8171 Adjust vec_construct for the fact we need additional higher latency
8172 128bit inserts for AVX256 and AVX512 vector builds.
8173 (ix86_add_stmt_cost): Scale vector construction cost for
8176 2018-02-16 Richard Biener <rguenther@suse.de>
8178 PR tree-optimization/84417
8179 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8180 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8181 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8183 2018-02-16 Martin Liska <mliska@suse.cz>
8186 * internal-fn.def (ASAN_CHECK): Set proper flags.
8187 (ASAN_MARK): Likewise.
8189 2018-02-16 Julia Koval <julia.koval@intel.com>
8191 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8192 from PTA_CANNONLAKE.
8194 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8197 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8198 Use ++iter rather than iter++ for std::list iterators.
8199 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8200 defer deleting them until all nodes in the forest are processed. Do
8201 free even leaf nodes. Change to_process into auto_vec.
8204 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8205 * vec.h (vec_default_construct): Use memset instead of placement new
8206 if BROKEN_VALUE_INITIALIZATION is defined.
8207 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8208 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8211 PR rtl-optimization/83723
8212 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8213 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8214 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8216 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8218 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8220 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8222 PR rtl-optimization/81443
8223 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8224 from inner REGs to paradoxical SUBREGs.
8226 2018-02-16 Richard Biener <rguenther@suse.de>
8228 PR tree-optimization/84399
8229 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8230 For operands we can analyze at their definition make sure we can
8231 analyze them at each use as well.
8233 2018-02-16 Richard Biener <rguenther@suse.de>
8235 PR tree-optimization/84190
8236 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8237 volatile accesses if the decl isn't volatile.
8239 2018-02-15 Jason Merrill <jason@redhat.com>
8241 PR c++/84314 - ICE with templates and fastcall attribute.
8242 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8243 TYPE_CANONICAL on an existing type.
8245 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8247 PR tree-optimization/84383
8248 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8249 dstoff nor call operand_equal_p if dstbase is NULL.
8251 PR tree-optimization/84334
8252 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8253 also a CONSTANT_CLASS_P, punt.
8255 2018-02-14 Jim Wilson <jimw@sifive.com>
8257 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8258 first SMALL_OPERAND check. New local min_second_step. Move assert
8259 to where locals are set. Add TARGET_RVC support.
8260 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8262 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8264 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8266 2018-02-14 Martin Sebor <msebor@redhat.com>
8268 PR tree-optimization/83698
8269 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8270 arrays constrain the offset range to their bounds.
8271 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8272 (builtin_access::overlap): Avoid setting the size of overlap if it's
8274 (maybe_diag_overlap): Also consider arrays when deciding what values
8275 of offsets to include in diagnostics.
8277 2018-02-14 Martin Sebor <msebor@redhat.com>
8280 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8281 that correspond to the kind of a declaration.
8283 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8286 * config/pa/pa.md: Load address of PIC label using the linkage table
8287 if the label is nonlocal.
8289 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8291 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8292 warning message if user requests -maltivec=be.
8293 * doc/invoke.texi: Document deprecation of -maltivec=be.
8295 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8298 * config/rs6000/rs6000-c.c: Update definitions for
8299 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8300 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8302 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8305 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8306 add _get_ssp intrinsics. Remove argument from
8307 __builtin_ia32_rdssp[d|q].
8308 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8309 * config/i386/i386-builtin.def: Remove argument from
8310 __builtin_ia32_rdssp[d|q].
8311 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8312 ix86_expand_special_args_builtin for _rdssp[d|q].
8313 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8314 Clear register before usage.
8315 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8316 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8318 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8320 PR tree-optimization/84357
8321 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8322 operand 1 of an ARRAY_REF too.
8324 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8327 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8328 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8330 (set_of_reg): New struct.
8331 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8332 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8333 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8335 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8336 Split into bitclr, bitset, bitinvert patterns if appropriate.
8337 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8338 use rx_fuse_in_memory_bitop.
8339 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8340 to named insn, correct maximum insn length.
8342 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8345 * machmode.def: Define a complex mode for PARTIAL_INT.
8346 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8348 * doc/rtl.texi: Document CSPImode.
8349 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8351 (msp430_hard_regno_nregs_with_padding): Likewise.
8353 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8356 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8358 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8360 PR rtl-optimization/84169
8361 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8362 we generated a parallel as new i3 and we split that to new i2 and i3
8363 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8364 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8365 those to i2, not i1. Partially rewrite this scan code.
8367 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8370 * stor-layout.c (place_field): For variable length fields, adjust
8371 offset_align afterwards not just based on the field's alignment,
8372 but also on the size.
8375 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8376 of exps and logs in the use_exp2 case.
8378 2018-02-13 Jeff Law <law@redhat.com>
8380 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8383 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8386 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8390 * common.opt (gas-loc-support, gas-locview-support): New.
8391 (ginline-points, ginternal-reset-location-views): New.
8392 * doc/invoke.texi: Document them. Use @itemx where intended.
8393 (gvariable-location-views): Adjust.
8394 * target.def (reset_location_view): New.
8395 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8396 (TARGET_RESET_LOCATION_VIEW): New.
8397 * doc/tm.texi: Rebuilt.
8398 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8399 (dwarf2out_default_as_locview_support): New.
8400 (output_asm_line_debug_info): Use option variables.
8401 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8402 (output_loc_list): Likewise.
8403 (add_high_low_attributes): Check option variables.
8404 Don't output entry view attribute in strict mode.
8405 (gen_inlined_subroutine_die): Check option variables.
8406 (dwarf2out_inline_entry): Likewise.
8407 (init_sections_and_labels): Likewise.
8408 (dwarf2out_early_finish): Likewise.
8409 (maybe_reset_location_view): New, from...
8410 (dwarf2out_var_location): ... here. Call it.
8411 * debug.h (dwarf2out_default_as_loc_support): Declare.
8412 (dwarf2out_default_as_locview_support): Declare.
8413 * hooks.c (hook_int_rtx_insn_0): New.
8414 * hooks.h (hook_int_rtx_insn_0): Declare.
8415 * toplev.c (process_options): Take -gas-loc-support and
8416 -gas-locview-support from dwarf2out. Enable
8417 -gvariable-location-views by default only with locview
8418 assembler support. Enable -ginternal-reset-location-views by
8419 default only if the target defines the corresponding hook.
8420 Enable -ginline-points by default if location views are
8421 enabled; force it disabled if statement frontiers are
8423 * tree-inline.c (expand_call_inline): Check option variables.
8424 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8426 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8428 PR tree-optimization/84321
8429 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8430 handling. Also check whether the anti-range contains any values
8431 that satisfy the mask; switch to a VR_RANGE if not.
8433 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8436 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8438 2018-02-13 Martin Jambor <mjambor@suse.cz>
8441 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8442 of call statements, also set location of a load to a temporary.
8444 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8446 * config/rl78/rl78.c (add_vector_labels): New function.
8447 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8448 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8449 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8450 which checks that no arguments are passed.
8451 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8452 * doc/extend.texi: Documentation for the new attribute.
8454 2018-02-13 Andreas Schwab <schwab@suse.de>
8456 * config/riscv/linux.h (CPP_SPEC): Define.
8458 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8461 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8462 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8463 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8464 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8465 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8466 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8467 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8468 temporarily for AES and PCLMUL builtins.
8470 PR tree-optimization/84339
8471 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8472 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8476 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8477 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8478 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8480 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8482 * omp-simd-clone.h: New file.
8483 * omp-simd-clone.c: Include omp-simd-clone.h.
8484 (expand_simd_clones): No longer static.
8485 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8486 cgraph.h and omp-simd-clone.h.
8487 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8488 (vect_recog_widen_shift_pattern): Formatting fix.
8489 (vect_pattern_recog_1): Don't check optab for calls.
8492 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8493 operands[2] into a REG before using gen_lowpart on it.
8495 2018-02-12 Jeff Law <law@redhat.com>
8498 * config/sh/sh.c (find_barrier): Consider a sibling call
8501 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8502 successfully back substituting a reg.
8504 2018-02-12 Richard Biener <rguenther@suse.de>
8506 PR tree-optimization/84037
8507 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8508 parameter, move visited init to caller.
8509 (vect_slp_analyze_operations): Separate cost from validity
8510 check, initialize visited once for all instances.
8511 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8513 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8515 (vect_model_promotion_demotion_cost): Likewise.
8516 (vectorizable_bswap): Guard cost modeling with !slp_node
8517 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8519 (vectorizable_call): Likewise.
8520 (vectorizable_conversion): Likewise.
8521 (vectorizable_assignment): Likewise.
8522 (vectorizable_shift): Likewise.
8523 (vectorizable_operation): Likewise.
8524 (vectorizable_store): Likewise.
8525 (vectorizable_load): Likewise.
8526 (vectorizable_condition): Likewise.
8527 (vectorizable_comparison): Likewise.
8529 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8532 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8533 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8534 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8536 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8539 * params.def (inline-min-speedup): Increase from 8 to 15.
8540 (max-inline-insns-auto): Decrease from 40 to 30.
8541 * ipa-split.c (consider_split): Add some buffer for function to
8542 be considered inlining candidate.
8543 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8546 2018-02-12 Richard Biener <rguenther@suse.de>
8548 PR tree-optimization/84037
8549 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8550 matched stmts if we cannot swap the non-matched ones.
8552 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8554 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8555 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8556 _mm_maskz_scalef_round_ss): New intrinsics.
8557 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8558 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8559 __builtin_ia32_scalefss_round): Remove.
8560 (__builtin_ia32_scalefsd_mask_round,
8561 __builtin_ia32_scalefss_mask_round): New intrinsics.
8562 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8563 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8564 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8565 "<round_constraint>")): Changed to ...
8566 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8567 "<round_scalar_constraint>")): ... this.
8568 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8569 %0, %1, %2<round_op3>}"): Changed to ...
8570 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8571 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8572 %2<round_scalar_mask_op3>}"): ... this.
8573 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8575 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8577 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8578 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8579 (_mm_maskz_sqrt_round_ss): New intrinsics.
8580 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8581 (__builtin_ia32_sqrtsd_mask_round)
8582 (__builtin_ia32_sqrtss_mask_round): New builtins.
8583 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8584 (__builtin_ia32_sqrtss_round): Remove.
8585 (__builtin_ia32_sqrtsd_mask_round)
8586 (__builtin_ia32_sqrtss_mask_round): New builtins.
8587 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8588 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8589 ((match_operand:VF_128 1 "vector_operand"
8590 "xBm,<round_constraint>")): Changed to ...
8591 ((match_operand:VF_128 1 "vector_operand"
8592 "xBm,<round_scalar_constraint>")): ... this.
8593 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8594 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8595 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8596 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8597 %<iptr>1<round_scalar_mask_op3>}): ... this.
8598 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8599 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8601 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8604 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8605 Cast vec_cmpeq result to correct type.
8606 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8607 Cast vec_cmpgt result to correct type.
8609 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8611 * final.c (final_scan_insn_1): Renamed from...
8612 (final_scan_insn): ... this. New wrapper, to recover
8613 seen from the outermost call in recursive ones.
8614 * config/sparc/sparc.c (output_return): Drop seen from call.
8615 (output_sibcall): Likewise.
8616 * config/visium/visium.c (output_branch): Likewise.
8618 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8620 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8623 2018-02-10 Alan Modra <amodra@gmail.com>
8626 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8627 Specify LR as an input.
8629 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8632 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8633 remove_member_access_dummy_vars): New functions.
8634 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8635 lower_omp_1, execute_lower_omp): Use them.
8637 PR rtl-optimization/84308
8638 * shrink-wrap.c (spread_components): Release todo vector.
8640 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8642 PR rtl-optimization/57193
8643 * ira-color.c (struct allocno_color_data): Add member
8644 conflict_allocno_hard_prefs.
8645 (update_conflict_allocno_hard_prefs): New.
8646 (bucket_allocno_compare_func): Add a preference based on
8647 conflict_allocno_hard_prefs.
8648 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8649 (color_allocnos): Remove a dead code. Initiate
8650 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8652 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8655 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8656 constraint from =wa to wa. Avoid a subreg on the output operand,
8657 instead use a pseudo and subreg it in a move.
8658 (p9_xxbrd_<mode>): Changed to ...
8659 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8660 (p9_xxbrd_v2df): New expander.
8661 (p9_xxbrw_<mode>): Changed to ...
8662 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8663 (p9_xxbrw_v4sf): New expander.
8665 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8667 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8669 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8672 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8673 multiply in 32-bit mode.
8674 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8675 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8678 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8680 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8681 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8682 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8683 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8685 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8688 * dwarf2out.c (is_trivial_indirect_ref): New function.
8689 (dwarf2out_late_global_decl): Do not generate a location
8690 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8691 and that are not defined in the current unit.
8693 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8695 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8696 instead of a libcall for UNORDERED.
8698 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8701 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8702 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8704 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8707 * config/s390/s390.c (s390_set_current_function): Invoke
8708 s390_indirect_branch_settings also if fndecl didn't change.
8710 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8712 * config/rs6000/rs6000.md (blockage): Set length to zero.
8714 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8716 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8718 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8721 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8722 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8726 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8727 PARALLEL incoming that failed vt_get_decl_and_offset check.
8730 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8731 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8733 (get_variable_section): For decls in named .bss* sections pass true as
8734 second argument to bss_initializer_p.
8736 2018-02-09 Marek Polacek <polacek@redhat.com>
8737 Jakub Jelinek <jakub@redhat.com>
8740 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8741 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8742 Sync some changes from cxx_fold_indirect_ref.
8744 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8746 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8748 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8749 (BLOCK_INLINE_ENTRY_LABEL): New.
8750 (dwarf2out_var_location): Disregard inline entry markers.
8751 (inline_entry_data): New struct.
8752 (inline_entry_data_hasher): New hashtable type.
8753 (inline_entry_data_hasher::hash): New.
8754 (inline_entry_data_hasher::equal): New.
8755 (inline_entry_data_table): New variable.
8756 (add_high_low_attributes): Add DW_AT_entry_pc and
8757 DW_AT_GNU_entry_view attributes if a pending entry is found
8758 in inline_entry_data_table. Add old entry_pc attribute only
8759 if debug nonbinding markers are disabled.
8760 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8761 markers are enabled.
8762 (block_within_block_p, dwarf2out_inline_entry): New.
8763 (dwarf2out_finish): Check that no entries remained in
8764 inline_entry_data_table.
8765 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8766 (final_scan_insn, notice_source_line): Likewise.
8767 (rest_of_clean_state): Skip inline entry markers.
8768 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8770 * gimple.c (gimple_build_debug_inline_entry): New.
8771 * gimple.h (enum gimple_debug_subcode): Add
8772 GIMPLE_DEBUG_INLINE_ENTRY.
8773 (gimple_build_debug_inline_entry): Declare.
8774 (gimple_debug_inline_entry_p): New.
8775 (gimple_debug_nonbind_marker_p): Adjust.
8776 * insn-notes.def (INLINE_ENTRY): New.
8777 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8778 inline entry marker notes.
8779 (print_insn): Likewise.
8780 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8781 (INSN_DEBUG_MARKER_KIND): Likewise.
8782 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8783 * tree-inline.c (expand_call_inline): Build and insert
8784 debug_inline_entry stmt.
8785 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8786 inline entry blocks early, if nonbind markers are enabled.
8787 (dump_scope_block): Dump fragment info.
8788 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8789 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8790 (gimple_build_debug_inline_entry): New.
8791 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8792 Enable/disable inline entry points too.
8793 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8794 (DEBUG_INSN): Describe inline entry markers.
8796 * common.opt (gvariable-location-views): New.
8797 (gvariable-location-views=incompat5): New.
8798 * config.in: Rebuilt.
8799 * configure: Rebuilt.
8800 * configure.ac: Test assembler for view support.
8801 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8802 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8803 * dwarf2out.c (var_loc_view): New typedef.
8804 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8805 (dwarf2out_locviews_in_attribute): New.
8806 (dwarf2out_locviews_in_loclist): New.
8807 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8808 (enum dw_line_info_opcode): Add LI_adv_address.
8809 (struct dw_line_info_table): Add view.
8810 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8811 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8812 (zero_view_p): New variable.
8813 (ZERO_VIEW_P): New macro.
8814 (output_asm_line_debug_info): New.
8815 (struct var_loc_node): Add view.
8816 (add_AT_view_list, AT_loc_list): New.
8817 (add_var_loc_to_decl): Add view param. Test it against last.
8818 (new_loc_list): Add view params. Record them.
8819 (AT_loc_list_ptr): Handle loc and view lists.
8820 (view_list_to_loc_list_val_node): New.
8821 (print_dw_val): Handle dw_val_class_view_list.
8822 (size_of_die): Likewise.
8823 (value_format): Likewise.
8824 (loc_list_has_views): New.
8825 (gen_llsym): Set vl_symbol too.
8826 (maybe_gen_llsym, skip_loc_list_entry): New.
8827 (dwarf2out_maybe_output_loclist_view_pair): New.
8828 (output_loc_list): Output view list or entries too.
8829 (output_view_list_offset): New.
8830 (output_die): Handle dw_val_class_view_list.
8831 (output_dwarf_version): New.
8832 (output_compilation_unit_header): Use it.
8833 (output_skeleton_debug_sections): Likewise.
8834 (output_rnglists, output_line_info): Likewise.
8835 (output_pubnames, output_aranges): Update version comments.
8836 (output_one_line_info_table): Output view numbers in asm comments.
8837 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8838 Call maybe_gen_llsym.
8839 (loc_list_from_tree_1): Adjust.
8840 (add_AT_location_description): Create view list attribute if
8841 needed, check it's absent otherwise.
8842 (convert_cfa_to_fb_loc_list): Adjust.
8843 (maybe_emit_file): Call output_asm_line_debug_info for test.
8844 (dwarf2out_var_location): Reset views as needed. Precompute
8845 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8846 attribute. Set view.
8847 (new_line_info_table): Reset next view.
8848 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8849 (dwarf2out_source_line): Likewise. Output view resets and labels to
8850 the assembler, or select appropriate line info opcodes.
8851 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8852 (optimize_string_length): Catch it. Adjust.
8853 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8854 dw_val_class_view_list, and remove it if no longer needed.
8855 (hash_loc_list): Hash view numbers.
8856 (loc_list_hasher::equal): Compare them.
8857 (optimize_location_lists): Check whether a view list symbol is
8858 needed, and whether the locview attribute is present, and
8859 whether they match. Remove the locview attribute if no longer
8861 (index_location_lists): Call skip_loc_list_entry for test.
8862 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8863 Use output_dwarf_version.
8864 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8865 (struct dw_val_node): Add val_view_list.
8866 * final.c (SEEN_NEXT_VIEW): New.
8867 (set_next_view_needed): New.
8868 (clear_next_view_needed): New.
8869 (maybe_output_next_view): New.
8870 (final_start_function): Rename to...
8871 (final_start_function_1): ... this. Take pointer to FIRST,
8872 add SEEN parameter. Emit param bindings in the initial view.
8873 (final_start_function): Reintroduce SEEN-less interface.
8874 (final): Rename to...
8875 (final_1): ... this. Take SEEN parameter. Output final pending
8876 next view at the end.
8877 (final): Reintroduce seen-less interface.
8878 (final_scan_insn): Output pending next view before switching
8879 sections or ending a block. Mark the next view as needed when
8880 outputting variable locations. Notify debug backend of section
8881 changes, and of location view changes.
8882 (rest_of_handle_final): Adjust.
8883 * toplev.c (process_options): Autodetect value for debug variable
8884 location views option. Warn on incompat5 without -gdwarf-5.
8885 * doc/invoke.texi (gvariable-location-views): New.
8886 (gvariable-location-views=incompat5): New.
8887 (gno-variable-location-views): New.
8889 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8891 PR tree-optimization/84136
8892 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8893 that the result of find_edge is non-NULL.
8895 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8898 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8899 storing integer register in SImode. Fix cost of 256 and 512
8900 byte aligned SSE register store.
8902 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8904 * config/i386/i386.c (ix86_multiplication_cost): Fix
8905 multiplication cost for TARGET_AVX512DQ.
8907 2018-02-08 Marek Polacek <polacek@redhat.com>
8909 PR tree-optimization/84238
8910 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8913 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8915 PR tree-optimization/84265
8916 * tree-vect-stmts.c (vectorizable_store): Don't treat
8917 VMAT_CONTIGUOUS accesses as grouped.
8918 (vectorizable_load): Likewise.
8920 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8922 PR tree-optimization/81635
8923 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8924 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8925 (test_round_for_mask): New functions.
8926 (wide_int_cc_tests): Call test_round_for_mask.
8927 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8928 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8929 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8930 range returned by get_range_info.
8932 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8935 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8936 * symtab.c: Include builtins.h
8937 (symtab_node::output_to_lto_symbol_table_p): Move here
8938 from lto-streamer-out.c:output_symbol_p.
8939 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8940 (output_symbol_p): Move all logic to symtab.c
8941 (produce_symtab): Update.
8943 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8945 * config/s390/s390-opts.h (enum indirect_branch): Define.
8946 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8947 (s390_indirect_branch_via_thunk)
8948 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8949 (enum s390_indirect_branch_type): Define.
8950 * config/s390/s390.c (struct s390_frame_layout, struct
8951 machine_function): Remove.
8952 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8953 (indirect_branch_table_label_no, indirect_branch_table_name):
8955 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8956 (enum s390_indirect_branch_option): Define.
8957 (s390_return_addr_from_memory): New function.
8958 (s390_handle_string_attribute): New function.
8959 (s390_attribute_table): Add new attribute handler.
8960 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8961 (s390_indirect_branch_via_thunk): New function.
8962 (s390_indirect_branch_via_inline_thunk): New function.
8963 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8964 sibling call optimization for non z10 compiles.
8965 (s390_emit_call): Force indirect branch target to be a single
8966 register. Add r1 clobber for non-z10 compiles.
8967 (s390_emit_epilogue): Emit return jump via return_use expander.
8968 (s390_reorg): Handle JUMP_INSNs as execute targets.
8969 (s390_option_override_internal): Perform validity checks for the
8970 new command line options.
8971 (s390_indirect_branch_attrvalue): New function.
8972 (s390_indirect_branch_settings): New function.
8973 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8974 (s390_output_indirect_thunk_function): New function.
8975 (s390_code_end): Implement target hook.
8976 (s390_case_values_threshold): Implement target hook.
8977 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8979 * config/s390/s390.h (struct s390_frame_layout)
8980 (struct machine_function): Move here from s390.c.
8981 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8982 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8983 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8984 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8985 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8986 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8987 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8988 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8989 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8990 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8991 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8992 (mnemonic attribute): Add values which aren't recognized
8994 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8995 pattern for branch conversion. Fix mnemonic attribute.
8996 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8997 indirect branch via thunk if requested.
8998 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8999 ("*indirect_jump"): Disable for branch conversion using out of
9001 ("indirect_jump_via_thunk<mode>_z10")
9002 ("indirect_jump_via_thunk<mode>")
9003 ("indirect_jump_via_inlinethunk<mode>_z10")
9004 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9005 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9006 ("casesi_jump_via_inlinethunk<mode>_z10")
9007 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9008 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9009 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9010 ("*indirect2_jump"): Disable for branch conversion.
9011 ("casesi_jump"): Turn into expander and expand patterns for branch
9013 ("return_use"): New expander.
9014 ("*return"): Emit return via thunk and rename it to ...
9015 ("*return<mode>"): ... this one.
9016 * config/s390/s390.opt: Add new options and and enum for the
9019 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9021 * lra-constraints.c (match_reload): Unconditionally use
9022 gen_lowpart_SUBREG, rather than selecting between that
9023 and equivalent gen_rtx_SUBREG code.
9025 2018-02-08 Richard Biener <rguenther@suse.de>
9027 PR tree-optimization/84233
9028 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9029 changed flag instead of boguously re-using phi_inserted.
9031 2018-02-08 Martin Jambor <mjambor@suse.cz>
9033 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9034 static local variables.
9036 2018-02-08 Richard Biener <rguenther@suse.de>
9038 PR tree-optimization/84278
9039 * tree-vect-stmts.c (vectorizable_store): When looking for
9040 smaller vector types to perform grouped strided loads/stores
9041 make sure the mode is supported by the target.
9042 (vectorizable_load): Likewise.
9044 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9046 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9047 Increase LDP/STP opportunities by adding adjacent callee-saves.
9049 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9051 PR rtl-optimization/84068
9052 PR rtl-optimization/83459
9053 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9055 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9057 PR tree-optimization/84224
9058 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9059 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9062 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9065 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9066 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9067 count, remove one USE.
9069 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9071 * doc/install.texi (Configuration): Document the
9072 --with-long-double-format={ibm,ieee} PowerPC configuration
9076 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9077 Convert from define_expand to be define_insn_and_split. Rework
9078 float/double/_Float128 conversions to QI/HI/SImode to work with
9079 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9080 conversions to QI/HImode types did a store and then a load to
9081 truncate the value. For conversions to VSX registers, don't split
9082 the insn, instead emit the code directly. Use the code iterator
9083 any_fix to combine signed and unsigned conversions.
9084 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9085 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9086 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9087 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9088 (fix_<mode>di2_hw): Likewise.
9089 (fixuns_<mode>di2_hw): Likewise.
9090 (fix_<mode>si2_hw): Likewise.
9091 (fixuns_<mode>si2_hw): Likewise.
9092 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9093 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9094 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9095 fix<uns>_trunc<SFDF:mode>si2_p8.
9096 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9097 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9098 (fix<uns>_<mode>_mem): Likewise.
9099 (fctiw<u>z_<mode>_mem): Likewise.
9100 (fix<uns>_<mode>_mem): Likewise.
9101 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9102 the register allocator from doing a direct move to the GPRs to do
9103 a store, and instead use the ISA 3.0 store byte/half-word from
9104 vector register instruction. For IEEE 128-bit floating point,
9105 also optimize stores of 32-bit ints.
9106 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9108 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9110 * genextract.c (push_pathstr_operand): New function to support
9112 (walk_rtx): Call push_pathstr_operand.
9113 (print_path): Support [a-zA-Z].
9115 2018-02-07 Richard Biener <rguenther@suse.de>
9117 PR tree-optimization/84037
9118 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9119 (cse_and_gimplify_to_preheader): Declare.
9120 (vect_get_place_in_interleaving_chain): Likewise.
9121 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9123 (_loop_vec_info::~_loop_vec_info): Delete it.
9124 (cse_and_gimplify_to_preheader): New function.
9125 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9126 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9127 (vectorizable_load): Likewise. For grouped stores always base
9128 the IV on the first element.
9129 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9130 condition before gimplifying.
9132 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9134 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9135 *DIV_EXPR and *MOD_EXPR.
9137 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9140 * config/i386/i386.c (ix86_option_override_internal): Mask out
9141 the CF_SET bit when checking -fcf-protection.
9143 2018-02-07 Tom de Vries <tom@codesourcery.com>
9146 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9149 2018-02-07 Richard Biener <rguenther@suse.de>
9151 PR tree-optimization/84204
9152 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9155 PR tree-optimization/84205
9156 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9157 special-case isl_ast_op_zdiv_r.
9159 PR tree-optimization/84223
9160 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9161 Only add conditions from within the region.
9162 (gather_bbs::after_dom_children): Adjust.
9164 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9167 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9168 * config/avr/avr.md: Only post-reload split REG-REG moves if
9169 either register is GENERAL_REG_P.
9171 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9173 PR tree-optimization/84235
9174 * tree-ssa-scopedtables.c
9175 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9176 if the subtraction is performed in floating point type where NaNs are
9177 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9178 build 1. Formatting fix.
9180 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9183 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9184 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9185 and skip it regardless of bb boundaries. Use CALL_P macro,
9186 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9187 unnecessarily, formatting fix.
9189 2018-02-06 Michael Collison <michael.collison@arm.com>
9191 * config/arm/thumb2.md:
9192 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9193 (*thumb_mov_notscc): Ditto.
9195 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9198 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9199 unsigned_fix, not "s".
9201 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9203 * configure.ac (gcc_fn_eh_frame_ro): New function.
9204 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9205 correct .eh_frame permissions.
9206 * configure: Regenerate.
9208 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9210 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9213 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9215 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9216 Display warning message for -mno-speculate-indirect-jumps.
9218 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9220 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9222 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9224 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9226 PR tree-optimization/84225
9227 * tree-eh.c (find_trapping_overflow): Only call
9228 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9230 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9233 * config/i386/i386.c: Reimplement the check of possible options
9234 -mibt/-mshstk conbination. Change error messages.
9235 * doc/invoke.texi: Fix a typo: remove extra '='.
9237 2018-02-06 Marek Polacek <polacek@redhat.com>
9239 PR tree-optimization/84228
9240 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9242 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9245 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9246 emitted arch directives.
9247 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9248 __ARM_FEATURE_COPROC before changing architectures.
9250 2018-02-06 Richard Biener <rguenther@suse.de>
9252 * config/i386/i386.c (print_reg): Fix typo.
9253 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9255 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9257 * configure: Regenerate.
9259 2018-02-05 Martin Sebor <msebor@redhat.com>
9261 PR tree-optimization/83369
9262 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9265 2018-02-05 Martin Liska <mliska@suse.cz>
9267 * doc/invoke.texi: Cherry-pick upstream r323995.
9269 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9271 * ira.c (ira_init_register_move_cost): Adjust comment.
9273 2018-02-05 Martin Liska <mliska@suse.cz>
9275 PR gcov-profile/84137
9276 * doc/gcov.texi: Fix typo in documentation.
9278 2018-02-05 Martin Liska <mliska@suse.cz>
9280 PR gcov-profile/83879
9281 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9282 using dlopen functionality.
9284 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9286 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9287 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9288 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9289 _mm_maskz_range_round_ss): New intrinsics.
9290 (__builtin_ia32_rangesd128_round)
9291 (__builtin_ia32_rangess128_round): Remove.
9292 (__builtin_ia32_rangesd128_mask_round,
9293 __builtin_ia32_rangess128_mask_round): New builtins.
9294 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9295 __builtin_ia32_rangess128_round): Remove.
9296 (__builtin_ia32_rangesd128_mask_round,
9297 __builtin_ia32_rangess128_mask_round): New builtins.
9298 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9299 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9300 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9301 "<round_saeonly_constraint>")): Changed to ...
9302 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9303 "<round_saeonly_scalar_constraint>")): ... this.
9304 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9305 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9306 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9307 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9308 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9310 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9312 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9314 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9315 Remove all values except native, 8540 and 8548.
9317 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9319 * config/i386/i386.c (ix86_output_function_return): Pass
9320 INVALID_REGNUM, instead of -1, as invalid register number to
9321 indirect_thunk_name and output_indirect_thunk.
9323 2018-02-02 Julia Koval <julia.koval@intel.com>
9325 * config.gcc: Add -march=icelake.
9326 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9327 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9328 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9329 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9330 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9331 (processor_target_table): Add icelake.
9332 (ix86_option_override_internal): Handle new PTAs.
9333 (get_builtin_code_for_version): Handle icelake.
9334 (M_INTEL_COREI7_ICELAKE): New.
9335 (fold_builtin_cpu): Handle icelake.
9336 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9337 * doc/invoke.texi: Add -march=icelake.
9339 2018-02-02 Julia Koval <julia.koval@intel.com>
9341 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9342 to wide_int_bitmask.
9343 * wide-int-bitmask.h: New.
9345 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9348 * config/i386/i386.md: Replace Pmode with word_mode in
9349 builtin_setjmp_setup and builtin_longjmp to support x32.
9351 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9355 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9357 (rs6000_supported_cpu_names): New static variable.
9358 (linux_cpu_translation_table): Likewise.
9359 (elf_platform) <cpu>: Define new static variable and use it.
9360 Translate kernel AT_PLATFORM name to canonical name if needed.
9361 Error if platform name is unknown.
9363 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9366 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9368 2018-02-01 Jeff Law <law@redhat.com>
9371 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9372 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9373 the scratch if RELEASE_VIA_POP is false.
9374 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9375 If we have to save a temporary register, decrement SIZE appropriately.
9376 Pass new arguments to release_scratch_register_on_entry.
9377 (ix86_adjust_stack_and_probe): Likewise.
9378 (ix86_emit_probe_stack_range): Pass new arguments to
9379 release_scratch_register_on_entry.
9381 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9383 PR rtl-optimization/84157
9384 * combine.c (change_zero_ext): Use REG_P predicate in
9385 front of HARD_REGISTER_P predicate.
9387 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9389 * config/avr/avr.c (avr_option_override): Move disabling of
9390 -fdelete-null-pointer-checks to...
9391 * common/config/avr/avr-common.c (avr_option_optimization_table):
9394 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9396 PR tree-optimization/81635
9397 * tree-data-ref.c (split_constant_offset_1): For types that
9398 wrap on overflow, try to use range info to prove that wrapping
9401 2018-02-01 Renlin Li <renlin.li@arm.com>
9404 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9406 (aarch64_register_move_cost): Likewise.
9407 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9409 (REG_CLASS_NAMES): Likewise.
9410 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9411 TAILCALL_ADDR_REGS. Remove IP registers.
9412 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9414 2018-02-01 Richard Biener <rguenther@suse.de>
9416 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9417 for specifying RPO order and allow NULL for that.
9418 * domwalk.c (dom_walker::dom_walker): Likewise.
9419 (dom_walker::walk): Handle NULL RPO order.
9420 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9422 (rewrite_update_dom_walker): Likewise.
9423 (mark_def_dom_walker): Likewise.
9425 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9427 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9428 (aarch64_maybe_expand_sve_subreg_move): Declare.
9429 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9430 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9432 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9433 that are semantically a reverse operation.
9434 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9435 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9436 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9438 (aarch64_can_change_mode_class): For big-endian, forbid changes
9439 between two SVE modes if they have different element sizes.
9441 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9443 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9444 the TImode handling for big-endian targets.
9446 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9448 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9449 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9451 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9452 Remove BSWAP handing for big-endian targets and use the form of
9453 LD1RQ appropariate for the mode.
9455 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9457 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9458 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9461 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9464 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9465 check for operands that need to go through aarch64_sve_reload_be.
9467 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9469 PR tree-optimization/81661
9470 PR tree-optimization/84117
9471 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9472 * tree-eh.c: Include gimplify.h.
9473 (find_trapping_overflow, replace_trapping_overflow,
9474 rewrite_to_non_trapping_overflow): New functions.
9475 * tree-vect-loop.c: Include tree-eh.h.
9476 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9477 * tree-data-ref.c: Include tree-eh.h.
9478 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9480 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9482 PR rtl-optimization/84123
9483 * combine.c (change_zero_ext): Check if hard register satisfies
9484 can_change_dest_mode before calling gen_lowpart_SUBREG.
9486 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9489 * ira.c (ira_init_register_move_cost): Remove assert.
9491 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9493 PR rtl-optimization/84071
9494 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9495 * doc/tm.texi: Regenerate.
9497 2018-01-31 Richard Biener <rguenther@suse.de>
9499 PR tree-optimization/84132
9500 * tree-data-ref.c (analyze_miv_subscript): Properly
9501 check whether evolution_function_is_affine_multivariate_p
9502 before calling gcd_of_steps_may_divide_p.
9504 2018-01-31 Julia Koval <julia.koval@intel.com>
9507 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9508 * config/i386/i386.md (rdpid_rex64) New.
9509 (rdpid): Make 32bit only.
9511 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9514 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9515 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9517 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9520 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9522 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9524 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9526 PR rtl-optimization/84071
9527 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9528 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9530 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9532 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9533 (arc_attribute_table): Add 'aux' attribute.
9534 (arc_in_small_data_p): Consider aux like variables.
9535 (arc_is_aux_reg_p): New function.
9536 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9537 (arc_get_aux_arg): New function.
9538 (prepare_move_operands): Handle aux-register access.
9539 (arc_handle_aux_attribute): New function.
9540 * doc/extend.texi (ARC Variable attributes): Add subsection.
9542 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9544 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9545 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9546 (arc_attribute_table): Add 'uncached' attribute.
9547 (arc_print_operand): Print '.di' flag for uncached memory
9549 (arc_in_small_data_p): Do not consider for small data the uncached
9551 (arc_is_uncached_mem_p): New function.
9552 * config/arc/predicates.md (compact_store_memory_operand): Check
9553 for uncached memory accesses.
9554 (nonvol_nonimm_operand): Likewise.
9555 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9557 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9560 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9561 falign-loops=): Add Optimization flag.
9563 2018-01-30 Jeff Law <law@redhat.com>
9566 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9567 INT_REGISTERS_SAVED. Check it prior to calling
9568 get_scratch_register_on_entry.
9569 (ix86_adjust_stack_and_probe): Similarly.
9570 (ix86_emit_probe_stack_range): Similarly.
9571 (ix86_expand_prologue): Corresponding changes.
9573 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9576 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9577 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9579 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9582 * lra-constraints.c (curr_insn_transform): Process AND in the
9585 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9587 PR rtl-optimization/83986
9588 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9589 dependence against last_pending_memory_flush in addition to
9592 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9594 PR tree-optimization/81611
9595 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9598 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9601 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9604 2018-01-30 Richard Biener <rguenther@suse.de>
9605 Jakub Jelinek <jakub@redhat.com>
9607 PR tree-optimization/84111
9608 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9609 inner loops added during recursion, as they don't have up-to-date
9612 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9615 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9616 (can_inline_edge_by_limits_p): ... here.
9617 (can_early_inline_edge_p, check_callers,
9618 update_caller_keys, update_callee_keys, recursive_inlining,
9619 add_new_edges_to_heap, speculation_useful_p,
9620 inline_small_functions,
9621 inline_small_functions, flatten_function,
9622 inline_to_all_callers_1): Update.
9624 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9626 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9629 2018-01-30 Richard Biener <rguenther@suse.de>
9631 PR tree-optimization/83008
9632 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9633 invariant and constant vector uses in stmts when they need
9636 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9639 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9640 * configure: Regenerate.
9642 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9644 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9646 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9647 Use gen_rtx_REG rather than gen_lowpart.
9649 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9651 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9652 rather than 0 when creating partial subregs.
9654 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9656 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9659 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9662 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9663 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9664 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9665 flags. This restores the settings used before the 2017-07-24.
9666 Turning off pre increment/decrement/modify allows IVOPTS to
9667 optimize DF/SF loops where the index is an int.
9669 2018-01-29 Richard Biener <rguenther@suse.de>
9670 Kelvin Nilsen <kelvin@gcc.gnu.org>
9673 * tree-vect-stmts.c (vectorizable_call): Don't call
9674 targetm.vectorize_builtin_md_vectorized_function if callee is
9677 2018-01-22 Carl Love <cel@us.ibm.com>
9679 * doc/extend.tex: Fix typo in second arg in
9680 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9682 2018-01-29 Richard Biener <rguenther@suse.de>
9684 PR tree-optimization/84086
9685 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9686 (flush_ssaname_freelist): When SSA names were released reset
9687 the SCEV hash table.
9689 2018-01-29 Richard Biener <rguenther@suse.de>
9691 PR tree-optimization/84057
9692 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9693 removed paths when removing edges.
9695 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9697 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9698 -mfunction-return=@var{choice}.
9700 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9703 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9705 (layout::print_source_line): Likewise.
9706 (test_get_line_width_without_trailing_whitespace): Add test cases.
9708 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9711 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9714 2018-01-26 Jim Wilson <jimw@sifive.com>
9716 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9718 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9721 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9723 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9724 and CMP + SUB-immediate -> SUBS.
9726 2018-01-26 Martin Sebor <msebor@redhat.com>
9728 PR tree-optimization/83896
9729 * tree-ssa-strlen.c (get_string_len): Rename...
9730 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9731 Avoid assuming length is constant.
9732 (handle_char_store): Use HOST_WIDE_INT for string length.
9734 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9737 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9738 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9740 2018-01-26 Richard Biener <rguenther@suse.de>
9742 PR rtl-optimization/84003
9743 * dse.c (record_store): Only record redundant stores when
9744 the earlier store aliases at least all accesses the later one does.
9746 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9748 PR rtl-optimization/83985
9749 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9750 REG_CFA_RESTORE insns.
9751 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9752 REG_CFA_RESTORE insns here.
9755 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9756 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9758 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9760 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9761 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9762 (arc_init): Likewise.
9763 (arc_override_options): Likewise.
9764 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9766 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9768 * config/arc/arc.h (TARGET_DBNZ): Define.
9769 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9770 properly set the tune attribute.
9771 (dbnz): Use TARGET_DBNZ guard.
9772 * config/arc/arc.opt (mtune): Add core3 option.
9774 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9776 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9777 recognize new pic like addresses.
9778 (arc_delegitimize_address): Clean up.
9780 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9782 * config/arc/arc-arches.def: Option mrf16 valid for all
9784 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9785 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9786 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9787 * config/arc/arc-tables.opt: Regenerate.
9788 * config/arc/arc.c (arc_conditional_register_usage): Handle
9789 reduced register file case.
9790 (arc_file_start): Set must have build attributes.
9791 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9793 * config/arc/arc.opt (mrf16): Add new option.
9794 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9795 * config/arc/genmultilib.awk: Handle new mrf16 option.
9796 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9797 * config/arc/t-multilib: Regenerate.
9798 * doc/invoke.texi (ARC Options): Document mrf16 option.
9800 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9802 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9803 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9804 (arc_attribute_table): Add 'secure_call' attribute.
9805 (arc_print_operand): Print secure call operand.
9806 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9808 (arc_is_secure_call_p): New function. * config/arc/arc.md
9809 (call_i): Add support for sjli instruction.
9810 (call_value_i): Likewise.
9811 * config/arc/constraints.md (Csc): New constraint.
9813 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9814 John Eric Martin <John.Martin@emmicro-us.com>
9816 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9817 * config/arc/arc.c (_arc_jli_section): New struct.
9818 (arc_jli_section): New type.
9819 (rc_jli_sections): New static variable.
9820 (arc_handle_jli_attribute): New function.
9821 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9822 (arc_file_end): New function.
9823 (TARGET_ASM_FILE_END): Define.
9824 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9825 (arc_add_jli_section): New function.
9826 (jli_call_scan): Likewise.
9827 (arc_reorg): Call jli_call_scan.
9828 (arc_output_addsi): Remove 'S' from printing asm operand.
9829 (arc_is_jli_call_p): New function.
9830 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9832 (movhi_insn): Likewise.
9833 (movsi_insn): Likewise.
9834 (movsi_set_cc_insn): Likewise.
9835 (loadqi_update): Likewise.
9836 (load_zeroextendqisi_update): Likewise.
9837 (load_signextendqisi_update): Likewise.
9838 (loadhi_update): Likewise.
9839 (load_zeroextendhisi_update): Likewise.
9840 (load_signextendhisi_update): Likewise.
9841 (loadsi_update): Likewise.
9842 (loadsf_update): Likewise.
9843 (movsicc_insn): Likewise.
9844 (bset_insn): Likewise.
9845 (bxor_insn): Likewise.
9846 (bclr_insn): Likewise.
9847 (bmsk_insn): Likewise.
9848 (bicsi3_insn): Likewise.
9849 (cmpsi_cc_c_insn): Likewise.
9850 (movsi_ne): Likewise.
9851 (movsi_cond_exec): Likewise.
9852 (clrsbsi2): Likewise.
9863 (call_i): Remove 'S' asm letter, add jli instruction.
9864 (call_value_i): Likewise.
9865 * config/arc/arc.op (mjli-always): New option.
9866 * config/arc/constraints.md (Cji): New constraint.
9867 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9869 (subsf3_fpx): Likewise.
9870 (mulsf3_fpx): Likewise.
9871 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9873 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9875 * doc/invoke.texi (ARC): Document mjli-always option.
9877 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9879 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9880 avoid addition with 0 and use incw and decw where possible.
9882 2018-01-26 Richard Biener <rguenther@suse.de>
9884 PR tree-optimization/81082
9885 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9886 association if it requires casting to unsigned.
9887 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9888 from fold_plusminus_mult_expr to catch important cases late when
9889 range info is available.
9891 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9893 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9894 * configure.ac (hidden_linkonce): New test.
9895 * configure: Regenerate.
9896 * config.in: Regenerate.
9898 2018-01-26 Julia Koval <julia.koval@intel.com>
9900 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9901 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9902 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9903 _mm_mask_bitshuffle_epi64_mask): Fix type.
9904 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9905 USI_FTYPE_V4DI_V4DI_USI): Remove.
9906 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9907 __builtin_ia32_vpshufbitqmb256_mask,
9908 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9909 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9910 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9912 2018-01-26 Alan Modra <amodra@gmail.com>
9915 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9916 UNSPEC_VBPERMQ. Sort other unspecs.
9918 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9920 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9922 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9925 * predict.c (drop_profile): Do not push/pop cfun; update also
9927 (handle_missing_profiles): Fix logic looking for zero profiles.
9929 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9932 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9933 on functions with #pragma omp declare simd or functions with simd
9935 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9936 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9937 Remove trailing \n from warning_at calls.
9939 2018-01-25 Tom de Vries <tom@codesourcery.com>
9942 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9943 for neutered workers.
9945 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9948 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9949 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9951 2018-01-24 Jeff Law <law@redhat.com>
9954 * i386.c (get_probe_interval): Move to earlier point.
9955 (ix86_compute_frame_layout): If -fstack-clash-protection and
9956 the frame is larger than the probe interval, then use pushes
9957 to save registers rather than reg->mem moves.
9958 (ix86_expand_prologue): Remove conditional for int_registers_saved
9961 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9964 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9965 min/max for never referenced object.
9967 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9970 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9972 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9973 attributes from DECL_ATTRIBUTES (decl) without affecting
9974 DECL_ATTRIBUTES (current_function_decl).
9975 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9976 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9978 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9980 PR tree-optimization/83979
9981 * fold-const.c (fold_comparison): Use constant_boolean_node
9982 instead of boolean_{true,false}_node.
9984 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9986 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9989 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9991 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9992 Simplify the clause that sets the length attribute.
9993 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9994 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9995 clause that sets the length attribute.
9996 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9998 2018-01-24 Tom de Vries <tom@codesourcery.com>
10001 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10002 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10003 Add strict parameter.
10004 (prevent_branch_around_nothing): Insert dummy insn between branch to
10005 label and label with no ptx insn inbetween.
10006 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10008 2018-01-24 Tom de Vries <tom@codesourcery.com>
10011 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10012 for neutered threads in warp.
10013 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10015 2018-01-24 Richard Biener <rguenther@suse.de>
10017 PR tree-optimization/83176
10018 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10021 2018-01-24 Richard Biener <rguenther@suse.de>
10023 PR tree-optimization/82819
10024 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10025 code generating pluses that are no-ops in the target precision.
10027 2018-01-24 Richard Biener <rguenther@suse.de>
10029 PR middle-end/84000
10030 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10032 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10034 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10035 to merge probabilities.
10036 * predict.c (probably_never_executed): Also mark as cold functions
10037 with global 0 profile and guessed local profile.
10038 * profile-count.c (profile_probability::combine_with_count): New
10040 * profile-count.h (profile_probability::operator*,
10041 profile_probability::operator*=, profile_probability::operator/,
10042 profile_probability::operator/=): Reduce precision to adjusted
10043 and set value to guessed on contradictory divisions.
10044 (profile_probability::combine_with_freq): Remove.
10045 (profile_probability::combine_wiht_count): Declare.
10046 (profile_count::force_nonzero):: Set to adjusted.
10047 (profile_count::probability_in):: Set quality to adjusted.
10048 * tree-ssa-tail-merge.c (replace_block_by): Use
10049 combine_with_count.
10051 2018-01-23 Andrew Waterman <andrew@sifive.com>
10052 Jim Wilson <jimw@sifive.com>
10054 * config/riscv/riscv.c (riscv_stack_boundary): New.
10055 (riscv_option_override): Set riscv_stack_boundary. Handle
10056 riscv_preferred_stack_boundary_arg.
10057 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10058 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10059 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10060 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10061 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10062 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10064 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10067 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10068 of struct ix86_frame.
10069 (ix86_expand_epilogue): Likewise. Add a local variable for
10070 the reg_save_offset field in struct ix86_frame.
10072 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10074 PR tree-optimization/82604
10075 * tree-loop-distribution.c (enum partition_kind): New enum item
10076 PKIND_PARTIAL_MEMSET.
10077 (partition_builtin_p): Support above new enum item.
10078 (generate_code_for_partition): Ditto.
10079 (compute_access_range): Differentiate cases that equality can be
10080 proven at all loops, the innermost loops or no loops.
10081 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10082 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10083 (finalize_partitions, distribute_loop): Don't fuse partition of
10084 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10085 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10086 parloop is enabled.
10088 2018-01-23 Martin Liska <mliska@suse.cz>
10090 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10091 order to ignore the predictor.
10092 (PRED_POLYMORPHIC_CALL): Likewise.
10093 (PRED_RECURSIVE_CALL): Likewise.
10095 2018-01-23 Martin Liska <mliska@suse.cz>
10097 * tree-profile.c (tree_profiling): Print function header to
10098 aware reader which function we are working on.
10099 * value-prof.c (gimple_find_values_to_profile): Do not print
10100 not interesting value histograms.
10102 2018-01-23 Martin Liska <mliska@suse.cz>
10104 * profile-count.h (enum profile_quality): Add
10105 profile_uninitialized as the first value. Do not number values
10106 as they are zero based.
10107 (profile_count::verify): Update sanity check.
10108 (profile_probability::verify): Likewise.
10110 2018-01-23 Nathan Sidwell <nathan@acm.org>
10112 * doc/invoke.texi (ffor-scope): Deprecate.
10114 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10116 PR tree-optimization/83510
10117 * domwalk.c (set_all_edges_as_executable): New function.
10118 (dom_walker::dom_walker): Convert bool param
10119 "skip_unreachable_blocks" to enum reachability. Move setup of
10120 edge flags to set_all_edges_as_executable and only do it when
10121 reachability is REACHABLE_BLOCKS.
10122 * domwalk.h (enum dom_walker::reachability): New enum.
10123 (dom_walker::dom_walker): Convert bool param
10124 "skip_unreachable_blocks" to enum reachability.
10125 (set_all_edges_as_executable): New decl.
10126 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10127 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10129 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10130 but converting true to REACHABLE_BLOCKS.
10131 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10133 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10134 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10135 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10137 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10138 if check_all_array_refs will be called.
10140 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10142 * tree.c (selftest::test_location_wrappers): Add more test
10145 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10147 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10148 (selftest::test_bit_in_range): Likewise.
10150 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10153 * doc/sourcebuild.texi (vect_float): Say that the selector
10154 only describes the situation when -funsafe-math-optimizations is on.
10155 (vect_float_strict): Document.
10157 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10159 PR tree-optimization/83965
10160 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10161 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10162 instead of checking only for a reduction.
10163 (vect_recog_widen_sum_pattern): Likewise.
10165 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10167 * predict.c (probably_never_executed): Only use precise profile info.
10168 (compute_function_frequency): Skip after inlining hack since we now
10169 have quality checking.
10171 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10173 * profile-count.h (profile_probability::very_unlikely,
10174 profile_probability::unlikely, profile_probability::even): Set
10175 precision to guessed.
10177 2018-01-23 Richard Biener <rguenther@suse.de>
10179 PR tree-optimization/83963
10180 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10181 Properly terminate dominator walk when crossing the exit edge not
10182 when visiting its source block.
10184 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10187 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10188 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10190 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10192 PR tree-optimization/83957
10193 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10194 semicolon after for body surrounded by braces.
10196 PR tree-optimization/83081
10197 * profile-count.h (profile_probability::split): New method.
10198 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10199 Use profile_probability::split.
10200 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10201 when splitting a single conditional jump into 2.
10203 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10205 PR tree-optimization/69452
10206 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10209 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10211 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10212 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10213 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10215 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10217 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10219 * config/rl78/rl78.md (movdi): New define_expand.
10220 * config/rl78/rl78.c (rl78_split_movdi): New function.
10222 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10225 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10227 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10228 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10229 128-bit to produce an UNSPEC move to get the double word with the
10230 signbit and then a shift directly to do signbit.
10231 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10232 implementation with a new version that just does either a direct
10233 move or a regular move. Move memory interface to separate insns.
10234 Move insns so they are next to the expander.
10235 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10236 with signbit move. Split big and little endian case.
10237 (signbit<mode>2_dm_mem_le): Likewise.
10238 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10239 (signbit<mode>2_dm2): Likewise.
10241 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10243 * config/rl78/rl78.md (anddi3): New define_expand.
10245 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10247 * config/rl78/rl78.md (umindi3): New define_expand.
10249 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10251 * config/rl78/rl78.md (smindi3): New define_expand.
10253 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10255 * config/rl78/rl78.md (smaxdi3): New define_expand.
10257 2018-01-22 Carl Love <cel@us.ibm.com>
10259 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10260 LVX_V1TI): Add macro expansion.
10261 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10262 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10263 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10264 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10265 Change check to determine if the instruction is a byte reversing
10266 entry. Fix typo in comment.
10267 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10268 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10269 Add def_builtin calls for new builtins.
10270 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10271 Add define_insn expansion.
10273 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10275 * config/rl78/rl78.md (umaxdi3): New define_expand.
10277 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10279 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10280 for non-QImode registers.
10282 2018-01-22 Richard Biener <rguenther@suse.de>
10284 PR tree-optimization/83963
10285 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10286 including the loop exit block.
10287 (scop_detection::merge_sese): Likewise.
10288 (scop_detection::add_scop): Do it here instead.
10290 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10292 * doc/sourcebuild.texi (arm_softfloat): Document.
10294 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10297 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10298 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10299 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10301 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10302 David Edelsohn <dje.gcc@gmail.com>
10305 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10306 Change "crset eq" to "crset 2".
10307 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10308 (*call_indirect_aix<mode>_nospec): Likewise.
10309 (*call_value_indirect_aix<mode>_nospec): Likewise.
10310 (*call_indirect_elfv2<mode>_nospec): Likewise.
10311 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10312 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10313 change assembly output from . to $.
10314 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10315 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10316 (*tablejump<mode>_internal1_nospec): Likewise.
10318 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10321 * config/sh/sh_optimize_sett_clrt.cc:
10322 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10324 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10326 PR tree-optimization/83940
10327 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10328 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10329 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10330 use it to pass back the definition type.
10331 (vect_check_store_rhs): Likewise rhs_dt_out.
10332 (vect_build_gather_load_calls): Add a mask_dt argument and use
10333 it instead of a call to vect_is_simple_use.
10334 (vectorizable_store): Update calls to vect_check_load_store_mask
10335 and vect_check_store_rhs. Use the dt returned by the latter instead
10336 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10337 instead of calls to vect_is_simple_use. Pass the scalar rather
10338 than the vector operand to vect_is_simple_use when handling
10339 second and subsequent copies of an rhs value.
10340 (vectorizable_load): Update calls to vect_check_load_store_mask
10341 and vect_build_gather_load_calls. Use the cached mask_dt and
10342 gs_info.offset_dt instead of calls to vect_is_simple_use.
10344 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10346 PR middle-end/83945
10347 * tree-emutls.c: Include gimplify.h.
10348 (lower_emutls_2): New function.
10349 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10350 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10351 it before further processing.
10354 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10355 UINTVAL (trueop1) instead of INTVAL (op1).
10357 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10361 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10362 INCOMING_FRAME_SP_OFFSET if not defined.
10363 (scan_trace): Add ENTRY argument. If true and
10364 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10365 emit a note to adjust the CFA offset.
10366 (create_cfi_notes): Adjust scan_trace callers.
10367 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10368 INCOMING_FRAME_SP_OFFSET in the CIE.
10369 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10370 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10372 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10373 * doc/tm.texi: Regenerated.
10375 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10377 PR rtl-optimization/83147
10378 * lra-constraints.c (remove_inheritance_pseudos): Use
10379 lra_substitute_pseudo_within_insn.
10381 2018-01-19 Tom de Vries <tom@codesourcery.com>
10382 Cesar Philippidis <cesar@codesourcery.com>
10385 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10387 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10390 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10391 spaces for function labels.
10393 2018-01-19 Martin Liska <mliska@suse.cz>
10395 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10396 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10397 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10398 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10399 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10400 (PRED_CONST_RETURN): Change from 69 to 65.
10401 (PRED_NULL_RETURN): Change from 91 to 71.
10402 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10403 (PRED_LOOP_GUARD): Change from 66 to 73.
10405 2018-01-19 Martin Liska <mliska@suse.cz>
10407 * predict.c (predict_insn_def): Add new assert.
10408 (struct branch_predictor): Change type to signed integer.
10409 (test_prediction_value_range): Amend test to cover
10410 PROB_UNINITIALIZED.
10411 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10412 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10413 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10414 (PRED_LOOP_IV_COMPARE): Likewise.
10415 * predict.h (PROB_UNINITIALIZED): Define new constant.
10417 2018-01-19 Martin Liska <mliska@suse.cz>
10419 * predict.c (dump_prediction): Add new format for
10420 analyze_brprob.py script which is enabled with -details
10422 * profile-count.h (precise_p): New function.
10424 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10426 PR tree-optimization/83922
10427 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10428 there are no statements that need masking.
10429 (vect_active_double_reduction_p): New function.
10430 (vect_analyze_loop_operations): Use it when handling phis that
10431 are not in the loop header.
10433 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10435 PR tree-optimization/83914
10436 * tree-vect-loop.c (vectorizable_induction): Don't convert
10437 init_expr or apply the peeling adjustment for inductions
10438 that are nested within the vectorized loop.
10440 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10442 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10445 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10449 * function.h (gimplify_parameters): Add gimple_seq * argument.
10450 * function.c: Include gimple.h and options.h.
10451 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10452 for the added local temporaries if needed.
10453 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10454 if there are any parameter cleanups, wrap whole body into a
10455 try/finally with the cleanups.
10457 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10460 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10461 Use GET_MODE_CLASS for scalar floating point.
10463 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10467 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10468 Fix call of call_cgraph_insertion_hooks.
10470 2018-01-18 Martin Sebor <msebor@redhat.com>
10472 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10474 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10477 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10480 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10483 * common.opt: (-ffile-prefix-map): New option.
10484 * opts.c (common_handle_option): Defer it.
10485 * opts-global.c (handle_common_deferred_options): Handle it.
10486 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10487 * file-prefix-map.h: New file.
10488 (remap_debug_filename, add_debug_prefix_map): ...here.
10489 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10490 * final.c (debug_prefix_map, add_debug_prefix_map
10491 remap_debug_filename): Move to...
10492 * file-prefix-map.c: New file.
10493 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10494 generalize, get rid of alloca(), use strrchr() instead of strchr().
10495 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10496 Implement in terms of add_prefix_map().
10497 (remap_macro_filename, remap_debug_filename): Implement in term of
10499 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10500 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10501 * dbxout.c: Include file-prefix-map.h.
10502 * varasm.c: Likewise.
10503 * vmsdbgout.c: Likewise.
10504 * xcoffout.c: Likewise.
10505 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10506 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10507 * doc/invoke.texi (-ffile-prefix-map): Document.
10508 (-fdebug-prefix-map): Update description.
10510 2018-01-18 Martin Liska <mliska@suse.cz>
10512 * config/i386/i386.c (indirect_thunk_name): Document that also
10514 (output_indirect_thunk): Document why both instructions
10515 (pause and lfence) are generated.
10517 2018-01-18 Richard Biener <rguenther@suse.de>
10519 PR tree-optimization/83887
10520 * graphite-scop-detection.c
10521 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10522 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10523 (scop_detection::merge_sese): Re-implement with a flood-fill
10524 algorithm that properly finds a SESE region if it exists.
10526 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10529 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10530 pointer_diff optimizations use view_convert instead of convert.
10532 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10534 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10535 Generate different code for -mno-speculate-indirect-jumps.
10536 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10537 (*call_indirect_aix<mode>): Disable for
10538 -mno-speculate-indirect-jumps.
10539 (*call_indirect_aix<mode>_nospec): New define_insn.
10540 (*call_value_indirect_aix<mode>): Disable for
10541 -mno-speculate-indirect-jumps.
10542 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10543 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10544 -mno-speculate-indirect-jumps.
10545 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10547 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10549 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10550 long double type, set the flags for noting the default long double
10551 type, even if we don't pass or return a long double type.
10553 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10556 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10559 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10561 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10562 support for merge[hl].
10563 (fold_mergehl_helper): New helper function.
10564 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10565 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10566 (altivec_vmrglw_direct): Add xxmrglw insn.
10568 2018-01-17 Andrew Waterman <andrew@sifive.com>
10570 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10571 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10573 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10576 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10577 call the lto_location_cache before reading the
10578 DECL_SOURCE_LOCATION of the types.
10580 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10581 Richard Sandiford <richard.sandiford@linaro.org>
10583 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10584 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10585 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10586 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10587 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10589 * config/aarch64/constraints.md (aarch64_movti_operand):
10591 * config/aarch64/predicates.md (Uti): Add new constraint.
10593 2018-01-17 Carl Love <cel@us.ibm.com>
10595 * config/rs6000/vsx.md (define_expand xl_len_r,
10596 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10597 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10599 (define_expand, define_insn): Move the shift left from the
10600 define_insn to the define_expand for lxvl and stxvl instructions.
10601 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10602 and XL_LEN_R definitions to PURE.
10604 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10606 * config/i386/i386.c (indirect_thunk_name): Declare regno
10607 as unsigned int. Compare regno with INVALID_REGNUM.
10608 (output_indirect_thunk): Ditto.
10609 (output_indirect_thunk_function): Ditto.
10610 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10611 in the call to output_indirect_thunk_function.
10613 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10615 PR middle-end/83884
10616 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10617 rather than the size of inner_type to determine the stack slot size
10618 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10620 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10623 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10626 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10628 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10629 endian Linux systems to optionally enable multilibs for selecting
10630 the long double type if the user configured an explicit type.
10631 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10632 have no long double multilibs if not defined.
10633 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10634 warn if the user used -mabi={ieee,ibm}longdouble and we built
10635 multilibs for long double.
10636 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10637 appropriate multilib option.
10638 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10640 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10641 for building long double multilibs.
10642 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10644 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10646 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10649 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10651 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10654 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10657 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10660 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10662 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10663 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10665 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10667 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10668 different rtl trees depending on TARGET_64BIT.
10669 (rs6000_gen_lvx): Likewise.
10671 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10673 * config/visium/visium.md (nop): Tweak comment.
10674 (hazard_nop): Likewise.
10676 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10678 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10679 -mspeculate-indirect-jumps.
10680 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10681 for -mno-speculate-indirect-jumps.
10682 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10683 (*call_value_indirect_elfv2<mode>): Disable for
10684 -mno-speculate-indirect-jumps.
10685 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10686 (indirect_jump): Emit different RTL for
10687 -mno-speculate-indirect-jumps.
10688 (*indirect_jump<mode>): Disable for
10689 -mno-speculate-indirect-jumps.
10690 (*indirect_jump<mode>_nospec): New define_insn.
10691 (tablejump): Emit different RTL for
10692 -mno-speculate-indirect-jumps.
10693 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10694 (tablejumpsi_nospec): New define_expand.
10695 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10696 (tablejumpdi_nospec): New define_expand.
10697 (*tablejump<mode>_internal1): Disable for
10698 -mno-speculate-indirect-jumps.
10699 (*tablejump<mode>_internal1_nospec): New define_insn.
10700 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10703 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10705 * caller-save.c (insert_save): Drop unnecessary parameter. All
10708 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10709 Richard Biener <rguenth@suse.de>
10712 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10713 return early, inline manually is_gimple_sizepos. Make sure if we
10714 call gimplify_expr we don't end up with a gimple constant.
10715 * tree.c (variably_modified_type_p): Don't return true for
10716 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10717 * gimplify.h (is_gimple_sizepos): Remove.
10719 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10721 PR tree-optimization/83857
10722 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10723 vectorizable_live_operation for pure SLP statements.
10724 (vectorizable_live_operation): Handle PHIs.
10726 2018-01-16 Richard Biener <rguenther@suse.de>
10728 PR tree-optimization/83867
10729 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10730 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10732 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10735 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10736 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10737 If off is not INTEGER_CST, issue a may not be aligned warning
10738 rather than isn't aligned. Use isn%'t rather than isn't.
10739 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10741 <case MULT_EXPR>: Improve the case when bottom and one of the
10742 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10743 operand, in that case check if the other operand is multiple of
10744 bottom divided by the INTEGER_CST operand.
10746 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10749 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10750 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10751 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10752 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10753 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10754 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10755 (pa_function_arg_size): New function.
10757 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10759 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10760 in a separate statement.
10762 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10764 PR tree-optimization/83847
10765 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10766 group gathers and scatters.
10768 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10770 PR rtl-optimization/86620
10771 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10773 PR rtl-optimization/83213
10774 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10775 to last if both are JUMP_INSNs.
10777 PR tree-optimization/83843
10778 * gimple-ssa-store-merging.c
10779 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10780 store_immediate_info for bswap/nop orig_stores.
10782 2018-01-15 Andrew Waterman <andrew@sifive.com>
10784 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10786 <UDIV>: Increase cost if !TARGET_DIV.
10788 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10790 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10791 (define_attr "cr_logical_3op"): New.
10792 (cceq_ior_compare): Adjust.
10793 (cceq_ior_compare_complement): Adjust.
10794 (*cceq_rev_compare): Adjust.
10795 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10796 (is_cracked_insn): Adjust.
10797 (insn_must_be_first_in_group): Adjust.
10798 * config/rs6000/40x.md: Adjust.
10799 * config/rs6000/440.md: Adjust.
10800 * config/rs6000/476.md: Adjust.
10801 * config/rs6000/601.md: Adjust.
10802 * config/rs6000/603.md: Adjust.
10803 * config/rs6000/6xx.md: Adjust.
10804 * config/rs6000/7450.md: Adjust.
10805 * config/rs6000/7xx.md: Adjust.
10806 * config/rs6000/8540.md: Adjust.
10807 * config/rs6000/cell.md: Adjust.
10808 * config/rs6000/e300c2c3.md: Adjust.
10809 * config/rs6000/e500mc.md: Adjust.
10810 * config/rs6000/e500mc64.md: Adjust.
10811 * config/rs6000/e5500.md: Adjust.
10812 * config/rs6000/e6500.md: Adjust.
10813 * config/rs6000/mpc.md: Adjust.
10814 * config/rs6000/power4.md: Adjust.
10815 * config/rs6000/power5.md: Adjust.
10816 * config/rs6000/power6.md: Adjust.
10817 * config/rs6000/power7.md: Adjust.
10818 * config/rs6000/power8.md: Adjust.
10819 * config/rs6000/power9.md: Adjust.
10820 * config/rs6000/rs64.md: Adjust.
10821 * config/rs6000/titan.md: Adjust.
10823 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10825 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10826 ix86_indirect_branch_register logic.
10828 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10830 * config/i386/constraints.md (Bs): Update
10831 ix86_indirect_branch_register check. Don't check
10832 ix86_indirect_branch_register with GOT_memory_operand.
10834 * config/i386/predicates.md (GOT_memory_operand): Don't check
10835 ix86_indirect_branch_register here.
10836 (GOT32_symbol_operand): Likewise.
10838 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10840 * config/i386/predicates.md (constant_call_address_operand):
10841 Rewrite ix86_indirect_branch_register logic.
10842 (sibcall_insn_operand): Likewise.
10844 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10846 * config/i386/constraints.md (Bs): Replace
10847 ix86_indirect_branch_thunk_register with
10848 ix86_indirect_branch_register.
10850 * config/i386/i386.md (indirect_jump): Likewise.
10851 (tablejump): Likewise.
10852 (*sibcall_memory): Likewise.
10853 (*sibcall_value_memory): Likewise.
10854 Peepholes of indirect call and jump via memory: Likewise.
10855 * config/i386/i386.opt: Likewise.
10856 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10857 (GOT_memory_operand): Likewise.
10858 (call_insn_operand): Likewise.
10859 (sibcall_insn_operand): Likewise.
10860 (GOT32_symbol_operand): Likewise.
10862 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10864 PR middle-end/83837
10865 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10866 type rather than type addr's type points to.
10867 (expand_omp_atomic_mutex): Likewise.
10868 (expand_omp_atomic): Likewise.
10870 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10873 * config/i386/i386.c (output_indirect_thunk_function): Use
10874 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10875 for __x86_return_thunk.
10877 2018-01-15 Richard Biener <rguenther@suse.de>
10879 PR middle-end/83850
10880 * expmed.c (extract_bit_field_1): Fix typo.
10882 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10885 * config/arm/iterators.md (VF): New mode iterator.
10886 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10887 Remove integer-related logic from pattern.
10888 (neon_vabd<mode>_3): Likewise.
10890 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10892 PR middle-end/82694
10893 * common.opt (fstrict-overflow): No longer an alias.
10894 (fwrapv-pointer): New option.
10895 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10896 also for pointer types based on flag_wrapv_pointer.
10897 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10898 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10899 opts->x_flag_wrapv got set.
10900 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10901 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10902 POINTER_TYPE_OVERFLOW_UNDEFINED.
10903 * match.pd: Likewise in address comparison pattern.
10904 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10906 2018-01-15 Richard Biener <rguenther@suse.de>
10909 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10910 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10911 Reset type names to their identifier if their TYPE_DECL doesn't
10912 have linkage (and thus is used for ODR and devirt).
10913 (save_debug_info_for_decl): Remove.
10914 (save_debug_info_for_type): Likewise.
10915 (add_tree_to_fld_list): Adjust.
10916 * tree-pretty-print.c (dump_generic_node): Make dumping of
10917 type names more robust.
10919 2018-01-15 Richard Biener <rguenther@suse.de>
10921 * BASE-VER: Bump to 8.0.1.
10923 2018-01-14 Martin Sebor <msebor@redhat.com>
10926 * builtins.c (check_access): Avoid warning when the no-warning bit
10929 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10931 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10932 * ira-color (allocno_hard_regs_compare): Likewise.
10934 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10937 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10938 Use .pushsection/.popsection.
10940 2018-01-14 Martin Sebor <msebor@redhat.com>
10943 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10945 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10947 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10948 entry from extra_headers.
10949 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10950 extra_headers, make the list bitwise identical to the i?86-*-* one.
10952 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10954 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10955 -mcmodel=large with -mindirect-branch=thunk,
10956 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10957 -mfunction-return=thunk-extern.
10958 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10959 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10960 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10962 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10964 * config/i386/i386.c (print_reg): Print the name of the full
10965 integer register without '%'.
10966 (ix86_print_operand): Handle 'V'.
10967 * doc/extend.texi: Document 'V' modifier.
10969 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10971 * config/i386/constraints.md (Bs): Disallow memory operand for
10972 -mindirect-branch-register.
10974 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10975 (GOT_memory_operand): Likewise.
10976 (call_insn_operand): Likewise.
10977 (sibcall_insn_operand): Likewise.
10978 (GOT32_symbol_operand): Likewise.
10979 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10980 for -mindirect-branch-register.
10981 (tablejump): Likewise.
10982 (*sibcall_memory): Likewise.
10983 (*sibcall_value_memory): Likewise.
10984 Disallow peepholes of indirect call and jump via memory for
10985 -mindirect-branch-register.
10986 (*call_pop): Replace m with Bw.
10987 (*call_value_pop): Likewise.
10988 (*sibcall_pop_memory): Replace m with Bs.
10989 * config/i386/i386.opt (mindirect-branch-register): New option.
10990 * doc/invoke.texi: Document -mindirect-branch-register option.
10992 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10994 * config/i386/i386-protos.h (ix86_output_function_return): New.
10995 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10996 set function_return_type.
10997 (indirect_thunk_name): Add ret_p to indicate thunk for function
10999 (output_indirect_thunk_function): Pass false to
11000 indirect_thunk_name.
11001 (ix86_output_indirect_branch_via_reg): Likewise.
11002 (ix86_output_indirect_branch_via_push): Likewise.
11003 (output_indirect_thunk_function): Create alias for function
11004 return thunk if regno < 0.
11005 (ix86_output_function_return): New function.
11006 (ix86_handle_fndecl_attribute): Handle function_return.
11007 (ix86_attribute_table): Add function_return.
11008 * config/i386/i386.h (machine_function): Add
11009 function_return_type.
11010 * config/i386/i386.md (simple_return_internal): Use
11011 ix86_output_function_return.
11012 (simple_return_internal_long): Likewise.
11013 * config/i386/i386.opt (mfunction-return=): New option.
11014 (indirect_branch): Mention -mfunction-return=.
11015 * doc/extend.texi: Document function_return function attribute.
11016 * doc/invoke.texi: Document -mfunction-return= option.
11018 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11020 * config/i386/i386-opts.h (indirect_branch): New.
11021 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11022 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11023 with local indirect jump when converting indirect call and jump.
11024 (ix86_set_indirect_branch_type): New.
11025 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11026 (indirectlabelno): New.
11027 (indirect_thunk_needed): Likewise.
11028 (indirect_thunk_bnd_needed): Likewise.
11029 (indirect_thunks_used): Likewise.
11030 (indirect_thunks_bnd_used): Likewise.
11031 (INDIRECT_LABEL): Likewise.
11032 (indirect_thunk_name): Likewise.
11033 (output_indirect_thunk): Likewise.
11034 (output_indirect_thunk_function): Likewise.
11035 (ix86_output_indirect_branch_via_reg): Likewise.
11036 (ix86_output_indirect_branch_via_push): Likewise.
11037 (ix86_output_indirect_branch): Likewise.
11038 (ix86_output_indirect_jmp): Likewise.
11039 (ix86_code_end): Call output_indirect_thunk_function if needed.
11040 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11042 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11043 (ix86_attribute_table): Add indirect_branch.
11044 * config/i386/i386.h (machine_function): Add indirect_branch_type
11045 and has_local_indirect_jump.
11046 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11048 (tablejump): Likewise.
11049 (*indirect_jump): Use ix86_output_indirect_jmp.
11050 (*tablejump_1): Likewise.
11051 (simple_return_indirect_internal): Likewise.
11052 * config/i386/i386.opt (mindirect-branch=): New option.
11053 (indirect_branch): New.
11056 (thunk-inline): Likewise.
11057 (thunk-extern): Likewise.
11058 * doc/extend.texi: Document indirect_branch function attribute.
11059 * doc/invoke.texi: Document -mindirect-branch= option.
11061 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11064 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11066 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11068 * ipa-inline.c (want_inline_small_function_p): Return false if
11069 inlining has already failed with CIF_FINAL_ERROR.
11070 (update_caller_keys): Call want_inline_small_function_p before
11072 (update_callee_keys): Likewise.
11074 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11076 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11078 (rs6000_quadword_masked_address_p): Likewise.
11079 (quad_aligned_load_p): Likewise.
11080 (quad_aligned_store_p): Likewise.
11081 (const_load_sequence_p): Add comment to describe the outer-most loop.
11082 (mimic_memory_attributes_and_flags): New function.
11083 (rs6000_gen_stvx): Likewise.
11084 (replace_swapped_aligned_store): Likewise.
11085 (rs6000_gen_lvx): Likewise.
11086 (replace_swapped_aligned_load): Likewise.
11087 (replace_swapped_load_constant): Capitalize argument name in
11088 comment describing this function.
11089 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11090 and stores that access quad-word aligned addresses and replace
11091 with stvx or lvx instructions when appropriate.
11092 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11093 New function prototype.
11094 (rs6000_quadword_masked_address_p): Likewise.
11095 (rs6000_gen_lvx): Likewise.
11096 (rs6000_gen_stvx): Likewise.
11097 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11098 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11099 when memory address is aligned.
11100 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11101 this split to select lvx instruction when memory address is aligned.
11102 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11103 instruction when memory address is aligned.
11104 (*vsx_le_perm_load_v16qi): Likewise.
11105 (four unnamed splitters): Modify to select the stvx instruction
11106 when memory is aligned.
11108 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11110 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11111 which appears in the queue multiple times.
11113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11114 Alan Hayward <alan.hayward@arm.com>
11115 David Sherwood <david.sherwood@arm.com>
11117 * tree-vectorizer.h (vec_lower_bound): New structure.
11118 (_loop_vec_info): Add check_nonzero and lower_bounds.
11119 (LOOP_VINFO_CHECK_NONZERO): New macro.
11120 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11121 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11122 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11123 fields. Make seg_len the distance travelled, not including the
11125 (dr_direction_indicator): Declare.
11126 (dr_zero_step_indicator): Likewise.
11127 (dr_known_forward_stride_p): Likewise.
11128 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11130 (runtime_alias_check_p): Allow runtime alias checks with
11132 (operator ==): Compare access_size and align.
11133 (prune_runtime_alias_test_list): Rework for new distinction between
11134 the access_size and seg_len.
11135 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11137 (get_segment_min_max): New function.
11138 (create_intersect_range_checks): Use it.
11139 (dr_step_indicator): New function.
11140 (dr_direction_indicator): Likewise.
11141 (dr_zero_step_indicator): Likewise.
11142 (dr_known_forward_stride_p): Likewise.
11143 * tree-loop-distribution.c (data_ref_segment_size): Return
11144 DR_STEP * (niters - 1).
11145 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11147 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11148 (vect_preserves_scalar_order_p): New function, split out from...
11149 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11150 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11151 (vect_vfa_access_size): New function.
11152 (vect_vfa_align): Likewise.
11153 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11154 (dump_lower_bound): New function.
11155 (vect_check_lower_bound): Likewise.
11156 (vect_small_gap_p): Likewise.
11157 (vectorizable_with_step_bound_p): Likewise.
11158 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11159 depencies if the vectorization factor is 1. Convert the checks
11160 for nonzero steps into checks on the bounds of DR_STEP. Try using
11161 a bunds check for variable steps if the minimum required step is
11162 relatively small. Update calls to the dr_with_seg_len
11163 constructor and to vect_compile_time_alias.
11164 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11166 (vect_loop_versioning): Call it.
11167 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11169 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11172 Alan Hayward <alan.hayward@arm.com>
11173 David Sherwood <david.sherwood@arm.com>
11175 * doc/sourcebuild.texi (vect_scatter_store): Document.
11176 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11178 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11180 * genopinit.c (main): Add supports_vec_scatter_store and
11181 supports_vec_scatter_store_cached to target_optabs.
11182 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11183 IFN_MASK_SCATTER_STORE.
11184 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11186 * internal-fn.h (internal_store_fn_p): Declare.
11187 (internal_fn_stored_value_index): Likewise.
11188 * internal-fn.c (scatter_store_direct): New macro.
11189 (expand_scatter_store_optab_fn): New function.
11190 (direct_scatter_store_optab_supported_p): New macro.
11191 (internal_store_fn_p): New function.
11192 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11193 IFN_MASK_SCATTER_STORE.
11194 (internal_fn_mask_index): Likewise.
11195 (internal_fn_stored_value_index): New function.
11196 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11197 for scatter stores.
11198 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11199 * optabs-query.c (supports_vec_scatter_store_p): New function.
11200 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11201 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11202 true for scatter stores.
11203 (vect_gather_scatter_fn_p): Handle scatter stores too.
11204 (vect_check_gather_scatter): Consider using scatter stores if
11205 supports_vec_scatter_store_p.
11206 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11207 scatter stores too.
11208 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11209 internal_fn_stored_value_index.
11210 (check_load_store_masking): Handle scatter stores too.
11211 (vect_get_store_rhs): Make public.
11212 (vectorizable_call): Use internal_store_fn_p.
11213 (vectorizable_store): Handle scatter store internal functions.
11214 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11215 when deciding whether the end of the group has been reached.
11216 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11217 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11218 (mask_scatter_store<mode>): New insns.
11220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11221 Alan Hayward <alan.hayward@arm.com>
11222 David Sherwood <david.sherwood@arm.com>
11224 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11225 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11226 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11228 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11229 Use vect_truncate_gather_scatter_offset if we can't treat the
11230 operation as a normal gather load or scatter store.
11231 (get_group_load_store_type): Take the gather_scatter_info
11232 as argument. Try using a gather load or scatter store for
11233 single-element groups.
11234 (get_load_store_type): Update calls to get_group_load_store_type
11235 and vect_use_strided_gather_scatters_p.
11237 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11238 Alan Hayward <alan.hayward@arm.com>
11239 David Sherwood <david.sherwood@arm.com>
11241 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11242 optional tree argument.
11243 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11245 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11246 but continue to use the current value as a fallback.
11247 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11248 to compare the updates.
11249 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11250 (get_load_store_type): Use it when handling a strided access.
11251 (vect_get_strided_load_store_ops): New function.
11252 (vect_get_data_ptr_increment): Likewise.
11253 (vectorizable_load): Handle strided gather loads. Always pass
11254 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11257 Alan Hayward <alan.hayward@arm.com>
11258 David Sherwood <david.sherwood@arm.com>
11260 * doc/md.texi (gather_load@var{m}): Document.
11261 (mask_gather_load@var{m}): Likewise.
11262 * genopinit.c (main): Add supports_vec_gather_load and
11263 supports_vec_gather_load_cached to target_optabs.
11264 * optabs-tree.c (init_tree_optimization_optabs): Use
11265 ggc_cleared_alloc to allocate target_optabs.
11266 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11267 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11269 * internal-fn.h (internal_load_fn_p): Declare.
11270 (internal_gather_scatter_fn_p): Likewise.
11271 (internal_fn_mask_index): Likewise.
11272 (internal_gather_scatter_fn_supported_p): Likewise.
11273 * internal-fn.c (gather_load_direct): New macro.
11274 (expand_gather_load_optab_fn): New function.
11275 (direct_gather_load_optab_supported_p): New macro.
11276 (direct_internal_fn_optab): New function.
11277 (internal_load_fn_p): Likewise.
11278 (internal_gather_scatter_fn_p): Likewise.
11279 (internal_fn_mask_index): Likewise.
11280 (internal_gather_scatter_fn_supported_p): Likewise.
11281 * optabs-query.c (supports_at_least_one_mode_p): New function.
11282 (supports_vec_gather_load_p): Likewise.
11283 * optabs-query.h (supports_vec_gather_load_p): Declare.
11284 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11285 and memory_type field.
11286 (NUM_PATTERNS): Bump to 15.
11287 * tree-vect-data-refs.c: Include internal-fn.h.
11288 (vect_gather_scatter_fn_p): New function.
11289 (vect_describe_gather_scatter_call): Likewise.
11290 (vect_check_gather_scatter): Try using internal functions for
11291 gather loads. Recognize existing calls to a gather load function.
11292 (vect_analyze_data_refs): Consider using gather loads if
11293 supports_vec_gather_load_p.
11294 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11295 (vect_get_gather_scatter_offset_type): Likewise.
11296 (vect_convert_mask_for_vectype): Likewise.
11297 (vect_add_conversion_to_patterm): Likewise.
11298 (vect_try_gather_scatter_pattern): Likewise.
11299 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11300 (vect_vect_recog_func_ptrs): Add it.
11301 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11302 internal_fn_mask_index and internal_gather_scatter_fn_p.
11303 (check_load_store_masking): Take the gather_scatter_info as an
11304 argument and handle gather loads.
11305 (vect_get_gather_scatter_ops): New function.
11306 (vectorizable_call): Check internal_load_fn_p.
11307 (vectorizable_load): Likewise. Handle gather load internal
11309 (vectorizable_store): Update call to check_load_store_masking.
11310 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11311 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11312 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11313 (aarch64_gather_scale_operand_d): New predicates.
11314 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11315 (mask_gather_load<mode>): New insns.
11317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * optabs.def (fold_left_plus_optab): New optab.
11322 * doc/md.texi (fold_left_plus_@var{m}): Document.
11323 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11324 * internal-fn.c (fold_left_direct): Define.
11325 (expand_fold_left_optab_fn): Likewise.
11326 (direct_fold_left_optab_supported_p): Likewise.
11327 * fold-const-call.c (fold_const_fold_left): New function.
11328 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11329 * tree-parloops.c (valid_reduction_p): New function.
11330 (gather_scalar_reductions): Use it.
11331 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11332 (vect_finish_replace_stmt): Declare.
11333 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11334 (needs_fold_left_reduction_p): New function, split out from...
11335 (vect_is_simple_reduction): ...here. Accept reductions that
11336 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11337 (vect_force_simple_reduction): Also store the reduction type in
11338 the assignment's STMT_VINFO_REDUC_TYPE.
11339 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11340 (merge_with_identity): New function.
11341 (vect_expand_fold_left): Likewise.
11342 (vectorize_fold_left_reduction): Likewise.
11343 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11344 scalar phi in place for it. Check for target support and reject
11345 cases that would reassociate the operation. Defer the transform
11346 phase to vectorize_fold_left_reduction.
11347 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11348 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11349 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11353 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11354 call to ifc_temp_var.
11356 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11357 Alan Hayward <alan.hayward@arm.com>
11358 David Sherwood <david.sherwood@arm.com>
11360 * target.def (legitimize_address_displacement): Take the original
11361 offset as a poly_int.
11362 * targhooks.h (default_legitimize_address_displacement): Update
11364 * targhooks.c (default_legitimize_address_displacement): Likewise.
11365 * doc/tm.texi: Regenerate.
11366 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11367 as an argument, moving assert of ad->disp == ad->disp_term to...
11368 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11369 Try calling targetm.legitimize_address_displacement before expanding
11370 the address rather than afterwards, and adjust for the new interface.
11371 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11372 Match the new hook interface. Handle SVE addresses.
11373 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11374 new hook interface.
11376 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11378 * Makefile.in (OBJS): Add early-remat.o.
11379 * target.def (select_early_remat_modes): New hook.
11380 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11381 * doc/tm.texi: Regenerate.
11382 * targhooks.h (default_select_early_remat_modes): Declare.
11383 * targhooks.c (default_select_early_remat_modes): New function.
11384 * timevar.def (TV_EARLY_REMAT): New timevar.
11385 * passes.def (pass_early_remat): New pass.
11386 * tree-pass.h (make_pass_early_remat): Declare.
11387 * early-remat.c: New file.
11388 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11390 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11393 Alan Hayward <alan.hayward@arm.com>
11394 David Sherwood <david.sherwood@arm.com>
11396 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11397 vfm1 with a bound_epilog parameter.
11398 (vect_do_peeling): Update calls accordingly, and move the prologue
11399 call earlier in the function. Treat the base bound_epilog as 0 for
11400 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11401 this base when peeling for gaps.
11402 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11403 with fully-masked loops.
11404 (vect_estimate_min_profitable_iters): Handle the single peeled
11405 iteration in that case.
11407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11408 Alan Hayward <alan.hayward@arm.com>
11409 David Sherwood <david.sherwood@arm.com>
11411 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11412 single-element interleaving even if the size is not a power of 2.
11413 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11414 accesses for single-element interleaving if the group size is
11417 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11418 Alan Hayward <alan.hayward@arm.com>
11419 David Sherwood <david.sherwood@arm.com>
11421 * doc/md.texi (fold_extract_last_@var{m}): Document.
11422 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11423 * optabs.def (fold_extract_last_optab): New optab.
11424 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11425 * internal-fn.c (fold_extract_direct): New macro.
11426 (expand_fold_extract_optab_fn): Likewise.
11427 (direct_fold_extract_optab_supported_p): Likewise.
11428 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11429 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11430 EXTRACT_LAST_REDUCTION.
11431 (get_initial_def_for_reduction): Do not create an initial vector
11432 for EXTRACT_LAST_REDUCTION reductions.
11433 (vectorizable_reduction): Leave the scalar phi in place for
11434 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11435 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11436 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11437 transform phase to vectorizable_condition.
11438 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11440 (vect_finish_stmt_generation): ...here.
11441 (vect_finish_replace_stmt): New function.
11442 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11443 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11445 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11448 Alan Hayward <alan.hayward@arm.com>
11449 David Sherwood <david.sherwood@arm.com>
11451 * doc/md.texi (extract_last_@var{m}): Document.
11452 * optabs.def (extract_last_optab): New optab.
11453 * internal-fn.def (EXTRACT_LAST): New internal function.
11454 * internal-fn.c (cond_unary_direct): New macro.
11455 (expand_cond_unary_optab_fn): Likewise.
11456 (direct_cond_unary_optab_supported_p): Likewise.
11457 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11458 loops using EXTRACT_LAST.
11459 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11460 (extract_last_<mode>): ...this optab.
11461 (vec_extract<mode><Vel>): Update accordingly.
11463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11464 Alan Hayward <alan.hayward@arm.com>
11465 David Sherwood <david.sherwood@arm.com>
11467 * target.def (empty_mask_is_expensive): New hook.
11468 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11469 * doc/tm.texi: Regenerate.
11470 * targhooks.h (default_empty_mask_is_expensive): Declare.
11471 * targhooks.c (default_empty_mask_is_expensive): New function.
11472 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11473 if the target says that empty masks are expensive.
11474 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11476 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11479 Alan Hayward <alan.hayward@arm.com>
11480 David Sherwood <david.sherwood@arm.com>
11482 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11483 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11484 (vect_use_loop_mask_for_alignment_p): New function.
11485 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11486 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11487 niters_skip argument. Make sure that the first niters_skip elements
11488 of the first iteration are inactive.
11489 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11490 Update call to vect_set_loop_masks_directly.
11491 (get_misalign_in_elems): New function, split out from...
11492 (vect_gen_prolog_loop_niters): ...here.
11493 (vect_update_init_of_dr): Take a code argument that specifies whether
11494 the adjustment should be added or subtracted.
11495 (vect_update_init_of_drs): Likewise.
11496 (vect_prepare_for_masked_peels): New function.
11497 (vect_do_peeling): Skip prologue peeling if we're using a mask
11498 instead. Update call to vect_update_inits_of_drs.
11499 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11501 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11502 alignment. Do not include the number of peeled iterations in
11503 the minimum threshold in that case.
11504 (vectorizable_induction): Adjust the start value down by
11505 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11506 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11507 Take the number of skipped iterations into account when calculating
11509 * tree-vect-stmts.c (vect_gen_while_not): New function.
11511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11512 Alan Hayward <alan.hayward@arm.com>
11513 David Sherwood <david.sherwood@arm.com>
11515 * doc/sourcebuild.texi (vect_fully_masked): Document.
11516 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11517 default value to 0.
11518 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11520 (vect_analyze_loop_2): ...here. Don't check the vectorization
11521 factor against the number of loop iterations if the loop is
11524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11525 Alan Hayward <alan.hayward@arm.com>
11526 David Sherwood <david.sherwood@arm.com>
11528 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11529 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11530 (dump_groups): Update accordingly.
11531 (iv_use::mem_type): New member variable.
11532 (address_p): New function.
11533 (record_use): Add a mem_type argument and initialize the new
11535 (record_group_use): Add a mem_type argument. Use address_p.
11536 Remove obsolete null checks of base_object. Update call to record_use.
11537 (find_interesting_uses_op): Update call to record_group_use.
11538 (find_interesting_uses_cond): Likewise.
11539 (find_interesting_uses_address): Likewise.
11540 (get_mem_type_for_internal_fn): New function.
11541 (find_address_like_use): Likewise.
11542 (find_interesting_uses_stmt): Try find_address_like_use before
11543 calling find_interesting_uses_op.
11544 (addr_offset_valid_p): Use the iv mem_type field as the type
11545 of the addressed memory.
11546 (add_autoinc_candidates): Likewise.
11547 (get_address_cost): Likewise.
11548 (split_small_address_groups_p): Use address_p.
11549 (split_address_groups): Likewise.
11550 (add_iv_candidate_for_use): Likewise.
11551 (autoinc_possible_for_pair): Likewise.
11552 (rewrite_groups): Likewise.
11553 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11554 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11555 (get_alias_ptr_type_for_ptr_address): New function.
11556 (rewrite_use_address): Rewrite address uses in calls that were
11557 identified by find_address_like_use.
11559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11560 Alan Hayward <alan.hayward@arm.com>
11561 David Sherwood <david.sherwood@arm.com>
11563 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11565 * gimple-expr.h (is_gimple_addressable: Likewise.
11566 * gimple-expr.c (is_gimple_address): Likewise.
11567 * internal-fn.c (expand_call_mem_ref): New function.
11568 (expand_mask_load_optab_fn): Use it.
11569 (expand_mask_store_optab_fn): Likewise.
11571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11572 Alan Hayward <alan.hayward@arm.com>
11573 David Sherwood <david.sherwood@arm.com>
11575 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11576 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11577 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11578 (cond_umax@var{mode}): Document.
11579 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11580 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11581 (cond_umin_optab, cond_umax_optab): New optabs.
11582 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11583 (COND_IOR, COND_XOR): New internal functions.
11584 * internal-fn.h (get_conditional_internal_fn): Declare.
11585 * internal-fn.c (cond_binary_direct): New macro.
11586 (expand_cond_binary_optab_fn): Likewise.
11587 (direct_cond_binary_optab_supported_p): Likewise.
11588 (get_conditional_internal_fn): New function.
11589 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11590 Cope with reduction statements that are vectorized as calls rather
11592 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11593 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11594 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11595 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11596 (UNSPEC_COND_EOR): New unspecs.
11597 (optab): Add mappings for them.
11598 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11599 (sve_int_op, sve_fp_op): New int attributes.
11601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11602 Alan Hayward <alan.hayward@arm.com>
11603 David Sherwood <david.sherwood@arm.com>
11605 * optabs.def (while_ult_optab): New optab.
11606 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11607 * internal-fn.def (WHILE_ULT): New internal function.
11608 * internal-fn.h (direct_internal_fn_supported_p): New override
11609 that takes two types as argument.
11610 * internal-fn.c (while_direct): New macro.
11611 (expand_while_optab_fn): New function.
11612 (convert_optab_supported_p): Likewise.
11613 (direct_while_optab_supported_p): New macro.
11614 * wide-int.h (wi::udiv_ceil): New function.
11615 * tree-vectorizer.h (rgroup_masks): New structure.
11616 (vec_loop_masks): New typedef.
11617 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11618 and fully_masked_p.
11619 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11620 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11621 (vect_max_vf): New function.
11622 (slpeel_make_loop_iterate_ntimes): Delete.
11623 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11624 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11625 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11626 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11627 internal-fn.h, stor-layout.h and optabs-query.h.
11628 (vect_set_loop_mask): New function.
11629 (add_preheader_seq): Likewise.
11630 (add_header_seq): Likewise.
11631 (interleave_supported_p): Likewise.
11632 (vect_maybe_permute_loop_masks): Likewise.
11633 (vect_set_loop_masks_directly): Likewise.
11634 (vect_set_loop_condition_masked): Likewise.
11635 (vect_set_loop_condition_unmasked): New function, split out from
11636 slpeel_make_loop_iterate_ntimes.
11637 (slpeel_make_loop_iterate_ntimes): Rename to..
11638 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11639 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11640 (vect_do_peeling): Update call accordingly.
11641 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11643 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11644 mask_compare_type, can_fully_mask_p and fully_masked_p.
11645 (release_vec_loop_masks): New function.
11646 (_loop_vec_info): Use it to free the loop masks.
11647 (can_produce_all_loop_masks_p): New function.
11648 (vect_get_max_nscalars_per_iter): Likewise.
11649 (vect_verify_full_masking): Likewise.
11650 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11651 retries, and free the mask rgroups before retrying. Check loop-wide
11652 reasons for disallowing fully-masked loops. Make the final decision
11653 about whether use a fully-masked loop or not.
11654 (vect_estimate_min_profitable_iters): Do not assume that peeling
11655 for the number of iterations will be needed for fully-masked loops.
11656 (vectorizable_reduction): Disable fully-masked loops.
11657 (vectorizable_live_operation): Likewise.
11658 (vect_halve_mask_nunits): New function.
11659 (vect_double_mask_nunits): Likewise.
11660 (vect_record_loop_mask): Likewise.
11661 (vect_get_loop_mask): Likewise.
11662 (vect_transform_loop): Handle the case in which the final loop
11663 iteration might handle a partial vector. Call vect_set_loop_condition
11664 instead of slpeel_make_loop_iterate_ntimes.
11665 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11666 (check_load_store_masking): New function.
11667 (prepare_load_store_mask): Likewise.
11668 (vectorizable_store): Handle fully-masked loops.
11669 (vectorizable_load): Likewise.
11670 (supportable_widening_operation): Use vect_halve_mask_nunits for
11672 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11673 (vect_gen_while): New function.
11674 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11675 (aarch64_uqdec<mode>): New insn.
11677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11678 Alan Hayward <alan.hayward@arm.com>
11679 David Sherwood <david.sherwood@arm.com>
11681 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11682 (reduc_xor_scal_optab): New optabs.
11683 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11684 (reduc_xor_scal_@var{m}): Document.
11685 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11686 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11687 internal functions.
11688 * fold-const-call.c (fold_const_call): Handle them.
11689 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11690 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11691 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11692 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11693 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11694 (UNSPEC_XORV): New unspecs.
11695 (optab): Add entries for them.
11696 (BITWISEV): New int iterator.
11697 (bit_reduc_op): New int attributes.
11699 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11700 Alan Hayward <alan.hayward@arm.com>
11701 David Sherwood <david.sherwood@arm.com>
11703 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11704 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11705 * optabs.def (vec_shl_insert_optab): New optab.
11706 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11707 (duplicate_and_interleave): Likewise.
11708 * tree-vect-loop.c: Include internal-fn.h.
11709 (neutral_op_for_slp_reduction): New function, split out from
11710 get_initial_defs_for_reduction.
11711 (get_initial_def_for_reduction): Handle option 2 for variable-length
11712 vectors by loading the neutral value into a vector and then shifting
11713 the initial value into element 0.
11714 (get_initial_defs_for_reduction): Replace the code argument with
11715 the neutral value calculated by neutral_op_for_slp_reduction.
11716 Use gimple_build_vector for constant-length vectors.
11717 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11718 but the first group_size elements have a neutral value.
11719 Use duplicate_and_interleave otherwise.
11720 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11721 Update call to get_initial_defs_for_reduction. Handle SLP
11722 reductions for variable-length vectors by creating one vector
11723 result for each scalar result, with the elements associated
11724 with other scalar results stubbed out with the neutral value.
11725 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11726 Require IFN_VEC_SHL_INSERT for double reductions on
11727 variable-length vectors, or SLP reductions that have
11728 a neutral value. Require can_duplicate_and_interleave_p
11729 support for variable-length unchained SLP reductions if there
11730 is no neutral value, such as for MIN/MAX reductions. Also require
11731 the number of vector elements to be a multiple of the number of
11732 SLP statements when doing variable-length unchained SLP reductions.
11733 Update call to vect_create_epilog_for_reduction.
11734 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11735 and remove initial values.
11736 (duplicate_and_interleave): Make public.
11737 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11738 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11741 Alan Hayward <alan.hayward@arm.com>
11742 David Sherwood <david.sherwood@arm.com>
11744 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11745 (can_duplicate_and_interleave_p): New function.
11746 (vect_get_and_check_slp_defs): Take the vector of statements
11747 rather than just the current one. Remove excess parentheses.
11748 Restriction rejectinon of vect_constant_def and vect_external_def
11749 for variable-length vectors to boolean types, or types for which
11750 can_duplicate_and_interleave_p is false.
11751 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11752 (duplicate_and_interleave): New function.
11753 (vect_get_constant_vectors): Use gimple_build_vector for
11754 constant-length vectors and suitable variable-length constant
11755 vectors. Use duplicate_and_interleave for other variable-length
11756 vectors. Don't defer the update when inserting new statements.
11758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11759 Alan Hayward <alan.hayward@arm.com>
11760 David Sherwood <david.sherwood@arm.com>
11762 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11763 min_profitable_iters doesn't go negative.
11765 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11766 Alan Hayward <alan.hayward@arm.com>
11767 David Sherwood <david.sherwood@arm.com>
11769 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11770 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11771 * optabs.def (vec_mask_load_lanes_optab): New optab.
11772 (vec_mask_store_lanes_optab): Likewise.
11773 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11774 (MASK_STORE_LANES): Likewise.
11775 * internal-fn.c (mask_load_lanes_direct): New macro.
11776 (mask_store_lanes_direct): Likewise.
11777 (expand_mask_load_optab_fn): Handle masked operations.
11778 (expand_mask_load_lanes_optab_fn): New macro.
11779 (expand_mask_store_optab_fn): Handle masked operations.
11780 (expand_mask_store_lanes_optab_fn): New macro.
11781 (direct_mask_load_lanes_optab_supported_p): Likewise.
11782 (direct_mask_store_lanes_optab_supported_p): Likewise.
11783 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11785 (vect_load_lanes_supported): Likewise.
11786 * tree-vect-data-refs.c (strip_conversion): New function.
11787 (can_group_stmts_p): Likewise.
11788 (vect_analyze_data_ref_accesses): Use it instead of checking
11789 for a pair of assignments.
11790 (vect_store_lanes_supported): Take a masked_p parameter.
11791 (vect_load_lanes_supported): Likewise.
11792 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11793 vect_store_lanes_supported and vect_load_lanes_supported.
11794 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11795 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11796 parameter. Don't allow gaps for masked accesses.
11797 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11798 and vect_load_lanes_supported.
11799 (get_load_store_type): Take a masked_p parameter and update
11800 call to get_group_load_store_type.
11801 (vectorizable_store): Update call to get_load_store_type.
11802 Handle IFN_MASK_STORE_LANES.
11803 (vectorizable_load): Update call to get_load_store_type.
11804 Handle IFN_MASK_LOAD_LANES.
11806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11812 * config/aarch64/aarch64-protos.h
11813 (aarch64_sve_struct_memory_operand_p): Declare.
11814 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11815 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11816 (VPRED, vpred): Handle SVE structure modes.
11817 * config/aarch64/constraints.md (Utx): New constraint.
11818 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11819 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11820 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11821 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11822 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11823 structure modes. Split into pieces after RA.
11824 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11825 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11827 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11828 SVE structure modes.
11829 (aarch64_classify_address): Likewise.
11830 (sizetochar): Move earlier in file.
11831 (aarch64_print_operand): Handle SVE register lists.
11832 (aarch64_array_mode): New function.
11833 (aarch64_sve_struct_memory_operand_p): Likewise.
11834 (TARGET_ARRAY_MODE): Redefine.
11836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11837 Alan Hayward <alan.hayward@arm.com>
11838 David Sherwood <david.sherwood@arm.com>
11840 * target.def (array_mode): New target hook.
11841 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11842 * doc/tm.texi: Regenerate.
11843 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11844 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11845 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11846 targetm.array_mode.
11847 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11851 Alan Hayward <alan.hayward@arm.com>
11852 David Sherwood <david.sherwood@arm.com>
11854 * fold-const.c (fold_binary_loc): Check the argument types
11855 rather than the result type when testing for a vector operation.
11857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11859 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11860 * doc/tm.texi: Regenerate.
11862 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11863 Alan Hayward <alan.hayward@arm.com>
11864 David Sherwood <david.sherwood@arm.com>
11866 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11867 (sve): Document new AArch64 extension.
11868 * doc/md.texi (w): Extend the description of the AArch64
11869 constraint to include SVE vectors.
11870 (Upl, Upa): Document new AArch64 predicate constraints.
11871 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11873 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11874 (msve-vector-bits=): New option.
11875 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11876 SVE when these are disabled.
11877 (sve): New extension.
11878 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11879 modes. Adjust their number of units based on aarch64_sve_vg.
11880 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11881 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11882 aarch64_addr_query_type.
11883 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11884 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11885 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11886 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11887 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11888 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11889 (aarch64_simd_imm_zero_p): Delete.
11890 (aarch64_check_zero_based_sve_index_immediate): Declare.
11891 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11892 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11893 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11894 (aarch64_sve_float_mul_immediate_p): Likewise.
11895 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11896 rather than an rtx.
11897 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11898 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11899 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11900 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11901 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11902 (aarch64_regmode_natural_size): Likewise.
11903 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11904 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11906 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11907 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11908 for VG and the SVE predicate registers.
11909 (V_ALIASES): Add a "z"-prefixed alias.
11910 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11911 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11912 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11913 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11914 (REG_CLASS_NAMES): Add entries for them.
11915 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11916 and the predicate registers.
11917 (aarch64_sve_vg): Declare.
11918 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11919 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11920 (REGMODE_NATURAL_SIZE): Define.
11921 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11923 * config/aarch64/aarch64.c: Include cfgrtl.h.
11924 (simd_immediate_info): Add a constructor for series vectors,
11925 and an associated step field.
11926 (aarch64_sve_vg): New variable.
11927 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11928 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11929 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11930 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11931 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11932 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11933 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11934 (aarch64_get_mask_mode): New functions.
11935 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11936 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11937 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11938 predicate modes and predicate registers. Explicitly restrict
11939 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11940 to store a vector mode if it is recognized by
11941 aarch64_classify_vector_mode.
11942 (aarch64_regmode_natural_size): New function.
11943 (aarch64_hard_regno_caller_save_mode): Return the original mode
11945 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11946 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11947 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11948 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11950 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11951 does not overlap dest if the function is frame-related. Handle
11953 (aarch64_split_add_offset): New function.
11954 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11955 them aarch64_add_offset.
11956 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11957 and update call to aarch64_sub_sp.
11958 (aarch64_add_cfa_expression): New function.
11959 (aarch64_expand_prologue): Pass extra temporary registers to the
11960 functions above. Handle the case in which we need to emit new
11961 DW_CFA_expressions for registers that were originally saved
11962 relative to the stack pointer, but now have to be expressed
11963 relative to the frame pointer.
11964 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11966 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11967 IP0 and IP1 values for SVE frames.
11968 (aarch64_expand_vec_series): New function.
11969 (aarch64_expand_sve_widened_duplicate): Likewise.
11970 (aarch64_expand_sve_const_vector): Likewise.
11971 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11972 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11973 into the register, rather than emitting a SET directly.
11974 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11975 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11976 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11977 (offset_9bit_signed_scaled_p): New functions.
11978 (aarch64_replicate_bitmask_imm): New function.
11979 (aarch64_bitmask_imm): Use it.
11980 (aarch64_cannot_force_const_mem): Reject expressions involving
11981 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11982 (aarch64_classify_index): Handle SVE indices, by requiring
11983 a plain register index with a scale that matches the element size.
11984 (aarch64_classify_address): Handle SVE addresses. Assert that
11985 the mode of the address is VOIDmode or an integer mode.
11986 Update call to aarch64_classify_symbol.
11987 (aarch64_classify_symbolic_expression): Update call to
11988 aarch64_classify_symbol.
11989 (aarch64_const_vec_all_in_range_p): New function.
11990 (aarch64_print_vector_float_operand): Likewise.
11991 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11992 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11993 and the FP immediates 1.0 and 0.5.
11994 (aarch64_print_address_internal): Handle SVE addresses.
11995 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11996 (aarch64_regno_regclass): Handle predicate registers.
11997 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11999 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12000 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12001 (aarch64_convert_sve_vector_bits): New function.
12002 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12003 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12004 rather than an rtx.
12005 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12006 Handle SVE vector and predicate modes. Accept VL-based constants
12007 that need only one temporary register, and VL offsets that require
12008 no temporary registers.
12009 (aarch64_conditional_register_usage): Mark the predicate registers
12010 as fixed if SVE isn't available.
12011 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12012 Return true for SVE vector and predicate modes.
12013 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12014 rather than an unsigned int. Handle SVE modes.
12015 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12017 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12019 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12020 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12021 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12022 (aarch64_sve_float_mul_immediate_p): New functions.
12023 (aarch64_sve_valid_immediate): New function.
12024 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12025 Explicitly reject structure modes. Check for INDEX constants.
12026 Handle PTRUE and PFALSE constants.
12027 (aarch64_check_zero_based_sve_index_immediate): New function.
12028 (aarch64_simd_imm_zero_p): Delete.
12029 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12030 vector modes. Accept constants in the range of CNT[BHWD].
12031 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12032 ask for an Advanced SIMD mode.
12033 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12034 (aarch64_simd_vector_alignment): Handle SVE predicates.
12035 (aarch64_vectorize_preferred_vector_alignment): New function.
12036 (aarch64_simd_vector_alignment_reachable): Use it instead of
12038 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12039 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12041 (MAX_VECT_LEN): Delete.
12042 (expand_vec_perm_d): Add a vec_flags field.
12043 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12044 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12045 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12047 (aarch64_evpc_rev): Rename to...
12048 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12049 (aarch64_evpc_rev_global): New function.
12050 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12051 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12053 (aarch64_evpc_sve_tbl): New function.
12054 (aarch64_expand_vec_perm_const_1): Update after rename of
12055 aarch64_evpc_rev. Handle SVE permutes too, trying
12056 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12057 than aarch64_evpc_tbl.
12058 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12059 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12060 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12061 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12062 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12063 (aarch64_expand_sve_vcond): New functions.
12064 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12065 of aarch64_vector_mode_p.
12066 (aarch64_dwarf_poly_indeterminate_value): New function.
12067 (aarch64_compute_pressure_classes): Likewise.
12068 (aarch64_can_change_mode_class): Likewise.
12069 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12070 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12071 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12072 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12073 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12074 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12075 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12076 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12078 (Dn, Dl, Dr): Accept const as well as const_vector.
12079 (Dz): Likewise. Compare against CONST0_RTX.
12080 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12081 of "vector" where appropriate.
12082 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12083 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12084 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12085 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12086 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12087 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12088 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12089 (v_int_equiv): Extend to SVE modes.
12090 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12092 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12093 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12094 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12095 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12096 (SVE_COND_FP_CMP): New int iterators.
12097 (perm_hilo): Handle the new unpack unspecs.
12098 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12100 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12101 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12102 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12103 (aarch64_equality_operator, aarch64_constant_vector_operand)
12104 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12105 (aarch64_sve_nonimmediate_operand): Likewise.
12106 (aarch64_sve_general_operand): Likewise.
12107 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12108 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12109 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12110 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12111 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12112 (aarch64_sve_float_arith_immediate): Likewise.
12113 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12114 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12115 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12116 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12117 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12118 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12119 (aarch64_sve_float_arith_operand): Likewise.
12120 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12121 (aarch64_sve_float_mul_operand): Likewise.
12122 (aarch64_sve_vec_perm_operand): Likewise.
12123 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12124 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12125 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12126 as well as const_vector.
12127 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12128 in file. Use CONST0_RTX and CONSTM1_RTX.
12129 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12130 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12131 Use aarch64_simd_imm_zero.
12132 * config/aarch64/aarch64-sve.md: New file.
12133 * config/aarch64/aarch64.md: Include it.
12134 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12135 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12136 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12137 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12138 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12139 (sve): New attribute.
12140 (enabled): Disable instructions with the sve attribute unless
12142 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12143 aarch64_expand_mov_immediate.
12144 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12145 CNT[BHSD] immediates.
12146 (movti): Split CONST_POLY_INT moves into two halves.
12147 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12148 Split additions that need a temporary here if the destination
12149 is the stack pointer.
12150 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12151 (*add<mode>3_poly_1): New instruction.
12152 (set_clobber_cc): New expander.
12154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12156 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12157 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12158 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12159 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12160 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12161 Change innermode from fixed_mode_size to machine_mode.
12162 (simplify_subreg): Update call accordingly. Handle a constant-sized
12163 subreg of a variable-length CONST_VECTOR.
12165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12166 Alan Hayward <alan.hayward@arm.com>
12167 David Sherwood <david.sherwood@arm.com>
12169 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12170 (add_offset_to_base): New function, split out from...
12171 (create_mem_ref): ...here. When handling a scale other than 1,
12172 check first whether the address is valid without the offset.
12173 Add it into the base if so, leaving the index and scale as-is.
12175 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12178 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12179 fold_for_warn before checking if arg2 is INTEGER_CST.
12181 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12183 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12184 (store_multiple_operation): Delete.
12185 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12186 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12187 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12188 guarded by TARGET_STRING.
12189 (rs6000_output_load_multiple): Delete.
12190 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12191 OPTION_MASK_STRING / TARGET_STRING handling.
12192 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12193 (const rs6000_opt_masks) <"string">: Change mask to 0.
12194 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12195 (MASK_STRING): Delete.
12196 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12198 (load_multiple): Delete.
12205 (store_multiple): Delete.
12212 (movmemsi_8reg): Delete.
12213 (corresponding unnamed define_insn): Delete.
12214 (movmemsi_6reg): Delete.
12215 (corresponding unnamed define_insn): Delete.
12216 (movmemsi_4reg): Delete.
12217 (corresponding unnamed define_insn): Delete.
12218 (movmemsi_2reg): Delete.
12219 (corresponding unnamed define_insn): Delete.
12220 (movmemsi_1reg): Delete.
12221 (corresponding unnamed define_insn): Delete.
12222 * config/rs6000/rs6000.opt (mno-string): New.
12223 (mstring): Replace by deprecation warning stub.
12224 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12226 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12228 * regrename.c (regrename_do_replace): If replacing the same
12229 reg multiple times, try to reuse last created gen_raw_REG.
12232 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12233 main to workaround a bug in GDB.
12235 2018-01-12 Tom de Vries <tom@codesourcery.com>
12238 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12240 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12242 PR rtl-optimization/80481
12243 * ira-color.c (get_cap_member): New function.
12244 (allocnos_conflict_by_live_ranges_p): Use it.
12245 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12246 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12248 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12251 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12252 (*saddl_se_1): Ditto.
12253 (*ssubsi_1): Ditto.
12254 (*ssubl_se_1): Ditto.
12256 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12258 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12259 rather than wi::to_widest for DR_INITs.
12260 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12261 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12262 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12264 (vect_analyze_group_access_1): Note that here.
12266 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12268 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12269 polynomial type sizes.
12271 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12273 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12274 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12275 (gimple_add_tmp_var): Likewise.
12277 2018-01-12 Martin Liska <mliska@suse.cz>
12279 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12280 (gimple_alloc_sizes): Likewise.
12281 (dump_gimple_statistics): Use PRIu64 in printf format.
12282 * gimple.h: Change uint64_t to int.
12284 2018-01-12 Martin Liska <mliska@suse.cz>
12286 * tree-core.h: Use uint64_t instead of int.
12287 * tree.c (tree_node_counts): Likewise.
12288 (tree_node_sizes): Likewise.
12289 (dump_tree_statistics): Use PRIu64 in printf format.
12291 2018-01-12 Martin Liska <mliska@suse.cz>
12293 * Makefile.in: As qsort_chk is implemented in vec.c, add
12294 vec.o to linkage of gencfn-macros.
12295 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12296 passing the info to record_node_allocation_statistics.
12297 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12299 * ggc-common.c (struct ggc_usage): Add operator== and use
12300 it in operator< and compare function.
12301 * mem-stats.h (struct mem_usage): Likewise.
12302 * vec.c (struct vec_usage): Remove operator< and compare
12303 function. Can be simply inherited.
12305 2018-01-12 Martin Jambor <mjambor@suse.cz>
12308 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12309 * tree-ssa-math-opts.c: Include domwalk.h.
12310 (convert_mult_to_fma_1): New function.
12311 (fma_transformation_info): New type.
12312 (fma_deferring_state): Likewise.
12313 (cancel_fma_deferring): New function.
12314 (result_of_phi): Likewise.
12315 (last_fma_candidate_feeds_initial_phi): Likewise.
12316 (convert_mult_to_fma): Added deferring logic, split actual
12317 transformation to convert_mult_to_fma_1.
12318 (math_opts_dom_walker): New type.
12319 (math_opts_dom_walker::after_dom_children): New method, body moved
12320 here from pass_optimize_widening_mul::execute, added deferring logic
12322 (pass_optimize_widening_mul::execute): Moved most of code to
12323 math_opts_dom_walker::after_dom_children.
12324 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12325 * config/i386/i386.c (ix86_option_override_internal): Added
12326 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12328 2018-01-12 Richard Biener <rguenther@suse.de>
12331 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12332 inline instance vars.
12334 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12337 * config/rx/rx.c (rx_is_restricted_memory_address):
12338 Handle SUBREG case.
12340 2018-01-12 Richard Biener <rguenther@suse.de>
12342 PR tree-optimization/80846
12343 * target.def (split_reduction): New target hook.
12344 * targhooks.c (default_split_reduction): New function.
12345 * targhooks.h (default_split_reduction): Declare.
12346 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12347 target requests first reduce vectors by combining low and high
12349 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12350 (get_vectype_for_scalar_type_and_size): Export.
12351 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12352 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12353 * doc/tm.texi: Regenerate.
12354 * config/i386/i386.c (ix86_split_reduction): Implement
12355 TARGET_VECTORIZE_SPLIT_REDUCTION.
12357 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12360 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12361 in PIC mode except for TARGET_VXWORKS_RTP.
12362 * config/sparc/sparc.c: Include cfgrtl.h.
12363 (TARGET_INIT_PIC_REG): Define.
12364 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12365 (sparc_pic_register_p): New predicate.
12366 (sparc_legitimate_address_p): Use it.
12367 (sparc_legitimize_pic_address): Likewise.
12368 (sparc_delegitimize_address): Likewise.
12369 (sparc_mode_dependent_address_p): Likewise.
12370 (gen_load_pcrel_sym): Remove 4th parameter.
12371 (load_got_register): Adjust call to above. Remove obsolete stuff.
12372 (sparc_expand_prologue): Do not call load_got_register here.
12373 (sparc_flat_expand_prologue): Likewise.
12374 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12375 (sparc_use_pseudo_pic_reg): New function.
12376 (sparc_init_pic_reg): Likewise.
12377 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12378 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12380 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12382 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12383 Add item for branch_cost.
12385 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12387 PR rtl-optimization/83565
12388 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12389 not extend the result to a larger mode for rotate operations.
12390 (num_sign_bit_copies1): Likewise.
12392 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12395 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12397 Use values-Xc.o for -pedantic.
12398 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12400 2018-01-12 Martin Liska <mliska@suse.cz>
12403 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12405 (possible_polymorphic_call_targets): Use it.
12406 (ipa_devirt): Likewise.
12408 2018-01-12 Martin Liska <mliska@suse.cz>
12410 * profile-count.h (enum profile_quality): Use 0 as invalid
12411 enum value of profile_quality.
12413 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12415 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12416 -mext-string options.
12418 2018-01-12 Richard Biener <rguenther@suse.de>
12420 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12421 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12422 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12424 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12426 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12428 * configure.ac (--with-long-double-format): Add support for the
12429 configuration option to change the default long double format on
12431 * config.gcc (powerpc*-linux*-*): Likewise.
12432 * configure: Regenerate.
12433 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12434 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12435 used without modification.
12437 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12439 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12440 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12441 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12442 MISC_BUILTIN_SPEC_BARRIER.
12443 (rs6000_init_builtins): Likewise.
12444 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12446 (speculation_barrier): New define_insn.
12447 * doc/extend.texi: Document __builtin_speculation_barrier.
12449 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12452 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12453 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12454 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12456 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12457 integral modes instead of "ss" and "sd".
12458 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12459 vectors with 32-bit and 64-bit elements.
12460 (vecdupssescalarmodesuffix): New mode attribute.
12461 (vec_dup<mode>): Use it.
12463 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12466 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12467 frame if argument is passed on stack.
12469 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12472 * ree.c (combine_reaching_defs): Optimize also
12473 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12474 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12476 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12478 PR middle-end/83189
12479 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12481 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12483 PR middle-end/83718
12484 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12485 after they are computed.
12487 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12489 PR tree-optimization/83695
12490 * gimple-loop-linterchange.cc
12491 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12492 reset cached scev information after interchange.
12493 (pass_linterchange::execute): Remove call to scev_reset_htab.
12495 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12497 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12498 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12499 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12500 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12501 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12502 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12503 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12504 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12505 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12506 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12507 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12508 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12509 (V_lane_reg): Likewise.
12510 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12512 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12513 (vfmal_lane_low<mode>_intrinsic,
12514 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12515 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12516 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12517 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12518 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12519 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12521 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12523 * config/arm/arm-cpus.in (fp16fml): New feature.
12524 (ALL_SIMD): Add fp16fml.
12525 (armv8.2-a): Add fp16fml as an option.
12526 (armv8.3-a): Likewise.
12527 (armv8.4-a): Add fp16fml as part of fp16.
12528 * config/arm/arm.h (TARGET_FP16FML): Define.
12529 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12531 * config/arm/arm-modes.def (V2HF): Define.
12532 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12533 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12534 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12535 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12536 vfmsl_low, vfmsl_high): New set of builtins.
12537 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12538 (vfml_op): New code attribute.
12539 (VFMLHALVES): New int iterator.
12540 (VFML, VFMLSEL): New mode attributes.
12541 (V_reg): Define mapping for V2HF.
12542 (V_hi, V_lo): New mode attributes.
12543 (VF_constraint): Likewise.
12544 (vfml_half, vfml_half_selector): New int attributes.
12545 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12547 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12548 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12550 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12551 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12552 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12553 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12555 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12556 Document new effective target and option set.
12558 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12560 * config/arm/arm-cpus.in (armv8_4): New feature.
12561 (ARMv8_4a): New fgroup.
12562 (armv8.4-a): New arch.
12563 * config/arm/arm-tables.opt: Regenerate.
12564 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12565 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12566 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12567 Add matching rules for -march=armv8.4-a and extensions.
12568 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12570 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12573 * config/rx/rx.md (BW): New mode attribute.
12574 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12576 2018-01-11 Richard Biener <rguenther@suse.de>
12578 PR tree-optimization/83435
12579 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12580 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12581 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12583 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12584 Alan Hayward <alan.hayward@arm.com>
12585 David Sherwood <david.sherwood@arm.com>
12587 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12589 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12590 (aarch64_print_address_internal): Use it to check for a zero offset.
12592 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12593 Alan Hayward <alan.hayward@arm.com>
12594 David Sherwood <david.sherwood@arm.com>
12596 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12597 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12598 Return a poly_int64 rather than a HOST_WIDE_INT.
12599 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12600 rather than a HOST_WIDE_INT.
12601 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12602 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12603 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12604 final_offset from HOST_WIDE_INT to poly_int64.
12605 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12606 to_constant when getting the number of units in an Advanced SIMD
12608 (aarch64_builtin_vectorized_function): Check for a constant number
12610 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12612 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12613 attribute instead of GET_MODE_NUNITS.
12614 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12615 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12616 GET_MODE_SIZE for fixed-size registers.
12617 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12618 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12619 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12620 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12621 (aarch64_print_operand, aarch64_print_address_internal)
12622 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12623 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12624 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12625 Handle polynomial GET_MODE_SIZE.
12626 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12627 wider than SImode without modification.
12628 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12629 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12630 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12631 passing and returning SVE modes.
12632 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12633 rather than GEN_INT.
12634 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12635 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12636 (aarch64_allocate_and_probe_stack_space): Likewise.
12637 (aarch64_layout_frame): Cope with polynomial offsets.
12638 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12639 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12640 polynomial offsets.
12641 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12642 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12643 poly_int64 rather than a HOST_WIDE_INT.
12644 (aarch64_get_separate_components, aarch64_process_components)
12645 (aarch64_expand_prologue, aarch64_expand_epilogue)
12646 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12647 (aarch64_anchor_offset): New function, split out from...
12648 (aarch64_legitimize_address): ...here.
12649 (aarch64_builtin_vectorization_cost): Handle polynomial
12650 TYPE_VECTOR_SUBPARTS.
12651 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12653 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12654 number of elements from the PARALLEL rather than the mode.
12655 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12656 rather than GET_MODE_BITSIZE.
12657 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12658 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12659 (aarch64_expand_vec_perm_const_1): Handle polynomial
12660 d->perm.length () and d->perm elements.
12661 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12662 Apply to_constant to d->perm elements.
12663 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12664 polynomial CONST_VECTOR_NUNITS.
12665 (aarch64_move_pointer): Take amount as a poly_int64 rather
12667 (aarch64_progress_pointer): Avoid temporary variable.
12668 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12669 the mode attribute instead of GET_MODE.
12671 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12672 Alan Hayward <alan.hayward@arm.com>
12673 David Sherwood <david.sherwood@arm.com>
12675 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12676 x exists before using it.
12677 (aarch64_add_constant_internal): Rename to...
12678 (aarch64_add_offset_1): ...this. Replace regnum with separate
12679 src and dest rtxes. Handle the case in which they're different,
12680 including when the offset is zero. Replace scratchreg with an rtx.
12681 Use 2 additions if there is no spare register into which we can
12682 move a 16-bit constant.
12683 (aarch64_add_constant): Delete.
12684 (aarch64_add_offset): Replace reg with separate src and dest
12685 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12686 Use aarch64_add_offset_1.
12687 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12688 an rtx rather than an int. Take the delta as a poly_int64
12689 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12690 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12691 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12692 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12693 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12694 and aarch64_add_sp.
12695 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12696 aarch64_add_constant.
12698 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12700 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12701 Use scalar_float_mode.
12703 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12705 * config/aarch64/aarch64-simd.md
12706 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12707 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12708 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12709 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12710 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12711 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12712 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12713 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12714 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12715 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12717 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12720 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12721 targ_options->x_arm_arch_string is non NULL.
12723 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12725 * config/aarch64/aarch64.h
12726 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12728 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12731 * expmed.c (emit_store_flag_force): Swap if const op0
12732 and change VOIDmode to mode of op0.
12734 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12736 PR rtl-optimization/83761
12737 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12738 than bytes to mode_for_size.
12740 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12742 PR middle-end/83189
12743 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12744 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12747 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12749 PR middle-end/83575
12750 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12751 when in layout mode.
12752 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12753 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12756 2018-01-10 Michael Collison <michael.collison@arm.com>
12758 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12759 * config/aarch64/aarch64-option-extension.def: Add
12760 AARCH64_OPT_EXTENSION of 'fp16fml'.
12761 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12762 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12763 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12764 * config/aarch64/constraints.md (Ui7): New constraint.
12765 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12766 (VFMLA_SEL_W): Ditto.
12769 (VFMLA16_LOW): New int iterator.
12770 (VFMLA16_HIGH): Ditto.
12771 (UNSPEC_FMLAL): New unspec.
12772 (UNSPEC_FMLSL): Ditto.
12773 (UNSPEC_FMLAL2): Ditto.
12774 (UNSPEC_FMLSL2): Ditto.
12775 (f16mac): New code attribute.
12776 * config/aarch64/aarch64-simd-builtins.def
12777 (aarch64_fmlal_lowv2sf): Ditto.
12778 (aarch64_fmlsl_lowv2sf): Ditto.
12779 (aarch64_fmlalq_lowv4sf): Ditto.
12780 (aarch64_fmlslq_lowv4sf): Ditto.
12781 (aarch64_fmlal_highv2sf): Ditto.
12782 (aarch64_fmlsl_highv2sf): Ditto.
12783 (aarch64_fmlalq_highv4sf): Ditto.
12784 (aarch64_fmlslq_highv4sf): Ditto.
12785 (aarch64_fmlal_lane_lowv2sf): Ditto.
12786 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12787 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12788 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12789 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12790 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12791 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12792 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12793 (aarch64_fmlal_lane_highv2sf): Ditto.
12794 (aarch64_fmlsl_lane_highv2sf): Ditto.
12795 (aarch64_fmlal_laneq_highv2sf): Ditto.
12796 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12797 (aarch64_fmlalq_lane_highv4sf): Ditto.
12798 (aarch64_fmlsl_lane_highv4sf): Ditto.
12799 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12800 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12801 * config/aarch64/aarch64-simd.md:
12802 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12803 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12804 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12805 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12806 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12807 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12808 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12809 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12810 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12811 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12812 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12813 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12814 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12815 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12816 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12817 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12818 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12819 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12820 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12821 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12822 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12823 (vfmlsl_low_u32): Ditto.
12824 (vfmlalq_low_u32): Ditto.
12825 (vfmlslq_low_u32): Ditto.
12826 (vfmlal_high_u32): Ditto.
12827 (vfmlsl_high_u32): Ditto.
12828 (vfmlalq_high_u32): Ditto.
12829 (vfmlslq_high_u32): Ditto.
12830 (vfmlal_lane_low_u32): Ditto.
12831 (vfmlsl_lane_low_u32): Ditto.
12832 (vfmlal_laneq_low_u32): Ditto.
12833 (vfmlsl_laneq_low_u32): Ditto.
12834 (vfmlalq_lane_low_u32): Ditto.
12835 (vfmlslq_lane_low_u32): Ditto.
12836 (vfmlalq_laneq_low_u32): Ditto.
12837 (vfmlslq_laneq_low_u32): Ditto.
12838 (vfmlal_lane_high_u32): Ditto.
12839 (vfmlsl_lane_high_u32): Ditto.
12840 (vfmlal_laneq_high_u32): Ditto.
12841 (vfmlsl_laneq_high_u32): Ditto.
12842 (vfmlalq_lane_high_u32): Ditto.
12843 (vfmlslq_lane_high_u32): Ditto.
12844 (vfmlalq_laneq_high_u32): Ditto.
12845 (vfmlslq_laneq_high_u32): Ditto.
12846 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12847 (AARCH64_FL_FOR_ARCH8_4): New.
12848 (AARCH64_ISA_F16FML): New ISA flag.
12849 (TARGET_F16FML): New feature flag for fp16fml.
12850 (doc/invoke.texi): Document new fp16fml option.
12852 2018-01-10 Michael Collison <michael.collison@arm.com>
12854 * config/aarch64/aarch64-builtins.c:
12855 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12856 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12857 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12858 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12859 (AARCH64_ISA_SHA3): New ISA flag.
12860 (TARGET_SHA3): New feature flag for sha3.
12861 * config/aarch64/iterators.md (sha512_op): New int attribute.
12862 (CRYPTO_SHA512): New int iterator.
12863 (UNSPEC_SHA512H): New unspec.
12864 (UNSPEC_SHA512H2): Ditto.
12865 (UNSPEC_SHA512SU0): Ditto.
12866 (UNSPEC_SHA512SU1): Ditto.
12867 * config/aarch64/aarch64-simd-builtins.def
12868 (aarch64_crypto_sha512hqv2di): New builtin.
12869 (aarch64_crypto_sha512h2qv2di): Ditto.
12870 (aarch64_crypto_sha512su0qv2di): Ditto.
12871 (aarch64_crypto_sha512su1qv2di): Ditto.
12872 (aarch64_eor3qv8hi): Ditto.
12873 (aarch64_rax1qv2di): Ditto.
12874 (aarch64_xarqv2di): Ditto.
12875 (aarch64_bcaxqv8hi): Ditto.
12876 * config/aarch64/aarch64-simd.md:
12877 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12878 (aarch64_crypto_sha512su0qv2di): Ditto.
12879 (aarch64_crypto_sha512su1qv2di): Ditto.
12880 (aarch64_eor3qv8hi): Ditto.
12881 (aarch64_rax1qv2di): Ditto.
12882 (aarch64_xarqv2di): Ditto.
12883 (aarch64_bcaxqv8hi): Ditto.
12884 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12885 (vsha512h2q_u64): Ditto.
12886 (vsha512su0q_u64): Ditto.
12887 (vsha512su1q_u64): Ditto.
12888 (veor3q_u16): Ditto.
12889 (vrax1q_u64): Ditto.
12890 (vxarq_u64): Ditto.
12891 (vbcaxq_u16): Ditto.
12892 * config/arm/types.md (crypto_sha512): New type attribute.
12893 (crypto_sha3): Ditto.
12894 (doc/invoke.texi): Document new sha3 option.
12896 2018-01-10 Michael Collison <michael.collison@arm.com>
12898 * config/aarch64/aarch64-builtins.c:
12899 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12900 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12901 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12902 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12903 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12904 (AARCH64_ISA_SM4): New ISA flag.
12905 (TARGET_SM4): New feature flag for sm4.
12906 * config/aarch64/aarch64-simd-builtins.def
12907 (aarch64_sm3ss1qv4si): Ditto.
12908 (aarch64_sm3tt1aq4si): Ditto.
12909 (aarch64_sm3tt1bq4si): Ditto.
12910 (aarch64_sm3tt2aq4si): Ditto.
12911 (aarch64_sm3tt2bq4si): Ditto.
12912 (aarch64_sm3partw1qv4si): Ditto.
12913 (aarch64_sm3partw2qv4si): Ditto.
12914 (aarch64_sm4eqv4si): Ditto.
12915 (aarch64_sm4ekeyqv4si): Ditto.
12916 * config/aarch64/aarch64-simd.md:
12917 (aarch64_sm3ss1qv4si): Ditto.
12918 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12919 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12920 (aarch64_sm4eqv4si): Ditto.
12921 (aarch64_sm4ekeyqv4si): Ditto.
12922 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12923 (sm3part_op): Ditto.
12924 (CRYPTO_SM3TT): Ditto.
12925 (CRYPTO_SM3PART): Ditto.
12926 (UNSPEC_SM3SS1): New unspec.
12927 (UNSPEC_SM3TT1A): Ditto.
12928 (UNSPEC_SM3TT1B): Ditto.
12929 (UNSPEC_SM3TT2A): Ditto.
12930 (UNSPEC_SM3TT2B): Ditto.
12931 (UNSPEC_SM3PARTW1): Ditto.
12932 (UNSPEC_SM3PARTW2): Ditto.
12933 (UNSPEC_SM4E): Ditto.
12934 (UNSPEC_SM4EKEY): Ditto.
12935 * config/aarch64/constraints.md (Ui2): New constraint.
12936 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12937 * config/arm/types.md (crypto_sm3): New type attribute.
12938 (crypto_sm4): Ditto.
12939 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12940 (vsm3tt1aq_u32): Ditto.
12941 (vsm3tt1bq_u32): Ditto.
12942 (vsm3tt2aq_u32): Ditto.
12943 (vsm3tt2bq_u32): Ditto.
12944 (vsm3partw1q_u32): Ditto.
12945 (vsm3partw2q_u32): Ditto.
12946 (vsm4eq_u32): Ditto.
12947 (vsm4ekeyq_u32): Ditto.
12948 (doc/invoke.texi): Document new sm4 option.
12950 2018-01-10 Michael Collison <michael.collison@arm.com>
12952 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12953 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12954 (AARCH64_FL_FOR_ARCH8_4): New.
12955 (AARCH64_FL_V8_4): New flag.
12956 (doc/invoke.texi): Document new armv8.4-a option.
12958 2018-01-10 Michael Collison <michael.collison@arm.com>
12960 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12961 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12962 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12963 * config/aarch64/aarch64-option-extension.def: Add
12964 AARCH64_OPT_EXTENSION of 'sha2'.
12965 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12966 (crypto): Disable sha2 and aes if crypto disabled.
12967 (crypto): Enable aes and sha2 if enabled.
12968 (simd): Disable sha2 and aes if simd disabled.
12969 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12971 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12972 (TARGET_SHA2): New feature flag for sha2.
12973 (TARGET_AES): New feature flag for aes.
12974 * config/aarch64/aarch64-simd.md:
12975 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12976 conditional on TARGET_AES.
12977 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12978 (aarch64_crypto_sha1hsi): Make pattern conditional
12980 (aarch64_crypto_sha1hv4si): Ditto.
12981 (aarch64_be_crypto_sha1hv4si): Ditto.
12982 (aarch64_crypto_sha1su1v4si): Ditto.
12983 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12984 (aarch64_crypto_sha1su0v4si): Ditto.
12985 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12986 (aarch64_crypto_sha256su0v4si): Ditto.
12987 (aarch64_crypto_sha256su1v4si): Ditto.
12988 (doc/invoke.texi): Document new aes and sha2 options.
12990 2018-01-10 Martin Sebor <msebor@redhat.com>
12992 PR tree-optimization/83781
12993 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12996 2018-01-11 Martin Sebor <msebor@gmail.com>
12997 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12999 PR tree-optimization/83501
13000 PR tree-optimization/81703
13002 * tree-ssa-strlen.c (get_string_cst): Rename...
13003 (get_string_len): ...to this. Handle global constants.
13004 (handle_char_store): Adjust.
13006 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13007 Jim Wilson <jimw@sifive.com>
13009 * config/riscv/riscv-protos.h (riscv_output_return): New.
13010 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13011 (riscv_attribute_table, riscv_output_return),
13012 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13013 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13014 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13016 (riscv_expand_prologue): Add early return for naked function.
13017 (riscv_expand_epilogue): Likewise.
13018 (riscv_function_ok_for_sibcall): Return false for naked function.
13019 (riscv_set_current_function): New.
13020 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13021 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13022 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13023 * doc/extend.texi (RISC-V Function Attributes): New.
13025 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13027 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13028 check for 128-bit long double before checking TCmode.
13029 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13030 128-bit long doubles before checking TFmode or TCmode.
13031 (FLOAT128_IBM_P): Likewise.
13033 2018-01-10 Martin Sebor <msebor@redhat.com>
13035 PR tree-optimization/83671
13036 * builtins.c (c_strlen): Unconditionally return zero for the empty
13038 Use -Warray-bounds for warnings.
13039 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13040 for non-constant array indices with COMPONENT_REF, arrays of
13041 arrays, and pointers to arrays.
13042 (gimple_fold_builtin_strlen): Determine and set length range for
13043 non-constant character arrays.
13045 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13047 PR middle-end/81897
13048 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13051 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13053 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13055 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13058 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13059 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13060 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13061 indexed_or_indirect_operand predicate.
13062 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13063 (*vsx_le_perm_load_v8hi): Likewise.
13064 (*vsx_le_perm_load_v16qi): Likewise.
13065 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13066 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13067 (*vsx_le_perm_store_v8hi): Likewise.
13068 (*vsx_le_perm_store_v16qi): Likewise.
13069 (eight unnamed splitters): Likewise.
13071 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13073 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13074 * config/rs6000/emmintrin.h: Likewise.
13075 * config/rs6000/mmintrin.h: Likewise.
13076 * config/rs6000/xmmintrin.h: Likewise.
13078 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13081 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13083 * tree.c (tree_nop_conversion): Return true for location wrapper
13085 (maybe_wrap_with_location): New function.
13086 (selftest::check_strip_nops): New function.
13087 (selftest::test_location_wrappers): New function.
13088 (selftest::tree_c_tests): Call it.
13089 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13090 (maybe_wrap_with_location): New decl.
13091 (EXPR_LOCATION_WRAPPER_P): New macro.
13092 (location_wrapper_p): New inline function.
13093 (tree_strip_any_location_wrapper): New inline function.
13095 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13098 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13099 stack_realign_offset for the largest alignment of stack slot
13101 (ix86_find_max_used_stack_alignment): New function.
13102 (ix86_finalize_stack_frame_flags): Use it. Set
13103 max_used_stack_alignment if we don't realign stack.
13104 * config/i386/i386.h (machine_function): Add
13105 max_used_stack_alignment.
13107 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13109 * config/arm/arm.opt (-mbranch-cost): New option.
13110 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13113 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13116 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13117 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13119 2018-01-10 Richard Biener <rguenther@suse.de>
13122 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13123 early out so it also covers the case where we have a non-NULL
13126 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13128 PR tree-optimization/83753
13129 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13130 for non-strided grouped accesses if the number of elements is 1.
13132 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13135 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13136 * i386.h (TARGET_USE_GATHER): Define.
13137 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13139 2018-01-10 Martin Liska <mliska@suse.cz>
13142 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13143 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13145 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13146 CLEANUP_NO_PARTITIONING is not set.
13148 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13150 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13151 for vectors, as a partial revert of r254296.
13152 * rtl.h (const_vec_p): Delete.
13153 (const_vec_duplicate_p): Don't test for vector CONSTs.
13154 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13155 * expmed.c (make_tree): Likewise.
13158 * common.md (E, F): Use CONSTANT_P instead of checking for
13160 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13161 checking for CONST_VECTOR.
13163 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13165 PR middle-end/83575
13166 * predict.c (force_edge_cold): Handle in more sane way edges
13167 with no prediction.
13169 2018-01-09 Carl Love <cel@us.ibm.com>
13171 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13173 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13174 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13175 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13176 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13177 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13178 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13179 * config/rs6000/rs6000-protos.h: Add extern defition for
13180 rs6000_generate_float2_double_code.
13181 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13183 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13184 (float2_v2df): Add define_expand.
13186 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13189 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13190 op_mode in the force_to_mode call.
13192 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13194 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13195 instead of checking each element individually.
13196 (aarch64_evpc_uzp): Likewise.
13197 (aarch64_evpc_zip): Likewise.
13198 (aarch64_evpc_ext): Likewise.
13199 (aarch64_evpc_rev): Likewise.
13200 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13201 instead of checking each element individually. Return true without
13203 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13204 whether all selected elements come from the same input, instead of
13205 checking each element individually. Remove calls to gen_rtx_REG,
13206 start_sequence and end_sequence and instead assert that no rtl is
13209 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13211 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13212 order of HIGH and CONST checks.
13214 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13216 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13217 if the destination isn't an SSA_NAME.
13219 2018-01-09 Richard Biener <rguenther@suse.de>
13221 PR tree-optimization/83668
13222 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13224 (canonicalize_loop_form): ... here, renamed from ...
13225 (canonicalize_loop_closed_ssa_form): ... this and amended to
13226 swap successor edges for loop exit blocks to make us use
13227 the RPO order we need for initial schedule generation.
13229 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13231 PR tree-optimization/64811
13232 * match.pd: When optimizing comparisons with Inf, avoid
13233 introducing or losing exceptions from comparisons with NaN.
13235 2018-01-09 Martin Liska <mliska@suse.cz>
13238 * asan.c (shadow_mem_size): Add gcc_assert.
13240 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13242 Don't save registers in main().
13245 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13246 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13247 * config/avr/avr.c (avr_set_current_function): Don't error if
13248 naked, OS_task or OS_main are specified at the same time.
13249 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13251 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13253 * common/config/avr/avr-common.c (avr_option_optimization_table):
13254 Switch on -mmain-is-OS_task for optimizing compilations.
13256 2018-01-09 Richard Biener <rguenther@suse.de>
13258 PR tree-optimization/83572
13259 * graphite.c: Include cfganal.h.
13260 (graphite_transform_loops): Connect infinite loops to exit
13261 and remove fake edges at the end.
13263 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13265 * ipa-inline.c (edge_badness): Revert accidental checkin.
13267 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13270 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13271 symbols; not inline clones.
13273 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13276 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13277 hard registers. Formatting fixes.
13279 PR preprocessor/83722
13280 * gcc.c (try_generate_repro): Pass
13281 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13282 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13285 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13286 Kito Cheng <kito.cheng@gmail.com>
13288 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13289 (riscv_leaf_function_p): Delete.
13290 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13292 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13294 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13296 (do_ifelse): New function.
13297 (do_isel): New function.
13298 (do_sub3): New function.
13299 (do_add3): New function.
13300 (do_load_mask_compare): New function.
13301 (do_overlap_load_compare): New function.
13302 (expand_compare_loop): New function.
13303 (expand_block_compare): Call expand_compare_loop() when appropriate.
13304 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13305 option description.
13306 (-mblock-compare-inline-loop-limit): New option.
13308 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13311 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13312 Reverse order of second and third operands in first alternative.
13313 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13314 of first and second elements in UNSPEC_VPERMR vector.
13315 (altivec_expand_vec_perm_le): Likewise.
13317 2018-01-08 Jeff Law <law@redhat.com>
13319 PR rtl-optimizatin/81308
13320 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13321 (process_switch): If group_case_labels makes a change, then set
13323 (pass_convert_switch::execute): If a switch is converted, then
13324 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13326 PR rtl-optimization/81308
13327 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13330 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13332 PR target/83663 - Revert r255946
13333 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13334 generation for cases where splatting a value is not useful.
13335 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13336 across a vec_duplicate and a paradoxical subreg forming a vector
13337 mode to a vec_concat.
13339 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13341 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13342 -march=armv8.3-a variants.
13343 * config/arm/t-multilib: Likewise.
13344 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13346 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13348 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13350 (cceq_ior_compare_complement): Give it a name so I can use it, and
13351 change boolean_or_operator predicate to boolean_operator so it can
13352 be used to generate a crand.
13353 (eqne): New code iterator.
13354 (bd/bd_neg): New code_attrs.
13355 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13356 a single define_insn.
13357 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13358 decrement (bdnzt/bdnzf/bdzt/bdzf).
13359 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13360 with the new names of the branch decrement patterns, and added the
13361 names of the branch decrement conditional patterns.
13363 2018-01-08 Richard Biener <rguenther@suse.de>
13365 PR tree-optimization/83563
13366 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13369 2018-01-08 Richard Biener <rguenther@suse.de>
13371 PR middle-end/83713
13372 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13374 2018-01-08 Richard Biener <rguenther@suse.de>
13376 PR tree-optimization/83685
13377 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13378 references to abnormals.
13380 2018-01-08 Richard Biener <rguenther@suse.de>
13383 * dwarf2out.c (output_indirect_strings): Handle empty
13384 skeleton_debug_str_hash.
13385 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13387 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13389 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13390 (emit_store_direct): Likewise.
13391 (arc_trampoline_adjust_address): Likewise.
13392 (arc_asm_trampoline_template): New function.
13393 (arc_initialize_trampoline): Use asm_trampoline_template.
13394 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13395 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13396 * config/arc/arc.md (flush_icache): Delete pattern.
13398 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13400 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13401 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13404 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13407 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13408 by not USED_FOR_TARGET.
13409 (make_pass_resolve_sw_modes): Likewise.
13411 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13413 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13416 2018-01-08 Richard Biener <rguenther@suse.de>
13418 PR middle-end/83580
13419 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13421 2018-01-08 Richard Biener <rguenther@suse.de>
13423 PR middle-end/83517
13424 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13426 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13428 PR middle-end/81897
13429 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13430 basic blocks with a small number of successors.
13431 (convert_control_dep_chain_into_preds): Improve handling of
13433 (dump_predicates): Split apart into...
13434 (dump_pred_chain): ...here...
13435 (dump_pred_info): ...and here.
13436 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13437 (can_chain_union_be_invalidated_p): Improve check for invalidation
13439 (uninit_uses_cannot_happen): Avoid unnecessary if
13440 convert_control_dep_chain_into_preds yielded nothing.
13442 2018-01-06 Martin Sebor <msebor@redhat.com>
13444 PR tree-optimization/83640
13445 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13446 subtracting negative offset from size.
13447 (builtin_access::overlap): Adjust offset bounds of the access to fall
13448 within the size of the object if possible.
13450 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13452 PR rtl-optimization/83699
13453 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13454 extract_bit_field_as_subreg to cases in which the extracted
13455 value is also a vector.
13457 * lra-constraints.c (process_alt_operands): Test for the equivalence
13458 substitutions when detecting a possible reload cycle.
13460 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13463 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13464 by default if flag_selective_schedling{,2}. Formatting fixes.
13466 PR rtl-optimization/83682
13467 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13468 if it has non-VECTOR_MODE element mode.
13469 (vec_duplicate_p): Likewise.
13471 PR middle-end/83694
13472 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13473 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13475 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13478 * config/i386/i386-builtin.def
13479 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13480 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13481 Require also OPTION_MASK_ISA_AVX512F in addition to
13482 OPTION_MASK_ISA_GFNI.
13483 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13484 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13485 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13486 to OPTION_MASK_ISA_GFNI.
13487 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13488 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13489 OPTION_MASK_ISA_AVX512BW.
13490 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13491 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13492 addition to OPTION_MASK_ISA_GFNI.
13493 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13494 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13495 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13496 to OPTION_MASK_ISA_GFNI.
13497 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13498 a requirement for all ISAs rather than any of them with a few
13500 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13502 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13503 bitmasks to be enabled with 3 exceptions, instead of requiring any
13504 enabled ISA with lots of exceptions.
13505 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13506 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13507 Change avx512bw in isa attribute to avx512f.
13508 * config/i386/sgxintrin.h: Add license boilerplate.
13509 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13510 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13511 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13512 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13514 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13515 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13516 temporarily sse2 rather than sse if not enabled already.
13519 * config/i386/sse.md (VI248_VLBW): Rename to ...
13520 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13521 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13522 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13523 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13524 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13525 mode iterator instead of VI248_VLBW.
13527 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13529 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13530 (record_modified): Skip clobbers; add debug output.
13531 (param_change_prob): Use sreal frequencies.
13533 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13535 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13536 punt for user-aligned variables.
13538 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13540 * tree-chrec.c (chrec_contains_symbols): Return true for
13543 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13546 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13547 of (x|y) == x for BICS pattern.
13549 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13551 PR tree-optimization/83605
13552 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13553 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13556 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13558 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13559 * config/epiphany/rtems.h: New file.
13561 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13562 Uros Bizjak <ubizjak@gmail.com>
13565 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13566 QIreg_operand instead of register_operand predicate.
13567 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13568 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13569 comments instead of -fmitigate[-_]rop.
13571 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13574 * cgraphunit.c (symbol_table::compile): Switch to text_section
13575 before calling assembly_start debug hook.
13576 * run-rtl-passes.c (run_rtl_passes): Likewise.
13579 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13581 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13582 range_int_cst_p rather than !symbolic_range_p before calling
13583 extract_range_from_multiplicative_op_1.
13585 2018-01-04 Jeff Law <law@redhat.com>
13587 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13588 redundant test in assertion.
13590 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13592 * doc/rtl.texi: Document machine_mode wrapper classes.
13594 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13596 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13597 using tree_to_uhwi.
13599 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13601 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13602 the VEC_PERM_EXPR fold to fail.
13604 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13607 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13608 to switched_sections.
13610 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13613 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13614 test for d.testing.
13616 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13619 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13620 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13622 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13625 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13626 is BLKmode and bitpos not zero or mode change is needed.
13628 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13631 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13634 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13637 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13638 instead of MULT rtx. Update all corresponding splitters.
13639 (*saddl_se): Ditto.
13640 (*ssub<modesuffix>): Ditto.
13641 (*ssubl_se): Ditto.
13642 (*cmp_sadd_di): Update split patterns.
13643 (*cmp_sadd_si): Ditto.
13644 (*cmp_sadd_sidi): Ditto.
13645 (*cmp_ssub_di): Ditto.
13646 (*cmp_ssub_si): Ditto.
13647 (*cmp_ssub_sidi): Ditto.
13648 * config/alpha/predicates.md (const23_operand): New predicate.
13649 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13650 Look for ASHIFT, not MULT inner operand.
13651 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13653 2018-01-04 Martin Liska <mliska@suse.cz>
13655 PR gcov-profile/83669
13656 * gcov.c (output_intermediate_file): Add version to intermediate
13658 * doc/gcov.texi: Document new field 'version' in intermediate
13659 file format. Fix location of '-k' option of gcov command.
13661 2018-01-04 Martin Liska <mliska@suse.cz>
13664 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13666 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13668 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13670 2018-01-03 Martin Sebor <msebor@redhat.com>
13672 PR tree-optimization/83655
13673 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13674 checking calls with invalid arguments.
13676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13678 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13679 (vectorizable_mask_load_store): Delete.
13680 (vectorizable_call): Return false for masked loads and stores.
13681 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13682 instead of gimple_assign_rhs1.
13683 (vectorizable_load): Handle IFN_MASK_LOAD.
13684 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13688 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13690 (vectorizable_mask_load_store): ...here.
13691 (vectorizable_load): ...and here.
13693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13695 * tree-vect-stmts.c (vect_build_all_ones_mask)
13696 (vect_build_zero_merge_argument): New functions, split out from...
13697 (vectorizable_load): ...here.
13699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13701 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13703 (vectorizable_mask_load_store): ...here.
13704 (vectorizable_store): ...and here.
13706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13708 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13710 (vectorizable_mask_load_store): ...here.
13712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13714 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13715 (vect_model_store_cost): Take a vec_load_store_type instead of a
13717 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13718 (vect_model_store_cost): Take a vec_load_store_type instead of a
13720 (vectorizable_mask_load_store): Update accordingly.
13721 (vectorizable_store): Likewise.
13722 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13726 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13727 IFN_MASK_LOAD calls here rather than...
13728 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13731 Alan Hayward <alan.hayward@arm.com>
13732 David Sherwood <david.sherwood@arm.com>
13734 * expmed.c (extract_bit_field_1): For vector extracts,
13735 fall back to extract_bit_field_as_subreg if vec_extract
13738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 Alan Hayward <alan.hayward@arm.com>
13740 David Sherwood <david.sherwood@arm.com>
13742 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13743 they are variable or constant sized.
13744 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13745 slots for constant-sized data.
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13752 handling COND_EXPRs with boolean comparisons, try to find a better
13753 basis for the mask type than the boolean itself.
13755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13757 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13758 is calculated and how it can be overridden.
13759 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13760 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13762 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13766 Alan Hayward <alan.hayward@arm.com>
13767 David Sherwood <david.sherwood@arm.com>
13769 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13770 Remove the mode argument.
13771 (aarch64_simd_valid_immediate): Remove the mode and inverse
13773 * config/aarch64/iterators.md (bitsize): New iterator.
13774 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13775 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13776 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13777 aarch64_simd_valid_immediate.
13778 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13779 (aarch64_reg_or_bic_imm): Likewise.
13780 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13781 with an insn_type enum and msl with a modifier_type enum.
13782 Replace element_width with a scalar_mode. Change the shift
13783 to unsigned int. Add constructors for scalar_float_mode and
13784 scalar_int_mode elements.
13785 (aarch64_vect_float_const_representable_p): Delete.
13786 (aarch64_can_const_movi_rtx_p)
13787 (aarch64_simd_scalar_immediate_valid_for_move)
13788 (aarch64_simd_make_constant): Update call to
13789 aarch64_simd_valid_immediate.
13790 (aarch64_advsimd_valid_immediate_hs): New function.
13791 (aarch64_advsimd_valid_immediate): Likewise.
13792 (aarch64_simd_valid_immediate): Remove mode and inverse
13793 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13794 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13795 and aarch64_float_const_representable_p on the result.
13796 (aarch64_output_simd_mov_immediate): Remove mode argument.
13797 Update call to aarch64_simd_valid_immediate and use of
13798 simd_immediate_info.
13799 (aarch64_output_scalar_simd_mov_immediate): Update call
13802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13803 Alan Hayward <alan.hayward@arm.com>
13804 David Sherwood <david.sherwood@arm.com>
13806 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13807 (mode_nunits): Likewise CONST_MODE_NUNITS.
13808 * machmode.def (ADJUST_NUNITS): Document.
13809 * genmodes.c (mode_data::need_nunits_adj): New field.
13810 (blank_mode): Update accordingly.
13811 (adj_nunits): New variable.
13812 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13814 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13815 listed in adj_nunits.
13816 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13817 listed in adj_nunits. Don't emit case statements for such modes.
13818 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13819 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13820 nothing if adj_nunits is nonnull.
13821 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13822 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13823 (emit_mode_fbit): Update use of print_maybe_const_decl.
13824 (emit_move_size): Likewise. Treat the array as non-const
13826 (emit_mode_adjustments): Handle adj_nunits.
13828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13830 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13831 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13832 (VECTOR_MODES): Use it.
13833 (make_vector_modes): Take the prefix as an argument.
13835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13836 Alan Hayward <alan.hayward@arm.com>
13837 David Sherwood <david.sherwood@arm.com>
13839 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13840 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13841 for MODE_VECTOR_BOOL.
13842 * machmode.def (VECTOR_BOOL_MODE): Document.
13843 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13844 (make_vector_bool_mode): New function.
13845 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13847 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13848 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13850 * stor-layout.c (int_mode_for_mode): Likewise.
13851 * tree.c (build_vector_type_for_mode): Likewise.
13852 * varasm.c (output_constant_pool_2): Likewise.
13853 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13854 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13855 for MODE_VECTOR_BOOL.
13856 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13857 of mode class checks.
13858 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13859 instead of a list of mode class checks.
13860 (expand_vector_scalar_condition): Likewise.
13861 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13864 Alan Hayward <alan.hayward@arm.com>
13865 David Sherwood <david.sherwood@arm.com>
13867 * machmode.h (mode_size): Change from unsigned short to
13869 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13870 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13871 or if measurement_type is not polynomial.
13872 (fixed_size_mode::includes_p): Check for constant-sized modes.
13873 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13874 return a poly_uint16 rather than an unsigned short.
13875 (emit_mode_size): Change the type of mode_size from unsigned short
13876 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13877 (emit_mode_adjustments): Cope with polynomial vector sizes.
13878 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13880 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13882 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13883 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13884 * caller-save.c (setup_save_areas): Likewise.
13885 (replace_reg_with_saved_mem): Likewise.
13886 * calls.c (emit_library_call_value_1): Likewise.
13887 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13888 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13889 (gen_lowpart_for_combine): Likewise.
13890 * convert.c (convert_to_integer_1): Likewise.
13891 * cse.c (equiv_constant, cse_insn): Likewise.
13892 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13893 (cselib_subst_to_values): Likewise.
13894 * dce.c (word_dce_process_block): Likewise.
13895 * df-problems.c (df_word_lr_mark_ref): Likewise.
13896 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13897 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13898 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13899 (rtl_for_decl_location): Likewise.
13900 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13901 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13902 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13903 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13904 (expand_expr_real_1): Likewise.
13905 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13906 (pad_below): Likewise.
13907 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13908 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13909 * ira.c (get_subreg_tracking_sizes): Likewise.
13910 * ira-build.c (ira_create_allocno_objects): Likewise.
13911 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13912 (ira_sort_regnos_for_alter_reg): Likewise.
13913 * ira-costs.c (record_operand_costs): Likewise.
13914 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13915 (resolve_simple_move): Likewise.
13916 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13917 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13918 (lra_constraints): Likewise.
13919 (CONST_POOL_OK_P): Reject variable-sized modes.
13920 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13921 (add_pseudo_to_slot, lra_spill): Likewise.
13922 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13923 * optabs-query.c (get_best_extraction_insn): Likewise.
13924 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13925 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13926 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13927 * recog.c (offsettable_address_addr_space_p): Likewise.
13928 * regcprop.c (maybe_mode_change): Likewise.
13929 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13930 * regrename.c (build_def_use): Likewise.
13931 * regstat.c (dump_reg_info): Likewise.
13932 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13933 (find_reloads, find_reloads_subreg_address): Likewise.
13934 * reload1.c (eliminate_regs_1): Likewise.
13935 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13936 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13937 (simplify_binary_operation_1, simplify_subreg): Likewise.
13938 * targhooks.c (default_function_arg_padding): Likewise.
13939 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13940 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13941 (verify_gimple_assign_ternary): Likewise.
13942 * tree-inline.c (estimate_move_cost): Likewise.
13943 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13944 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13945 (get_address_cost_ainc): Likewise.
13946 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13947 (vect_supportable_dr_alignment): Likewise.
13948 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13949 (vectorizable_reduction): Likewise.
13950 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13951 (vectorizable_operation, vectorizable_load): Likewise.
13952 * tree.c (build_same_sized_truth_vector_type): Likewise.
13953 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13954 * var-tracking.c (emit_note_insn_var_location): Likewise.
13955 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13956 (ADDR_VEC_ALIGN): Likewise.
13958 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13959 Alan Hayward <alan.hayward@arm.com>
13960 David Sherwood <david.sherwood@arm.com>
13962 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13964 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13965 or if measurement_type is polynomial.
13966 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13967 * combine.c (make_extraction): Likewise.
13968 * dse.c (find_shift_sequence): Likewise.
13969 * dwarf2out.c (mem_loc_descriptor): Likewise.
13970 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13971 (extract_bit_field, extract_low_bits): Likewise.
13972 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13973 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13974 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13975 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13976 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13977 * reload.c (find_reloads): Likewise.
13978 * reload1.c (alter_reg): Likewise.
13979 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13980 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13981 * tree-if-conv.c (predicate_mem_writes): Likewise.
13982 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13983 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13984 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13985 * valtrack.c (dead_debug_insert_temp): Likewise.
13986 * varasm.c (mergeable_constant_section): Likewise.
13987 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13990 Alan Hayward <alan.hayward@arm.com>
13991 David Sherwood <david.sherwood@arm.com>
13993 * expr.c (expand_assignment): Cope with polynomial mode sizes
13994 when assigning to a CONCAT.
13996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13997 Alan Hayward <alan.hayward@arm.com>
13998 David Sherwood <david.sherwood@arm.com>
14000 * machmode.h (mode_precision): Change from unsigned short to
14002 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14004 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14005 or if measurement_type is not polynomial.
14006 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14007 in which the mode is already known to be a scalar_int_mode.
14008 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14009 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14011 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14012 for GET_MODE_PRECISION.
14013 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14014 for GET_MODE_PRECISION.
14015 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14017 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14018 (expand_field_assignment, make_extraction): Likewise.
14019 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14020 (get_last_value): Likewise.
14021 * convert.c (convert_to_integer_1): Likewise.
14022 * cse.c (cse_insn): Likewise.
14023 * expr.c (expand_expr_real_1): Likewise.
14024 * lra-constraints.c (simplify_operand_subreg): Likewise.
14025 * optabs-query.c (can_atomic_load_p): Likewise.
14026 * optabs.c (expand_atomic_load): Likewise.
14027 (expand_atomic_store): Likewise.
14028 * ree.c (combine_reaching_defs): Likewise.
14029 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14030 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14031 * tree.h (type_has_mode_precision_p): Likewise.
14032 * ubsan.c (instrument_si_overflow): Likewise.
14034 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14035 Alan Hayward <alan.hayward@arm.com>
14036 David Sherwood <david.sherwood@arm.com>
14038 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14039 polynomial numbers of units.
14040 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14041 (valid_vector_subparts_p): New function.
14042 (build_vector_type): Remove temporary shim and take the number
14043 of units as a poly_uint64 rather than an int.
14044 (build_opaque_vector_type): Take the number of units as a
14045 poly_uint64 rather than an int.
14046 * tree.c (build_vector_from_ctor): Handle polynomial
14047 TYPE_VECTOR_SUBPARTS.
14048 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14049 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14050 (build_vector_from_val): If the number of units is variable,
14051 use build_vec_duplicate_cst for constant operands and
14052 VEC_DUPLICATE_EXPR otherwise.
14053 (make_vector_type): Remove temporary is_constant ().
14054 (build_vector_type, build_opaque_vector_type): Take the number of
14055 units as a poly_uint64 rather than an int.
14056 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14058 * cfgexpand.c (expand_debug_expr): Likewise.
14059 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14060 (store_constructor, expand_expr_real_1): Likewise.
14061 (const_scalar_mask_from_tree): Likewise.
14062 * fold-const-call.c (fold_const_reduction): Likewise.
14063 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14064 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14065 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14066 (fold_relational_const): Likewise.
14067 (native_interpret_vector): Likewise. Change the size from an
14068 int to an unsigned int.
14069 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14070 TYPE_VECTOR_SUBPARTS.
14071 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14072 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14073 duplicating a non-constant operand into a variable-length vector.
14074 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14075 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14076 * ipa-icf.c (sem_variable::equals): Likewise.
14077 * match.pd: Likewise.
14078 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14079 * print-tree.c (print_node): Likewise.
14080 * stor-layout.c (layout_type): Likewise.
14081 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14082 * tree-cfg.c (verify_gimple_comparison): Likewise.
14083 (verify_gimple_assign_binary): Likewise.
14084 (verify_gimple_assign_ternary): Likewise.
14085 (verify_gimple_assign_single): Likewise.
14086 * tree-pretty-print.c (dump_generic_node): Likewise.
14087 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14088 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14089 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14090 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14091 (vect_shift_permute_load_chain): Likewise.
14092 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14093 (expand_vector_condition, optimize_vector_constructor): Likewise.
14094 (lower_vec_perm, get_compute_type): Likewise.
14095 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14096 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14097 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14098 (vect_recog_mask_conversion_pattern): Likewise.
14099 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14100 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14101 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14102 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14103 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14104 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14105 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14106 (supportable_widening_operation): Likewise.
14107 (supportable_narrowing_operation): Likewise.
14108 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14110 * varasm.c (output_constant): Likewise.
14112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14113 Alan Hayward <alan.hayward@arm.com>
14114 David Sherwood <david.sherwood@arm.com>
14116 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14117 so that both the length == 3 and length != 3 cases set up their
14118 own permute vectors. Add comments explaining why we know the
14119 number of elements is constant.
14120 (vect_permute_load_chain): Likewise.
14122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14123 Alan Hayward <alan.hayward@arm.com>
14124 David Sherwood <david.sherwood@arm.com>
14126 * machmode.h (mode_nunits): Change from unsigned char to
14128 (ONLY_FIXED_SIZE_MODES): New macro.
14129 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14130 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14131 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14133 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14134 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14135 or if measurement_type is not polynomial.
14136 * genmodes.c (ZERO_COEFFS): New macro.
14137 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14139 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14140 Use ZERO_COEFFS when emitting initializers.
14141 * data-streamer.h (bp_pack_poly_value): New function.
14142 (bp_unpack_poly_value): Likewise.
14143 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14144 for GET_MODE_NUNITS.
14145 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14146 for GET_MODE_NUNITS.
14147 * tree.c (make_vector_type): Remove temporary shim and make
14148 the real function take the number of units as a poly_uint64
14149 rather than an int.
14150 (build_vector_type_for_mode): Handle polynomial nunits.
14151 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14152 * emit-rtl.c (const_vec_series_p_1): Likewise.
14153 (gen_rtx_CONST_VECTOR): Likewise.
14154 * fold-const.c (test_vec_duplicate_folding): Likewise.
14155 * genrecog.c (validate_pattern): Likewise.
14156 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14157 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14158 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14159 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14160 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14161 * rtlanal.c (subreg_get_info): Likewise.
14162 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14163 (vect_grouped_load_supported): Likewise.
14164 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14165 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14166 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14167 (simplify_const_unary_operation, simplify_binary_operation_1)
14168 (simplify_const_binary_operation, simplify_ternary_operation)
14169 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14170 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14171 instead of CONST_VECTOR_NUNITS.
14172 * varasm.c (output_constant_pool_2): Likewise.
14173 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14174 explicit-encoded elements in the XVEC for variable-length vectors.
14176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14178 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14181 Alan Hayward <alan.hayward@arm.com>
14182 David Sherwood <david.sherwood@arm.com>
14184 * coretypes.h (fixed_size_mode): Declare.
14185 (fixed_size_mode_pod): New typedef.
14186 * builtins.h (target_builtins::x_apply_args_mode)
14187 (target_builtins::x_apply_result_mode): Change type to
14188 fixed_size_mode_pod.
14189 * builtins.c (apply_args_size, apply_result_size, result_vector)
14190 (expand_builtin_apply_args_1, expand_builtin_apply)
14191 (expand_builtin_return): Update accordingly.
14193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14195 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14196 * cselib.c (cselib_hash_rtx): Likewise.
14197 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14198 CONST_VECTOR encoding.
14200 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14201 Jeff Law <law@redhat.com>
14204 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14205 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14206 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14207 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14210 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14211 explicitly probe *sp in a noreturn function if there were any callee
14212 register saves or frame pointer is needed.
14214 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14217 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14218 BLKmode for ternary, binary or unary expressions.
14221 * var-tracking.c (delete_vta_debug_insn): New inline function.
14222 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14223 insns from get_insns () to NULL instead of each bb separately.
14224 Use delete_vta_debug_insn. No longer static.
14225 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14226 delete_vta_debug_insns callers.
14227 * rtl.h (delete_vta_debug_insns): Declare.
14228 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14229 instead of variable_tracking_main.
14231 2018-01-03 Martin Sebor <msebor@redhat.com>
14233 PR tree-optimization/83603
14234 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14235 arguments past the endof the argument list in functions declared
14236 without a prototype.
14237 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14238 Avoid checking when arguments are null.
14240 2018-01-03 Martin Sebor <msebor@redhat.com>
14243 * doc/extend.texi (attribute const): Fix a typo.
14244 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14245 issuing -Wsuggest-attribute for void functions.
14247 2018-01-03 Martin Sebor <msebor@redhat.com>
14249 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14250 offset_int::from instead of wide_int::to_shwi.
14251 (maybe_diag_overlap): Remove assertion.
14252 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14253 * gimple-ssa-sprintf.c (format_directive): Same.
14254 (parse_directive): Same.
14255 (sprintf_dom_walker::compute_format_length): Same.
14256 (try_substitute_return_value): Same.
14258 2018-01-03 Jeff Law <law@redhat.com>
14260 PR middle-end/83654
14261 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14262 non-constant residual for zero at runtime and avoid probing in
14263 that case. Reorganize code for trailing problem to mirror handling
14266 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14268 PR tree-optimization/83501
14269 * tree-ssa-strlen.c (get_string_cst): New.
14270 (handle_char_store): Call get_string_cst.
14272 2018-01-03 Martin Liska <mliska@suse.cz>
14274 PR tree-optimization/83593
14275 * tree-ssa-strlen.c: Include tree-cfg.h.
14276 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14277 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14278 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14280 (strlen_dom_walker::before_dom_children): Call
14281 gimple_purge_dead_eh_edges. Dump tranformation with details
14283 (strlen_dom_walker::before_dom_children): Update call by adding
14284 new argument cleanup_eh.
14285 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14287 2018-01-03 Martin Liska <mliska@suse.cz>
14290 * cif-code.def (VARIADIC_THUNK): New enum value.
14291 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14294 2018-01-03 Jan Beulich <jbeulich@suse.com>
14296 * sse.md (mov<mode>_internal): Tighten condition for when to use
14297 vmovdqu<ssescalarsize> for TI and OI modes.
14299 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14301 Update copyright years.
14303 2018-01-03 Martin Liska <mliska@suse.cz>
14306 * ipa-visibility.c (function_and_variable_visibility): Skip
14307 functions with noipa attribure.
14309 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14311 * gcc.c (process_command): Update copyright notice dates.
14312 * gcov-dump.c (print_version): Ditto.
14313 * gcov.c (print_version): Ditto.
14314 * gcov-tool.c (print_version): Ditto.
14315 * gengtype.c (create_file): Ditto.
14316 * doc/cpp.texi: Bump @copying's copyright year.
14317 * doc/cppinternals.texi: Ditto.
14318 * doc/gcc.texi: Ditto.
14319 * doc/gccint.texi: Ditto.
14320 * doc/gcov.texi: Ditto.
14321 * doc/install.texi: Ditto.
14322 * doc/invoke.texi: Ditto.
14324 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14326 * vector-builder.h (vector_builder::m_full_nelts): Change from
14327 unsigned int to poly_uint64.
14328 (vector_builder::full_nelts): Update prototype accordingly.
14329 (vector_builder::new_vector): Likewise.
14330 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14331 (vector_builder::operator ==): Likewise.
14332 (vector_builder::finalize): Likewise.
14333 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14334 Take the number of elements as a poly_uint64 rather than an
14336 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14337 from unsigned int to poly_uint64.
14338 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14339 (vec_perm_indices::new_vector): Likewise.
14340 (vec_perm_indices::length): Likewise.
14341 (vec_perm_indices::nelts_per_input): Likewise.
14342 (vec_perm_indices::input_nelts): Likewise.
14343 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14344 number of elements per input as a poly_uint64 rather than an
14345 unsigned int. Use the original encoding for variable-length
14346 vectors, rather than clamping each individual element.
14347 For the second and subsequent elements in each pattern,
14348 clamp the step and base before clamping their sum.
14349 (vec_perm_indices::series_p): Handle polynomial element counts.
14350 (vec_perm_indices::all_in_range_p): Likewise.
14351 (vec_perm_indices_to_tree): Likewise.
14352 (vec_perm_indices_to_rtx): Likewise.
14353 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14354 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14355 (tree_vector_builder::new_binary_operation): Handle polynomial
14356 element counts. Return false if we need to know the number
14357 of elements at compile time.
14358 * fold-const.c (fold_vec_perm): Punt if the number of elements
14359 isn't known at compile time.
14361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14363 * vec-perm-indices.h (vec_perm_builder): Change element type
14364 from HOST_WIDE_INT to poly_int64.
14365 (vec_perm_indices::element_type): Update accordingly.
14366 (vec_perm_indices::clamp): Handle polynomial element_types.
14367 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14368 (vec_perm_indices::all_in_range_p): Likewise.
14369 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14371 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14372 polynomial vec_perm_indices element types.
14373 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14374 * fold-const.c (fold_vec_perm): Likewise.
14375 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14376 * tree-vect-generic.c (lower_vec_perm): Likewise.
14377 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14378 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14379 element type to HOST_WIDE_INT.
14381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14382 Alan Hayward <alan.hayward@arm.com>
14383 David Sherwood <david.sherwood@arm.com>
14385 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14386 rather than an int. Use plus_constant.
14387 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14388 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14391 Alan Hayward <alan.hayward@arm.com>
14392 David Sherwood <david.sherwood@arm.com>
14394 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14395 a HOST_WIDE_INT to a poly_int64.
14397 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14398 Alan Hayward <alan.hayward@arm.com>
14399 David Sherwood <david.sherwood@arm.com>
14401 * calls.c (load_register_parameters): Cope with polynomial
14402 mode sizes. Require a constant size for BLKmode parameters
14403 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14404 forces a parameter to be padded at the lsb end in order to
14405 fill a complete number of words, require the parameter size
14406 to be ordered wrt UNITS_PER_WORD.
14408 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14409 Alan Hayward <alan.hayward@arm.com>
14410 David Sherwood <david.sherwood@arm.com>
14412 * reload1.c (spill_stack_slot_width): Change element type
14413 from unsigned int to poly_uint64_pod.
14414 (alter_reg): Treat mode sizes as polynomial.
14416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14417 Alan Hayward <alan.hayward@arm.com>
14418 David Sherwood <david.sherwood@arm.com>
14420 * reload.c (complex_word_subreg_p): New function.
14421 (reload_inner_reg_of_subreg, push_reload): Use it.
14423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14424 Alan Hayward <alan.hayward@arm.com>
14425 David Sherwood <david.sherwood@arm.com>
14427 * lra-constraints.c (process_alt_operands): Reject matched
14428 operands whose sizes aren't ordered.
14429 (match_reload): Refer to this check here.
14431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14432 Alan Hayward <alan.hayward@arm.com>
14433 David Sherwood <david.sherwood@arm.com>
14435 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14436 that the mode size is in the set {1, 2, 4, 8, 16}.
14438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14439 Alan Hayward <alan.hayward@arm.com>
14440 David Sherwood <david.sherwood@arm.com>
14442 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14443 Use plus_constant instead of gen_rtx_PLUS.
14445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14446 Alan Hayward <alan.hayward@arm.com>
14447 David Sherwood <david.sherwood@arm.com>
14449 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14450 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14451 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14452 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14453 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14454 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14455 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14456 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14457 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14458 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14460 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14461 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14462 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14463 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14464 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14465 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14466 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14467 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14468 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14469 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14471 * expr.c (emit_move_resolve_push): Treat the input and result
14472 of PUSH_ROUNDING as a poly_int64.
14473 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14474 (emit_push_insn): Likewise.
14475 * lra-eliminations.c (mark_not_eliminable): Likewise.
14476 * recog.c (push_operand): Likewise.
14477 * reload1.c (elimination_effects): Likewise.
14478 * rtlanal.c (nonzero_bits1): Likewise.
14479 * calls.c (store_one_arg): Likewise. Require the padding to be
14480 known at compile time.
14482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14483 Alan Hayward <alan.hayward@arm.com>
14484 David Sherwood <david.sherwood@arm.com>
14486 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14487 Use plus_constant instead of gen_rtx_PLUS.
14489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14490 Alan Hayward <alan.hayward@arm.com>
14491 David Sherwood <david.sherwood@arm.com>
14493 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14494 rather than an int.
14496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14497 Alan Hayward <alan.hayward@arm.com>
14498 David Sherwood <david.sherwood@arm.com>
14500 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14501 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14502 via stack temporaries. Treat the mode size as polynomial too.
14504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14505 Alan Hayward <alan.hayward@arm.com>
14506 David Sherwood <david.sherwood@arm.com>
14508 * expr.c (expand_expr_real_2): When handling conversions involving
14509 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14510 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14511 as a poly_uint64 too.
14513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14514 Alan Hayward <alan.hayward@arm.com>
14515 David Sherwood <david.sherwood@arm.com>
14517 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14520 Alan Hayward <alan.hayward@arm.com>
14521 David Sherwood <david.sherwood@arm.com>
14523 * combine.c (can_change_dest_mode): Handle polynomial
14524 REGMODE_NATURAL_SIZE.
14525 * expmed.c (store_bit_field_1): Likewise.
14526 * expr.c (store_constructor): Likewise.
14527 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14528 and polynomial REGMODE_NATURAL_SIZE.
14529 (gen_lowpart_common): Likewise.
14530 * reginfo.c (record_subregs_of_mode): Likewise.
14531 * rtlanal.c (read_modify_subreg_p): Likewise.
14533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14534 Alan Hayward <alan.hayward@arm.com>
14535 David Sherwood <david.sherwood@arm.com>
14537 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14538 numbers of elements.
14540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14541 Alan Hayward <alan.hayward@arm.com>
14542 David Sherwood <david.sherwood@arm.com>
14544 * match.pd: Cope with polynomial numbers of vector elements.
14546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14547 Alan Hayward <alan.hayward@arm.com>
14548 David Sherwood <david.sherwood@arm.com>
14550 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14551 in a POINTER_PLUS_EXPR.
14553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14554 Alan Hayward <alan.hayward@arm.com>
14555 David Sherwood <david.sherwood@arm.com>
14557 * omp-simd-clone.c (simd_clone_subparts): New function.
14558 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14559 (ipa_simd_modify_function_body): Likewise.
14561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14562 Alan Hayward <alan.hayward@arm.com>
14563 David Sherwood <david.sherwood@arm.com>
14565 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14566 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14567 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14568 (expand_vector_condition, vector_element): Likewise.
14569 (subparts_gt): New function.
14570 (get_compute_type): Use subparts_gt.
14571 (count_type_subparts): Delete.
14572 (expand_vector_operations_1): Use subparts_gt instead of
14573 count_type_subparts.
14575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14576 Alan Hayward <alan.hayward@arm.com>
14577 David Sherwood <david.sherwood@arm.com>
14579 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14580 (vect_compile_time_alias): ...this new function. Do the calculation
14581 on poly_ints rather than trees.
14582 (vect_prune_runtime_alias_test_list): Update call accordingly.
14584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14585 Alan Hayward <alan.hayward@arm.com>
14586 David Sherwood <david.sherwood@arm.com>
14588 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14590 (vect_schedule_slp_instance): Likewise.
14592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14593 Alan Hayward <alan.hayward@arm.com>
14594 David Sherwood <david.sherwood@arm.com>
14596 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14597 constant and extern definitions for variable-length vectors.
14598 (vect_get_constant_vectors): Note that the number of units
14599 is known to be constant.
14601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14602 Alan Hayward <alan.hayward@arm.com>
14603 David Sherwood <david.sherwood@arm.com>
14605 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14606 of units as polynomial. Choose between WIDE and NARROW based
14609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14610 Alan Hayward <alan.hayward@arm.com>
14611 David Sherwood <david.sherwood@arm.com>
14613 * tree-vect-stmts.c (simd_clone_subparts): New function.
14614 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14617 Alan Hayward <alan.hayward@arm.com>
14618 David Sherwood <david.sherwood@arm.com>
14620 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14621 vectors as polynomial. Use build_index_vector for
14622 IFN_GOMP_SIMD_LANE.
14624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14625 Alan Hayward <alan.hayward@arm.com>
14626 David Sherwood <david.sherwood@arm.com>
14628 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14629 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14630 for variable-length vectors.
14631 (vectorizable_mask_load_store): Treat the number of units as
14632 polynomial, asserting that it is constant if the condition has
14633 already been enforced.
14634 (vectorizable_store, vectorizable_load): Likewise.
14636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14637 Alan Hayward <alan.hayward@arm.com>
14638 David Sherwood <david.sherwood@arm.com>
14640 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14641 of units as polynomial. Punt if we can't tell at compile time
14642 which vector contains the final result.
14644 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14645 Alan Hayward <alan.hayward@arm.com>
14646 David Sherwood <david.sherwood@arm.com>
14648 * tree-vect-loop.c (vectorizable_induction): Treat the number
14649 of units as polynomial. Punt on SLP inductions. Use an integer
14650 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14651 cast of such a series for variable-length floating-point
14654 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14655 Alan Hayward <alan.hayward@arm.com>
14656 David Sherwood <david.sherwood@arm.com>
14658 * tree.h (build_index_vector): Declare.
14659 * tree.c (build_index_vector): New function.
14660 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14661 of units as polynomial, forcibly converting it to a constant if
14662 vectorizable_reduction has already enforced the condition.
14663 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14664 to create a {1,2,3,...} vector.
14665 (vectorizable_reduction): Treat the number of units as polynomial.
14666 Choose vectype_in based on the largest scalar element size rather
14667 than the smallest number of units. Enforce the restrictions
14670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14671 Alan Hayward <alan.hayward@arm.com>
14672 David Sherwood <david.sherwood@arm.com>
14674 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14675 number of units as polynomial.
14677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14678 Alan Hayward <alan.hayward@arm.com>
14679 David Sherwood <david.sherwood@arm.com>
14681 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14682 * target.def (autovectorize_vector_sizes): Return the vector sizes
14683 by pointer, using vector_sizes rather than a bitmask.
14684 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14685 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14686 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14688 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14689 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14690 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14691 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14692 * omp-general.c (omp_max_vf): Likewise.
14693 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14694 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14695 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14696 * tree-vect-slp.c (vect_slp_bb): Likewise.
14697 * doc/tm.texi: Regenerate.
14698 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14700 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14701 the vector size as a poly_uint64 rather than an unsigned int.
14702 (current_vector_size): Change from an unsigned int to a poly_uint64.
14703 (get_vectype_for_scalar_type): Update accordingly.
14704 * tree.h (build_truth_vector_type): Take the size and number of
14705 units as a poly_uint64 rather than an unsigned int.
14706 (build_vector_type): Add a temporary overload that takes
14707 the number of units as a poly_uint64 rather than an unsigned int.
14708 * tree.c (make_vector_type): Likewise.
14709 (build_truth_vector_type): Take the number of units as a poly_uint64
14710 rather than an unsigned int.
14712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14713 Alan Hayward <alan.hayward@arm.com>
14714 David Sherwood <david.sherwood@arm.com>
14716 * target.def (get_mask_mode): Take the number of units and length
14717 as poly_uint64s rather than unsigned ints.
14718 * targhooks.h (default_get_mask_mode): Update accordingly.
14719 * targhooks.c (default_get_mask_mode): Likewise.
14720 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14721 * doc/tm.texi: Regenerate.
14723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14724 Alan Hayward <alan.hayward@arm.com>
14725 David Sherwood <david.sherwood@arm.com>
14727 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14728 * omp-general.c (omp_max_vf): Likewise.
14729 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14730 (expand_omp_simd): Handle polynomial safelen.
14731 * omp-low.c (omplow_simd_context): Add a default constructor.
14732 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14733 (lower_rec_simd_input_clauses): Update accordingly.
14734 (lower_rec_input_clauses): Likewise.
14736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14737 Alan Hayward <alan.hayward@arm.com>
14738 David Sherwood <david.sherwood@arm.com>
14740 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14741 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14742 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14743 (vect_analyze_slp_cost): Likewise.
14744 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14745 (vect_model_load_cost): Likewise.
14747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14748 Alan Hayward <alan.hayward@arm.com>
14749 David Sherwood <david.sherwood@arm.com>
14751 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14752 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14753 from an unsigned int * to a poly_uint64_pod *.
14754 (calculate_unrolling_factor): New function.
14755 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14758 Alan Hayward <alan.hayward@arm.com>
14759 David Sherwood <david.sherwood@arm.com>
14761 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14762 from an unsigned int to a poly_uint64.
14763 (_loop_vec_info::slp_unrolling_factor): Likewise.
14764 (_loop_vec_info::vectorization_factor): Change from an int
14766 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14767 (vect_get_num_vectors): New function.
14768 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14769 (vect_get_num_copies): Use vect_get_num_vectors.
14770 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14771 to an unsigned int *.
14772 (vect_analyze_data_refs): Change min_vf from an int * to a
14774 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14775 than an unsigned HOST_WIDE_INT.
14776 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14777 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14778 to an unsigned int *.
14779 (vect_analyze_data_ref_dependences): Likewise.
14780 (vect_compute_data_ref_alignment): Handle polynomial vf.
14781 (vect_enhance_data_refs_alignment): Likewise.
14782 (vect_prune_runtime_alias_test_list): Likewise.
14783 (vect_shift_permute_load_chain): Likewise.
14784 (vect_supportable_dr_alignment): Likewise.
14785 (dependence_distance_ge_vf): Take the vectorization factor as a
14786 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14787 (vect_analyze_data_refs): Change min_vf from an int * to a
14789 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14790 vfm1 as a poly_uint64 rather than an int. Make the same change
14791 for the returned bound_scalar.
14792 (vect_gen_vector_loop_niters): Handle polynomial vf.
14793 (vect_do_peeling): Likewise. Update call to
14794 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14795 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14797 * tree-vect-loop.c (vect_determine_vectorization_factor)
14798 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14799 (vect_get_known_peeling_cost): Likewise.
14800 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14801 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14802 (vect_transform_loop): Likewise. Use the lowest possible VF when
14803 updating the upper bounds of the loop.
14804 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14805 rather than an int.
14806 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14807 polynomial unroll factors.
14808 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14809 (vect_make_slp_decision): Likewise.
14810 (vect_supported_load_permutation_p): Likewise, and polynomial
14812 (vect_analyze_slp_cost): Handle polynomial vf.
14813 (vect_slp_analyze_node_operations): Likewise.
14814 (vect_slp_analyze_bb_1): Likewise.
14815 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14816 than an unsigned HOST_WIDE_INT.
14817 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14818 (vectorizable_load): Handle polynomial vf.
14819 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14821 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14823 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14824 Alan Hayward <alan.hayward@arm.com>
14825 David Sherwood <david.sherwood@arm.com>
14827 * match.pd: Handle bit operations involving three constants
14828 and try to fold one pair.
14830 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14832 * tree-vect-loop-manip.c: Include gimple-fold.h.
14833 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14834 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14835 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14836 Add a path that uses a step of VF instead of 1, but disable it
14838 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14839 and niters_no_overflow parameters. Update calls to
14840 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14841 Create a new SSA name if the latter choses to use a ste other
14842 than zero, and return it via niters_vector_mult_vf_var.
14843 * tree-vect-loop.c (vect_transform_loop): Update calls to
14844 vect_do_peeling, vect_gen_vector_loop_niters and
14845 slpeel_make_loop_iterate_ntimes.
14846 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14847 (vect_gen_vector_loop_niters): Update declarations after above changes.
14849 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14851 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14852 128-bit round to integer instructions.
14853 (ceil<mode>2): Likewise.
14854 (btrunc<mode>2): Likewise.
14855 (round<mode>2): Likewise.
14857 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14859 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14860 unaligned VSX load/store on P8/P9.
14861 (expand_block_clear): Allow the use of unaligned VSX
14862 load/store on P8/P9.
14864 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14866 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14868 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14869 swap associated with both a load and a store.
14871 2018-01-02 Andrew Waterman <andrew@sifive.com>
14873 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14874 * config/riscv/riscv.md (clear_cache): Use it.
14876 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14878 * web.c: Remove out-of-date comment.
14880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14882 * expr.c (fixup_args_size_notes): Check that any existing
14883 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14884 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14885 (emit_single_push_insn): ...here.
14887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14889 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14890 (const_vector_encoded_nelts): New function.
14891 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14892 (const_vector_int_elt, const_vector_elt): Declare.
14893 * emit-rtl.c (const_vector_int_elt_1): New function.
14894 (const_vector_elt): Likewise.
14895 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14896 of CONST_VECTOR_ELT.
14898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14900 * expr.c: Include rtx-vector-builder.h.
14901 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14902 directly on the tree encoding.
14903 (const_vector_from_tree): Likewise.
14904 * optabs.c: Include rtx-vector-builder.h.
14905 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14906 sequence of "u" values.
14907 * vec-perm-indices.c: Include rtx-vector-builder.h.
14908 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14909 directly on the vec_perm_indices encoding.
14911 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14913 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14914 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14915 * rtx-vector-builder.h: New file.
14916 * rtx-vector-builder.c: Likewise.
14917 * rtl.h (rtx_def::u2): Add a const_vector field.
14918 (CONST_VECTOR_NPATTERNS): New macro.
14919 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14920 (CONST_VECTOR_DUPLICATE_P): Likewise.
14921 (CONST_VECTOR_STEPPED_P): Likewise.
14922 (CONST_VECTOR_ENCODED_ELT): Likewise.
14923 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14924 (unwrap_const_vec_duplicate): Likewise.
14925 (const_vec_series_p): Check for a non-duplicated vector encoding.
14926 Say that the function only returns true for integer vectors.
14927 * emit-rtl.c: Include rtx-vector-builder.h.
14928 (gen_const_vec_duplicate_1): Delete.
14929 (gen_const_vector): Call gen_const_vec_duplicate instead of
14930 gen_const_vec_duplicate_1.
14931 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14932 (gen_const_vec_duplicate): Use rtx_vector_builder.
14933 (gen_const_vec_series): Likewise.
14934 (gen_rtx_CONST_VECTOR): Likewise.
14935 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14936 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14937 Build a new vector rather than modifying a CONST_VECTOR in-place.
14938 (handle_special_swappables): Update call accordingly.
14939 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14940 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14941 Build a new vector rather than modifying a CONST_VECTOR in-place.
14942 (handle_special_swappables): Update call accordingly.
14944 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14946 * simplify-rtx.c (simplify_const_binary_operation): Use
14947 CONST_VECTOR_ELT instead of XVECEXP.
14949 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14951 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14952 the selector elements to be different from the data elements
14953 if the selector is a VECTOR_CST.
14954 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14955 ssizetype for the selector.
14957 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14959 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14960 before testing each element individually.
14961 * tree-vect-generic.c (lower_vec_perm): Likewise.
14963 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14965 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14966 * selftest-run-tests.c (selftest::run_tests): Call it.
14967 * vector-builder.h (vector_builder::operator ==): New function.
14968 (vector_builder::operator !=): Likewise.
14969 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14970 (vec_perm_indices::all_from_input_p): New function.
14971 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14972 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14973 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14974 instead of reading the VECTOR_CST directly. Detect whether both
14975 vector inputs are the same before constructing the vec_perm_indices,
14976 and update the number of inputs argument accordingly. Use the
14977 utility functions added above. Only construct sel2 if we need to.
14979 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14981 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14982 the broadcast of the low byte.
14983 (expand_mult_highpart): Use an explicit encoding for the permutes.
14984 * optabs-query.c (can_mult_highpart_p): Likewise.
14985 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14986 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14987 (vectorizable_bswap): Likewise.
14988 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14989 explicit encoding for the power-of-2 permutes.
14990 (vect_permute_store_chain): Likewise.
14991 (vect_grouped_load_supported): Likewise.
14992 (vect_permute_load_chain): Likewise.
14994 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14996 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14997 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14998 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14999 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15000 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15001 (vect_gen_perm_mask_any): Likewise.
15003 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15005 * int-vector-builder.h: New file.
15006 * vec-perm-indices.h: Include int-vector-builder.h.
15007 (vec_perm_indices): Redefine as an int_vector_builder.
15008 (auto_vec_perm_indices): Delete.
15009 (vec_perm_builder): Redefine as a stand-alone class.
15010 (vec_perm_indices::vec_perm_indices): New function.
15011 (vec_perm_indices::clamp): Likewise.
15012 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15013 (vec_perm_indices::new_vector): New function.
15014 (vec_perm_indices::new_expanded_vector): Update for new
15015 vec_perm_indices class.
15016 (vec_perm_indices::rotate_inputs): New function.
15017 (vec_perm_indices::all_in_range_p): Operate directly on the
15018 encoded form, without computing elided elements.
15019 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15020 encoding. Update for new vec_perm_indices class.
15021 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15022 the given vec_perm_builder.
15023 (expand_vec_perm_var): Update vec_perm_builder constructor.
15024 (expand_mult_highpart): Use vec_perm_builder instead of
15025 auto_vec_perm_indices.
15026 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15027 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15028 or double series encoding as appropriate.
15029 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15030 vec_perm_indices instead of auto_vec_perm_indices.
15031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15032 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15033 (vect_permute_store_chain): Likewise.
15034 (vect_grouped_load_supported): Likewise.
15035 (vect_permute_load_chain): Likewise.
15036 (vect_shift_permute_load_chain): Likewise.
15037 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15038 (vect_transform_slp_perm_load): Likewise.
15039 (vect_schedule_slp_instance): Likewise.
15040 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15041 (vectorizable_mask_load_store): Likewise.
15042 (vectorizable_bswap): Likewise.
15043 (vectorizable_store): Likewise.
15044 (vectorizable_load): Likewise.
15045 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15046 vec_perm_indices instead of auto_vec_perm_indices. Use
15047 tree_to_vec_perm_builder to read the vector from a tree.
15048 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15049 vec_perm_builder instead of a vec_perm_indices.
15050 (have_whole_vector_shift): Use vec_perm_builder and
15051 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15052 truncation to calc_vec_perm_mask_for_shift.
15053 (vect_create_epilog_for_reduction): Likewise.
15054 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15055 from auto_vec_perm_indices to vec_perm_indices.
15056 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15057 instead of changing individual elements.
15058 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15059 the vector in d.perm.
15060 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15061 from auto_vec_perm_indices to vec_perm_indices.
15062 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15063 instead of changing individual elements.
15064 (arm_vectorize_vec_perm_const): Use new_vector to install
15065 the vector in d.perm.
15066 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15067 Update vec_perm_builder constructor.
15068 (rs6000_expand_interleave): Likewise.
15069 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15070 (rs6000_expand_interleave): Likewise.
15072 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15074 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15075 to qimode could truncate the indices.
15076 * optabs.c (expand_vec_perm_var): Likewise.
15078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15080 * Makefile.in (OBJS): Add vec-perm-indices.o.
15081 * vec-perm-indices.h: New file.
15082 * vec-perm-indices.c: Likewise.
15083 * target.h (vec_perm_indices): Replace with a forward class
15085 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15086 * optabs.h: Include vec-perm-indices.h.
15087 (expand_vec_perm): Delete.
15088 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15089 (expand_vec_perm_const): Declare.
15090 * target.def (vec_perm_const_ok): Replace with...
15091 (vec_perm_const): ...this new hook.
15092 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15093 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15094 * doc/tm.texi: Regenerate.
15095 * optabs.def (vec_perm_const): Delete.
15096 * doc/md.texi (vec_perm_const): Likewise.
15097 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15098 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15099 expand_vec_perm for constant permutation vectors. Assert that
15100 the mode of variable permutation vectors is the integer equivalent
15101 of the mode that is being permuted.
15102 * optabs-query.h (selector_fits_mode_p): Declare.
15103 * optabs-query.c: Include vec-perm-indices.h.
15104 (selector_fits_mode_p): New function.
15105 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15106 is defined, instead of checking whether the vec_perm_const_optab
15107 exists. Use targetm.vectorize.vec_perm_const instead of
15108 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15109 fit in the vector mode before using a variable permute.
15110 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15111 vec_perm_indices instead of an rtx.
15112 (expand_vec_perm): Replace with...
15113 (expand_vec_perm_const): ...this new function. Take the selector
15114 as a vec_perm_indices rather than an rtx. Also take the mode of
15115 the selector. Update call to shift_amt_for_vec_perm_mask.
15116 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15117 Use vec_perm_indices::new_expanded_vector to expand the original
15118 selector into bytes. Check whether the indices fit in the vector
15119 mode before using a variable permute.
15120 (expand_vec_perm_var): Make global.
15121 (expand_mult_highpart): Use expand_vec_perm_const.
15122 * fold-const.c: Includes vec-perm-indices.h.
15123 * tree-ssa-forwprop.c: Likewise.
15124 * tree-vect-data-refs.c: Likewise.
15125 * tree-vect-generic.c: Likewise.
15126 * tree-vect-loop.c: Likewise.
15127 * tree-vect-slp.c: Likewise.
15128 * tree-vect-stmts.c: Likewise.
15129 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15131 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15132 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15133 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15134 (aarch64_vectorize_vec_perm_const): ...this new function.
15135 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15136 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15137 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15138 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15139 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15140 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15141 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15143 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15144 check for NEON modes.
15145 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15146 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15147 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15148 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15150 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15151 the old VEC_PERM_CONST conditions.
15152 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15153 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15154 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15155 (ia64_vectorize_vec_perm_const_ok): Merge into...
15156 (ia64_vectorize_vec_perm_const): ...this new function.
15157 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15158 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15159 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15160 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15161 * config/mips/mips.c (mips_expand_vec_perm_const)
15162 (mips_vectorize_vec_perm_const_ok): Merge into...
15163 (mips_vectorize_vec_perm_const): ...this new function.
15164 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15165 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15166 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15167 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15168 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15169 (rs6000_expand_vec_perm_const): Delete.
15170 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15172 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15173 (altivec_expand_vec_perm_const_le): Take each operand individually.
15174 Operate on constant selectors rather than rtxes.
15175 (altivec_expand_vec_perm_const): Likewise. Update call to
15176 altivec_expand_vec_perm_const_le.
15177 (rs6000_expand_vec_perm_const): Delete.
15178 (rs6000_vectorize_vec_perm_const_ok): Delete.
15179 (rs6000_vectorize_vec_perm_const): New function.
15180 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15181 an element count and rtx array.
15182 (rs6000_expand_extract_even): Update call accordingly.
15183 (rs6000_expand_interleave): Likewise.
15184 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15185 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15186 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15187 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15188 (rs6000_expand_vec_perm_const): Delete.
15189 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15190 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15191 (altivec_expand_vec_perm_const_le): Take each operand individually.
15192 Operate on constant selectors rather than rtxes.
15193 (altivec_expand_vec_perm_const): Likewise. Update call to
15194 altivec_expand_vec_perm_const_le.
15195 (rs6000_expand_vec_perm_const): Delete.
15196 (rs6000_vectorize_vec_perm_const_ok): Delete.
15197 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15198 reference to the SPE evmerge intructions.
15199 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15200 an element count and rtx array.
15201 (rs6000_expand_extract_even): Update call accordingly.
15202 (rs6000_expand_interleave): Likewise.
15203 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15204 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15206 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15210 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15211 vector mode and that that mode matches the mode of the data
15213 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15214 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15215 directly using expand_vec_perm_1 when forcing selectors into
15217 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15221 * optabs-query.h (can_vec_perm_p): Delete.
15222 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15223 * optabs-query.c (can_vec_perm_p): Split into...
15224 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15225 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15226 particular selector is valid.
15227 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15228 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15229 (vect_grouped_load_supported): Likewise.
15230 (vect_shift_permute_load_chain): Likewise.
15231 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15232 (vect_transform_slp_perm_load): Likewise.
15233 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15234 (vectorizable_bswap): Likewise.
15235 (vect_gen_perm_mask_checked): Likewise.
15236 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15237 implementations of variable permutation vectors into account
15238 when deciding which selector to use.
15239 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15240 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15241 with a false third argument.
15242 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15243 to test whether the constant selector is valid and can_vec_perm_var_p
15244 to test whether a variable selector is valid.
15246 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15248 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15249 * optabs-query.c (can_vec_perm_p): Likewise.
15250 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15251 instead of vec_perm_indices.
15252 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15253 (vect_gen_perm_mask_checked): Likewise,
15254 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15255 (vect_gen_perm_mask_checked): Likewise,
15257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15259 * optabs-query.h (qimode_for_vec_perm): Declare.
15260 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15261 (qimode_for_vec_perm): ...this new function.
15262 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15264 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15266 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15267 does not have a conditional at the top.
15269 2018-01-02 Richard Biener <rguenther@suse.de>
15271 * ipa-inline.c (big_speedup_p): Fix expression.
15273 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15276 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15279 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15283 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15284 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15285 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15286 cond_taken_branch_cost 3->4.
15288 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15290 PR tree-optimization/83581
15291 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15292 TODO_cleanup_cfg if any changes have been made.
15294 PR middle-end/83608
15295 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15296 convert_modes if target mode has the right side, but different mode
15299 PR middle-end/83609
15300 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15301 last argument when extracting from CONCAT. If either from_real or
15302 from_imag is NULL, use expansion through memory. If result is not
15303 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15304 the parts directly to inner mode, if even that fails, use expansion
15307 PR middle-end/83623
15308 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15309 check for bswap in mode rather than HImode and use that in expand_unop
15312 Copyright (C) 2018 Free Software Foundation, Inc.
15314 Copying and distribution of this file, with or without modification,
15315 are permitted in any medium without royalty provided the copyright
15316 notice and this notice are preserved.