1 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
3 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
6 2018-06-05 Steve Ellcey <sellcey@cavium.com>
9 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12 Remove second argument, change how error is called.
13 (aarch64_layout_arg): Remove second argument from
14 aarch64_err_no_fpadvsimd call.
15 (aarch64_init_cumulative_args): Ditto.
16 (aarch64_gimplify_va_arg_expr): Ditto.
17 * config/aarch64/aarch64.md (mov<mode>): Ditto.
19 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
21 * config/i386/i386.md (simple_return_indirect_internal): New expander.
22 (*simple_return_indirect_internal<mode>): Rename from
23 simple_return_indirect_internal. Use W mode iterator.
24 (rstorssp): New expander.
25 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
26 (clrssbsy): New expander.
27 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
29 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
31 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
33 (cmse_check_pointed_object): Likewise.
35 2018-06-05 Martin Liska <mliska@suse.cz>
38 * doc/invoke.texi: Document how -fprofile-dir format
41 2018-06-05 Richard Biener <rguenther@suse.de>
43 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
44 removal pretend DOM info isn't available so we do not update
45 it and only remove edges, not dominated blocks. Actually free
46 DOM info in case we removed something. Remove unreachable blocks.
47 (mfb_keep_latches): Work with either DOM info or marked backedges.
48 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
49 first. Mark backedges if DOM info isn't available.
50 (Re-)compute DOM info after cleanup_control_flow_pre.
52 2018-06-05 Richard Biener <rguenther@suse.de>
54 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
55 (locus_discrim_hasher::hash): Adjust.
56 (locus_discrim_hasher::equal): Likewise.
57 (next_discriminator_for_locus): Work on line directly.
58 (same_line_p): Pass in expanded locus1 as well.
59 (assign_discriminators): Avoid redundant location expansions.
61 2018-06-05 Richard Biener <rguenther@suse.de>
63 PR tree-optimization/86046
64 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
65 if required after clearing TREE_ADDRESSABLE.
67 2018-06-05 Richard Biener <rguenther@suse.de>
69 PR tree-optimization/86047
70 * tree-ssa-loop.c (for_each_index): Glob handling of all
71 decls and constants and really handle all of them.
73 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
76 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
77 qualifier_void_pointer and qualifier_const_void_pointer.
78 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
79 (arm_init_builtins): Handle the above.
80 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
81 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
82 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
85 2018-06-05 Martin Liska <mliska@suse.cz>
87 * auto-profile.c (read_autofdo_file): Do not use
88 gcov_ctr_summary struct.
89 (afdo_callsite_hot_enough_for_early_inline): Likewise.
90 * coverage.c (struct counts_entry): Likewise.
91 (read_counts_file): Read just single summary entry.
92 (get_coverage_counts): Use gcov_summary struct.
93 * coverage.h (get_coverage_counts): Likewise.
94 * gcov-dump.c (dump_working_sets): Likewise.
95 (tag_summary): Dump just single summary.
96 * gcov-io.c (gcov_write_summary): Write just histogram
98 (gcov_read_summary): Read just single summary.
99 (compute_working_sets): Use gcov_summary struct.
100 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
101 of GCOV_COUNTERS_SUMMABLE.
102 (GCOV_COUNTERS_SUMMABLE): Remove.
103 (GCOV_FIRST_VALUE_COUNTER): Replace with
104 GCOV_COUNTER_V_INTERVAL.
105 (struct gcov_ctr_summary): Remove.
106 (struct gcov_summary): Directly use fields of former
108 (compute_working_sets): Use gcov_summary struct.
109 * gcov.c (read_count_file): Do not use ctrs fields.
110 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
112 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
114 * profile.c: Likewise.
115 * profile.h: Likewise.
117 2018-06-05 Martin Liska <mliska@suse.cz>
119 PR gcov-profile/84846
120 * gcov.c (output_lines): Print working directory only
121 in intermediate format.
123 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
125 * config/s390/s390-builtin-types.def: Add void function type.
126 * config/s390/s390-builtins.def: Use the function type for the
129 2018-06-04 Jim Wilson <jimw@sifive.com>
131 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
133 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
134 and maybe_eh_return. Change regno to unsigned int. Use new args to
135 handle EH_RETURN_DATA_REGNO registers properly.
136 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
137 (riscv_expand_epilogue): Update comment. Change argument name and
138 type. Update code to use new name and type. Pass new args to
139 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
141 * config/riscv/riscv.md (NORMAL_RETURN): New.
142 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
143 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
144 (eh_return): Call gen_eh_return_internal and emit barrier.
145 (eh_return_internal): Call riscv_expand_epilogue.
147 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
149 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
150 bit_insertion field and declare can_be_merged_into method.
151 (merged_store_group::can_be_merged_into): New method.
152 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
153 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
154 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
156 2018-06-04 Richard Biener <rguenther@suse.de>
158 PR tree-optimization/85955
159 * builtins.c (fold_builtin_sincos): Convert pointers to
160 destination to appropriate type before dereferencing.
162 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
164 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
166 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
168 * expr.c (expand_expr_real_1): Force the operand into memory if
169 its TYPE_MODE is BLKmode and if there is no integer mode for
170 the number of bits being extracted.
172 2018-06-04 Jakub Jelinek <jakub@redhat.com>
176 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
177 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
179 2018-06-04 Richard Biener <rguenther@suse.de>
181 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
182 (cleanup_tree_cfg_noloop): ... single caller. Do
183 start_recording_case_labels later.
185 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
187 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
188 to _IMMINTRIN_H_INCLUDED.
189 * config/i386/pconfigintrin.h: Ditto.
190 * config/i386/waitpkgintrin.h: Ditto.
191 * config/i386/immintrin.h: Add includes for sgxintrin.h,
192 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
193 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
194 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
195 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
196 waitpkgintrin.h and cldemoteintrin.h.
198 2018-06-04 Richard Biener <rguenther@suse.de>
200 PR tree-optimization/86038
201 * tracer.c (find_best_successor): Check probability for
202 being initialized, bail out if not.
204 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
207 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
208 of bits to ignore when comparing architectures.
210 2018-06-04 Jakub Jelinek <jakub@redhat.com>
212 PR tree-optimization/69615
213 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
214 maximum or minimum of the type, try to merge it also as if
215 range1 is + [-, x - 1] or + [x + 1, -].
218 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
220 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
222 PR tree-optimization/86034
223 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
224 the unsigned bitfield type in a bit insertion sequence if it does not
225 have a larger precision than the bitfield size.
226 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
228 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
230 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
232 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
234 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
235 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
236 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
237 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
239 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
241 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
242 Disable -fdelete-null-pointer-checks for ELF toolchain.
244 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
245 Kito Cheng <kito.cheng@gmail.com>
247 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
248 (nds32le-*-*, nds32be-*-*): Integrate checking process.
249 (nds32*-*-*): Add glibc and uclibc conditions.
250 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
251 (TARGET_EXCEPT_UNWIND_INFO): Define.
252 * config/nds32/elf.h: New file.
253 * config/nds32/linux.h: New file.
254 * config/nds32/nds32-elf.opt: New file.
255 * config/nds32/nds32-linux.opt: New file.
256 * config/nds32/nds32-fp-as-gp.c
257 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
258 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
260 (nds32_asm_file_end): Ditto.
261 (nds32_print_operand): Ditto.
262 (nds32_insert_attributes): Ditto.
263 (nds32_init_libfuncs): New function.
264 (TARGET_HAVE_TLS): Define.
265 (TARGET_INIT_LIBFUNCS): Define.
266 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
268 (TARGET_ELF): Apply different mcmodel setting.
269 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
270 been migrated into elf.h and linux.h files.
271 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
272 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
273 (mcmodel): The content has been migrated into nds32-elf.opt and
274 nds32-linux.opt files.
275 * config/nds32/t-elf: New file.
276 * config/nds32/t-linux: New file.
278 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
279 Shiva Chen <shiva0217@gmail.com>
281 * config/nds32/constants.md (unspec_volatile_element): Add
282 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
283 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
285 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
286 (make_pass_nds32_fp_as_gp): Declare.
287 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
289 (nds32_asm_function_end_prologue): Remove unused asm output.
290 (nds32_asm_function_begin_epilogue): Remove unused asm output.
291 (nds32_asm_file_start): Output necessary fp_as_gp information.
292 (nds32_option_override): Adjust register usage.
293 (nds32_expand_prologue): Consider fp_as_gp situation.
294 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
295 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
298 (simple_return): Ditto.
299 (omit_fp_begin): Output special directive for fp_as_gp.
300 (omit_fp_end): Output special directive for fp_as_gp.
301 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
302 mforbid-fp-as-gp): New options.
304 2018-06-01 Mark Wielaard <mark@klomp.org>
306 * dwarf2out.c (dwarf2out_finish): Remove generation of
309 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
311 * gimple-ssa-store-merging.c: Include gimple-fold.h.
312 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
313 (struct merged_store_group): Add bit_insertion field.
314 (dump_char_array): Use standard hexadecimal format.
315 (merged_store_group::merged_store_group): Set bit_insertion to false.
316 (merged_store_group::apply_stores): Use optimal buffer size. Deal
317 with BIT_INSERT_EXPR stores. Move up code updating the mask and
318 also print the mask in the dump file.
319 (pass_store_merging::gate): Minor tweak.
320 (imm_store_chain_info::coalesce_immediate): Fix wrong association
321 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
322 stores with INTEGER_CST stores.
323 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
324 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
325 and use it throughout. Generate bit insertion sequences if need be.
326 (pass_store_merging::process_store): Remove redundant condition.
327 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
329 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
331 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
332 the 128-bit floating point types. Fix function comment.
334 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
336 * config/aarch64/aarch64-simd.md
337 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
339 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
342 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
344 PR tree-optimization/85989
345 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
347 (backprop::intersect_uses): Check it when deciding whether this
348 is a backedge reference.
349 (backprop::process_block): Add each phi to m_visited_phis
350 after visiting it, then clear it at the end.
352 2018-06-01 Richard Biener <rguenther@suse.de>
354 * tree-vectorizer.h (vect_dr_stmt): New function.
355 (vect_get_load_cost): Adjust.
356 (vect_get_store_cost): Likewise.
357 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
358 Use vect_dr_stmt instead of DR_SMTT.
359 (vect_record_base_alignments): Likewise.
360 (vect_calculate_target_alignment): Likewise.
361 (vect_compute_data_ref_alignment): Likewise and make static.
362 (vect_update_misalignment_for_peel): Likewise.
363 (vect_verify_datarefs_alignment): Likewise.
364 (vector_alignment_reachable_p): Likewise.
365 (vect_get_data_access_cost): Likewise. Pass down
366 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
367 (vect_get_peeling_costs_all_drs): Likewise.
368 (vect_peeling_hash_get_lowest_cost): Likewise.
369 (vect_enhance_data_refs_alignment): Likewise.
370 (vect_find_same_alignment_drs): Likewise.
371 (vect_analyze_data_refs_alignment): Likewise.
372 (vect_analyze_group_access_1): Likewise.
373 (vect_analyze_group_access): Likewise.
374 (vect_analyze_data_ref_access): Likewise.
375 (vect_analyze_data_ref_accesses): Likewise.
376 (vect_vfa_segment_size): Likewise.
377 (vect_small_gap_p): Likewise.
378 (vectorizable_with_step_bound_p): Likewise.
379 (vect_prune_runtime_alias_test_list): Likewise.
380 (vect_analyze_data_refs): Likewise.
381 (vect_supportable_dr_alignment): Likewise.
382 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
383 (vect_gen_prolog_loop_niters): Likewise.
384 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
385 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
387 (vect_recog_mask_conversion_pattern): Likewise.
388 (vect_try_gather_scatter_pattern): Likewise.
389 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
390 to vect_get_store_cost.
391 (vect_get_store_cost): Get stmt_info instead of DR.
392 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
393 (vect_get_load_cost): Get stmt_info instead of DR.
395 2018-06-01 Richard Biener <rguenther@suse.de>
398 * gimple-fold.c (var_decl_component_p): Also allow offsetted
399 vars wrapped in MEM_REFs.
401 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
403 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
404 Fix subreg tests so that we only return a choice between
405 GENERAL_REGS and FP_REGS if the original classes included both.
407 2018-06-01 Richard Biener <rguenther@suse.de>
410 * tree-ssa-structalias.c (get_function_part_constraint):
411 Handle NULL fi->decl.
412 (find_func_aliases_for_call): Properly handle indirect
414 (find_func_clobbers): Likewise.
415 (ipa_pta_execute): Likewise.
416 (create_variable_info_for): For functions that are ifunc_resolver
417 resolve to a varinfo that contains the result of the resolver call.
418 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
421 2018-05-31 Michael Collison <michael.collison@arm.com>
423 * config/aarch64/aarch64.md:
424 (*fix_to_zero_extenddfdi2): New pattern.
425 * gcc.target/aarch64/fix_extend1.c: New testcase.
427 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
431 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
432 and BUILT_IN_STRNCMP_EQ.
433 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
435 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
436 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
437 (gimple_fold_builtin): Likewise.
438 * tree-ssa-strlen.c (compute_string_length): New function.
439 (determine_min_obsize): New function.
440 (handle_builtin_string_cmp): New function to handle calls to
441 string compare functions.
442 (strlen_optimize_stmt): Add handling to builtin string compare
444 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
445 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
446 * tree.c (build_common_builtin_nodes): Add new defines of
447 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
449 2018-05-31 Jakub Jelinek <jakub@redhat.com>
452 * bb-reorder.c (pass_partition_blocks::gate): Return false for
453 functions with naked attribute.
455 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
457 * config/i386/sse.md (avx_vec_concat<mode>):
458 Substitute concat_tg_mode mode attribute with xtg_mode.
459 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
460 (concat_tg_mode): Remove mode attribute.
462 2018-05-31 Martin Sebor <msebor@redhat.com>
465 * calls.c (alloc_max_size): Correct a logic error/typo.
466 Treat excessive arguments as infinite. Warn for invalid arguments.
467 * doc/invoke.texi (-Walloc-size-larger-than): Update.
469 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
472 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
473 and movx for Haswell.
475 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
476 Cesar Philippidis <cesar@codesourcery.com>
479 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
480 when emitting error on private/firstprivate reductions.
481 * omp-low.c (lower_omp_target): Avoid reference-type processing
482 on pointers for firstprivate clause.
484 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
486 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
489 * config/aarch64/aarch64-simd.md
490 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
491 (aarch64_ld1_x3_<mode>): Likewise
492 (aarch64_st1x2<VALLDIF:mode>): Likewise
493 (aarch64_st1_x2_<mode>): Likewise
494 (aarch64_st1x3<VALLDIF:mode>): Likewise
495 (aarch64_st1_x3_<mode>): Likewise
496 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
497 (vld1_s8_x3): Likewise.
498 (vld1_u16_x3): Likewise.
499 (vld1_s16_x3): Likewise.
500 (vld1_u32_x3): Likewise.
501 (vld1_s32_x3): Likewise.
502 (vld1_u64_x3): Likewise.
503 (vld1_s64_x3): Likewise.
504 (vld1_f16_x3): Likewise.
505 (vld1_f32_x3): Likewise.
506 (vld1_f64_x3): Likewise.
507 (vld1_p8_x3): Likewise.
508 (vld1_p16_x3): Likewise.
509 (vld1_p64_x3): Likewise.
510 (vld1q_u8_x3): Likewise.
511 (vld1q_s8_x3): Likewise.
512 (vld1q_u16_x3): Likewise.
513 (vld1q_s16_x3): Likewise.
514 (vld1q_u32_x3): Likewise.
515 (vld1q_s32_x3): Likewise.
516 (vld1q_u64_x3): Likewise.
517 (vld1q_s64_x3): Likewise.
518 (vld1q_f16_x3): Likewise.
519 (vld1q_f32_x3): Likewise.
520 (vld1q_f64_x3): Likewise.
521 (vld1q_p8_x3): Likewise.
522 (vld1q_p16_x3): Likewise.
523 (vld1q_p64_x3): Likewise.
524 (vst1_s64_x2): Likewise.
525 (vst1_u64_x2): Likewise.
526 (vst1_f64_x2): Likewise.
527 (vst1_s8_x2): Likewise.
528 (vst1_p8_x2): Likewise.
529 (vst1_s16_x2): Likewise.
530 (vst1_p16_x2): Likewise.
531 (vst1_s32_x2): Likewise.
532 (vst1_u8_x2): Likewise.
533 (vst1_u16_x2): Likewise.
534 (vst1_u32_x2): Likewise.
535 (vst1_f16_x2): Likewise.
536 (vst1_f32_x2): Likewise.
537 (vst1_p64_x2): Likewise.
538 (vst1q_s8_x2): Likewise.
539 (vst1q_p8_x2): Likewise.
540 (vst1q_s16_x2): Likewise.
541 (vst1q_p16_x2): Likewise.
542 (vst1q_s32_x2): Likewise.
543 (vst1q_s64_x2): Likewise.
544 (vst1q_u8_x2): Likewise.
545 (vst1q_u16_x2): Likewise.
546 (vst1q_u32_x2): Likewise.
547 (vst1q_u64_x2): Likewise.
548 (vst1q_f16_x2): Likewise.
549 (vst1q_f32_x2): Likewise.
550 (vst1q_f64_x2): Likewise.
551 (vst1q_p64_x2): Likewise.
552 (vst1_s64_x3): Likewise.
553 (vst1_u64_x3): Likewise.
554 (vst1_f64_x3): Likewise.
555 (vst1_s8_x3): Likewise.
556 (vst1_p8_x3): Likewise.
557 (vst1_s16_x3): Likewise.
558 (vst1_p16_x3): Likewise.
559 (vst1_s32_x3): Likewise.
560 (vst1_u8_x3): Likewise.
561 (vst1_u16_x3): Likewise.
562 (vst1_u32_x3): Likewise.
563 (vst1_f16_x3): Likewise.
564 (vst1_f32_x3): Likewise.
565 (vst1_p64_x3): Likewise.
566 (vst1q_s8_x3): Likewise.
567 (vst1q_p8_x3): Likewise.
568 (vst1q_s16_x3): Likewise.
569 (vst1q_p16_x3): Likewise.
570 (vst1q_s32_x3): Likewise.
571 (vst1q_s64_x3): Likewise.
572 (vst1q_u8_x3): Likewise.
573 (vst1q_u16_x3): Likewise.
574 (vst1q_u32_x3): Likewise.
575 (vst1q_u64_x3): Likewise.
576 (vst1q_f16_x3): Likewise.
577 (vst1q_f32_x3): Likewise.
578 (vst1q_f64_x3): Likewise.
579 (vst1q_p64_x3): Likewise.
581 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
583 * config/msp430/msp430.c (msp430_output_labelref): Prepend
584 user_label_prefix to name.
586 * tree-core.h: Update comment about the format of NAME string
587 passed to handler in attribute_spec.
589 * config/msp430/msp430.md: Remove erroneous subreg expression from
590 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
591 zero_extend{q,h}isi2.
593 2018-05-30 Borislav Petkov <bp@suse.de>
595 * doc/extend.texi: Document some architecture specific
596 constraints and sort entries.
598 2018-05-30 Martin Sebor <msebor@redhat.com>
601 * builtins.c (expand_builtin_stpcpy_1): New function.
602 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
603 only if the former succeeds.
605 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
607 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
610 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
612 * doc/invoke.texi (-flinker-output): Document
614 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
616 * passes.c (ipa_write_summaries): Only modify statements if body
618 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
619 incrementally linking.
620 (ipa_passes): Likewise.
621 * lto-cgraph.c (lto_output_node): When incrementally linking do not
622 pass down resolution info.
623 * common.opt (flag_incremental_link): Update info.
624 * gcc.c (plugin specs): Turn flinker-output=* to
625 -plugin-opt=-linker-output-known
626 * toplev.c (compile_file): Also cut compilation when doing incremental
628 * flag-types. (enum lto_partition_model): Add
629 LTO_LINKER_OUTPUT_NOLTOREL.
630 (invoke.texi): Add -flinker-output docs.
631 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
632 link same way as WPA; do not stream in dead initializers.
634 * dwarf2out.c (dwarf2out_die_ref_for_decl,
635 darf2out_register_external_decl): Support incremental link.
637 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
639 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
641 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
643 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
644 it down to simple_object_copy_lto_debug_sections.
645 (run_gcc): Determine incremental LTO link time and configure
646 lto1 into non-wpa mode, disable renaming of debug sections.
648 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
650 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
651 descriptions of various incorrectly documented functions.
653 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
656 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
657 address check not strict.
659 2018-05-30 Richard Biener <rguenther@suse.de>
661 PR tree-optimization/85964
662 * tracer.c (better_p): Drop initialized count check, we only
663 call the function with initialized counts now.
664 (find_best_successor): Do find a best edge if one
665 has uninitialized count.
666 (find_best_predecessor): Likewise. Do BB frequency check only
667 if count is initialized.
669 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
671 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
672 (aarch64_ldrstr_offset_compare): New.
673 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
674 load/store orderings.
675 (aarch64_gen_adjusted_ldpstp): Likewise.
677 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
679 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
680 Check for subset of GENERAL_REGS and FP_REGS.
681 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
684 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
686 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
687 and wi::to_poly_offset. Add the current offset and then check
688 whether the sum fits, rather than using an unchecked addition of
689 a checked term. Check for a shwi rather than a uhwi.
690 * expr.c (get_bit_range): Use tree_to_poly_uint64.
691 (store_constructor): Use poly_int_tree_p.
692 (expand_expr_real_1): Likewise.
693 * function.c (assign_temp): Likewise.
694 * fold-const.c (const_binop): Use poly_int_tree_p and
696 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
698 * ipa-icf-gimple.c (func_checker::compare_operand): Use
699 to_poly_offset for MEM offsets.
700 * ipa-icf.c (sem_variable::equals): Likewise.
701 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
702 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
703 wi::to_poly_offset for BIT_FIELD_REF offsets.
704 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
706 * var-tracking.c (emit_note_insn_var_location): Use
709 2018-05-29 Jim Wilson <jimw@sifive.com>
711 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
713 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
716 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
717 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
719 (sse4_1_round<mode>2): Use nonimmediate_operand
720 for operand 1 predicate.
722 2018-05-29 Martin Sebor <msebor@redhat.com>
723 Richard Biener <rguenther@suse.de>
726 * calls.c (get_size_range): Call determine_value_range instead
728 * tree-vrp.h (determine_value_range): Declared new function.
729 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
731 2018-05-29 Richard Biener <rguenther@suse.de>
733 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
734 sure to use non-pattern stmts for get_earlier_stmt arguments.
735 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
736 called on pattern stmts.
737 (get_later_stmt): Likewise.
739 2018-05-29 Martin Liska <mliska@suse.cz>
741 PR gcov-profile/85759
742 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
745 2018-05-29 Jakub Jelinek <jakub@redhat.com>
747 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
749 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
753 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
754 VEC_PACK_FLOAT_EXPR): New tree codes.
755 * tree-pretty-print.c (op_code_prio): Handle
756 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
757 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
758 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
759 * tree-inline.c (estimate_operator_cost): Likewise.
760 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
761 * fold-const.c (const_binop): Likewise.
762 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
763 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
764 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
765 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
766 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
767 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
768 * expr.c (expand_expr_real_2): Likewise.
769 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
770 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
771 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
773 * optabs.c (expand_widen_pattern_expr): For
774 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
775 sign from result type rather than operand's type.
776 (expand_binop_directly): For vec_packu_float_optab and
777 vec_packs_float_optab allow result type to be different from operand's
779 * optabs-tree.c (optab_for_tree_code): Handle
780 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
781 VEC_PACK_FLOAT_EXPR. Formatting fixes.
782 * tree-vect-generic.c (expand_vector_operations_1): Handle
783 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
785 * tree-vect-stmts.c (supportable_widening_operation): Handle
787 (supportable_narrowing_operation): Handle FLOAT_EXPR.
788 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
789 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
790 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
791 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
793 (vec_pack<floatprefix>_float_<mode>): New expander.
794 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
796 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
797 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
798 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
799 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
800 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
802 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
803 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
804 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
805 VEC_PACK_FLOAT_EXPR): Document.
807 2018-05-29 Richard Biener <rguenther@suse.de>
809 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
811 (stmt_vec_info_vec): Make pointer.
812 (init_stmt_vec_info_vec): Remove.
813 (free_stmt_vec_info_vec): Likewise.
814 (set_stmt_vec_info_vec): New function.
815 (free_stmt_vec_infos): Likewise.
816 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
817 (set_vinfo_for_stmt): Likewise.
818 (get_earlier_stmt): Likewise.
819 (get_later_stmt): Likewise.
820 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
821 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
822 (vec_info::~vec_info): Free stmt_vec_infos.
823 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
824 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
825 (pass_slp_vectorize::execute): Likewise.
826 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
827 (free_stmt_vec_info_vec): Likewise.
828 (set_stmt_vec_info_vec): New function.
829 (free_stmt_vec_infos): Likewise.
830 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
831 the global stmt_vec_info_vec.
832 * tree-parloops.c (gather_scalar_reductions): Use
833 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
836 2018-05-29 Richard Biener <rguenther@suse.de>
838 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
840 2018-05-29 Martin Liska <mliska@suse.cz>
841 David Malcolm <dmalcolm@redhat.com>
843 * vec.c (test_reverse): New.
844 (vec_c_tests): Add new test.
845 * vec.h (vl_ptr>::reverse): New function.
847 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
849 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
851 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
854 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
856 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
858 2018-05-28 Richard Biener <rguenther@suse.de>
860 PR tree-optimization/85933
861 * tree-vect-data-refs.c (vect_record_base_alignments): Only
862 look at stmts marked as vectorizable.
864 2018-05-28 Richard Biener <rguenther@suse.de>
866 PR tree-optimization/85934
867 * tree-vect-generic.c (expand_vector_operations_1): Hoist
868 vector boolean check before scalar optimization.
870 2018-05-28 Jakub Jelinek <jakub@redhat.com>
872 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
875 2018-05-28 Mark Wielaard <mark@klomp.org>
877 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
878 if it is an expression containing a minus sign.
880 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
882 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
884 2018-05-27 Paul Koning <ni1d@arrl.net>
886 * config/pdp11/pdp11.md (truncsihi2): Remove.
888 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
889 Chung-Ju Wu <jasonwucj@gmail.com>
891 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
893 (unaligned_store_dw): Ditto.
894 * config/nds32/nds32-memory-manipulation.c
895 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
896 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
897 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
898 (emit_setmem_word_loop): Rename to ...
899 (emit_setmem_doubleword_loop): ... this.
900 (nds32_gen_dup_4_byte_to_word_value): New function.
901 (nds32_gen_dup_8_byte_to_double_word_value): New function.
902 (nds32_expand_setmem_loop): Refine implementation.
903 (nds32_expand_setmem_loop_v3m): Ditto.
904 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
907 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
909 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
911 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
913 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
914 (nds32_init_machine_status): Initialize machine->attr_naked_p and
915 machine->attr_no_prologue_p.
916 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
917 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
918 (nds32_expand_epilogue): Consider attr_naked_p.
919 (nds32_expand_epilogue_v3pop): Likewise.
920 (nds32_can_use_return_insn): Likewise.
921 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
922 attr_no_prologue_p fields.
923 * config/nds32/nds32.opt (mret-in-naked-func): New option.
925 2018-05-27 Jakub Jelinek <jakub@redhat.com>
928 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
931 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
933 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
935 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
937 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
939 (*<floatsuffix>floatv2div2sf2): Rename to ...
940 (*float<floatunssuffix>v2div2sf2): ... this.
941 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
942 (float<floatunssuffix>v2div2sf2_mask): ... this.
943 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
944 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
945 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
947 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
949 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
951 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
953 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
955 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
957 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
958 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
959 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
960 gen_ufix_truncv8dfv8si2.
961 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
962 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
963 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
964 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
965 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
966 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
967 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
968 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
970 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
974 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
976 2018-05-25 Jim Wilson <jimw@sifive.com>
978 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
979 * config/riscv/riscv.c (struct machine_function): Add
980 interrupt_handler_p and attribute_checked_p fields.
981 (riscv_attribute_table): Add interrupt.
982 (riscv_interrupt_type_p): New.
983 (riscv_save_reg_p): Save extra regs for interrupt handler.
984 (riscv_use_save_libcall): Return false for interrupt handler.
985 (riscv_first_stack_step): Add forward declaration.
986 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
987 for interrupt handler with large frame. Use it for saved reg list.
988 (riscv_expand_prologue): Move flag_stack_usage_info support to
989 eliminate duplication.
990 (riscv_expand_epilogue): Generate mret for interrupt handler.
991 (riscv_epilogue_uses): New.
992 (riscv_can_use_return_insn): Return false for interrupt handler.
993 (riscv_function_ok_for_sibcall): Likewise.
994 (riscv_set_current_function): Add interrupt handler support.
995 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
996 * config/riscv/riscv.md (UNSPECV_MRET): New.
998 (riscv_frflags, riscv_fsflags): Use tab after opcode.
1000 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
1002 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
1004 PR tree-optimization/85712
1005 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
1006 this candidate has already been replaced in-situ by a copy.
1008 2018-05-25 Jason Merrill <jason@redhat.com>
1010 PR c++/80485 - inline function non-zero address.
1011 * symtab.c (nonzero_address): Check DECL_COMDAT.
1013 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
1016 * config/alpha/alpha.md (ashlsi3): New insn pattern.
1017 (*ashlsi_se): Rename from *ashldi_se. Define as sign
1018 extension of SImode operation. Use const123_operand predicate.
1019 (*saddsi_1): Remove.
1020 (*saddl_se_1): Ditto.
1022 (*ssubl_se_1): Ditto.
1023 * config/alpha/predicates.md (const123_operand): New predicate.
1024 * config/alpha/constraints.md (P): Use IN_RANGE.
1026 2018-05-25 Richard Biener <rguenther@suse.de>
1028 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
1030 (ref_maybe_used_by_stmt_p): Likewise.
1031 (stmt_may_clobber_ref_p): Likewise.
1032 (stmt_may_clobber_ref_p_1): Likewise.
1033 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1035 (ref_maybe_used_by_stmt_p): Likewise.
1036 (stmt_may_clobber_ref_p): Likewise.
1037 (stmt_may_clobber_ref_p_1): Likewise.
1038 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1039 the alias oracle to disambiguate DRs with stmts DR analysis
1041 (vect_analyze_data_refs): Do not give up on not analyzable
1042 DRs for BB vectorization. Remove code truncating the dataref
1045 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1048 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1049 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1050 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1052 2018-05-25 Richard Biener <rguenther@suse.de>
1054 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1055 function, combining stmt data ref gathering and fatal analysis
1057 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1058 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1059 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1060 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1061 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1063 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1065 PR tree-optimization/85720
1066 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1067 SCC if all partitions are builtins.
1068 (version_loop_by_alias_check): New parameter. Generate cancelable
1069 runtime alias check if all partitions are builtins.
1070 (distribute_loop): Update call to above function.
1072 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1074 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1075 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1076 (parm_default_def_partition_arg): Ditto.
1077 (set_parm_default_def_partition): Ditto.
1078 (get_parm_default_def_partitions): Ditto and make it static.
1079 (get_undefined_value_partitions): Ditto and make it static.
1080 (remove_ssa_form): Refactor call to init_var_map here.
1081 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1082 computation for loop region.
1083 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1084 (register_default_def): Delete.
1085 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1086 (parm_default_def_partition_arg): Ditto.
1087 (set_parm_default_def_partition): Ditto.
1088 (get_parm_default_def_partitions): Ditto and make it static.
1089 (get_undefined_value_partitions): Ditto and make it static.
1090 (coalesce_with_default, coalesce_with_default): Update comment.
1091 (create_coalesce_list_for_region): New func factored out from
1092 create_outofssa_var_map.
1093 (populate_coalesce_list_for_outofssa): New func factored out from
1094 create_outofssa_var_map and coalesce_ssa_name.
1095 (create_outofssa_var_map): Delete.
1096 (coalesce_ssa_name): Refactor to support live range computation.
1097 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1098 (get_parm_default_def_partitions): Delete.
1099 (get_undefined_value_partitions): Ditto.
1100 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1101 computation for loop region.
1102 (new_tree_live_info, loe_visit_block): Ditto.
1103 (live_worklist, set_var_live_on_entry): Ditto.
1104 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1105 * tree-ssa-live.h (struct _var_map): New fields.
1106 (init_var_map): Change decl.
1107 (region_contains_p): New.
1109 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1111 * tree-ssa-live.h (live_merge_and_clear): Delete.
1113 2018-05-25 Richard Biener <rguenther@suse.de>
1116 * tree-dump.c (dequeue_and_dump): Remove access to removed
1117 operand 2 of a SWITCH_EXPR.
1119 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1121 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1122 multiplication and division.
1123 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1124 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1125 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1126 (cond_udiv_optab, cond_umod_optab): New optabs.
1127 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1128 (IFN_COND_RDIV): New internal functions.
1129 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1130 TRUNC_MOD_EXPR and RDIV_EXPR.
1131 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1132 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1134 (SVE_INT_BINARY): Include mult.
1135 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1136 (optab, sve_int_op): Handle mult.
1137 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1139 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1140 for SVE_INT_BINARY_SD.
1142 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1144 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1145 (optab, sve_int_op): Handle div and udiv.
1146 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1147 for SVE_INT_BINARY_SD.
1148 (*<optab><mode>3): New insn for the same.
1150 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1152 * tree-vect-patterns.c: Include predict.h.
1153 (vect_recog_divmod_pattern): Restrict check for division support
1154 to when optimizing for size.
1156 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1158 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1159 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1160 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1161 (gimple_match_op::set_op): Likewise.
1162 (gimple_resimplify4): Declare.
1163 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1164 (expr::gen_transform): Likewise.
1165 (decision_tree::gen): Generate a simplification routine for 4 operands.
1166 * gimple-match-head.c (gimple_simplify): Add an overload for
1167 4 operands. In the top-level function, handle up to 4 call
1168 arguments and call gimple_resimplify4.
1169 (gimple_resimplify4): New function.
1170 (build_call_internal): Pass a fourth operand.
1171 (maybe_push_to_seq): Likewise.
1172 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1173 Fold VEC_COND_EXPRs of an operation and a default value into
1174 an IFN_COND_* function if possible.
1175 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1177 (SVE_COND_FP_BINARY): Include them.
1178 (optab, sve_fp_op): Handle them.
1179 (SVE_INT_BINARY_REV): New code iterator.
1180 (SVE_COND_FP_BINARY_REV): New int iterator.
1181 (commutative): New int attribute.
1182 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1184 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1186 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1187 (*cond_<optab><mode>): New patterns for reversed operands.
1189 2018-05-25 Richard Biener <rguenther@suse.de>
1191 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1192 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1193 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1194 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1195 * tree-vect-data-refs.c (everywhere): Adjust users.
1196 * tree-vect-loop.c (everywhere): Likewise.
1197 * tree-vect-slp.c (everywhere): Likewise.
1198 * tree-vect-stmts.c (everywhere): Likewise.
1199 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1201 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1203 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1205 (gcc_cv_as_section_exclude): ... this.
1206 Try Solaris as #exclude syntax.
1207 * configure: Regenerate.
1208 * config.in: Regenerate.
1209 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1211 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1212 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1214 * varasm.c (default_elf_asm_named_section): Don't check if
1215 HAVE_GAS_SECTION_EXCLUDE is defined.
1217 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1219 * doc/md.texi: Update the documentation of the cond_* optabs
1220 to mention the new final operand. Fix GET_MODE_NUNITS call.
1221 Describe the scalar case too.
1222 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1223 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1225 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1226 (get_conditional_internal_fn): Update comment.
1227 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1228 accumulator value as a final argument to conditional functions.
1229 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1230 a define_expand and add an "else" operand. Assert for now that
1231 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1232 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1233 (*cond_<optab><mode>): New patterns.
1234 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1235 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1236 (UNSPEC_COND_EOR): Delete.
1237 (optab): Remove associated mappings.
1238 (SVE_INT_BINARY): New code iterator.
1239 (sve_int_op): Remove int attribute and add "minus" to the code
1241 (SVE_COND_INT_OP): Delete.
1242 (SVE_COND_FP_OP): Rename to...
1243 (SVE_COND_FP_BINARY): ...this.
1245 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1247 * optabs.c (can_reuse_operands_p): New function.
1248 (maybe_legitimize_operands): Try to reuse the results for
1251 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1253 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1254 Add {q} suffix to insn mnemonic.
1256 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1258 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1259 (msp430_warn_func_return): New.
1261 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1263 * fold-const.c (tree_nonzero_bits): New function.
1264 * fold-const.h (tree_nonzero_bits): Likewise.
1265 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1266 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1268 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1272 * varasm.c (assemble_alias): Check ifunc_resolver only on
1275 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1278 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1279 when memory input operand is handled.
1281 2018-05-24 Luis Machado <luis.machado@linaro.org>
1283 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1285 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1287 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1289 * match.pd: Delay FMA folds until after vectorization.
1291 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1294 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1295 address check not strict.
1297 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1299 * gimple-match.h (gimple_match_op): New class.
1300 (mprts_hook): Replace parameters with a gimple_match_op *.
1301 (maybe_build_generic_op): Likewise.
1302 (gimple_simplified_result_is_gimple_val): Replace parameters with
1303 a const gimple_match_op *.
1304 (gimple_simplify): Replace code_helper * and tree * parameters with
1305 a gimple_match_op * parameter.
1306 (gimple_resimplify1): Replace code_helper *, tree and tree *
1307 parameters with a gimple_match_op * parameter.
1308 (gimple_resimplify2): Likewise.
1309 (gimple_resimplify3): Likewise.
1310 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1311 parameters with a gimple_match_op * parameter.
1312 * gimple-match-head.c (gimple_simplify): Change prototypes of
1313 auto-generated functions to take a gimple_match_op * instead of
1314 separate code_helper * and tree * parameters. Make the same
1315 change in the top-level overload and update calls to the
1316 gimple_resimplify routines. Update calls to the auto-generated
1317 functions and to maybe_push_res_to_seq in the publicly-facing
1318 operation-specific gimple_simplify overloads.
1319 (gimple_match_op::MAX_NUM_OPS): Define.
1320 (gimple_resimplify1): Replace rcode and ops with a single res_op
1321 parameter. Update call to gimple_simplify.
1322 (gimple_resimplify2): Likewise.
1323 (gimple_resimplify3): Likewise.
1324 (mprts_hook): Replace parameters with a gimple_match_op *.
1325 (maybe_build_generic_op): Likewise.
1326 (build_call_internal): Replace type, nargs and ops with
1327 a gimple_match_op *.
1328 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1329 with a single gimple_match_op *. Update calls to mprts_hook,
1330 build_call_internal and gimple_simplified_result_is_gimple_val.
1331 Factor out code that is common to the tree_code and combined_fn cases.
1332 * genmatch.c (expr::gen_transform): Replace tem_code and
1333 tem_ops with a gimple_match_op called tem_op. Update calls
1334 to the gimple_resimplify functions and maybe_push_res_to_seq.
1335 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1336 res_ops. Update call to the gimple_resimplify functions.
1337 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1338 (decision_tree::gen): Make the functions take a gimple_match_op *
1339 called res_op instead of separate res_code and res_ops parameters.
1340 Update call accordingly.
1341 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1342 and ops with a single res_op parameter. Update calls to
1343 maybe_build_generic_op and maybe_push_res_to_seq.
1344 (fold_stmt_1): Update calls to gimple_simplify and
1345 replace_stmt_with_simplification.
1346 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1347 and gimple_simplified_result_is_gimple_val.
1348 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1350 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1351 with a gimple_match_op *.
1352 (vn_nary_build_or_lookup): Likewise. Update call to
1353 vn_nary_build_or_lookup_1.
1354 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1355 gimple_match_op *. Update calls to the gimple_resimplify routines
1356 and to gimple_simplified_result_is_gimple_val.
1357 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1358 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1359 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1360 (visit_nary_op): Likewise.
1361 (visit_reference_op_load): Likewise.
1363 2018-05-23 Luis Machado <luis.machado@linaro.org>
1365 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1366 modifier for printing the step amount.
1368 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1371 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1374 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1376 * doc/sourcebuild.texi (Endianness): New subsubsection.
1378 2018-05-23 Luis Machado <luis.machado@linaro.org>
1380 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1381 <prefetch_dynamic_strides>: New const bool field.
1382 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1383 prefetch_dynamic_strides.
1384 (exynosm1_prefetch_tune): Likewise.
1385 (thunderxt88_prefetch_tune): Likewise.
1386 (thunderx_prefetch_tune): Likewise.
1387 (thunderx2t99_prefetch_tune): Likewise.
1388 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1390 (aarch64_override_options_internal): Update to set
1391 PARAM_PREFETCH_DYNAMIC_STRIDES.
1392 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1393 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1394 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1395 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1396 prefetch-dynamic-strides setting.
1398 2018-05-23 Luis Machado <luis.machado@linaro.org>
1400 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1401 <minimum_stride>: New const int field.
1402 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1403 minimum_stride field defaulting to -1.
1404 (exynosm1_prefetch_tune): Likewise.
1405 (thunderxt88_prefetch_tune): Likewise.
1406 (thunderx_prefetch_tune): Likewise.
1407 (thunderx2t99_prefetch_tune): Likewise.
1408 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1409 <default_opt_level>: Set to 3.
1410 (aarch64_override_options_internal): Update to set
1411 PARAM_PREFETCH_MINIMUM_STRIDE.
1412 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1413 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1414 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1415 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1416 stride is constant and is below the minimum stride threshold.
1418 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1420 * config/arm/arm-cpus.in (mode26): Delete.
1421 (armv4): Delete mode26 reference.
1422 * config/arm/arm.c (arm_configure_build_target): Delete use of
1425 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1427 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1429 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1430 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1432 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1433 (floatunsdidf2): Ditto.
1435 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1437 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1438 (fixuns_trunc<mode>si2_avx512f): Ditto.
1439 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1440 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1441 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1443 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1445 PR rtl-optimization/79985
1446 * df-scan.c (df_insn_refs_collect): Remove special case for
1447 global registers and asm statements.
1449 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1451 * extend.texi (Global Register Variables): Rewrite the bullet list.
1452 Note that the register is available for allocation. Note that access
1453 via inline asm must use constraints. Add note about async-signal
1454 handlers. Remove paragraph about automagic register selection.
1456 2018-05-23 Richard Biener <rguenther@suse.de>
1458 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1459 of fixed offset from memset VN.
1461 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1463 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1465 (alloc_cand_and_find_basis): Initialize first_interp field.
1466 (slsr_process_mul): Modify first_interp field.
1467 (slsr_process_add): Likewise.
1468 (slsr_process_cast): Modify first_interp field for each new
1470 (slsr_process_copy): Likewise.
1471 (dump_candidate): Dump first_interp field.
1472 (replace_mult_candidate): Process all interpretations, not just
1474 (replace_rhs_if_not_dup): Likewise.
1475 (replace_one_candidate): Likewise.
1477 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1479 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1481 (aarch64_needs_frame_chain): New function.
1482 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1484 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1487 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1488 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1489 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1490 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1491 as true for strict-align.
1492 (aarch64_can_inline_p): Perform checks even when callee has no
1493 attributes to check for strict alignment.
1494 * doc/extend.texi (AArch64 Function Attributes): Document
1496 * doc/invoke.texi: (AArch64 Options): Likewise.
1498 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1500 PR tree-optimization/85853
1501 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1502 the handling of the root of the node to...
1503 (vect_slp_analyze_node_operations_1): ...this new function,
1504 and run the whole thing with the child nodes' def types
1505 set according to their SLP node's def type.
1507 2018-05-23 Richard Biener <rguenther@suse.de>
1510 * tree-data-ref.c (create_runtime_alias_checks): Defer
1511 and ignore overflow warnings.
1513 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1515 PR tree-optimization/85822
1516 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1519 2018-05-23 Richard Biener <rguenther@suse.de>
1521 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1522 memset constants via native_interpret_expr.
1524 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1527 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1529 (cgraph_node::create_alias): Likewise.
1530 (cgraph_node::get_availability): Check ifunc_resolver instead
1531 of looking up ifunc attribute.
1532 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1533 * varasm.c (do_assemble_alias): Likewise.
1534 (assemble_alias): Likewise.
1535 (default_binds_local_p_3): Likewise.
1536 * cgraph.h (cgraph_node): Add ifunc_resolver.
1537 (cgraph_node::only_called_directly_or_aliased_p): Return false
1539 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1541 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1542 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1543 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1544 instead of looking up ifunc attribute.
1546 2018-05-22 Luis Machado <luis.machado@linaro.org>
1548 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1550 2018-05-22 Martin Sebor <msebor@redhat.com>
1553 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1554 only when expasion succeeds.
1555 (expand_builtin_strcmp): Same.
1556 (expand_builtin_strncmp): Same.
1558 2018-05-22 Martin Sebor <msebor@redhat.com>
1560 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1562 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1563 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1565 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1566 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1568 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1569 Add check for legitimate_address.
1570 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1571 (aarch64_swap_ldrstr_operands): New.
1572 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1575 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1576 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1578 * config/aarch64/aarch64.md: New patterns to generate stp
1580 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1581 single words and double words.
1582 (load_pair_sw, load_pair_dw): Likewise.
1583 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1585 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1587 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1588 for different mode ldpstp and add peephole for merged zero stores.
1590 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1592 (aarch64_gen_store_pair): Rename calls to match new patterns.
1593 (aarch64_gen_load_pair): Rename calls to match new patterns.
1594 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1595 (load_pair<DREG:mode><DREG2:mode>): ... This.
1596 (store_pair<mode>): Rename to...
1597 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1598 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1600 (V_INT_EQUIV): Handle SImode.
1601 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1604 2018-05-22 Martin Sebor <msebor@redhat.com>
1607 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1608 or ajust the presumed bound on an operation to avoid unnecessary
1611 2018-05-22 Martin Sebor <msebor@redhat.com>
1613 PR tree-optimization/85826
1614 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1615 assuming that a DECL necesarily has a constant size.
1617 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1620 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1622 2018-05-22 Richard Biener <rguenther@suse.de>
1624 PR tree-optimization/85834
1625 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1626 non-constant and non-zero memset arguments.
1628 2018-05-22 Martin Liska <mliska@suse.cz>
1631 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1633 2018-05-22 Richard Biener <rguenther@suse.de>
1635 PR tree-optimization/85863
1636 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1637 comparisons when vectype is specified.
1638 (vectorizable_condition): Do not specify vectype for
1639 vect_is_simple_cond when SLP vectorizing.
1641 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1644 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1645 define __ibm128 as long double.
1646 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1647 as a distinct type when IEEE 128-bit support is enabled.
1648 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1649 128-bit types to use the correct functions.
1650 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1651 convert between 128-bit floating point types that have different
1652 modes but the same representation, instead of using gen_lowpart to
1654 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1656 (IFKF_reg): New attributes to give the register constraints for
1658 (extend<mode>tf2_internal): New insns to mark an explicit
1659 conversion between 128-bit floating point types that have a
1660 different mode but share the same representation.
1662 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1664 PR tree-optimization/85814
1665 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1666 a null return from get_strinfo when unsharing the next
1667 strinfo in the chain.
1669 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1672 * varasm.c (weak_finish): Clean up weak_decls.
1674 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1676 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1677 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1678 UNSPEC_UADALP values.
1679 * config/aarch64/iterators.md (ABAL): New int iterator.
1682 (sur): Add mappings for the above.
1683 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1685 (aarch64_<sur>abal<mode>_4): Likewise.
1686 (aarch64_<sur>adalp<mode>_3): Likewise.
1687 (<sur>sadv16qi): New define_expand.
1689 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1691 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1692 (*movdf_internal): Ditto.
1693 (*rcpsf2_sse): Ditto.
1694 (*rsqrtsf2_sse): Ditto.
1695 (*sqrt<mode>2_sse): Ditto.
1697 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1699 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1701 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1702 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1703 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1704 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1706 * config/aarch64/arm_neon.h: Likewise.
1707 * config/aarch64/iterators.md (VQ_I): New.
1709 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1711 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1713 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1716 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1718 * config/nds32/constraints.md (S): New constraint.
1719 * config/nds32/nds32.md (call_internal): Use constraint S.
1720 (call_value_internal): Likewise.
1721 (sibcall_internal): Likewise.
1722 (sibcall_value_internal): Likewise.
1724 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1725 Chung-Ju Wu <jasonwucj@gmail.com>
1727 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1730 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1731 Chung-Ju Wu <jasonwucj@gmail.com>
1733 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1734 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1735 (nds32_rtx_costs_impl): Simplify.
1736 (nds32_address_cost_impl): Simplify.
1737 (nds32_init_rtx_costs): New function.
1738 (nds32_rtx_costs_speed_prefer): Likewise.
1739 (nds32_rtx_costs_size_prefer): Likewise.
1740 (nds32_address_cost_speed_prefer): Likewise.
1741 (nds32_address_cost_speed_fwprop): Likewise.
1742 (nds32_address_cost_size_prefer): Likewise.
1743 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1744 * config/nds32/nds32.c (nds32_option_override): Use
1745 nds32_init_rtx_costs function.
1747 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1749 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1750 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1751 (TARGET_PIPELINE_N8): Likewise.
1752 (TARGET_PIPELINE_N10): Likewise.
1753 (TARGET_PIPELINE_N13): Likewise.
1754 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1756 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1758 * config/nds32/nds32-fpu.md: Update copyright year.
1760 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1762 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1764 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1766 * config/nds32/nds32.c
1767 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1768 * config/nds32/nds32.opt (minline-asm-r15): New option.
1770 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1772 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1774 * config/nds32/nds32.md (abssi2): New pattern.
1776 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1778 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1779 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1780 Merge insn pattern from sse_cvtsi2ss<round_name> and
1781 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1782 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1783 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1784 using SWI48 mode iterator.
1785 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1786 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1787 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1788 pattern from sse_cvttss2si<round_saeonly_name>
1789 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1790 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1791 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1792 using SWI48 mode iterator.
1793 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1794 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1795 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1796 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1797 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1798 using SWI48 mode iterator.
1799 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1800 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1801 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1802 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1803 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1804 SWI48 mode iterator.
1805 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1806 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1807 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1808 pattern from sse_cvttsd2si<round_saeonly_name>
1809 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1811 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1813 * config/nds32/nds32-md-auxiliary.c
1814 (nds32_valid_smw_lwm_base_p): Refine.
1815 (nds32_output_smw_single_word): Refine.
1816 (nds32_output_smw_double_word): New.
1817 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1819 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1821 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1822 (nds32_output_stack_pop): Refine.
1823 (nds32_expand_unaligned_load): Refine.
1824 (nds32_expand_unaligned_store): Refine.
1826 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1827 Chung-Ju Wu <jasonwucj@gmail.com>
1829 * config/nds32/constants.md: Add TP_REGNUM constant.
1830 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1831 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1833 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1834 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1835 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1836 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1837 and PIC code generation.
1838 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1840 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1842 * config/nds32/nds32.md: Support TLS and PIC.
1843 * config/nds32/nds32.c: Support TLS and PIC.
1844 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1845 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1848 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1850 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1851 mode with E_ prefix.
1853 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1854 Chung-Ju Wu <jasonwucj@gmail.com>
1856 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1857 * config/nds32/nds32-md-auxiliary.c
1858 (symbolic_reference_mentioned_p): New.
1859 (nds32_legitimize_ict_address): New.
1860 (nds32_expand_ict_move): New.
1861 (nds32_indirect_call_referenced_p): New.
1862 (nds32_symbol_binds_local_p): Delete.
1863 (nds32_long_call_p): Modify.
1864 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1865 * config/nds32/nds32-protos.h
1866 (symbolic_reference_mentioned_p): Declare.
1867 (nds32_legitimize_ict_address): Declare.
1868 (nds32_expand_ict_move): Declare.
1869 (nds32_indirect_call_referenced_p): Declare.
1870 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1871 (nds32_relax_group): Use nds32_ict_const_p as condition.
1872 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1873 (nds32_asm_file_start): Output ict_model directive in asm code.
1874 (nds32_legitimate_address_p): Consider indirect call.
1875 (nds32_print_operand): Consider indirect call.
1876 (nds32_print_operand_address): Consider indirect call.
1877 (nds32_insert_attributes): Handle "indirect_call" attribute.
1878 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1879 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1880 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1881 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1882 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1883 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1884 (TARGET_ICT_MODEL_SMALL): Define.
1885 (TARGET_ICT_MODEL_LARGE): Define.
1886 * config/nds32/nds32.md (movsi): Consider ict model.
1887 (call, call_value): Consider ict model.
1888 (sibcall, sibcall_value): Consider ict model.
1889 * config/nds32/nds32.opt (mict-model): New option.
1890 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1893 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1894 Monk Chiang <sh.chiang04@gmail.com>
1895 Jim Wilson <jimw@sifive.com>
1897 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1898 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1899 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1900 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1901 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1902 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1903 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1904 compute save_libcall_adjustment properly.
1905 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1906 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1907 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1908 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1909 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1910 (ABI_SPEC): Handle mabi=ilp32e.
1911 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1912 (RVE): Add RVE mask.
1913 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1914 <-march>: Add rv32e as an example.
1916 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1919 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1920 (intra_create_variable_infos): Handle C++ constructors.
1922 2018-05-18 Martin Liska <mliska@suse.cz>
1924 * passes.def: Remove a redundant pass.
1926 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1929 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1931 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1933 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1935 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1937 (armv2, armv2a, armv3, armv3m): Delete architectures.
1938 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1939 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1940 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1942 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1943 (*mulsidi3adddi): Likewise.
1944 (mulsidi3): Likewise.
1945 (*mulsidi3_nov6): Likewise.
1946 (umulsidi3): Likewise.
1947 (umulsidi3_nov6): Likewise.
1948 (umaddsidi4): Likewise.
1949 (*umulsidi3adddi): Likewise.
1950 (smulsi3_highpart): Likewise.
1951 (*smulsi3_highpart_nov6): Likewise.
1952 (umulsi3_highpart): Likewise.
1953 (*umulsi3_highpart_nov6): Likewise.
1954 * config/arm/arm.h (arm_arch3m): Delete.
1955 * config/arm/arm.c (arm_arch3m): Delete.
1956 (arm_option_override_internal): Update armv3-related comment.
1957 (arm_configure_build_target): Delete use of isa_bit_mode32.
1958 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1959 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1960 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1962 (mulusa3): Likewise.
1963 * config/arm/arm-protos.h (arm_arch3m): Delete.
1964 * config/arm/arm-tables.opt: Regenerate.
1965 * config/arm/arm-tune.md: Likewise.
1966 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1967 deleted architectures.
1969 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1971 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1972 (armv5t, armv5te): New features.
1973 (ARMv5, ARMv5e): Delete fgroups.
1974 (ARMv5t, ARMv5te): Adjust for above changes.
1976 (armv5, armv5e): Delete arches.
1977 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1979 (*call_reg_arm): Likewise.
1980 (*call_value_reg_armv5): Likewise.
1981 (*call_value_reg_arm): Likewise.
1982 (*call_symbol): Likewise.
1983 (*call_value_symbol): Likewise.
1984 (*sibcall_insn): Likewise.
1985 (*sibcall_value_insn): Likewise.
1987 (prefetch): Likewise.
1988 (define_split and define_peephole2 dependent on arm_arch5):
1990 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1992 (TARGET_ARM_QBIT): Likewise.
1993 (TARGET_DSP_MULTIPLY): Likewise.
1994 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1995 (arm_arch5, arm_arch5e): Delete.
1996 (arm_arch5t, arm_arch5te): Declare.
1997 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1998 (arm_arch5t): Declare.
1999 (arm_option_reconfigure_globals): Update for the above.
2000 (arm_options_perform_arch_sanity_checks): Update comment, replace
2001 use of arm_arch5 with arm_arch5t.
2002 (use_return_insn): Likewise.
2003 (arm_emit_call_insn): Likewise.
2004 (output_return_instruction): Likewise.
2005 (arm_final_prescan_insn): Likewise.
2006 (arm_coproc_builtin_available): Likewise.
2007 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
2008 arm_arch5e with arm_arch5t and arm_arch5te.
2009 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
2010 (arm_arch5t, arm_arch5te): Declare.
2011 * config/arm/arm-tables.opt: Regenerate.
2012 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
2013 * config/arm/t-multilib: Likewise.
2014 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
2015 instead of arm_arch5.
2016 (*call_reg_thumb1): Likewise.
2017 (*call_value_reg_thumb1_v5): Likewise.
2018 (*call_value_reg_thumb1): Likewise.
2019 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
2021 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
2023 2018-05-18 Martin Liska <mliska@suse.cz>
2025 PR gcov-profile/84846
2026 * doc/gcov.texi: Document -t option of gcov tool.
2028 2018-05-18 Martin Liska <mliska@suse.cz>
2030 PR gcov-profile/84846
2031 * gcov.c (print_usage): Add new -t option.
2032 (process_args): Handle the option.
2033 (generate_results): Use stdout as output when requested by
2036 2018-05-18 Martin Liska <mliska@suse.cz>
2038 PR gcov-profile/84846
2039 * coverage.c (coverage_init): Write PWD to .gcno file.
2040 * doc/gcov.texi: Document how working directory is printed.
2041 * gcov-dump.c (dump_gcov_file): Print PWD.
2042 * gcov.c (output_intermediate_file): Likewise.
2043 (read_graph_file): Read PWD string.
2044 (output_lines): Print PWD.
2046 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2049 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2050 for retval and return false if all args to phi are zero.
2052 2018-05-18 Richard Biener <rguenther@suse.de>
2054 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2056 (evrp_dom_walker::before_dom_children): Call it.
2058 2018-05-18 Richard Biener <rguenther@suse.de>
2060 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2061 results when processing array refs with variable index.
2063 2018-05-18 Toon Moene <toon@moene.org>
2065 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2066 directly after that of -floop-interchange. Indicate that both
2067 options are enabled by default when specifying -O3.
2069 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2071 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2072 iterator. Delete separate integer-mode vec_set<mode> expander.
2073 (aarch64_simd_vec_setv2di): Delete.
2074 (vec_setv2di): Delete.
2075 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2076 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2077 the "w, r" alternative.
2079 2018-05-18 Martin Liska <mliska@suse.cz>
2081 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2082 * tree-pass.h (make_pass_lower_switch_O0): New function.
2083 * tree-switch-conversion.c (node_has_low_bound): Remove.
2084 (node_has_high_bound): Likewise.
2085 (node_is_bounded): Likewise.
2086 (class pass_lower_switch): Make it a template type and create
2088 (pass_lower_switch::execute): Add template argument.
2089 (make_pass_lower_switch): New function.
2090 (make_pass_lower_switch_O0): New function.
2091 (do_jump_if_equal): Remove.
2092 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2093 all the hard work to tree optimization passes.
2095 2018-05-18 Martin Liska <mliska@suse.cz>
2097 * dbgcnt.c (limit_low): Renamed from limit.
2098 (limit_high): New variable.
2099 (dbg_cnt_is_enabled): Check for upper limit.
2100 (dbg_cnt): Adjust dumping.
2101 (dbg_cnt_set_limit_by_index): Add new argument for high
2103 (dbg_cnt_set_limit_by_name): Likewise.
2104 (dbg_cnt_process_single_pair): Parse new format.
2105 (dbg_cnt_process_opt): Use strtok.
2106 (dbg_cnt_list_all_counters): Remove 'value' and add
2108 * doc/invoke.texi: Document changes.
2110 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2112 * doc/sourcebuild.texi (scalar_all_fma): Document.
2113 * tree.def (FMA_EXPR): Delete.
2114 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2115 * internal-fn.c (ternary_direct): New macro.
2116 (expand_ternary_optab_fn): Likewise.
2117 (direct_ternary_optab_supported_p): Likewise.
2118 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2119 * builtins.c (fold_builtin_fma): Delete.
2120 (fold_builtin_3): Don't call it.
2121 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2122 * expr.c (expand_expr_real_2): Likewise.
2123 * fold-const.c (operand_equal_p): Likewise.
2124 (fold_ternary_loc): Likewise.
2125 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2126 * gimple.c (DEFTREECODE): Likewise.
2127 * gimplify.c (gimplify_expr): Likewise.
2128 * optabs-tree.c (optab_for_tree_code): Likewise.
2129 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2130 * tree-eh.c (operation_could_trap_p): Likewise.
2131 (stmt_could_throw_1_p): Likewise.
2132 * tree-inline.c (estimate_operator_cost): Likewise.
2133 * tree-pretty-print.c (dump_generic_node): Likewise.
2134 (op_code_prio): Likewise.
2135 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2136 * tree-ssa-operands.c (get_expr_operands): Likewise.
2137 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2138 * fold-const-call.h (fold_fma): Delete.
2139 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2140 CFN_FNMA and CFN_FNMS.
2142 * genmatch.c (combined_fn): New enum.
2143 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2144 (commutative_op): New function.
2145 (commutate): Use it. Handle more than 2 operands.
2146 (dt_operand::gen_gimple_expr): Use commutative_op.
2147 (parser::parse_expr): Allow :c to be used with non-binary
2148 operators if the commutative operand is known.
2149 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2150 CFN_FMS, CFN_FNMA and CFN_FNMS.
2151 (backprop::process_assign_use): Remove FMA_EXPR handling.
2152 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2153 (gen_hsa_fma): New function.
2154 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2155 IFN_FNMA and IFN_FNMS.
2156 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2157 * gimple-fold.h (follow_all_ssa_edges): Declare.
2158 * gimple-fold.c (follow_all_ssa_edges): New function.
2159 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2160 gimple_build interface and use follow_all_ssa_edges to fold the result.
2161 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2162 instead of checking for optabs directly.
2163 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2164 rather than FMA_EXPRs.
2165 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2166 call to IFN_FMA instead of an FMA_EXPR.
2168 2018-05-17 Jim Wilson <jimw@sifive.com>
2170 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2171 sign extended promoted subreg, and the range does not have the sign bit
2172 set, then do a sign extend.
2174 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2175 test, check for sign extended subreg and/or constant operands, and
2176 do a sign extend in that case.
2178 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2180 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2181 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2183 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2184 Change logics_shift_reg to logics_shift_imm.
2185 (thunderx2t99_fp_loadpair_basic): Delete.
2186 (thunderx2t99_fp_storepair_basic): Delete.
2187 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2188 (thunderx2t99_asimd_polynomial): Delete.
2189 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2190 and neon_fp_mul_d_scalar_q.
2191 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2192 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2193 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2194 (thunderx2t99_asimd_lut): Add missing tbl types.
2195 (thunderx2t99_asimd_ext): Delete.
2196 (thunderx2t99_asimd_load1_1_mult): Delete.
2197 (thunderx2t99_asimd_load1_2_mult): Delete.
2198 (thunderx2t99_asimd_load1_ldp): New.
2199 (thunderx2t99_asimd_load1): New.
2200 (thunderx2t99_asimd_load2): Add missing *load2* types.
2201 (thunderx2t99_asimd_load3): New.
2202 (thunderx2t99_asimd_load4): New.
2203 (thunderx2t99_asimd_store1_1_mult): Delete.
2204 (thunderx2t99_asimd_store1_2_mult): Delete.
2205 (thunderx2t99_asimd_store2_mult): Delete.
2206 (thunderx2t99_asimd_store2_onelane): Delete.
2207 (thunderx2t99_asimd_store_stp): New.
2208 (thunderx2t99_asimd_store1): New.
2209 (thunderx2t99_asimd_store2): New.
2210 (thunderx2t99_asimd_store3): New.
2211 (thunderx2t99_asimd_store4): New.
2213 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2215 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2216 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2218 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2219 Segher Boessenkool <segher@kernel.crashing.org>
2222 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2225 2018-05-17 Richard Biener <rguenther@suse.de>
2227 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2228 for pruning loop and prune defs feeding only already visited PHIs.
2230 2018-05-17 Richard Biener <rguenther@suse.de>
2232 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2234 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2235 Richard Biener <rguenther@suse.de>
2237 PR tree-optimization/85793
2238 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2239 for VMAT_ELEMENTWISE.
2241 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2243 * internal-fn.h (lookup_internal_fn): Declare
2244 * internal-fn.c (lookup_internal_fn): New function.
2245 * gimple.c (gimple_build_call_from_tree): Handle calls to
2247 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2248 internal function names.
2249 * tree-pretty-print.c (dump_generic_node): Likewise.
2250 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2252 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2254 * gimple-fold.h (gimple_build): Make the function forms take
2255 combined_fn rather than built_in_function.
2256 (gimple_simplify): Likewise.
2257 * gimple-match-head.c (gimple_simplify): Likewise.
2258 * gimple-fold.c (gimple_build): Likewise.
2259 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2260 rather than gimple_build_call_internal.
2261 (get_initial_defs_for_reduction): Likewise.
2262 (vect_create_epilog_for_reduction): Likewise.
2263 (vectorizable_live_operation): Likewise.
2265 2018-05-17 Martin Liska <mliska@suse.cz>
2267 * gimple-ssa-sprintf.c (format_directive): Do not use
2268 space in between 'G_' and '('.
2270 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2273 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2274 even if the mask is not all ones.
2277 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2279 (ix86_gimple_fold_builtin): Likewise.
2282 * config/i386/i386.c: Include tree-vector-builder.h.
2283 (ix86_vector_shift_count): New function.
2284 (ix86_fold_builtin): Fold shift builtins by scalar count.
2285 (ix86_gimple_fold_builtin): Likewise.
2287 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2288 _mm512_setzero): New intrinsics.
2290 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2291 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2293 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2294 code generation for cases where splatting a value is not useful.
2295 * simplify-rtx.c (simplify_ternary_operation): Simplify
2296 vec_merge across a vec_duplicate and a paradoxical subreg forming
2297 a vector mode to a vec_concat.
2299 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2301 * config.gcc: Support "goldmont-plus".
2302 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2304 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2305 PROCESSOR_GOLDMONT_PLUS.
2306 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2307 (processor_target_table): Add "goldmont-plus".
2308 (PTA_GOLDMONT_PLUS): Define.
2309 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2310 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2311 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2312 (fold_builtin_cpu): Add "goldmont-plus".
2313 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2314 (ix86_option_override_internal): Add "goldmont-plus".
2315 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2316 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2317 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2318 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2320 2018-05-17 Richard Biener <rguenther@suse.de>
2322 PR tree-optimization/85757
2323 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2324 remove defs that only feed that PHI from further processing.
2326 2018-05-16 Jim Wilson <jimw@sifive.com>
2328 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2330 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2332 2018-05-16 Mark Wielaard <mark@klomp.org>
2334 * dwarf2out.c (count_index_strings): New function.
2335 (output_indirect_strings): Call count_index_strings and generate
2336 header for dwarf_version >= 5.
2338 2018-05-16 Mark Wielaard <mark@klomp.org>
2340 * dwarf2out.c (dwarf_FORM): New function.
2341 (set_indirect_string): Use dwarf_FORM.
2342 (reset_indirect_string): Likewise.
2343 (size_of_die): Likewise.
2344 (value_format): Likewise.
2345 (output_die): Likewise.
2346 (add_skeleton_AT_string): Likewise.
2347 (output_macinfo_op): Likewise.
2348 (index_string): Likewise.
2349 (output_index_string_offset): Likewise.
2350 (output_index_string): Likewise.
2351 (count_index_strings): Likewise.
2353 2018-05-16 Carl Love <cel@us.ibm.com>
2355 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2356 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2358 2018-05-16 Martin Jambor <mjambor@suse.cz>
2360 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2361 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2363 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2365 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2366 (fnma<mode>4): Likewise.
2367 (fms<mode>4): Likewise.
2368 (fnms<mode>4): Likewise.
2369 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2370 (aarch64_fnma<mode>4): Likewise.
2371 (aarch64_fms<mode>4): Likewise.
2372 (aarch64_fnms<mode>4): Likewise.
2373 (aarch64_fnmadd<mode>4): Likewise.
2375 2018-05-16 Jason Merrill <jason@redhat.com>
2377 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2379 2018-05-16 Richard Biener <rguenther@suse.de>
2381 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2382 (dump_stmt_cost): Declare.
2383 (add_stmt_cost): Dump cost we add.
2384 (add_stmt_costs): New function.
2385 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2387 (vect_analyze_stmt): Adjust prototype.
2388 (vectorizable_condition): Likewise.
2389 (vectorizable_live_operation): Likewise.
2390 (vectorizable_reduction): Likewise.
2391 (vectorizable_induction): Likewise.
2392 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2393 cost vector to pass to vectorizable_ and record afterwards.
2394 (vect_model_reduction_cost): Take cost vector argument and adjust.
2395 (vect_model_induction_cost): Likewise.
2396 (vectorizable_reduction): Likewise.
2397 (vectorizable_induction): Likewise.
2398 (vectorizable_live_operation): Likewise.
2399 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2400 SLP_TREE_NUMBER_OF_VEC_STMTS.
2401 (vect_analyze_slp_cost_1): Remove.
2402 (vect_analyze_slp_cost): Likewise.
2403 (vect_slp_analyze_node_operations): Take visited args and
2404 a target cost vector. Avoid processing already visited stmt sets.
2405 (vect_slp_analyze_operations): Use a local cost vector to gather
2406 costs and register those of non-discarded instances.
2407 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2408 (vect_schedule_slp_instance): Remove copying of
2409 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2411 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2412 adding cost. Record cost entry location.
2413 (vect_prologue_cost_for_slp_op): Function to compute cost of
2414 a constant or invariant generated for SLP vect in the prologue,
2415 split out from vect_analyze_slp_cost_1.
2416 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2417 (vect_model_promotion_demotion_cost): Likewise.
2418 (vect_model_store_cost): Likewise, make static.
2419 (vect_model_load_cost): Likewise.
2420 (vectorizable_bswap): Add cost vector arg and adjust.
2421 (vectorizable_call): Likewise.
2422 (vectorizable_simd_clone_call): Likewise.
2423 (vectorizable_conversion): Likewise.
2424 (vectorizable_assignment): Likewise.
2425 (vectorizable_shift): Likewise.
2426 (vectorizable_operation): Likewise.
2427 (vectorizable_store): Likewise.
2428 (vectorizable_load): Likewise.
2429 (vectorizable_condition): Likewise.
2430 (vectorizable_comparison): Likewise.
2431 (can_vectorize_live_stmts): Likewise.
2432 (vect_analyze_stmt): Likewise.
2433 (vect_transform_stmt): Adjust calls to vectorizable_*.
2434 * tree-vectorizer.c: Include gimple-pretty-print.h.
2435 (dump_stmt_cost): New function.
2437 2018-05-16 Richard Biener <rguenther@suse.de>
2439 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2440 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2441 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2442 (check_name): New callback.
2443 (dse_classify_store): Track cycles via a visited bitmap of PHI
2444 defs and simplify handling of in-loop and across loop dead stores
2445 and properly fail for loop-variant refs. Handle byte-tracking with
2446 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2449 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2451 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2452 (vect_get_mask_type_for_stmt): Likewise.
2453 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2455 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2456 to determine the statement's vector type and the vector type that
2457 should be used for calculating nunits. Deal with cases in which
2458 the type has to be deferred.
2459 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2460 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2461 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2462 (vect_determine_vf_for_stmt): New functions, split out from...
2463 (vect_determine_vectorization_factor): ...here.
2464 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2465 (vect_get_mask_type_for_stmt): New functions, split out from
2466 vect_determine_vectorization_factor.
2468 2018-05-16 Richard Biener <rguenther@suse.de>
2470 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2471 verify the [VEC_]COND_EXPR embedded comparison.
2473 2018-05-15 Martin Sebor <msebor@redhat.com>
2475 PR tree-optimization/85753
2476 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2477 RECORD_TYPE in addition to ARRAY_TYPE.
2479 2018-05-15 Martin Sebor <msebor@redhat.com>
2482 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2484 2018-05-15 Richard Biener <rguenther@suse.de>
2486 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2487 add by_clobber_p one. Change algorithm to collect all defs
2488 representing uses we need to walk and try reducing them to
2489 a single one before failing.
2490 (dse_dom_walker::dse_optimize_stmt): Adjust.
2492 2018-05-13 Mark Wielaard <mark@klomp.org>
2494 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2495 (size_of_loc_descr): Likewise.
2496 (output_loc_operands): Likewise.
2497 (output_loc_operands_raw): Likewise.
2498 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2499 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2500 (hash_loc_operands): Likewise.
2501 (compare_loc_operands): Likewise.
2503 2018-05-14 Mark Wielaard <mark@klomp.org>
2505 * dwarf2out.c (count_index_addrs): New function.
2506 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2508 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2510 PR tree-optimization/83648
2511 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2512 return value as malloc candidate.
2514 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2517 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
2518 param as true in call to suggest_attribute.
2520 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2522 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2523 -mreadonly-in-sdata.
2525 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2527 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2529 (aarch64_crypto_aesd_fused): Likewise.
2531 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2533 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2534 (movsi_aarch64): Likewise.
2535 (load_pairsi): Likewise.
2536 (load_pairdi): Likewise.
2537 (store_pairsi): Likewise.
2538 (store_pairdi): Likewise.
2539 (load_pairsf): Likewise.
2540 (load_pairdf): Likewise.
2541 (store_pairsf): Likewise.
2542 (store_pairdf): Likewise.
2543 (zero_extend): Likewise.
2544 (trunc): Swap alternatives.
2545 (fcvt_target): Add '?' to prefer w over r.
2547 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2550 * config/i386/i386.md: Disallow non-commutative arithmetics in
2551 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2552 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2553 in the peephole2 before it.
2555 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2557 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2558 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2559 (ix86_handle_option): Handle -mcldemote.
2560 * config.gcc: New header.
2561 * config/i386/cldemoteintrin.h: New file.
2562 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2563 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2565 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2566 OPTION_MASK_ISA_CLDEMOTE.
2567 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2568 (ix86_valid_target_attribute_inner_p): Ditto.
2569 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2570 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2571 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2572 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2573 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2575 * config/i386/i386.opt: Add -mcldemote.
2576 * config/i386/x86intrin.h: New header.
2577 * doc/invoke.texi: Add -mcldemote.
2579 2018-05-14 Richard Biener <rguenther@suse.de>
2581 * doc/match-and-simplify.texi: Adjust :s documentation.
2583 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2585 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2586 intended memcpy size.
2587 (REORDER_45): Likewise.
2589 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2591 * sort.cc: New file.
2592 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2593 * vec.c (qsort_chk): Use gcc_qsort.
2594 * Makefile.in (OBJS-libcommon): Add sort.o.
2595 (build/sort.o): New target. Use it...
2596 (BUILD_RTL): ... here, and...
2597 (build/gencfn-macros): ... here, and...
2598 (build/genmatch): ... here.
2600 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2601 Chung-Ju Wu <jasonwucj@gmail.com>
2603 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2604 * config/nds32/nds32-graywolf.md: New file.
2605 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2606 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2608 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2609 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2610 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2611 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2612 * config/nds32/pipelines.md: Include n15 settings.
2614 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2615 Chung-Ju Wu <jasonwucj@gmail.com>
2617 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2618 * config/nds32/nds32-n13.md: New file.
2619 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2620 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2622 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2623 * config/nds32/nds32.md (pipeline_model): Add n13.
2624 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2625 * config/nds32/pipelines.md: Include n13 settings.
2627 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2628 Chung-Ju Wu <jasonwucj@gmail.com>
2630 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2631 * config/nds32/nds32-n10.md: New file.
2632 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2633 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2635 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2636 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2637 * config/nds32/nds32.md (pipeline_model): Add n10.
2638 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2639 * config/nds32/pipelines.md: Include n10 settings.
2641 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2642 Kito Cheng <kito.cheng@gmail.com>
2643 Chung-Ju Wu <jasonwucj@gmail.com>
2645 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2646 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2647 Add enum values for DSP extension instructions.
2648 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2650 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2651 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2653 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2654 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2655 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2656 * config/nds32/nds32-intrinsic.md: Likewise.
2657 * config/nds32/nds32_intrinsic.h: Likewise.
2658 * config/nds32/nds32-md-auxiliary.c: Likewise.
2659 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2660 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2661 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2662 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2663 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2664 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2665 TYPE_DMAC in switch statement.
2666 * config/nds32/nds32.c: New checking and implementation for DSP
2667 extension instructions.
2668 * config/nds32/nds32.h: Likewise.
2669 * config/nds32/nds32.md: Likewise.
2670 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2671 * config/nds32/predicates.md: Implement new predicates for DSP
2674 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2676 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2677 Reformat alternatives and attributes so it is easier to identify
2678 which constraints/attributes go with which instruction.
2679 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2680 (mov<mode>_softfloat32, FMOVE64): Likewise.
2681 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2682 (mov<mode>_softfloat64, FMOVE64): Likewise.
2684 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2686 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2688 (Basic PowerPC Built-in Functions): The new name of the
2689 subsection previously known as "PowerPC Built-in Functions".
2690 (Basic PowerPC Built-in Functions Available on all Configurations):
2692 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2693 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2694 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2695 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2697 2018-05-11 Martin Jambor <mjambor@suse.cz>
2700 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2703 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2706 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2708 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2710 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2711 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2712 (ix86_handle_option): Handle -mwaitpkg.
2713 * config.gcc: New header.
2714 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2715 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2716 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2718 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2719 OPTION_MASK_ISA_WAITPKG.
2720 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2721 (ix86_option_override_internal): Add PTA_WAITPKG.
2722 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2723 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2724 IX86_BUILTIN_TPAUSE.
2725 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2726 __builtin_ia32_umwait and __builtin_ia32_tpause.
2727 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2728 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2729 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2730 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2731 UNSPECV_TPAUSE): New.
2732 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2733 * config/i386/i386.opt: Add -mwaitpkg.
2734 * config/i386/waitpkgintrin.h: New file.
2735 * config/i386/x86intrin.h: New header.
2736 * doc/invoke.texi: Add -mwaitpkg.
2738 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2741 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2743 (cortex-m0): Use armv6s-m isa.
2744 (cortex-m0plus): Likewise.
2745 (cortex-m1): Likewise.
2746 (cortex-m0.small-multiply): Likewise.
2747 (cortex-m0plus.small-multiply): Likewise.
2748 (cortex-m1.small-multiply): Likewise.
2750 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2751 Jakub Jelinek <jakub@redhat.com>
2753 PR tree-optimization/85692
2754 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2755 source permute as well.
2757 2018-05-11 Martin Liska <mliska@suse.cz>
2760 * doc/extend.texi: Document LLVM style format for no_sanitize
2763 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2765 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2766 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2767 (mode_supports_vsx_dform_quad): Likewise.
2768 (mode_supports_vmx_dform): Move these functions to be next to the
2769 other mode_supports functions.
2770 (mode_supports_dq_form): Likewise.
2771 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2772 mode_supports_dq_form.
2773 (reg_offset_addressing_ok_p): Likewise.
2774 (offsettable_ok_by_alignment): Likewise.
2775 (rs6000_legitimate_offset_address_p): Likewise.
2776 (legitimate_lo_sum_address_p): Likewise.
2777 (rs6000_legitimize_address): Likewise.
2778 (rs6000_legitimize_reload_address): Likewise.
2779 (rs6000_secondary_reload_inner): Likewise.
2780 (rs6000_preferred_reload_class): Likewise.
2781 (rs6000_output_move_128bit): Likewise.
2783 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2785 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2786 Generate SImode target register for null target.
2787 <case IX86_BUILTIN_XGETBV>: Ditto.
2788 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2789 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2791 2018-05-10 Carl Love <cel@us.ibm.com>
2793 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2794 dcbtt and dcbtstt if operands[2] is 0.
2796 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2799 * config/i386/sse.md (usadv64qi): New expander.
2801 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2803 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2804 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2805 -maltivec=be support.
2806 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2807 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2808 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2809 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2810 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2811 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2812 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2813 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2814 altivec_vsumsws): Adjust.
2815 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2816 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2817 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2819 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2820 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2821 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2822 (altivec_lve<VI_char>x): Delete expand.
2823 (*altivec_lve<VI_char>x_internal): Rename to...
2824 (altivec_lve<VI_char>x): ... this.
2825 (altivec_lvxl_<mode>): Delete expand.
2826 (*altivec_lvxl_<mode>_internal): Rename to ...
2827 (altivec_lvxl_<mode>): ... this.
2828 (altivec_stvxl_<mode>): Delete expand.
2829 (*altivec_stvxl_<mode>_internal): Rename to ...
2830 (altivec_stvxl_<mode>): ... this.
2831 (altivec_stve<VI_char>x): Delete expand.
2832 (*altivec_stve<VI_char>x_internal): Rename to ...
2833 (altivec_stve<VI_char>x): ... this.
2834 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2835 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2836 reduc_plus_scal_<mode>): Adjust.
2837 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2839 (rs6000_cpu_cpp_builtins): Adjust.
2840 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2841 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2842 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2843 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2844 -maltivec=be support.
2845 (rs6000_split_vec_extract_var): Adjust.
2846 (rs6000_split_v4si_init): Adjust.
2847 (swap_selector_for_mode): Delete.
2848 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2849 altivec_expand_stvex_be): Delete.
2850 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2851 -maltivec=be support.
2852 (rs6000_gimple_fold_builtin): Ditto.
2853 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2855 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2856 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2857 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2858 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2859 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2860 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2861 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2862 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2863 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2864 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2865 anonymous split): Adjust.
2866 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2867 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2869 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2871 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2872 when --with-gxx-include-dir is also specified.
2873 * configure: Regenerate.
2875 2018-05-09 Jim Wilson <jimw@sifive.com>
2878 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2879 * config/riscv/t-withmultilib: New.
2880 * config/riscv/withmultilib.h: New.
2881 * doc/install.texi: Document RISC-V --with-multilib-list support.
2883 2018-05-09 Richard Biener <rguenther@suse.de>
2885 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2887 (vect_bb_vectorization_profitable_p): Adjust. Compute
2888 actual scalar cost using the cost vector and the add_stmt_cost
2891 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2893 PR rtl-optimization/85645
2894 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2895 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2897 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2899 PR rtl-optimization/85645
2900 * shrink-wrap.c (spread_components): Return a boolean saying if
2901 anything was changed.
2902 (try_shrink_wrapping_separate): Iterate spread_components until
2903 nothing changes anymore.
2905 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2907 PR rtl-optimization/85645
2908 * regrename.c (build_def_use): Also kill the chains that include the
2909 destination of a REG_CFA_REGISTER note.
2911 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2913 PR rtl-optimization/85645
2914 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2915 insn that has a REG_CFA_REGISTER note.
2917 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2919 * cfgexpand.c (expand_clobber): New function.
2920 (expand_gimple_stmt_1): Use it.
2921 * tree-vect-stmts.c (vect_clobber_variable): New function,
2923 (vectorizable_simd_clone_call): ...here.
2924 (vectorizable_store): Emit a clobber either side of an
2925 IFN_STORE_LANES sequence.
2926 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2928 2018-05-09 Tom de Vries <tom@codesourcery.com>
2931 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2932 (define_insn "trap_if_false"): Add exit after trap.
2934 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2936 PR rtl-optimization/85638
2937 * bb-reorder.c: Include common/common-target.h.
2938 (create_forwarder_block): New function extracted from...
2939 (fix_up_crossing_landing_pad): ...here. Rename into...
2940 (dw2_fix_up_crossing_landing_pad): ...this.
2941 (sjlj_fix_up_crossing_landing_pad): New function.
2942 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2943 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2944 from both partitions and exit the loop after one iteration.
2946 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2949 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2951 (Basic PowerPC Built-in Functions): The new name of the
2952 subsection previously known as "PowerPC Built-in Functions".
2953 (Basic PowerPC Built-in Functions Available on all Configurations):
2955 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2957 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2959 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2961 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2964 2018-05-08 Jim Wilson <jimw@sifive.com>
2966 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2967 (LD_EMUL_SUFFIX): New.
2968 (LINK_SPEC): Use it.
2970 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2972 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2974 (Basic PowerPC Built-in Functions): The new name of the
2975 subsection previously known as "PowerPC Built-in Functions".
2976 (Basic PowerPC Built-in Functions Available on all Configurations):
2978 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2980 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2982 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2984 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2987 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2990 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2991 after cmpelim optimization.
2993 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2995 * config.gcc: Support "goldmont".
2996 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2997 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2999 * config/i386/i386.c (m_GOLDMONT): Define.
3000 (processor_target_table): Add "goldmont".
3001 (PTA_GOLDMONT): Define.
3002 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
3003 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
3004 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
3005 (fold_builtin_cpu): Add "goldmont".
3006 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
3007 (ix86_option_override_internal): Add "goldmont".
3008 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
3009 (processor_type): Add PROCESSOR_GOLDMONT.
3010 * config/i386/i386.md: Add CPU "glm".
3011 * config/i386/glm.md: New file.
3012 * config/i386/x86-tune.def: Add m_GOLDMONT.
3013 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
3015 2018-05-08 Jakub Jelinek <jakub@redhat.com>
3018 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
3020 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
3021 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
3022 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
3025 * config/i386/i386.c (ix86_fold_builtin): Handle
3026 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
3029 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
3030 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
3032 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3035 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3036 (check_arch): Likewise.
3037 (check_fpu): Return the result rather than printing it.
3038 (end arch): Fix operator precedence.
3039 (end cpu): Likewise.
3040 (END): Print the result from check_fpu.
3042 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3043 Alan Hayward <alan.hayward@arm.com>
3044 David Sherwood <david.sherwood@arm.com>
3046 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3047 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3048 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3049 (*fcmuo<mode>_and): New patterns.
3051 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3053 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3054 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3055 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3056 (cmp_op, sve_imm_con): New code attributes.
3057 (SVE_COND_INT_CMP, imm_con): Delete.
3058 (cmp_op): Remove above unspecs from int attribute.
3059 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3061 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3062 comparison-specific unspecs.
3063 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3064 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3065 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3066 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3067 (*vec_fcm<cmp_op><mode>): Rename to...
3068 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3069 (*vec_fcmuo<mode>): Rename to...
3070 (*fcmuo<mode>): ...this and adjust likewise.
3071 (*pred_fcm<cmp_op><mode>): New pattern.
3072 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3073 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3075 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3077 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3078 (aarch64_emit_sve_predicated_cond): New function.
3079 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3080 (aarch64_emit_unspec_cond_or): Replace with...
3081 (aarch64_emit_sve_or_conds): ...this new function. Use
3082 aarch64_emit_sve_ptrue_op for the individual comparisons and
3083 aarch64_emit_binop to OR them together.
3084 (aarch64_emit_inverted_unspec_cond): Replace with...
3085 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3086 aarch64_emit_sve_ptrue_op for the comparison and
3087 aarch64_emit_unop to invert the result.
3088 (aarch64_expand_sve_vec_cmp_float): Update after the above
3089 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3091 2018-05-07 Nathan Sidwell <nathan@acm.org>
3093 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3094 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3095 (Backwards Compatibility): Likewise.
3097 2018-05-07 Luis Machado <luis.machado@linaro.org>
3101 2018-05-07 Luis Machado <luis.machado@linaro.org>
3103 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3104 <prefetch_dynamic_strides>: New const bool field.
3105 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3106 prefetch_dynamic_strides.
3107 (exynosm1_prefetch_tune): Likewise.
3108 (thunderxt88_prefetch_tune): Likewise.
3109 (thunderx_prefetch_tune): Likewise.
3110 (thunderx2t99_prefetch_tune): Likewise.
3111 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3113 (aarch64_override_options_internal): Update to set
3114 PARAM_PREFETCH_DYNAMIC_STRIDES.
3115 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3116 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3117 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3118 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3119 prefetch-dynamic-strides setting.
3121 2018-05-07 Luis Machado <luis.machado@linaro.org>
3123 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3124 <minimum_stride>: New const int field.
3125 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3126 minimum_stride field.
3127 (exynosm1_prefetch_tune): Likewise.
3128 (thunderxt88_prefetch_tune): Likewise.
3129 (thunderx_prefetch_tune): Likewise.
3130 (thunderx2t99_prefetch_tune): Likewise.
3131 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3132 (aarch64_override_options_internal): Update to set
3133 PARAM_PREFETCH_MINIMUM_STRIDE.
3134 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3135 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3136 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3137 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3138 stride is constant and is below the minimum stride threshold.
3140 2018-05-07 Luis Machado <luis.machado@linaro.org>
3142 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3145 2018-05-07 Luis Machado <luis.machado@linaro.org>
3147 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3148 <prefetch_dynamic_strides>: New const bool field.
3149 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3150 prefetch_dynamic_strides.
3151 (exynosm1_prefetch_tune): Likewise.
3152 (thunderxt88_prefetch_tune): Likewise.
3153 (thunderx_prefetch_tune): Likewise.
3154 (thunderx2t99_prefetch_tune): Likewise.
3155 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
3157 (aarch64_override_options_internal): Update to set
3158 PARAM_PREFETCH_DYNAMIC_STRIDES.
3159 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3160 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3161 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3162 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3163 prefetch-dynamic-strides setting.
3165 2018-05-07 Luis Machado <luis.machado@linaro.org>
3167 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3168 <minimum_stride>: New const int field.
3169 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3170 minimum_stride field.
3171 (exynosm1_prefetch_tune): Likewise.
3172 (thunderxt88_prefetch_tune): Likewise.
3173 (thunderx_prefetch_tune): Likewise.
3174 (thunderx2t99_prefetch_tune): Likewise.
3175 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3176 (aarch64_override_options_internal): Update to set
3177 PARAM_PREFETCH_MINIMUM_STRIDE.
3178 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3179 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3180 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3181 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3182 stride is constant and is below the minimum stride threshold.
3184 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3187 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3188 the type is addressable. Don't force op into register if it has
3191 2018-05-05 Roland McGrath <mcgrathr@google.com>
3194 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3195 any section for which we don't know a specific type it should have,
3196 regardless of name. Previously this was done only for the exact
3197 names ".init_array", ".fini_array", and ".preinit_array".
3198 (default_elf_asm_named_section): Add comment about
3199 relationship with default_section_type_flags and SECTION_NOTYPE.
3200 (get_section): Don't consider it a type conflict if one side has
3201 SECTION_NOTYPE and the other doesn't, as long as neither has the
3202 SECTION_BSS et al used in the default_section_type_flags logic.
3204 2018-05-05 Tom de Vries <tom@codesourcery.com>
3207 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3208 (workaround_barsyncs): New function.
3209 (nvptx_reorg): Use workaround_barsyncs.
3210 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3211 (define_expand "nvptx_membar_cta"): New define_expand.
3212 (define_insn "*nvptx_membar_cta"): New insn.
3214 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3216 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3217 To improve optimization opportunities.
3218 * builtin-types.def: The new needed builtin types for the above.
3220 2018-05-04 Richard Biener <rguenther@suse.de>
3222 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3223 * gimple-ssa-store-merging.c
3224 (imm_store_chain_info::output_merged_store): Remove redundant create,
3225 release split_store vector contents on failure.
3226 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3227 scalar stmt vector on cache hit.
3229 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3231 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3233 * config.gcc (powerpc-xilinx-eabi*): Remove.
3234 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3236 (fusion_addis_mem_combo_load): Ditto.
3237 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3239 (rs6000_cpu_cpp_builtins): Ditto.
3240 * config/rs6000/rs6000-linux.c
3241 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3242 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3243 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3245 (rs6000_setup_reg_addr_masks): Ditto.
3246 (rs6000_init_hard_regno_mode_ok): Ditto.
3247 (rs6000_option_override_internal): Ditto.
3248 (legitimate_lo_sum_address_p): Ditto.
3249 (rs6000_legitimize_address): Ditto.
3250 (rs6000_legitimize_reload_address): Ditto.
3251 (rs6000_legitimate_address_p): Ditto.
3252 (abi_v4_pass_in_fpr): Ditto.
3253 (setup_incoming_varargs): Ditto.
3254 (rs6000_gimplify_va_arg): Ditto.
3255 (rs6000_split_multireg_move): Ditto.
3256 (rs6000_savres_strategy): Ditto.
3257 (rs6000_emit_prologue_components): Ditto.
3258 (rs6000_emit_epilogue_components): Ditto.
3259 (rs6000_emit_prologue): Ditto.
3260 (rs6000_emit_epilogue): Ditto.
3261 (rs6000_elf_file_end): Ditto.
3262 (rs6000_function_value): Ditto.
3263 (rs6000_libcall_value): Ditto.
3264 * config/rs6000/rs6000.h: Ditto.
3265 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3266 (TARGET_MINMAX): ... this. New.
3267 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3268 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3269 (*movsi_internal1_single): Delete.
3270 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3271 mfpu=, mxilinx-fpu): Delete.
3272 * config/rs6000/singlefp.h: Delete.
3273 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3274 * config/rs6000/t-rs6000: Ditto.
3275 * config/rs6000/t-xilinx: Delete.
3276 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3277 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3278 (VStype_simple): Delete.
3279 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3280 * config/rs6000/xfpu.h: Delete.
3281 * config/rs6000/xfpu.md: Delete.
3282 * config/rs6000/xilinx.h: Delete.
3283 * config/rs6000/xilinx.opt: Delete.
3284 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3285 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3287 2018-05-04 Tom de Vries <tom@codesourcery.com>
3290 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3293 2018-05-04 Richard Biener <rguenther@suse.de>
3296 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3297 (expand_complex_div_wide): Likewise.
3298 (expand_complex_operations_1): Likewise.
3299 (expand_complex_libcall): Preserve EH info of the original stmt.
3300 (tree_lower_complex): Handle removed blocks.
3301 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3302 on complex multiplication and division libcall builtins.
3304 2018-05-04 Richard Biener <rguenther@suse.de>
3307 * fold-const.c (negate_expr_p): Restrict negation of operand
3308 zero of a division to when we know that can happen without
3310 (fold_negate_expr_1): Likewise.
3312 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3315 * real.h (real_nextafter): Declare.
3316 * real.c (real_nextafter): New function.
3317 * fold-const-call.c (fold_const_nextafter): New function.
3318 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3319 CASE_CFN_NEXTTOWARD.
3320 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3321 even when arg1_mode is different from arg0_mode.
3323 2018-05-03 Nathan Sidwell <nathan@acm.org>
3325 * doc/extend.texi (Deprecated Features): Remove
3327 (Backwards Compatibility): Likewise.
3328 * doc/invoke.texi (C++ Language Options): Likewise.
3329 (C++ Dialect Options): Likewise.
3331 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3334 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3335 _mm512_mask_mullox_epi64): New intrinsics.
3337 2018-05-03 Tom de Vries <tom@codesourcery.com>
3340 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3341 dump files): Add offload-tree.
3343 2018-05-03 Richard Biener <rguenther@suse.de>
3345 PR tree-optimization/85615
3346 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3347 to loops not nested in BBs loop father to avoid creating multi-entry
3350 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3352 PR tree-optimization/70291
3353 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3354 arguments. Change return type to tree. Emit libcall as a new
3355 statement rather than replacing existing one when inplace_p is true.
3356 (expand_complex_multiplication_components): New function.
3357 (expand_complex_multiplication): Expand floating-point complex
3358 multiplication using the above.
3359 (expand_complex_division): Rename inner_type parameter to type.
3360 Update expand_complex_libcall call-site.
3361 (expand_complex_operations_1): Update expand_complex_multiplication
3362 and expand_complex_division call-sites.
3364 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3367 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3368 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3369 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3370 the highest significant bit of the shift count mask is clear. In
3371 check whether and[sq]i3 is needed verify that all significant bits
3372 of the shift count other than the highest are set.
3374 2018-05-02 Tom de Vries <tom@codesourcery.com>
3377 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3378 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3379 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3380 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3381 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3382 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3383 __builtin_goacc_parlevel_size.
3385 2018-05-02 Richard Biener <rguenther@suse.de>
3387 PR tree-optimization/85597
3388 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3389 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3392 2018-05-02 Tom de Vries <tom@codesourcery.com>
3395 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3396 dump files): Add ltrans-tree.
3398 2018-05-02 Tom de Vries <tom@codesourcery.com>
3401 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3402 dump files): Add wpa-ipa.
3404 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3406 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3407 powerpc*-*-linux*paired* target.
3408 * config/rs6000/750cl.h: Delete.
3409 * config/rs6000/paired.h: Delete.
3410 * config/rs6000/paired.md: Delete.
3411 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3413 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3414 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3415 comment. Remove paired float support.
3416 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3417 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3419 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3420 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3422 * config/rs6000/rs6000.c: Remove paired float support.
3423 (paired_expand_vector_init, paired_expand_vector_move,
3424 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3425 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3426 paired_expand_builtin, paired_expand_predicate_builtin,
3427 paired_init_builtins): Delete.
3428 * config/rs6000/rs6000.h: Remove paired float support.
3429 * config/rs6000/rs6000.md: Remove paired float support.
3430 (move_from_CR_ov_bit): Delete.
3431 * config/rs6000/rs6000.opt (mpaired): Delete.
3432 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3433 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3435 2018-05-02 Richard Biener <rguenther@suse.de>
3438 * gimplify.c (gimplify_save_expr): When in SSA form allow
3439 SAVE_EXPRs to compute to SSA vars.
3441 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3444 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3445 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3446 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3447 clobber operands[2], instead use a new pseudo. Formatting fixes.
3449 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3451 PR tree-optimization/85586
3452 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3453 exit early for statements in the same group if the accesses are
3456 2018-05-02 Tom de Vries <tom@codesourcery.com>
3459 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3462 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3464 PR tree-optimization/85143
3465 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3467 2018-05-01 Tom de Vries <tom@codesourcery.com>
3470 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3471 not found" error message.
3473 2018-05-01 Tom de Vries <tom@codesourcery.com>
3476 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3477 * vec.c (test_ordered_remove_if): New function.
3478 (vec_c_tests): Call test_ordered_remove_if.
3479 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3480 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3481 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3482 VEC_ORDERED_REMOVE_IF.
3484 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3486 PR tree-optimization/82665
3487 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3488 pointer subtraction where arguments come from a memchr call.
3490 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3492 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3493 --push-state --as-needed and --pop-state instead of --as-needed and
3494 --no-as-needed if ld supports it.
3495 * configure: Regenerated.
3498 * doc/install.texi2html: Replace _002d with - and _002a with * in
3499 generated html files using sed.
3501 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3504 * gcc-rich-location.c (blank_line_before_p): New function.
3505 (use_new_line): New function.
3506 (gcc_rich_location::add_fixit_insert_formatted): New function.
3507 * gcc-rich-location.h
3508 (gcc_rich_location::add_fixit_insert_formatted): New function.
3510 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3512 * selftest.c (assert_streq): Rename "expected" and "actual" to
3513 "val1" and "val2". Extend NULL-handling to cover both inputs
3514 symmetrically, while still requiring both to be non-NULL for a pass.
3515 * selftest.h (assert_streq): Rename "expected" and "actual" to
3517 (ASSERT_EQ): Likewise.
3518 (ASSERT_EQ_AT): Likewise.
3519 (ASSERT_KNOWN_EQ): Likewise.
3520 (ASSERT_KNOWN_EQ_AT): Likewise.
3521 (ASSERT_NE): Likewise.
3522 (ASSERT_MAYBE_NE): Likewise.
3523 (ASSERT_MAYBE_NE_AT): Likewise.
3524 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3525 the assertion to pass.
3526 (ASSERT_STREQ_AT): Likewise.
3528 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3530 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3531 interaction with -pie.
3533 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3535 * selftest.h: Fix alphabetization of per-source-file selftest
3538 2018-04-30 Jason Merrill <jason@redhat.com>
3540 PR c++/61982 - dead stores to destroyed objects.
3541 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3544 2018-04-30 Jason Merrill <jason@redhat.com>
3546 * tree.c (build_clobber): New.
3547 * tree.h: Declare it.
3548 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3550 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3552 * diagnostic-show-locus.c (layout::layout): Update for
3553 location_get_source_line returning a char_span.
3554 (struct char_span): Move to input.h.
3555 (struct correction): Update for fields in char_span becoming
3557 (struct source_line): Update for location_get_source_line
3558 returning a char_span.
3559 (layout::print_line): Likewise.
3560 * edit-context.c (edited_file::print_content): Likewise.
3561 (edited_file::print_diff_hunk): Likewise.
3562 (edited_file::print_run_of_changed_lines): Likewise.
3563 (edited_file::get_num_lines): Likewise.
3564 (edited_line::edited_line): Likewise.
3565 * final.c (asm_show_source): Likewise.
3566 * input.c (location_get_source_line): Convert return type
3567 from const char * to char_span, losing the final "line_len"
3569 (dump_location_info): Update for the above.
3570 (get_substring_ranges_for_loc): Likewise. Use a char_span
3571 when handling the literal within the line.
3572 (test_reading_source_line): Update for location_get_source_line
3573 returning a char_span.
3574 * input.h (class char_span): Move here from
3575 diagnostic-show-locus.c, converting from a struct to a class.
3576 Make data members private.
3577 (char_span::operator bool): New.
3578 (char_span::length): New.
3579 (char_span::get_buffer): New.
3580 (char_span::operator[]): New.
3581 (char_span::subspan): Make const.
3582 (char_span::xstrdup): New.
3583 (location_get_source_line): Convert return type from const char *
3584 to char_span, losing the final "line_size" param.
3586 2018-04-30 Jan Hubicka <jh@suse.cz>
3588 * lto-wrapper.c (ltrans_priorities): New static var.
3589 (cmp_priority): New.
3590 (run_gcc): Read priorities and if doing parallel build order
3591 the Makefile by them.
3593 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3595 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3597 2018-04-30 Richard Biener <rguenther@suse.de>
3599 * tree-cfg.c (verify_address): Remove base argument, add
3600 flag whether to check TREE_ADDRESSABLE and do that.
3601 (verify_expr): Remove.
3602 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3603 (verify_gimple_assign_single): Likewise.
3604 (verify_gimple_switch): Likewise.
3605 (verify_expr_location_1): Dereference tp once. Add (disabled)
3606 piece from verify_expr.
3607 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3609 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3611 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3613 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3615 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3616 (small_data_pattern): Likewise.
3617 (arc_rewrite_small_data): Likewise.
3618 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3619 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3620 (get_symbol_alignment): New function.
3621 (legitimate_small_data_address_p): Likewise.
3622 (legitimate_scaled_address): Update, call
3623 legitimate_small_data_address_p.
3624 (output_sdata): New static variable.
3625 (arc_print_operand): Update how we handle small data operands.
3626 (arc_print_operand_address): Likewise.
3627 (arc_legitimate_address_p): Update, use
3628 legitimate_small_data_address_p.
3629 (arc_rewrite_small_data_p): Remove.
3630 (arc_rewrite_small_data_1): Likewise.
3631 (arc_rewrite_small_data): Likewise.
3632 (small_data_pattern): Likewise.
3633 (compact_sda_memory_operand): Update to use
3634 legitimate_small_data_address_p and get_symbol_alignment.
3635 (prepare_move_operands): Don't rewite sdata pattern.
3636 (prepare_extend_operands): Remove.
3637 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3639 (zero_extendqisi2): Likewise.
3640 (zero_extendhisi2): Likewise.
3641 (extendqihi2): Likewise.
3642 (extendqisi2): Likewise.
3643 (extendhisi2): Likewise.
3647 * config/arc/constraints.md (Usd): Change it to memory constraint.
3649 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3651 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3652 as source of std instructions.
3653 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3654 allow 6-bit constants as source for store instructions.
3655 (movdi_insn): Update instruction pattern to allow 6-bit constants
3656 as source for store instructions.
3658 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3660 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3662 2018-04-30 Nathan Sidwell <nathan@acm.org>
3663 Sandra Loosemore <sandra@codesourcery.com>
3665 * dumpfile.c (dump_open): Allow '-' for stdout.
3666 * doc/invoke.texi (Developer Options): Document dump filename
3667 determination early. Document stdin/stdout selection.
3669 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3671 Microblaze Target: PIC data text relative
3673 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3674 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3676 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3677 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3678 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3679 New addressing mode for data-text relative position indepenedent code.
3680 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3681 'ADDRESS_SYMBOLIC_TXT_REL'.
3682 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3683 (microblaze_legitimate_pic_operand): Exclude function calls from
3684 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3685 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3687 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3688 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3689 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3690 for 'address + offset'.
3691 (microblaze_expand_prologue): Add new function prologue call for
3693 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3694 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3695 table in case of TARGET_PIC_DATA_TEXT_REL.
3696 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3697 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3698 Add new macros 'UNSPEC_TEXT',
3699 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3700 + exclude function calls from 'UNSPEC_PLT' in case of data text
3702 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3703 new target hook for generating address diff vector tables in case of
3705 * doc/tm.texi : Regenerate.
3706 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3707 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3708 of addr diff vector generation.
3709 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3710 target hook definition.
3711 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3712 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3713 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3714 option for data text relative.
3716 2018-04-30 Richard Biener <rguenther@suse.de>
3718 * tree-chrec.h (evolution_function_is_constant_p): Remove
3720 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3722 2018-04-30 Richard Biener <rguenther@suse.de>
3725 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3727 2018-04-30 Richard Biener <rguenther@suse.de>
3729 PR tree-optimization/28364
3730 PR tree-optimization/85275
3731 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3732 copying first exit test.
3734 2018-04-28 Mark Wielaard <mark@klomp.org>
3736 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3738 (dwarf_AT): Handle DW_AT_addr_base.
3739 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3741 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3744 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3745 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3746 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3747 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3749 2018-04-28 Richard Biener <rguenther@suse.de>
3751 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3752 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3753 to reflect use. Only add interesting stmts.
3755 2018-04-27 Martin Jambor <mjambor@suse.cz>
3758 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3759 the jump function allows for passing through aggregate values.
3761 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3763 * input.h (in_system_header_at): Convert from macro to inline
3765 (from_macro_expansion_at): Likewise.
3766 (from_macro_definition_at): Likewise.
3768 2018-04-27 Jeff Law <law@redhat.com>
3770 * config.gcc: Mark tile* targets as deprecated/obsolete.
3772 2018-04-27 Richard Biener <rguenther@suse.de>
3774 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3777 2018-04-27 Richard Biener <rguenther@suse.de>
3779 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3781 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3783 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3784 with Yd constraint. Set "preferred_for_speed" attribute from
3785 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3787 (*movdi_internal): Ditto.
3788 (movti_interunit splitters): Remove
3789 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3790 (movdi_interunit splitters): Ditto.
3791 * config/i386/constraints.md (Ye): Remove.
3792 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3794 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3797 * config/aarch64/constraints.md (Usg): Limit to 31.
3800 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3802 PR tree-optimization/85529
3803 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3804 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3805 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3806 zero extension or masking of the MSB bit.
3807 (optimize_range_tests): Add FIRST_BB argument, pass it through
3808 to optimize_range_tests_var_bound.
3809 (maybe_optimize_range_tests, reassociate_bb): Adjust
3810 optimize_range_tests callers.
3812 2018-04-26 Richard Biener <rguenther@suse.de>
3813 Jakub Jelinek <jakub@redhat.com>
3815 * cgraph.h (symbol_table): Just declare debug method here.
3816 * symtab.c (symbol_table::debug): Define.
3818 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3820 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3822 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3824 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3825 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3826 (*movdi_internal): Substitute Yi and Yj constraint with x
3827 and Ym and Yn constraint with y constraint. Update "isa"
3828 attribute and set "preferred_for_speed" attribute from
3829 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3830 (*movsi_internal): Ditto.
3831 (*movdf_internal): Ditto.
3832 (*movsf_internal): Ditto.
3833 (*zero_extendsidi2): Ditto.
3834 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3835 (sse2_loadld): Ditto.
3836 (*vec_extract<ssevecmodelower>_0): Ditto.
3837 (*vec_extractv4si_0_zext_sse4): Ditto.
3838 (vec_concatv2di): Ditto.
3839 (*vec_dup<mode>): Ditto.
3840 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3841 * config/i386/constraints.md (Yi): Remove.
3846 2018-04-26 Nathan Sidwell <nathan@acm.org>
3848 * dumpfile.c (dump_open): New.
3849 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3850 (dump_finish): Detect stdio/stderr by value not name.
3852 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3854 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3856 2018-04-26 Tom de Vries <tom@codesourcery.com>
3859 * config/nvptx/nvptx.c (verify_neutering_jumps)
3860 (verify_neutering_labels): New function
3861 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3863 2018-04-26 Tom de Vries <tom@codesourcery.com>
3866 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3867 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3870 2018-04-26 Richard Biener <rguenther@suse.de>
3871 Tom de Vries <tom@codesourcery.com>
3874 * lto-streamer-out.c (output_function): Fixup loops if required to match
3875 discovery done in the reader.
3877 2018-04-26 Richard Biener <rguenther@suse.de>
3879 PR tree-optimization/85116
3880 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3881 have a loop exit from the single latch predecessor. Remove
3882 case of header with just condition.
3883 (ch_base::copy_headers): Exclude infinite loops from any
3885 (pass_ch::execute): Record exits.
3887 2018-04-26 Richard Biener <rguenther@suse.de>
3889 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3890 prologue cost vector and pass it to vect_get_load_cost.
3891 (vect_get_peeling_costs_all_drs): Likewise.
3892 (vect_peeling_hash_get_lowest_cost): Likewise.
3893 (vect_enhance_data_refs_alignment): Likewise.
3895 2018-04-26 Richard Biener <rguenther@suse.de>
3898 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3899 checking of integer<->pointer conversions.
3900 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3901 sign-/zero-extending pointer types.
3902 (expand_omp_for_static_chunk): Likewise.
3904 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3905 Jean Lee <xiaoyur347@gmail.com>
3907 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3908 (TARGET_ASAN_SHADOW_OFFSET): Define.
3909 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3910 true for -fsanitize=address.
3912 2018-04-25 Mark Wielaard <mark@klomp.org>
3914 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3917 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3919 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3920 than "alu", remove explicit "memory" and "imm_disp" attributes.
3921 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3924 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3925 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3926 gen_lowpart_no_emit.
3928 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3931 * config/i386/i386.c (ix86_expand_builtin): Change memory
3932 operand to XI, extend p0 to Pmode.
3933 * config/i386/i386.md: Change unspec volatile and operand
3934 1 mode to XI, change operand 0 mode to P.
3936 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3938 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3939 GET_MODE_MASK before any checking.
3940 (nds32_can_use_bset_p): Likewise.
3941 (nds32_can_use_btgl_p): Likewise.
3943 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3945 * config/nds32/nds32-doubleword.md: New define_split pattern for
3946 illegal register number.
3948 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3950 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3952 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3954 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3956 2018-04-25 Richard Biener <rguenther@suse.de>
3958 * lto-streamer.h (LTO_major_version): Bump to 8.
3960 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3962 * BASE-VER: Set to 9.0.0.
3964 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3966 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3967 in __abskf2 and __powikf2.
3969 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3972 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3973 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3974 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3975 Use the above on operand 2. Reindent.
3976 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3978 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3981 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3983 * config/i386/i386.opt (mcet): Removed.
3984 * doc/install.texi: Remove -mcet documentation.
3985 * doc/invoke.texi: Likewise.
3987 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3990 * doc/install.texi: Remove -mcet from bootstrap-cet.
3992 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3995 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3996 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
4000 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
4001 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
4002 containing a CONST_VECTOR.
4004 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
4006 * doc/install.texi: Update newlib dependency for nvptx.
4008 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4011 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
4012 instead of INTVAL when shifting x left.
4014 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
4016 PR tree-optimization/85478
4017 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
4018 vect_grouped_store_supported for single element vectors.
4020 2018-04-24 Richard Biener <rguenther@suse.de>
4023 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
4024 load cost increase to the case of non-constant step.
4026 2018-04-24 Jakub Jelinek <jakub@redhat.com>
4029 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
4030 destination if any_malformed_asm.
4032 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4035 * expr.c (store_field): In the bitfield case, if the value comes from
4036 a function call and is returned in registers by means of a PARALLEL,
4037 do not change the mode of the temporary unless BLKmode and VOIDmode.
4039 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4041 PR rtl-optimization/85423
4042 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4043 dependencies to debug insns when the previous insn is non-debug.
4045 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4047 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4048 enums into a single definition.
4049 (fls): Fix predicates and printing.
4052 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4054 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4055 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4056 and short u6 immediate.
4057 (check_if_valid_sleep_operand): Remove.
4058 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4060 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4062 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4063 flag_always_save_lp condition.
4064 * config/nds32/nds32.opt (malways-save-lp): New option.
4066 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4068 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4069 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4070 * config/nds32/nds32.h
4071 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4072 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4074 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4076 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4077 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4079 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4080 Chung-Ju Wu <jasonwucj@gmail.com>
4082 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4084 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4085 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4087 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4089 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4091 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4093 * config/nds32/nds32-protos.h (nds32_data_alignment,
4094 nds32_local_alignment): Declare.
4095 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4096 nds32_local_alignment): New functions.
4097 (TARGET_CONSTANT_ALIGNMENT): Define.
4098 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4100 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4102 * config/nds32/nds32.c
4103 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4104 (TARGET_MODES_TIEABLE_P): Likewise.
4106 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4108 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4111 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4112 Chung-Ju Wu <jasonwucj@gmail.com>
4114 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4115 for unaligned access.
4116 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4118 * config/nds32/nds32-intrinsic.md: Likewise.
4119 * config/nds32/nds32_intrinsic.h: Likewise.
4120 * config/nds32/nds32.h (nds32_builtins): Likewise.
4121 * config/nds32/nds32.opt (munaligned-access): New option.
4122 * config/nds32/nds32.c (nds32_asm_file_start): Display
4123 flag_unaligned_access status.
4125 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4127 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4128 -mno-relax is present.
4129 * config/riscv/linux.h (LINK_SPEC): Ditto.
4131 2018-04-20 Martin Sebor <msebor@redhat.com>
4134 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4136 (gimple_fold_builtin_stxcpy_chk): Same.
4137 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4139 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4142 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4143 __powikf2 when long double is IEEE 128-bit.
4145 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4147 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4148 step to make sure stack always aligned.
4150 2018-04-20 Carl Love <cel@us.ibm.com>
4153 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4154 size check for arg0.
4156 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4157 Tom de Vries <tom@codesourcery.com>
4160 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4161 Emit insns for calls too.
4162 (nvptx_find_par): Always look for worker-level predecessor insn.
4163 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4165 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4166 (nvptx_process_pars): Propagate frames for calls.
4168 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4171 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4173 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4174 (ix86_handle_option): Don't handle OPT_mibt.
4175 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4177 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4179 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4180 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4181 (ix86_target_macros): Define __CET__ with flag_cf_protection
4182 for -fcf-protection.
4183 * config/i386/i386.c (isa2_opts): Remove -mibt.
4184 * config/i386/i386.h (TARGET_IBT): Removed.
4185 (TARGET_IBT_P): Likewise.
4186 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4187 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4188 * config/i386/i386.opt (mcet): Update help message.
4191 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4192 -mcet as an alias for -mshstk.
4194 2018-04-20 Richard Biener <rguenther@suse.de>
4197 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4198 complexity by forcing a single use of the multiply operand.
4200 2018-04-20 Martin Jambor <mjambor@suse.cz>
4203 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4204 recursion dependency to only apply to non-clones.
4206 2018-04-20 Martin Jambor <mjambor@suse.cz>
4209 * ipa-cp.c (create_specialized_node): Check that clones of
4210 self-recursive edges exist during IPA-CP.
4212 2018-04-19 Toon Moene <toon@moene.org>
4214 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4217 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4219 PR tree-optimization/85467
4220 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4221 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4222 VECTOR_CST element to type.
4224 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4227 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4228 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4229 (builtin_longjmp): Likewise.
4230 (save_stack_nonlocal): New pattern.
4231 (restore_stack_nonlocal): Likewise.
4233 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4236 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4237 Replace ASM_OUTPUT_LABEL with fprintf.
4239 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4242 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4243 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4244 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4245 define __IBT__ and __SHSTK__ for -fcf-protection.
4246 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4248 (ix86_trampoline_init): Likewise.
4249 (x86_output_mi_thunk): Likewise.
4250 (ix86_notrack_prefixed_insn_p): Likewise.
4251 (ix86_option_override_internal): Don't disallow -fcf-protection.
4252 * config/i386/i386.md (rdssp<mode>): Also enable for
4254 (incssp<mode>): Likewise.
4255 (nop_endbr): Likewise.
4256 * config/i386/i386.opt (mcet): Change help message to built-in
4260 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4261 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4262 enable CET built-in functions.
4264 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4266 * common/config/i386/i386-common.c
4267 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4268 OPTION_MASK_ISA_MOVDIRI_UNSET,
4269 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4270 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4271 * config.gcc (movdirintrin.h): New header.
4272 * config/i386/cpuid.h (bit_MOVDIRI,
4273 bit_MOVDIR64B): New bits.
4274 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4276 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4277 (VOID, PVOID, PCVOID)): New function types.
4278 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4279 __builtin_ia32_directstoreu_u64,
4280 __builtin_ia32_movdir64b): New builtins.
4281 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4282 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4284 (ix86_valid_target_attribute_inner_p): Ditto.
4285 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4286 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4287 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4288 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4289 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4290 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4291 (movdiri<mode>, movdir64b_<mode>): New.
4292 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4293 * config/i386/immintrin.h: Include movdirintrin.h.
4294 * config/i386/movdirintrin.h: New file.
4295 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4297 2018-04-19 Richard Biener <rguenther@suse.de>
4300 * cfg.c (clear_bb_flags): When loop state says we have
4301 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4303 2018-04-19 Richard Biener <rguenther@suse.de>
4305 PR tree-optimization/84737
4306 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4307 copying restrict info.
4308 (vect_setup_realignment): Use it.
4309 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4310 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4311 the first DR to all generated stores.
4312 (vectorizable_load): Likewise for loads.
4314 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4316 PR tree-optimization/85446
4317 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4318 the integral and pointer types to have the same precision.
4320 * doc/install.texi: Document --disable-cet being the default and
4323 2018-04-18 Martin Liska <mliska@suse.cz>
4325 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4328 2018-04-18 Martin Liska <mliska@suse.cz>
4331 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4334 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4335 arguments if they are comparable.
4337 2018-04-18 Martin Liska <mliska@suse.cz>
4340 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4343 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4346 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4349 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4350 ENDBR after calling __morestack.
4352 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4355 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4356 by using gcc_base_ver to generate a gcc_driver_version, and use
4357 it when generating GCC_DRIVER_NAME.
4358 * configure: Regenerate.
4360 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4363 * config.gcc: Obsolete powerpc*-*-*spe*.
4365 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4368 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4369 (stabstr_D): Change type of unum from unsigned int to
4370 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4373 2018-04-17 Jim Wilson <jimw@sifive.com>
4376 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4377 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4378 Set arg_pointer_offset after using pretend_args_size.
4380 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4382 PR rtl-optimization/85431
4383 * dse.c (record_store): Ignore zero width stores.
4386 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4387 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4388 __builtin_stack_restore rather than after it.
4389 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4390 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4391 argument instead of virtual_dynamic_stack_rtx.
4393 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4395 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4397 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4398 Add note to error message to explain internal mapping of overloaded
4399 built-in function name to non-overloaded built-in function name.
4400 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4403 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4406 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4407 where the inputs overlap with the output.
4409 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4411 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4412 (=v, v) alternative and explicit "memory" attribute.
4413 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4414 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4416 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4417 "sselog1" type instead of "sselog".
4418 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4419 "sselog". Remove explicit "memory" attribute.
4420 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4421 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4423 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4424 "sselog1" type instead of "sselog". Remove explicit "memory"
4426 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4427 use "sselog1" type instead of "sselog". Remove explicit "memory"
4429 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4430 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4432 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4433 "sselog1" type instead of "sselog". Remove explicit "memory"
4435 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4436 use "sselog1" type instead of "sselog". Remove explicit "memory"
4440 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4443 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4446 2018-04-17 Martin Jambor <mjambor@suse.cz>
4449 * ipa-cp.c (create_specialized_node): Call
4450 expand_all_artificial_thunks if necessary.
4452 2018-04-17 Martin Liska <mliska@suse.cz>
4455 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4456 in message, remote space in between '_G' and '('.
4458 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4461 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4462 avx512f_vmcmp<mode>3<round_saeonly_name>,
4463 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4464 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4465 avx512f_rndscale<mode><round_saeonly_name>,
4466 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4467 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4468 Use %<iptr>2 instead of %2 for -masm=intel.
4469 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4470 avx512f_vcvttss2usi<round_saeonly_name>,
4471 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4473 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4474 avx512f_vcvttsd2usi<round_saeonly_name>,
4475 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4476 Use %q1 instead of %1 for -masm=intel.
4477 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4478 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4479 of %3 for -masm=intel.
4480 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4482 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4484 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4486 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4487 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4489 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4491 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4492 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4493 %g1 and one with %0 and %1.
4494 (avx512er_vmrcp28<mode><round_saeonly_name>,
4495 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4497 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4498 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4499 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4500 of %0 and %{%4%} for -masm=intel.
4501 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4502 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4503 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4504 order of %0 and %{%5%}%{z%} for -masm=intel.
4506 2018-04-17 Jan Hubicka <jh@suse.cz>
4509 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4511 2018-04-17 Martin Liska <mliska@suse.cz>
4514 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4515 for target_clone error message. Make default implementation
4516 clone to be a local declaration.
4517 (separate_attrs): Add new argument and check for an empty
4519 (expand_target_clones): Handle it.
4520 (ipa_target_clone): Make redirection just for target_clones
4523 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4524 Tom de Vries <tom@codesourcery.com>
4527 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4528 tiled basic blocks without omp continue statements.
4530 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4533 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4534 vec_extract expression as having side effects to make sure it gets
4537 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4540 * config/i386/i386.c (get_builtin_code_for_version): Check
4543 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4546 * gcc/config.gcc: Support "skylake".
4547 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4549 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4550 (processor_target_table): Add "skylake".
4551 (ix86_option_override_internal): Add "skylake".
4552 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4553 PROCESSOR_CANNONLAKE.
4554 (get_builtin_code_for_version): Fix priority for
4555 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4556 PROCESSOR_SKYLAKE-AVX512.
4557 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4558 (processor_type): Add PROCESSOR_SKYLAKE.
4560 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4561 Jason Merrill <jason@redhat.com>
4564 * convert.c (convert_to_integer_1): Use direct recursion for
4565 enumeral types and types with a precision less than the number
4566 of bits in their mode.
4568 2018-04-16 Julia Koval <julia.koval@intel.com>
4571 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4572 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4574 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4577 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4578 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4579 and -mno-direct-move.
4581 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4584 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4585 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4586 Ensure negative shifts result in {0}.
4588 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4590 PR rtl-optimization/79916
4591 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4592 regs (if any) to define how to gnerate SD moves when LRA is in
4595 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4597 PR rtl-optimization/85393
4598 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4599 * except.c (expand_dw2_landing_pad_for_region): Make static.
4600 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4601 a label and unconditional jump to old_bb, rather than
4602 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4605 PR rtl-optimization/85376
4606 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4607 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4608 instead of a specific value.
4610 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4611 Bin Cheng <bin.cheng@arm.com>
4613 PR tree-optimization/82965
4614 PR tree-optimization/83991
4615 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4616 by_profile_only parameter.
4617 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4618 information if the loop was predicted to iterate too many times.
4619 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4621 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4624 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4627 2018-04-13 Martin Liska <mliska@suse.cz>
4628 Jakub Jelinek <jakub@redhat.com>
4631 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4632 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4633 * builtins.c (expand_builtin_memory_copy_args): Use
4634 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4635 handle dest_addr == pc_rtx.
4637 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4640 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4641 asked to not generate direct moves.
4642 (fix_trunc<mode>si2_stfiwx): Similar.
4643 (fix_trunc<mode>si2_internal): Similar.
4645 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4648 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4649 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4650 lookup if dest in some wider mode is known to be const0_rtx and
4651 if so, record permanent equivalence for it to be ZERO_EXTEND of
4652 the narrower mode destination.
4654 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4656 * lto-streamer-out.c (output_function): Revert 259346.
4657 * omp-expand.c (expand_oacc_for): Likewise.
4659 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4661 PR rtl-optimization/85354
4662 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4663 * sel-sched.c (sel_global_init): ... here.
4665 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4668 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4669 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4670 mode for PE-COFF targets.
4671 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4672 (i386_pe_asm_lto_end): Likewise.
4673 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4674 (TARGET_ASM_LTO_END): Likewise.
4675 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4676 (i386_pe_asm_lto_start): New function.
4677 (i386_pe_asm_lto_end): Likewise.
4679 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4680 Richard Biener <rguenther@suse.de>
4683 * lto-streamer-out.c (output_function): Fix CFG loop state before
4685 * omp-expand.c (expand_oacc_for): Handle calls to internal
4686 functions like regular functions.
4688 2018-04-12 Richard Biener <rguenther@suse.de>
4691 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4692 for the early LTO debug to properly generate references to it
4693 during DIE emission. Do not re-use that for the skeleton for
4695 (dwarf2out_early_finish): Likewise.
4697 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4700 * config/i386/sse.md
4701 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4702 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4703 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4704 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4705 and output is a reg, avoid creating invalid lowpart subreg, but
4706 instead split into a 512-bit move. Don't split if not AVX512VL,
4707 input is xmm16+ reg and output is a mem.
4708 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4709 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4710 xmm16+ reg and output is a mem.
4712 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4714 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4715 also for flag_dwarf2_cfi_asm.
4717 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4719 PR rtl-optimization/85342
4720 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4721 a bool scalar var inside of the loop instead. Don't try to update
4722 recog_data.operand after failed apply_change_group.
4724 2018-04-12 Tom de Vries <tom@codesourcery.com>
4727 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4728 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4729 array with flexible array member as array without given dimension.
4730 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4731 argument for undefined param to true.
4733 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4736 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4737 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4738 from PowerPC section.
4739 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4740 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4741 help text that is too long.
4742 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4743 help text that is too long.
4744 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4745 help text that is too long.
4747 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4749 * config/alpha/alpha.md (stack_probe_internal): Rename
4750 from "probe_stack". Update all callers.
4752 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4754 PR rtl-optimization/84566
4755 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4756 sched_macro_fuse_insns.
4758 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4761 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4762 (compute_block_dependences): ... from here.
4764 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4766 PR tree-optimization/85331
4767 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4768 from int to HOST_WIDE_INT.
4770 2018-04-11 Martin Jambor <mjambor@suse.cz>
4773 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4774 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4775 not the same as the source val.
4776 (cgraph_edge_brings_value_p): New parameter.
4777 (gather_edges_for_value): Pass destination value to
4778 cgraph_edge_brings_value_p.
4779 (perhaps_add_new_callers): Likewise.
4780 (get_info_about_necessary_edges): Likewise and exclude values brought
4781 only by self-recursive edges.
4782 (create_specialized_node): Redirect only clones of self-calling edges.
4783 (+self_recursive_pass_through_p): New function.
4784 (find_more_scalar_values_for_callers_subset): Use it.
4785 (find_aggregate_values_for_callers_subset): Likewise.
4786 (known_aggs_to_agg_replacement_list): Removed.
4787 (decide_whether_version_node): Re-calculate known constants for all
4788 remaining context clones.
4790 2018-04-11 Richard Biener <rguenther@suse.de>
4793 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4794 from early DWARF output.
4795 (dwarf2out_early_finish): Output line info unconditionally into
4796 early DWARF and add reference to it.
4798 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4801 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4802 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4803 other than V2DFmode using iptr mode attribute.
4804 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4806 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4808 PR rtl-optimization/84659
4809 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4811 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4814 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4816 (output_loc_list): Pass address of a dummy size variable even in the
4817 locview handling loop.
4818 (index_location_lists): Add comment on why skip_loc_list_entry can't
4821 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4824 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4827 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4830 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4831 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4832 and -mstring-compare-inline-limit.
4834 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4837 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4838 for stack clash protection in a register whenever we need it to be in
4841 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4843 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4844 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4846 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4849 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4851 (mlong-double-): Ditto.
4852 * config/rs6000/sysv4.opt (msdata=): Ditto.
4853 (mtls-size=): Ditto.
4855 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4857 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4858 erroneous entries for
4859 "vector int vec_ldl (int, long int *)", and
4860 "vector unsigned int vec_ldl (int, unsigned long int *)".
4861 Add comments and entries for
4862 "vector bool char vec_ldl (int, bool char *)",
4863 "vector bool short vec_ldl (int, bool short *)",
4864 "vector bool int vec_ldl (int, bool int *)",
4865 "vector bool long long vec_ldl (int, bool long long *)",
4866 "vector pixel vec_ldl (int, pixel *)",
4867 "vector long long vec_ldl (int, long long *)",
4868 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4869 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4870 type tree bool_long_long_type_node and correct definition of
4871 bool_V2DI_type_node to make reference to this new type tree.
4872 (rs6000_mangle_type): Replace erroneous reference to
4873 bool_long_type_node with bool_long_long_type_node.
4874 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4875 comments to emphasize sign distinctions for char and int types and
4876 replace RS6000_BTI_bool_long constant with
4877 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4878 use of RS6000_BTI_pixel.
4879 (bool_long_type_node): Remove this macro definition.
4880 (bool_long_long_type_node): New macro definition
4882 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4884 PR rtl-optimization/85300
4885 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4886 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4887 simplify_unary_operation fails.
4889 2018-04-10 Martin Liska <mliska@suse.cz>
4891 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4892 cgraph_edge and ipa_ref.
4894 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4898 * config/i386/sse.md
4899 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4900 computation of the VEC_MERGE selector from mask.
4901 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4902 Fix decoding of the VEC_MERGE selector into mask.
4904 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4906 PR tree-optimization/85286
4907 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4909 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4911 * final.c (final_1): Set insn_last_address as well as
4912 insn_current_address.
4914 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4917 * explow.c (emit_stack_probe): Call validize_mem on memory location
4918 before passing it to gen_probe_stack. Create address operand and
4919 legitimize it for the probe_stack_address case.
4921 2018-04-09 Jan Hubicka <jh@suse.cz>
4924 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4925 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4926 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4928 (free_lang_data): Rebuild type inheritance graph.
4930 2018-04-09 Martin Sebor <msebor@redhat.com>
4932 * invoke.texi (-finline-small-functions): Mention other optimization
4934 (-findirect-inlining, -fpartial-inlining): Same.
4935 (-finline-functions-called-once): Same.
4936 (-freorder-blocks-and-partition): Same.
4938 2018-04-09 Jan Hubicka <jh@suse.cz>
4941 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4942 jumps; choose last target that matches the criteria (i.e.
4943 no partition changes for non-crossing jumps).
4944 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4945 support for redirecting crossing jumps to non-crossing.
4947 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4949 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4950 also for naked functions.
4952 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4954 * config/arc/arc.md (add_shift): New pattern.
4955 (add_shift2): Likewise.
4956 (sub_shift): Likewise.
4957 (sub_shift_cmp0_noout): Likewise.
4958 (compare_si_ashiftsi): Likewise.
4959 (xbfu_cmp0_noout): New combine pattern.
4960 (xbfu_cmp0"): Likewise.
4961 (movsi_set_cc_insn): Place the predicable variant first.
4962 (commutative_binary_cmp0_noout): Remove clobber.
4963 (commutative_binary_cmp0): New pattern.
4964 (noncommutative_binary_cmp0): Likewise.
4965 (noncommutative_binary_cmp0_noout): Likewise.
4966 (noncommutative_binary_comparison_result_used): Removed.
4967 (rsub_cmp0): New pattern.
4968 (rsub_cmp0_noout): Likewise.
4969 (extzvsi): Changed, keep only meaningful variants.
4970 (SQH, SEZ): New iterators.
4971 (SQH_postfix): New mode attribute.
4972 (SEZ_prefix): New code attribute.
4973 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4974 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4975 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4977 (noncommutative_operator): Check the availability of barrel
4980 2018-04-09 Richard Biener <rguenther@suse.de>
4982 PR tree-optimization/85284
4983 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4984 Only use the niter constraining form of simple_iv when the exit
4987 2018-04-09 Tom de Vries <tom@codesourcery.com>
4990 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4991 (define_expand "*memory_barrier"): New define_expand.
4992 (define_insn "memory_barrier"): New insn.
4994 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4996 PR rtl-optimization/80463
4997 PR rtl-optimization/83972
4998 PR rtl-optimization/83480
5000 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
5001 correct producer for the insn.
5002 (tidy_control_flow): Fixup seqnos in case of debug insns.
5004 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5006 PR rtl-optimization/83913
5008 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
5009 different sched-times when merging exprs.
5011 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5013 PR rtl-optimization/83962
5015 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
5016 tidy_fallthru_edge and tidy_control_flow.
5018 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
5020 PR rtl-optimization/83530
5022 * sel-sched.c (force_next_insn): New global variable.
5023 (remove_insn_for_debug): When force_next_insn is true, also leave only
5024 next insn in the ready list.
5025 (sel_sched_region): When the region wasn't scheduled, make another pass
5026 over it with force_next_insn set to 1.
5028 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5030 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
5032 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5033 for interrupt control.
5034 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5035 functions for interrupt control.
5036 * config/nds32/nds32-intrinsic.md: Likewise.
5037 * config/nds32/nds32_intrinsic.h: Likewise.
5038 * config/nds32/nds32.h (nds32_builtins): Likewise.
5040 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5042 * config/nds32/nds32.c (nds32_init_machine_status,
5043 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5044 strict_aligned_p field.
5045 (nds32_expand_to_rtl_hook): New function.
5046 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5047 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5049 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5050 Chung-Ju Wu <jasonwucj@gmail.com>
5052 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5053 * config/nds32/nds32-n7.md: New file.
5054 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5055 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5057 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5058 * config/nds32/nds32.md (pipeline_model): Add n7.
5059 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5060 * config/nds32/pipelines.md: Include n7 settings.
5062 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5063 Chung-Ju Wu <jasonwucj@gmail.com>
5065 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5066 * config/nds32/nds32-e8.md: New file.
5067 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5068 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5070 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5071 * config/nds32/nds32.md (pipeline_model): Add e8.
5072 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5073 * config/nds32/pipelines.md: Include e8 settings.
5075 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5076 Chung-Ju Wu <jasonwucj@gmail.com>
5078 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5079 * config/nds32/nds32-n8.md: New file.
5080 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5081 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5083 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5084 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5085 * config/nds32/nds32.md (pipeline_model): Add n8.
5086 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5087 * config/nds32/pipelines.md: Include n8 settings.
5089 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5090 Chung-Ju Wu <jasonwucj@gmail.com>
5092 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5093 * config/nds32/nds32-n9-2r1w.md: New file.
5094 * config/nds32/nds32-n9-3r2w.md: New file.
5095 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5096 nds32_register_ports): New or modify for cpu n9.
5097 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5099 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5100 * config/nds32/nds32-utils.c: New file.
5101 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5102 TARGET_MUL_SLOW): Define.
5103 * config/nds32/nds32.md (pipeline_model): New attribute.
5104 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5105 New options that support cpu n9.
5106 * config/nds32/pipelines.md: Include n9 settings.
5107 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5109 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5111 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5112 information if necessary.
5113 (output_cond_branch_compare_zero): Likewise.
5114 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5115 (nds32_target_alignment): Refine for alignment.
5116 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5117 (FUNCTION_BOUNDARY): Modify.
5118 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5120 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5122 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5124 * config/nds32/constants.md (unspec_volatile_element): Add values for
5125 TLB operation and data prefetch.
5126 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5127 functions for TLB operation and data prefetch.
5128 * config/nds32/nds32-intrinsic.md: Likewise.
5129 * config/nds32/nds32_intrinsic.h: Likewise.
5130 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5131 (nds32_print_operand): Likewise.
5132 * config/nds32/nds32.h (nds32_builtins): Likewise.
5134 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5135 Andrew Pinski <pinsika@gcc.gnu.org>
5138 * match.pd: Use constant_boolean_node of correct type instead of
5139 boolean_true_node or boolean_false_node for simplifying
5140 pointer comparisons to zero.
5142 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5144 PR tree-optimization/80021
5145 * tree.c (verify_type_variant): Make error call in verify_variant_match
5146 translatable and remove final full stop.
5148 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/constants.md (unspec_volatile_element): Add
5151 UNSPEC_VOLATILE_EH_RETURN.
5152 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5153 nds32_output_stack_pop): Support dwarf exception handling process.
5154 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5155 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5156 exception handling process.
5157 (nds32_compute_stack_frame): Likewise.
5158 (nds32_return_addr_rtx): Likewise.
5159 (nds32_initial_elimination_offset): Likewise.
5160 (nds32_expand_prologue): Likewise.
5161 (nds32_expand_epilogue): Likewise.
5162 (nds32_dynamic_chain_address): New function.
5163 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5165 (DYNAMIC_CHAIN_ADDRESS): Define.
5166 (EH_RETURN_DATA_REGNO): Define.
5167 (EH_RETURN_STACKADJ_RTX): Define.
5168 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5169 patterns for dwarf exception handling.
5171 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5173 * config/nds32/nds32.h: Clean up obsolete macros.
5175 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5177 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5178 Add enum values for particular instructions.
5179 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5180 particular intrinsic functions.
5181 * config/nds32/nds32-intrinsic.md: Likewise.
5182 * config/nds32/nds32_intrinsic.h: Likewise.
5183 * config/nds32/nds32.h (nds32_builtins): Likewise.
5184 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5185 (btst, ave): New patterns for particular instructions.
5187 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5189 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5190 Add enum values for atomic load/store and memory sync.
5191 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5193 * config/nds32/nds32-intrinsic.md: Likewise.
5194 * config/nds32/nds32_intrinsic.h: Likewise.
5195 * config/nds32/nds32.h (nds32_builtins): Likewise.
5197 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5199 PR tree-optimization/85257
5200 * fold-const.c (native_encode_vector): If not all elts could fit
5201 and off is -1, return 0 rather than offset.
5202 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5203 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5204 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5205 adjust buffer in native_interpret_expr call.
5207 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5209 * config/nds32/constants.md (unspec_volatile_element): Add cache
5210 control enum values.
5211 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5212 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5213 * config/nds32/nds32.c (nds32_cctl_names): New.
5214 (nds32_print_operand): Handle cache control register names.
5215 * config/nds32/nds32.h (nds32_builtins): New enum values.
5216 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5218 * config/nds32/nds32.md (type): Add mmu.
5219 * config/nds32/pipelines.md (simple_insn): Add mmu.
5221 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5223 * config/nds32/nds32.md (type): Remove call.
5224 * config/nds32/pipelines.md (simple_insn): Likewise.
5226 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5228 * config/nds32/constants.md (unspec_volatile_element): Add
5229 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5230 UNSPEC_VOLATILE_FMFCFG.
5231 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5232 description for fmfcfg and fmfcsr.
5233 (bdesc_1arg): Add fmtcsr.
5234 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5235 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5236 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5237 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5238 unspec_fmfcfg): New patterns.
5239 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5240 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5241 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5242 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5243 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5244 __nds32__fmfcfg): Define.
5246 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5248 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5249 intrinsic register names.
5250 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5251 intrinsic register enum values and macros.
5253 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5255 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5256 for load/store addressing form.
5257 (nds32_print_operand_address): Likewise.
5259 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5262 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5263 based on LABEL_REF. Remove useless assertion.
5264 (pic_address_needs_scratch): Fix formatting.
5265 (sparc_legitimize_pic_address): Minor tweaks.
5266 (sparc_delegitimize_address): Adjust assertion accordingly.
5267 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5268 into symbolic_operand.
5269 (movsi_high_pic_label_ref): Likewise.
5270 (movsi_lo_sum_pic_label_ref): Likewise.
5271 (movdi_pic_label_ref): Likewise.
5272 (movdi_high_pic_label_ref): Likewise.
5273 (movdi_lo_sum_pic_label_ref): Likewise.
5275 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5277 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5278 custom LIB_SPEC setup.
5280 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5281 Kito Cheng <kito.cheng@gmail.com>
5283 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5284 * config/riscv/freebsd.h: New.
5286 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5288 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5289 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5292 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5293 Kito Cheng <kito.cheng@gmail.com>
5295 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5296 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5297 * config/nds32/nds32-protos.h (nds32_output_call,
5298 nds32_output_return): Declare.
5299 * config/nds32/nds32.md: Refine all the call and return patterns.
5301 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5304 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5305 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5307 PR rtl-optimization/84872
5308 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5309 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5312 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5314 * expr.c (copy_blkmode_to_reg): Revert 254862.
5315 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5317 2018-04-06 Richard Biener <rguenther@suse.de>
5320 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5321 after seeing a component reference with an adjacent field. Treat
5322 refs to arrays at struct end of external decls similar to
5323 refs to unconstrained commons.
5325 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5328 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5329 look through SAVE_EXPRs with non-side-effects argument. Adjust
5331 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5334 2018-04-06 Richard Biener <rguenther@suse.de>
5337 * alias.c (find_base_term): New wrapper around find_base_term
5338 unwinding CSELIB_VAL_PTR changes.
5339 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5342 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5344 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5346 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5347 constant definitions.
5348 ("nop"): lr 0,0 -> nopr r0
5349 ("nop_lr0", "nop_lr1"): New insn definitions.
5351 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5353 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5354 NDS32_V3PUSH_AVAILABLE_P macro.
5356 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5357 Chung-Ju Wu <jasonwucj@gmail.com>
5359 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5360 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5361 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5362 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5363 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5364 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5365 * config/nds32/constraints.md: New constraints and checking for hard
5366 float configuration.
5367 * config/nds32/iterators.md: New mode iterator and attribute for hard
5368 float configuration.
5369 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5371 * config/nds32/nds32-fpu.md: New file.
5372 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5373 deal with hard float code generation.
5374 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5376 (abi_type, float_reg_number): New enum type.
5377 * config/nds32/nds32-predicates.c: New predicates for hard float.
5378 * config/nds32/nds32-protos.h: Declare functions for hard float.
5379 * config/nds32/nds32.c: Implementation for hard float configuration.
5380 * config/nds32/nds32.h: Definitions for hard float configuration.
5381 * config/nds32/nds32.md: Include hard float machine description and
5382 modify patterns for hard float configuration.
5383 * config/nds32/nds32.opt: New options for hard float configuration.
5384 * config/nds32/predicates.md: New predicates for hard float
5387 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5389 * common/config/nds32/nds32-common.c
5390 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5392 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5395 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5396 CONSTRUCTOR_ELT (ctor, ...)->value.
5398 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5401 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5403 2018-04-05 Tom de Vries <tom@codesourcery.com>
5406 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5409 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5410 Kito Cheng <kito.cheng@gmail.com>
5412 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5413 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5414 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5415 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5417 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5418 Kito Cheng <kito.cheng@gmail.com>
5420 * config/nds32/constraints.md (Ufe): New memory constraint.
5421 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5422 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5423 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5425 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5426 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5428 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5430 * config/nds32/nds32.md: Use optimize_size in the condition for
5431 alu-shift instructions.
5433 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5435 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5437 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5439 * config/nds32/nds32.md (negsi2): Refine pattern.
5441 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5442 Chung-Ju Wu <jasonwucj@gmail.com>
5444 * config/nds32/iterators.md (shift_rotate): New code iterator.
5445 (shift): New code attribute.
5446 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5447 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5448 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5449 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5450 bit-wise operations.
5451 (andsi3, *andsi3): Ditto.
5452 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5453 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5454 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5455 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5456 nds32_ior_operand, nds32_xor_operand): New predicates.
5458 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5460 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5461 (addsi3, subsi3): ... this.
5463 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5465 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5467 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5469 * config/nds32/nds32.md: Adjust indention.
5471 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5473 * config/nds32/nds32.md (feature): New attribute.
5475 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5477 * config/nds32/nds32.md (subtype): New attribute.
5479 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5482 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5483 expansion to perform a bitwise AND of the argument followed by a
5484 boolean negation of the result.
5486 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5488 PR rtl-optimization/84878
5489 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5490 the basic block. Assert the use reference is not artificial and that
5491 it has an associated insn.
5493 2018-04-04 Michael Matz <matz@suse.de>
5495 * builtins.c (compute_objsize): Pass correct operand
5496 to array_at_struct_end_p.
5498 2018-04-04 Richard Biener <rguenther@suse.de>
5501 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5502 from contexts for DINFO_LEVEL_TERSE and below.
5504 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5506 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5507 resiter_operand condition.
5508 * config/nds32/nds32.md (*move<mode>): Ditto.
5510 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5511 Monk Chiang <sh.chiang04@gmail.com>
5513 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5515 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5517 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5519 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5520 Kito Cheng <kito.cheng@gmail.com>
5522 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5523 nds32_cond_code_str, output_cond_branch,
5524 output_cond_branch_compare_zero, nds32_expand_cbranch,
5525 nds32_expand_cstore, nds32_expand_movcc,
5526 nds32_output_cbranchsi4_equality_zero,
5527 nds32_output_cbranchsi4_equality_reg,
5528 nds32_output_cbranchsi4_equality_reg_or_const_int,
5529 nds32_output_cbranchsi4_greater_less_zero: New functions.
5530 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5531 nds32_expand_cstore, nds32_expand_movcc,
5532 nds32_output_cbranchsi4_equality_zero,
5533 nds32_output_cbranchsi4_equality_reg,
5534 nds32_output_cbranchsi4_equality_reg_or_const_int,
5535 nds32_output_cbranchsi4_greater_less_zero): Declare.
5536 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5537 nds32_rimm11s_operand): New predicates.
5538 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5539 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5542 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5544 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5545 * config/nds32/nds32.md: Ditto.
5546 * config/nds32/pipelines.md: Ditto.
5548 2018-04-04 Richard Biener <rguenther@suse.de>
5550 PR tree-optimization/85168
5551 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5552 propagating abnormals.
5554 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5556 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5558 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5559 Kito Cheng <kito.cheng@gmail.com>
5561 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5562 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5563 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5564 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5565 * config/nds32/nds32.md (sibcall_internal): New.
5566 (sibcall_register): Remove.
5567 (sibcall_immediate): Remove.
5568 (sibcall_value_internal): New.
5569 (sibcall_value_register): Remove.
5570 (sibcall_value_immediate): Remove.
5571 * config/nds32/predicates.md (nds32_general_register_operand): New.
5572 (nds32_call_address_operand): New.
5574 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5576 PR rtl-optimization/85167
5577 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5578 bb_defs if *split_p, instead preinitialize it to NULL.
5580 PR tree-optimization/85156
5581 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5582 evaluating the argument multiple times.
5584 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5586 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5588 (_mm_cvtpd_ps): Likewise.
5589 (_mm_cvttpd_epi32): Likewise.
5590 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5591 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5592 vector, pixel, and bool following altivec.h include.
5594 2018-04-03 Martin Sebor <msebor@redhat.com>
5596 * doc/extend.texi (Common Function Attributes): Clarify.
5597 (const attribute): Likewise.
5598 (pure attribute): Likewise.
5600 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5603 * config/i386/i386.c (ix86_expand_vector_set): Use
5604 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5606 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5608 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5609 instructions when changing rounding bits to preserve precision bits
5610 in the x87 control word.
5612 2018-04-03 Martin Liska <mliska@suse.cz>
5614 PR tree-optimization/82491
5615 * rtl.h (strip_offset_and_add): Replace += suboffset with
5616 poly_uint64 () + suboffset.
5618 2018-03-29 Martin Liska <mliska@suse.cz>
5619 Martin Jambor <mjambor@suse.cz>
5622 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5623 param_type is not an integral or pointer type.
5625 2018-04-03 Richard Biener <rguenther@suse.de>
5627 * sese.h (recompute_all_dominators): Remove.
5629 2018-04-02 Martin Sebor <msebor@redhat.com>
5631 * doc/invoke.texi (-Wrestrict): Fix typos.
5633 2018-04-02 Jim Wilson <jimw@sifive.com>
5635 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5636 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5637 (<optab>di3, <optab>si3_extend): Likewise.
5638 (<optab>si3_mask, <optab>si3_mask_1): New.
5639 (<optab>di3_mask, <optab>di3_mask_1): New.
5640 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5641 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5642 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5644 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5646 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5649 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5651 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5652 (nds32_canonicalize_comparison): New function.
5654 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5655 Kito Cheng <kito.cheng@gmail.com>
5656 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5658 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5659 * config/nds32/constants.md (unspec_volatile_element): Add
5660 UNSPEC_VOLATILE_RELAX_GROUP.
5661 * config/nds32/nds32-relax-opt.c: New file.
5662 * config/nds32/nds32-predicates.c
5663 (nds32_symbol_load_store_p): New function.
5664 * config/nds32/nds32-protos.h
5665 (nds32_symbol_load_store_p): Declare function.
5666 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5667 * config/nds32/nds32.c
5668 (nds32_register_pass): New function to register pass.
5669 (nds32_register_passes): New function to register passes.
5670 * config/nds32/nds32.md (relax_group): New pattern.
5671 * config/nds32/nds32.opt (mrelax-hint): New option.
5672 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5674 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5676 * config/nds32/t-nds32: Modify files dependency.
5678 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5680 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5681 (PROFILE_HOOK): Define its implementation.
5683 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5685 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5686 type and 32-bit size.
5688 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5691 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5692 (V_128_256): New mode iterator.
5693 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5694 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5695 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5697 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5698 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5700 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5703 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5704 NaN inputs correctly.
5706 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5709 * config/rs6000/vsx.md (??r): New mode attribute.
5710 (*vsx_mov<mode>_64bit): Use it.
5711 (*vsx_mov<mode>_32bit): Likewise.
5713 2018-03-30 Martin Sebor <msebor@redhat.com>
5715 PR tree-optimization/84818
5716 * builtins.c (check_access): Use warning_n.
5718 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5721 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5723 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5726 2018-03-30 Julia Koval <julia.koval@intel.com>
5729 * x86-tune.def (movx, partial_reg_dependency): Enable for
5732 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5735 * lra-constraints.c (process_alt_operands): Move setting
5736 this_alternative_matches below.
5738 2018-03-29 Martin Liska <mliska@suse.cz>
5741 * doc/invoke.texi: Document how LTO works with debug info.
5742 Describe auto-load support of binutils. Mention 'x86-64'
5743 as valid option value of -march option.
5745 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5747 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5750 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5751 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5752 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5755 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5758 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5759 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5760 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5761 for RS6000_BTM_POWERPC64.
5762 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5763 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5764 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5769 2018-03-28 Carl Love <cel@us.ibm.com>
5772 2017-09-27 Carl Love <cel@us.ibm.com>
5774 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5775 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5776 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5779 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5781 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5782 instead of __vector bool.
5783 (_mm_max_pu8): Likewise.
5784 (_mm_min_pi16): Likewise.
5786 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5789 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5790 (DIVWEUO): Likewise.
5792 (DIVDEUO): Likewise.
5793 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5794 DIVWEUO and DIVDEUO.
5795 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5796 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5797 (div_extend): Likewise.
5798 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5800 (__builtin_divweuo): Likewise.
5801 (__builtin_divdeo): Likewise.
5802 (__builtin_divdeuo): Likewise.
5804 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5807 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5808 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5810 PR tree-optimization/82004
5811 * gimple-match-head.c (optimize_pow_to_exp): New function.
5812 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5813 Don't fold to exp if optimize_pow_to_exp is false.
5815 2018-03-28 Martin Liska <mliska@suse.cz>
5818 * calls.c (initialize_argument_information): Fix trailing space.
5819 * common.opt: Fix typo and provide better explanation for
5820 -fsanitize-coverage option.
5821 * config/i386/i386.opt: Fix typo.
5823 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5824 Martin Liska <mliska@suse.cz>
5827 * gimplify.c (asan_poison_variable): Don't do the check for
5828 gimplify_omp_ctxp here.
5829 (gimplify_decl_expr): Do it here.
5830 (gimplify_target_expr): Likewise.
5832 2018-03-28 Martin Liska <mliska@suse.cz>
5835 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5836 chkp_type_bounds_count if MPX is not enabled.
5838 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5840 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5842 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5845 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5846 function to create the function decl for complex long double
5847 multiply and divide for -mabi=ieeelongdouble.
5848 (init_float128_ieee): Call it.
5850 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5853 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5854 -fcf-protection=branch -mibt.
5855 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5857 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5860 * config/arm/arm.c (arm_valid_symbolic_address): Handle
5861 arm_word_relocations.
5863 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5866 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5867 extern array declarations.
5869 2018-03-27 Richard Biener <rguenther@suse.de>
5872 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5873 explicit single_use checks.
5875 2018-03-27 Richard Biener <rguenther@suse.de>
5877 PR tree-optimization/85082
5878 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5881 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5883 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5884 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5885 Turn on fasynchronous-unwind-tables and funwind-tables.
5887 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5890 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5891 (*bmi_blsr_<mode>_ccz): Ditto.
5893 2018-03-26 Tom de Vries <tom@codesourcery.com>
5895 PR tree-optimization/85063
5896 * omp-general.c (offloading_function_p): New function. Factor out
5898 * omp-offload.c (pass_omp_target_link::gate): ... here.
5899 * omp-general.h (offloading_function_p): Declare.
5900 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5901 with attribute omp declare target for offloading functions.
5903 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5905 PR tree-optimization/84005
5906 * tree-data-ref.h (get_base_for_alignment): Declare.
5907 * tree-data-ref.c (get_base_for_alignment_1): New function.
5908 (get_base_for_alignment): Likewise.
5909 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5910 get_base_for_alignment to find a suitable base object, instead
5911 of always using drb->base_address.
5913 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5916 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5917 known size by default.
5919 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5922 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5923 and non BLKmode operands.
5925 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5928 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5929 Clean up attributes.
5931 2018-03-23 Richard Biener <rguenther@suse.de>
5934 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5935 we are going to emit early debug for LTO.
5937 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5940 * function.c (match_asm_constraints_1): Don't optimize if input
5941 doesn't satisfy general_operand predicate for output's mode.
5944 * alias.c (write_dependence_p): Don't require for x_canonicalized
5945 non-VOIDmode if x has VOIDmode.
5948 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5949 just don't try to optimize it rather than assert it never happens.
5951 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5953 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5954 macro expansions for definition of ST_INTERNAL_<mode> and
5955 LD_INTERNAL_<mode> builtins.
5956 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5958 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5960 (altivec_expand_st_builtin): Likewise.
5961 (altivec_expand_builtin): Remove calls to deleted functions.
5962 (rs6000_address_for_altivec): Delete this function.
5963 * config/rs6000/vector.md: Remove expands for
5964 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5966 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5969 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5970 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5971 re-computing once computed.
5972 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5973 (arm_init_machine_status): Initialize
5974 machine->static_chain_stack_bytes.
5976 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5979 * doc/extend.texi: Add four new prototypes for vec_ld.
5980 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5981 definitions for more logical presentation.
5982 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5983 entries for V1TI variants of __builtin_altivec_ld builtin.
5984 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5985 handling of V1TI variant of LVX icode pattern.
5986 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5987 (rs6000_gimple_fold_builtin): Likewise.
5988 (altivec_init_builtins): Add code to define
5989 __builtin_altivec_lvx_v1ti function.
5991 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5994 * function.c (match_asm_constraints_1): Don't do the optimization
5995 if input isn't a REG, SUBREG, MEM or constant.
5997 2018-03-22 Tom de Vries <tom@codesourcery.com>
5999 PR tree-optimization/84956
6000 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
6001 bb_has_abnormal_pred.
6003 2018-03-22 Jakub Jelinek <jakub@redhat.com>
6006 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
6007 DECL_INITIAL (decl) to decl at the end.
6008 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
6011 2018-03-21 Joseph Myers <joseph@codesourcery.com>
6013 * doc/extend.texi (__builtin_tgmath): Document when complex
6014 integer types are treated as _Complex _Float64.
6016 2018-03-21 Tom de Vries <tom@codesourcery.com>
6018 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
6020 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6022 PR tree-optimization/84960
6023 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
6024 if it is ENTRY block, move them into single succ of ENTRY in that case.
6026 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
6028 PR tree-optimization/84811
6029 * poly-int.h (poly_span_traits): Remove the T3 parameter and
6030 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
6031 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
6032 (known_subrange_p): Update accordingly. Cast each value involved
6033 in the size comparison, rather than casting the result of the
6036 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6038 PR tree-optimization/84982
6039 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6040 by flipping the least significant bit rather than all bits from
6041 bitpos to bitpos + bitsize - 1.
6043 2018-03-21 Nathan Sidwell <nathan@acm.org>
6045 * doc/extend.texi (Deprecated Features): Remove mention of
6046 long-deleted deprecations.
6048 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6051 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6052 * configure: Regenerate.
6054 2018-03-21 Tom de Vries <tom@codesourcery.com>
6056 PR tree-optimization/83126
6057 * tree-parloops.c (num_phis): New function.
6058 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6060 2018-03-21 Nathan Sidwell <nathan@acm.org>
6062 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6063 mention anon-struct/union members and trailing attributes.
6065 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6067 PR tree-optimization/84969
6068 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6069 builtin memset partitions if they set different rhs values.
6071 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6073 PR rtl-optimization/84989
6074 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6075 VEC_DUPLICATE with scalar result mode.
6077 2018-03-21 Martin Liska <mliska@suse.cz>
6080 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6081 not intended return statement.
6083 2018-03-21 Martin Liska <mliska@suse.cz>
6086 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6087 (chkp_find_bound_slots_1): Limit number of iterations.
6089 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6092 * Minor grammar fixes for x86 options.
6094 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6097 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6098 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6100 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6103 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6104 (altivec_lvx_<mode>_1op): Likewise.
6105 (altivec_stvx_<mode>_2op): Likewise.
6106 (altivec_stvx_<mode>_1op): Likewise.
6107 (altivec_lvx_<VM2:mode>): New define_expand.
6108 (altivec_stvx_<VM2:mode>): Likewise.
6109 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6110 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6111 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6112 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6113 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6114 (rs6000_gen_lvx): Likewise.
6115 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6116 (altivec_expand_stv_builtin): Likewise.
6117 (altivec_expand_builtin): Likewise.
6118 * config/rs6000/vector.md: Likewise.
6120 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6123 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6126 2018-03-20 Richard Biener <rguenther@suse.de>
6129 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6130 sign-conversions as zero, fall back to standard scalar_stmt
6133 2018-03-20 Martin Liska <mliska@suse.cz>
6136 * predict.c (rebuild_frequencies): Handle case when we have
6137 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6139 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6142 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6143 flag_section_anchors.
6144 * varasm.c (use_blocks_for_decl_p): Remove hack for
6145 dw2_force_const_mem.
6148 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6150 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6151 be created, use lowpart_subreg of operands[0] rather than operands[0]
6153 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6154 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6155 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6156 and n constraint instead of aarch64_shift_imm_di and Usd.
6157 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6158 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6160 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6163 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6164 to favor GPR over NEON registers.
6165 (<shift>di3_neon): Likewise.
6167 2018-03-20 Tom de Vries <tom@codesourcery.com>
6170 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6171 (nvptx_process_pars): Emit bar.sync asap and alap.
6173 2018-03-20 Tom de Vries <tom@codesourcery.com>
6176 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6177 seen_label if seen_label is already set.
6179 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6182 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6183 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6184 Use 1U instead of 1. Formatting fixes.
6187 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6188 instead of TREE_TYPE (s1) for the return value.
6190 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6192 PR tree-optimization/84946
6193 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6194 bitsize + bitsize in poly_uint64 rather than poly_int64.
6197 * dwarf2asm.c: Include fold-const.c.
6198 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6199 of decl rather than decl itself.
6201 PR rtl-optimization/84643
6202 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6204 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6207 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6208 calling assemble_variable.
6210 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6213 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6214 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6216 2018-03-19 Jim Wilson <jimw@sifive.com>
6219 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6220 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6221 (riscv_first_stack_step): Likewise.
6222 (riscv_option_override): Use STACK_BOUNDARY instead of
6224 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6226 (BIGGEST_ALIGNMENT): Set to 128.
6227 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6228 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6231 2018-03-19 Richard Biener <rguenther@suse.de>
6233 PR tree-optimization/84933
6234 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6235 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6237 2018-03-19 Richard Biener <rguenther@suse.de>
6239 PR tree-optimization/84859
6240 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6241 (cond_if_else_store_replacement): Perform sinking operation on
6242 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6243 Generalize what a BB with a single eligible store is.
6245 2018-03-19 Richard Biener <rguenther@suse.de>
6247 PR tree-optimization/84929
6248 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6249 chrec_is_positive against non-chrec arg.
6251 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6254 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6256 2018-03-18 Martin Liska <mliska@suse.cz>
6258 PR rtl-optimization/84635
6259 * regrename.c (build_def_use): Use matches_mode only when
6262 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6264 PR tree-optimization/84913
6265 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6266 vectorize chains of COND_EXPRs.
6268 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6270 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6272 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6274 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6276 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6278 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6280 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6281 Kito Cheng <kito.cheng@gmail.com>
6283 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6284 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6285 (nds32_adjust_reg_alloc_order): New function.
6286 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6288 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6290 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6291 nds32_print_operand, nds32_print_operand_address): Use
6292 HOST_WIDE_INT_PRINT_DEC instead.
6294 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6296 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6298 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6301 * config/i386/i386.c (initial_ix86_tune_features,
6302 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6304 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6305 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6306 rather than 1u << ix86_tune. Formatting fix.
6307 (ix86_option_override_internal): Change ix86_arch_mask from
6308 unsigned int to unsigned HOST_WIDE_INT, initialize to
6309 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6310 (ix86_function_specific_restore): Likewise.
6312 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6315 * postreload.c (reload_combine_recognize_pattern): Perform
6316 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6317 truncate_int_for_mode the result for the destination's mode.
6320 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6321 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6323 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6326 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6329 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6330 regno_allocno_class_array and sorted_pseudos.
6331 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6332 insns where regno is used.
6334 2018-03-16 Martin Liska <mliska@suse.cz>
6337 * multiple_target.c (create_dispatcher_calls): Redirect
6338 reference in the symbol table.
6340 2018-03-16 Martin Liska <mliska@suse.cz>
6343 * multiple_target.c (create_dispatcher_calls): Redirect also
6346 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6350 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6351 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6352 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6354 2018-03-16 Julia Koval <julia.koval@intel.com>
6356 * doc/invoke.texi (Skylake Server): Add CLWB.
6357 Cannonlake): Remove CLWB.
6359 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6361 PR tree-optimization/84841
6362 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6364 (FLOAT_ONE_CONST_TYPE): Define.
6365 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6366 (sort_by_operand_rank): Put entries with higher constant_type last
6367 rather than first to match comments.
6369 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6371 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6374 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6377 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6378 check and use address of target always.
6380 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6383 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6384 (indirect_thunk_bnd_needed): Likewise.
6385 (indirect_thunks_used): Likewise.
6386 (indirect_thunks_bnd_used): Likewise.
6387 (indirect_return_needed): New.
6388 (indirect_return_bnd_needed): Likewise.
6389 (output_indirect_thunk_function): Add a bool argument for
6391 (output_indirect_thunk_function): Don't generate alias for
6392 function return thunk.
6393 (ix86_code_end): Call output_indirect_thunk_function to generate
6394 function return thunks.
6395 (ix86_output_function_return): Set indirect_return_bnd_needed
6396 and indirect_return_needed instead of indirect_thunk_bnd_needed
6397 and indirect_thunk_needed.
6399 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6401 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6402 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6403 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6405 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6406 Paul Hua <paul.hua.gm@gmail.com>
6409 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6411 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6413 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6414 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6417 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6420 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6421 instead of GET_MODE_SIZE when comparing Units.
6423 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6426 * varasm.c (hash_section): Return an unchangeble hash value
6427 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6428 Return !aarch64_can_use_per_function_literal_pools_p ().
6430 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6433 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6434 rather than address of cmode as last argument to prepare_cmp_insn.
6436 2018-03-15 Julia Koval <julia.koval@intel.com>
6438 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6439 F_AVX512VNNI, F_AVX512BITALG): New.
6441 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6444 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6445 insn for floating-point loads and stores.
6447 2018-03-14 Carl Love <cel@us.ibm.com>
6449 * config/rs6000/rs6000-c.c: Add macro definitions for
6450 ALTIVEC_BUILTIN_VEC_PERMXOR.
6451 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6452 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6453 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6454 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6456 * config/doc/extend.texi: Add prototypes for vec_permxor.
6458 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6461 * diagnostic-show-locus.c (class layout_point): Convert m_line
6462 from int to linenum_type.
6463 (line_span::comparator): Use linenum "compare" function when
6464 comparing line numbers.
6465 (test_line_span): New function.
6466 (layout_range::contains_point): Convert param "row" from int to
6468 (layout_range::intersects_line_p): Likewise.
6469 (layout::will_show_line_p): Likewise.
6470 (layout::print_source_line): Likewise.
6471 (layout::should_print_annotation_line_p): Likewise.
6472 (layout::print_annotation_line): Likewise.
6473 (layout::print_leading_fixits): Likewise.
6474 (layout::annotation_line_showed_range_p): Likewise.
6475 (struct line_corrections): Likewise for field m_row.
6476 (line_corrections::line_corrections): Likewise for param "row".
6477 (layout::print_trailing_fixits): Likewise.
6478 (layout::get_state_at_point): Likewise.
6479 (layout::get_x_bound_for_row): Likewise.
6480 (layout::print_line): Likewise.
6481 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
6482 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6483 * input.c (selftest::test_linenum_comparisons): New function.
6484 (selftest::input_c_tests): Call it.
6485 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6486 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6487 * selftest.h (ASSERT_GT): New macro.
6488 (ASSERT_GT_AT): New macro.
6489 (ASSERT_LT): New macro.
6490 (ASSERT_LT_AT): New macro.
6492 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6494 PR rtl-optimization/84780
6495 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6497 2018-03-14 Martin Liska <mliska@suse.cz>
6499 * tree.c (record_node_allocation_statistics): Use
6500 get_stats_node_kind.
6501 (get_stats_node_kind): New function extracted from
6502 record_node_allocation_statistics.
6503 (free_node): Use get_stats_node_kind.
6505 2018-03-14 Richard Biener <rguenther@suse.de>
6507 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6508 that the value-set of ANTIC_IN doesn't grow.
6511 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6513 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6514 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6516 2018-03-14 Julia Koval <julia.koval@intel.com>
6518 * config.gcc (icelake-client, icelake-server): New.
6520 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6521 (initial_ix86_arch_features): Ditto.
6522 (PTA_SKYLAKE): Add SGX.
6523 (PTA_ICELAKE): Remove.
6524 (PTA_ICELAKE_CLIENT): New.
6525 (PTA_ICELAKE_SERVER): New.
6526 (ix86_option_override_internal): Split up icelake on icelake client and
6528 (get_builtin_code_for_version): Ditto.
6529 (fold_builtin_cpu): Ditto.
6530 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6531 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6532 * config/i386/i386.h (processor_type): Ditto.
6533 * doc/invoke.texi: Ditto.
6535 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6538 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6539 INTEGER_CST offset, add it together with bitpos / 8 and
6540 sign extend based on POINTER_SIZE.
6544 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6547 * config/i386/constraints.md (Yc): New register constraint.
6548 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6549 Use Yc constraint for alternative 2 of operand 0. Remove
6550 preferred_for_speed attribute.
6552 2018-03-14 Richard Biener <rguenther@suse.de>
6554 PR tree-optimization/84830
6555 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6556 with the old one to avoid oscillations.
6558 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6561 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6563 (assign_by_spills): Return a flag of reload assignment failure.
6564 Do not process the reload assignment failures. Do not spill other
6565 reload pseudos if they has the same reg class. Update n if
6567 (lra_assign): Add a return arg. Set up from the result of
6568 assign_by_spills call.
6569 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6570 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6571 usage_insns if it is not NULL.
6572 (spill_hard_reg_in_range): New function.
6573 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6574 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6575 function prototypes.
6576 (lra_assign): Change prototype.
6577 * lra.c (lra): Add code to deal with fails by splitting hard reg
6580 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6582 * config/riscv/riscv.opt (mrelax): New option.
6583 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6584 "norelax" when riscv_mrelax is disabled.
6585 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6587 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6590 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6591 reassociation for int modes.
6593 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6595 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6596 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6598 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6599 * config/aarch64/aarch64-sve.md
6600 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6601 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6602 (*extend<mode><Vwide>2): Rename to...
6603 (aarch64_sve_extend<mode><Vwide>2): ...this.
6604 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6605 renaming the old pattern to...
6606 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6608 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6609 define_expand, renaming the old pattern to...
6610 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6611 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6612 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6613 account when deciding which SVE instruction the optab should use.
6614 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6616 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6618 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6619 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6620 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6621 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6622 (tlsdesc_small_<mode>): Turn a define_expand and use
6623 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6624 (tlsdesc_small_advsimd_<mode>): ...this.
6625 (tlsdesc_small_sve_<mode>): New pattern.
6627 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6629 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6630 (UNSPEC_UMUL_HIGHPART): New constants.
6631 (MUL_HIGHPART): New int iteraor.
6632 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6633 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6635 (*<su>mul<mode>3_highpart): New define_insn.
6637 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6640 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6643 2018-03-13 Martin Liska <mliska@suse.cz>
6646 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6648 (sem_item_optimizer::~sem_item_optimizer): Release it.
6649 (sem_item_optimizer::merge_classes): Register variable aliases.
6650 (sem_item_optimizer::fixup_pt_set): New function.
6651 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6652 * ipa-icf.h: Declare new variables and functions.
6654 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6657 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6658 integer_pow2p@2 and test integer_pow2p in condition.
6659 (A < 0 ? C : 0): Similarly for @1.
6662 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6663 characters starting at p contain '\0' character, don't look beyond
6667 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6668 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6671 * reg-stack.c (change_stack): Change update_end var from int to
6672 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6673 also call set_block_for_insn on the newly added insns and rescan.
6676 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6677 on the last operand.
6680 * tree.c (stabilize_reference_1): Return save_expr (e) for
6681 STATEMENT_LIST even if it doesn't have side-effects.
6683 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6685 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6687 2018-03-12 Renlin Li <renlin.li@arm.com>
6689 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6690 aarch64_output_scalar_simd_mov_immediate.
6692 2018-03-12 Martin Sebor <msebor@redhat.com>
6694 PR tree-optimization/83456
6695 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6696 for perfectly overlapping calls to memcpy.
6697 (gimple_fold_builtin_memory_chk): Same.
6698 (gimple_fold_builtin_strcpy): Handle no-warning.
6699 (gimple_fold_builtin_stxcpy_chk): Same.
6700 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6702 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6704 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6705 parameter. Use it for SFmode.
6706 (rs6000_function_arg_advance_1): Adjust.
6707 (rs6000_function_arg): Adjust.
6708 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6710 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6712 PR rtl-optimization/84169
6713 PR rtl-optimization/84780
6714 * combine.c (can_combine_p): Check for a 2-insn combination whether
6715 the destination register is used between the two insns, too.
6717 2018-03-12 Richard Biener <rguenther@suse.de>
6719 PR tree-optimization/84803
6720 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6721 for refs DR analysis didn't process.
6723 2018-03-12 Richard Biener <rguenther@suse.de>
6725 PR tree-optimization/84777
6726 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6727 force-vectorize loops ignore whether we are optimizing for size.
6729 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6731 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6732 (TARGET_MD_ASM_ADJUST): Define.
6734 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6735 Kito Cheng <kito.cheng@gmail.com>
6736 Chung-Ju Wu <jasonwucj@gmail.com>
6738 * config/nds32/nds32.c (nds32_compute_stack_frame,
6739 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6740 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6741 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6742 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6743 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6744 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6745 * config/nds32/nds32.md (prologue, epilogue): Use macro
6746 NDS32_V3PUSH_AVAILABLE_P to do checking.
6748 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6751 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6752 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6753 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6754 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6755 addition of most attributes on !orig_type_die or the attribute not
6756 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6758 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6759 Chung-Ju Wu <jasonwucj@gmail.com>
6761 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6763 * config/nds32/nds32.opt (mvh): New option.
6765 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6766 Chung-Ju Wu <jasonwucj@gmail.com>
6768 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6770 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6771 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6774 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6775 Chung-Ju Wu <jasonwucj@gmail.com>
6777 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6779 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6780 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6782 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6783 Kito Cheng <kito.cheng@gmail.com>
6784 Chung-Ju Wu <jasonwucj@gmail.com>
6786 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6787 UNSPEC_FFMISM and UNSPEC_FLMISM.
6788 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6789 for ffb, ffmism and flmism.
6790 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6791 (unspec_ffmism): Ditto.
6792 (unspec_flmism): Ditto.
6793 (nds32_expand_builtin_impl): Check if string extension is available.
6794 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6795 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6797 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6800 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6803 * lra-assigns.c (assign_by_spills): Return a flag of reload
6804 assignment failure. Do not process the reload assignment
6805 failures. Do not spill other reload pseudos if they has the same
6807 (lra_assign): Add a return arg. Set up from the result of
6808 assign_by_spills call.
6809 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6810 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6811 usage_insns if it is not NULL.
6812 (spill_hard_reg_in_range): New function.
6813 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6814 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6815 function prototypes.
6816 (lra_assign): Change prototype.
6817 * lra.c (lra): Add code to deal with fails by splitting hard reg
6820 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6823 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6825 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6828 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6829 (dw_val_node): Add val_symbolic_view.
6830 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6831 (symview_upper_bound): New.
6832 (new_line_info_table): Initialize symviews_since_reset.
6833 (dwarf2out_source_line): Count symviews_since_reset and set
6834 symview_upper_bound.
6835 (dw_val_equal_p): Handle symview.
6836 (add_AT_symview): New.
6837 (print_dw_val): Handle symview.
6838 (attr_checksum, attr_checksum_ordered): Likewise.
6839 (same_dw_val_p, size_of_die): Likewise.
6840 (value_format, output_die): Likewise.
6841 (add_high_low_attributes): Use add_AT_symview for entry_view.
6842 (dwarf2out_finish): Reset symview_upper_bound, clear
6845 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6848 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6849 Add strict argument and use it.
6850 (rs6000_split_multireg_move): Update for new strict argument.
6851 (mem_operand_gpr): Disallow all non-offsettable addresses.
6852 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6854 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6857 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6858 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6859 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6862 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6863 decl, use remap_type if we want to use the type.
6865 2018-03-09 Martin Sebor <msebor@redhat.com>
6867 PR tree-optimization/84526
6868 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6870 (builtin_access::generic_overlap): Be prepared to handle non-array
6873 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6875 PR rtl-optimization/84682
6876 * lra-constraints.c (process_address_1): Check is_address flag
6877 for address constraints.
6878 (process_alt_operands): Likewise.
6879 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6880 preprocess_constraints.
6881 * recog.h (preprocess_constraints): Add oploc parameter.
6883 * recog.c (preprocess_constraints): Test address_operand for
6884 CT_ADDRESS constraints.
6886 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6889 * lra-assigns.c (assign_by_spills): Return a flag of reload
6890 assignment failure. Do not process the reload assignment
6891 failures. Do not spill other reload pseudos if they has the same
6893 (lra_assign): Add a return arg. Set up from the result of
6894 assign_by_spills call.
6895 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6896 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6897 usage_insns if it is not NULL.
6898 (spill_hard_reg_in_range): New function.
6899 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6900 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6901 function prototypes.
6902 (lra_assign): Change prototype.
6903 * lra.c (lra): Add code to deal with fails by splitting hard reg
6906 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6909 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6910 Accept complain bool parameter. Only emit errors if it is true.
6911 (arm_parse_cpu_option_name): Likewise.
6912 (arm_target_thumb_only): Adjust callers of the above.
6913 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6914 prototype to take a default true bool parameter.
6915 (arm_parse_arch_option_name): Likewise.
6917 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6918 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6922 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6923 * configure: Regenerate.
6924 * configure.ac ("linker --version-script option"): New.
6925 ("linker soname option"): New.
6927 2018-03-09 Richard Biener <rguenther@suse.de>
6929 PR tree-optimization/84775
6930 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6931 immediate uses of predicate stmts and mark them modified.
6934 PR tree-optimization/84178
6935 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6937 (version_loop_for_if_conversion): Delay update_ssa call.
6938 (tree_if_conversion): Delay update_ssa until after predicate
6941 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6944 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6945 when the function accesses prior frames.
6947 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6950 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6951 gen_llsym, otherwise call maybe_gen_llsym.
6954 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6955 has ',' character inside of it.
6957 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6960 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6961 as clobbering CC_REGNUM.
6963 2018-03-08 Richard Biener <rguenther@suse.de>
6966 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6967 (follow_copies_to_constant): Do not follow SSA names registered
6970 2018-03-08 Richard Biener <rguenther@suse.de>
6972 PR tree-optimization/84178
6973 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6975 (version_loop_for_if_conversion): Delay update_ssa call.
6976 (tree_if_conversion): Delay update_ssa until after predicate
6979 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6981 PR tree-optimization/84178
6982 * tree-if-conv.c (release_bb_predicate): Remove the
6983 the assertion that the stmts have NULL use_ops.
6984 Discard the statements, asserting that they haven't
6985 yet been added to a BB.
6987 2018-03-08 Richard Biener <rguenther@suse.de>
6989 PR tree-optimization/84746
6990 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6991 (phi_translate): Pass in destination ANTIC_OUT set.
6992 (phi_translate_1): Likewise. For a simplified result lookup
6993 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6994 (phi_translate_set): Adjust.
6995 (do_pre_regular_insertion): Likewise.
6996 (do_pre_partial_partial_insertion): Likewise.
6998 2018-03-08 Martin Liska <mliska@suse.cz>
7000 PR gcov-profile/84735
7001 * doc/gcov.texi: Document usage of profile files.
7002 * gcov-io.h: Document changes in the format.
7004 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
7008 * dwarf2out.c (struct dw_line_info_table): Update comments for
7010 (FORCE_RESET_NEXT_VIEW): New.
7011 (FORCE_RESETTING_VIEW_P): New.
7012 (RESETTING_VIEW_P): Check for -1 too.
7013 (ZERO_VIEW_P): Likewise.
7014 (new_line_info_table): Force-reset next view.
7015 (dwarf2out_begin_function): Likewise.
7016 (dwarf2out_source_line): Simplify zero_view_p initialization.
7017 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
7018 view directly. Omit view when omitting .loc at line 0.
7020 2018-03-08 Jakub Jelinek <jakub@redhat.com>
7022 PR tree-optimization/84740
7023 * tree-switch-conversion.c (process_switch): Call build_constructors
7024 only if info.phi_count is non-zero.
7026 PR tree-optimization/84739
7027 * tree-tailcall.c (find_tail_calls): Check call arguments against
7028 DECL_ARGUMENTS (current_function_decl) rather than
7029 DECL_ARGUMENTS (func) when checking for tail recursion.
7031 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7033 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
7034 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7035 Volker Reichelt's entry and add entries for people that perform
7036 GCC fuzzy testing and report numerous bugs.
7038 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7041 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7042 readonly data in sdata, if that is disabled.
7043 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7044 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7045 -mreadonly-in-sdata option.
7047 2018-03-07 Martin Sebor <msebor@redhat.com>
7049 PR tree-optimization/84468
7050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7051 basic block when looking for nul assignment.
7053 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7056 * except.h (output_function_exception_table): Adjust prototype.
7057 * except.c (output_function_exception_table): Remove FNNAME parameter
7058 and add SECTION parameter. Ouput one part of the table at a time.
7059 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7060 the first part of the exception table and emit unwind directives.
7061 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7062 (i386_pe_seh_cold_init): Likewise.
7063 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7064 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7065 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7066 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7067 * config/i386/winnt.c: Include except.h.
7068 (struct seh_frame_state): Add reg_offset, after_prologue and
7069 in_cold_section fields.
7070 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7071 (i386_pe_seh_cold_init): New function.
7072 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7073 to seh->in_cold_section.
7074 (seh_emit_push): Record the offset of the push.
7075 (seh_emit_save): Record the offet of the save.
7076 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7077 Test seh->after_prologue to disregard the epilogue.
7078 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7079 (i386_pe_end_cold_function): New function.
7081 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7084 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7085 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7088 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7090 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7091 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7094 * multiple_target.c: Include tree-inline.h and intl.h.
7095 (expand_target_clones): Diagnose and fail if node->definition and
7096 !tree_versionable_function_p (node->decl).
7098 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7100 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7102 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7103 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7104 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7106 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7109 * combine.c (try_combine): Use reg_or_subregno instead of handling
7110 just paradoxical SUBREGs and REGs.
7112 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7114 * config/arc/arc.c (arc_finalize_pic): Remove function.
7115 (arc_must_save_register): We use single base PIC register, remove
7116 checks to save/restore the PIC register.
7117 (arc_expand_prologue): Likewise.
7118 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7120 (arc_verify_short): Likewise.
7121 (arc_attr_type): Likewise.
7122 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7123 (walk_stores): Likewise.
7124 (arc_address_cost): Make it static.
7125 (arc_verify_short): Likewise.
7126 (branch_dest): Likewise.
7127 (arc_attr_type): Likewise.
7128 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7129 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7130 (arc_final_prescan_insn): Remove inserting the nops due to
7131 hardware hazards. It is done in reorg step.
7132 (insn_length_variant_t): Remove.
7133 (insn_length_parameters_t): Likewise.
7134 (arc_insn_length_parameters): Likewise.
7135 (arc_get_insn_variants): Likewise.
7136 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7138 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7141 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7144 PR tree-optimization/84687
7145 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7147 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7149 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7151 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7152 Rename to ppc_speculation_barrier.
7153 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7154 __builtin_ppc_speculation_barrier.
7156 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7159 * combine.c (combine_simplify_rtx): Don't try to simplify if
7160 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7163 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7165 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7166 to 32 bytes when compiling for POWER9.
7168 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7171 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7172 regparm >= 3 with no arg reg available also for calls with
7173 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7176 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7178 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7180 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7183 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7185 2018-03-05 Richard Biener <rguenther@suse.de>
7187 PR tree-optimization/84486
7188 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7189 When inserting a __builtin_assume_aligned call set the LHS
7190 SSA name alignment info accordingly.
7192 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7194 PR tree-optimization/84114
7195 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7196 Avoid reassociation of FLOAT_MODE addition.
7198 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7200 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7201 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7202 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7203 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7204 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7205 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7206 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7208 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7209 __builtin_ia32_wbinvd): New builtins.
7210 (SPECIAL_ARGS2): New.
7211 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7212 (SPECIAL_ARGS2): New.
7213 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7214 (ix86_valid_target_attribute_inner_p): Ditto.
7215 (ix86_init_mmx_sse_builtins): Add special_args2.
7216 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7217 TARGET_WBNOINVD_P): New.
7218 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7219 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7220 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7221 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7222 * config/i386/pconfigintrin.h: New file.
7223 * config/i386/wbnoinvdintrin.h: Ditto.
7224 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
7226 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7228 2018-03-05 Richard Biener <rguenther@suse.de>
7230 PR tree-optimization/84670
7231 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7233 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7234 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7235 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7236 doesn't grow if all successors (recursively) were visited at least
7239 2018-03-05 Richard Biener <rguenther@suse.de>
7241 PR tree-optimization/84650
7242 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7243 if executed in the loop pipeline.
7245 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7247 * doc/configfiles.texi (Configuration Files): Move info about
7248 conditionalizing $target-protos.h to...
7249 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7250 differs from $target-protos.h.
7252 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7253 Chung-Ju Wu <jasonwucj@gmail.com>
7255 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7256 * config/nds32/nds32-multiple.md (setmemsi): Define.
7257 * config/nds32/nds32-memory-manipulation.c
7258 (nds32_gen_dup_4_byte_to_word_value): New.
7259 (emit_setmem_word_loop): New.
7260 (emit_setmem_byte_loop): New.
7261 (nds32_expand_setmem_loop): New.
7262 (nds32_expand_setmem_loop_v3m): New.
7263 (nds32_expand_setmem_unroll): New.
7264 (nds32_expand_setmem): New.
7266 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7267 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config/nds32/nds32-memory-manipulation.c
7270 (nds32_emit_load_store): New.
7271 (nds32_emit_post_inc_load_store): New.
7272 (nds32_emit_mem_move): New.
7273 (nds32_emit_mem_move_block): New.
7274 (nds32_expand_movmemsi_loop_unknown_size): New.
7275 (nds32_expand_movmemsi_loop_known_size): New.
7276 (nds32_expand_movmemsi_loop): New.
7277 (nds32_expand_movmemsi_unroll): New.
7278 (nds32_expand_movmemqi): Rename ...
7279 (nds32_expand_movmemsi): ... to this.
7280 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7281 (movmemsi): ... to this.
7282 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7283 (nds32_expand_movmemsi): ... to this.
7285 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7286 Monk Chiang <sh.chiang04@gmail.com>
7287 Chung-Ju Wu <jasonwucj@gmail.com>
7289 * config/nds32/nds32-protos.h
7290 (nds32_expand_load_multiple): New arguments.
7291 (nds32_expand_store_multiple): Ditto.
7292 (nds32_valid_multiple_load_store): Rename ...
7293 (nds32_valid_multiple_load_store_p): ... to this.
7294 * config/nds32/nds32-memory-manipulation.c
7295 (nds32_expand_load_multiple): Refine implementation.
7296 (nds32_expand_store_multiple): Ditto.
7297 * config/nds32/nds32-multiple.md
7298 (load_multiple): Update nds32_expand_load_multiple interface.
7299 (store_multiple): Update nds32_expand_store_multiple interface.
7300 * config/nds32/nds32-predicates.c
7301 (nds32_valid_multiple_load_store): Rename ...
7302 (nds32_valid_multiple_load_store_p): ... to this and refine
7304 * config/nds32/predicates.md
7305 (nds32_load_multiple_and_update_address_operation): New predicate.
7306 (nds32_store_multiple_and_update_address_operation): New predicate.
7308 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7309 Chung-Ju Wu <jasonwucj@gmail.com>
7311 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7312 (combo): New attribute.
7313 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7315 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7317 * config/nds32/nds32.opt: Change -mcmodel= default value.
7319 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7320 Monk Chiang <sh.chiang04@gmail.com>
7321 Chung-Ju Wu <jasonwucj@gmail.com>
7323 * config/nds32/constants.md (unspec_element): New enum.
7324 * config/nds32/constraints.md (Umw): New constraint.
7325 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7326 * config/nds32/nds32-intrinsic.md: Likewise.
7327 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7328 (nds32_valid_smw_lwm_base_p): New.
7329 (nds32_output_smw_single_word): New.
7330 (nds32_output_lmw_single_word): New.
7331 (nds32_expand_unaligned_load): New.
7332 (nds32_expand_unaligned_store): New.
7333 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7334 (nds32_output_smw_single_word): Declare.
7335 (nds32_output_lmw_single_word): Declare.
7336 (nds32_expand_unaligned_load): Declare.
7337 (nds32_expand_unaligned_store): Declare.
7338 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7339 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7340 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7341 NDS32_BUILTIN_UASTORE_DW.
7342 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7345 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7346 Kito Cheng <kito.cheng@gmail.com>
7347 Chung-Ju Wu <jasonwucj@gmail.com>
7349 * config/nds32/nds32-intrinsic.c
7350 (nds32_expand_builtin_null_ftype_reg): Delete.
7351 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7352 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7353 (nds32_read_argument): New.
7354 (nds32_legitimize_target): Ditto.
7355 (nds32_legitimize_argument): Ditto.
7356 (nds32_check_constant_argument): Ditto.
7357 (nds32_expand_unop_builtin): Ditto.
7358 (nds32_expand_unopimm_builtin): Ditto.
7359 (nds32_expand_binop_builtin): Ditto.
7360 (nds32_builtin_decl_impl): Ditto.
7361 (builtin_description): Ditto.
7362 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7363 (nds32_init_builtins_impl): Ditto.
7364 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7365 (nds32_builtin_decl): New.
7366 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7367 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7369 2018-03-02 Jeff Law <law@redhat.com>
7371 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7372 (redundant_insn, fill_simple_delay_slots): Likewise.
7373 (fill_slots_from_thread): Likewise.
7374 * resource.c (mark_referenced_resources): Likewise.
7375 (mark_set_resources, find_dead_or_set_registers): Likewise.
7377 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7379 * substring-locations.h (format_warning_va): Formatting fix for
7381 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7383 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7384 * substring-locations.c: Include intl.h.
7385 (format_warning_va): Turned into small wrapper around
7386 format_warning_n_va, renamed to ...
7387 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7388 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7390 (format_warning_at_substring_n): New function.
7391 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7392 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7393 format_warning_at_substring with just a shorter name instead of
7394 const function pointer.
7395 (fmtwarn_n): New function.
7396 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7397 appropriate, get rid of all the fmtstr temporaries, move conditionals
7398 with G_() wrapped string literals directly into fmtwarn arguments,
7399 cast dir.len to (int), formatting fixes.
7401 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7403 * doc/invoke.texi: Remove "Cilk Plus" references.
7405 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7406 Richard Biener <rguenther@suse.de>
7409 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7410 for error or warning attributes if CALL_FROM_THUNK_P is set.
7413 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7416 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7417 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7420 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7421 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7423 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7424 instead of -1U in last predictors element's probability member.
7426 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7429 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7430 arguments if they are comparable.
7432 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7434 PR tree-optimization/84634
7435 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7436 masks and masked_loop_p with a single loop_masks, making sure it's
7437 null for bb vectorization.
7439 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7441 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7442 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7443 loop->force_vectorize to check whether there is no alias.
7445 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7448 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7450 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7452 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7453 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7454 instead of a loop around prev_real_insn.
7455 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7459 * config/i386/i386.c (ix86_print_operand): Use conditional
7460 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7463 2018-03-02 Richard Biener <rguenther@suse.de>
7465 PR tree-optimization/84427
7466 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7467 (bitmap_set_subtract_values): Rewrite to handle multiple
7470 (prune_clobbered_mems): Likewise.
7471 (phi_translate): Take edge instead of pred/phiblock.
7472 (phi_translate_1): Likewise.
7473 (phi_translate_set): Likewise. Insert all translated
7474 exprs for a value into the set, keeping possibly multiple
7475 expressions per value.
7476 (compute_antic_aux): Adjust for phi_translate changes.
7477 When intersecting union the expressions and prune those
7478 not in the final value set, keeping possibly multiple
7479 expressions per value. Do not use value-insertion
7480 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7481 all expressions. Add verification that the value-sets
7482 only shrink during iteration.
7483 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7484 (do_pre_regular_insertion): Likewise.
7485 (do_pre_partial_partial_insertion): Likewise.
7487 2018-03-02 Richard Biener <rguenther@suse.de>
7490 * config/darwin.c (saved_debug_info_level): New static global.
7491 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7492 (darwin_asm_lto_end): Restore debug info generation settings.
7494 2018-03-01 Martin Liska <mliska@suse.cz>
7497 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7500 2018-03-01 Richard Biener <rguenther@suse.de>
7503 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7504 type annotation with LTO when debug was disabled at compile-time.
7506 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7508 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7510 (mips_final_postscan_insn): Likewise.
7512 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7514 PR rtl-optimization/84528
7515 * alias.c (init_alias_target): Add commentary.
7516 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7517 a unique base value if the frame pointer is not eliminated
7518 to the stack pointer.
7520 2018-03-01 Tom de Vries <tom@codesourcery.com>
7522 PR rtl-optimization/83327
7523 * lra-int.h (hard_regs_spilled_into): Declare.
7524 * lra.c (hard_regs_spilled_into): Define.
7525 (init_reg_info): Init hard_regs_spilled_into.
7526 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7527 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7528 (process_bb_lives): Handle hard_regs_spilled_into.
7529 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7530 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7532 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7534 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7535 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7536 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7537 * config/rs6000/aix72.h: New file.
7539 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7541 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7542 instead of warning_at with conditional singular and plural messages
7546 * stor-layout.c (update_alignment_for_field): For
7547 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7548 && !DECL_PACKED (field), do the alignment update, just use
7549 only desired_align instead of MAX (type_align, desired_align)
7551 (place_field): Don't do known_align < desired_align handling
7552 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7553 is non-NULL, instead do it after rli->prev_field handling and
7554 only if not within a bitfield word. For DECL_PACKED (field)
7555 use type_align of BITS_PER_UNIT.
7557 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7559 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7560 superfluous parentheses and trailing spaces.
7562 2018-02-28 Richard Biener <rguenther@suse.de>
7564 PR tree-optimization/84584
7565 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7566 SCoPs with fake exit edge.
7568 2018-02-28 Martin Liska <mliska@suse.cz>
7571 * timevar.c (timer::print): Fix format to properly print 100%
7574 2018-02-28 Richard Biener <rguenther@suse.de>
7577 * genmatch.c (capture_info::walk_match): Do not mark
7578 captured expressions without operands as expr_p given
7579 they act more like predicates and should be subject to
7580 "lost tail" side-effect preserving.
7582 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7584 PR rtl-optimization/81611
7585 * auto-inc-dec.c (attempt_change): Move dead note from
7586 mem_insn if it's the next use of regno
7587 (find_address): Take address use of reg holding
7588 non-incremented value. Add parm to limit search to the named
7590 (merge_in_block): Attempt to use a mem insn that is the next
7591 use of the original regno.
7593 2018-02-27 Martin Sebor <msebor@redhat.com>
7596 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7597 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7599 2018-02-27 Martin Sebor <msebor@redhat.com>
7601 PR translation/84207
7602 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7603 n argument to unsigned HOST_WIDE_INT.
7604 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7605 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7606 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7607 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7609 2018-02-27 Richard Biener <rguenther@suse.de>
7611 PR tree-optimization/84512
7612 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7613 Do not use the estimate returned from record_stmt_cost for
7614 the scalar iteration cost but sum properly using add_stmt_cost.
7616 2018-02-27 Richard Biener <rguenther@suse.de>
7618 PR tree-optimization/84466
7619 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7620 Adjust last change to less strictly validate use operands.
7622 2018-02-27 Martin Liska <mliska@suse.cz>
7624 PR gcov-profile/84548
7625 * gcov.c (process_file): Allow partial overlap and consider it
7626 also as group functions.
7627 (output_lines): Properly calculate range of lines for a group.
7629 2018-02-27 Martin Liska <mliska@suse.cz>
7631 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7632 'ggc' suffixes. Change first column width.
7633 (timer::print): Fix formatting of the column.
7635 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7637 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7638 preserve inline entry blocks for the sake of debug inline
7639 entry point markers alone.
7640 (remove_unused_locals): Suggest in comments a better place to
7641 force the preservation of inline entry blocks that are
7642 otherwise unused, but do not preserve them.
7644 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7646 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7648 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7651 * config/i386/constraints.md (Bs): Replace
7652 ix86_indirect_branch_register with
7653 TARGET_INDIRECT_BRANCH_REGISTER.
7655 * config/i386/i386.md (indirect_jump): Likewise.
7656 (tablejump): Likewise.
7657 (*sibcall_memory): Likewise.
7658 (*sibcall_value_memory): Likewise.
7659 Peepholes of indirect call and jump via memory: Likewise.
7660 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7661 (*sibcall_value_GOT_32): Likewise.
7662 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7663 (GOT_memory_operand): Likewise.
7664 (call_insn_operand): Likewise.
7665 (sibcall_insn_operand): Likewise.
7666 (GOT32_symbol_operand): Likewise.
7667 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7669 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7671 PR rtl-optimization/83496
7672 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7673 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7674 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7675 redundant insn, if any.
7676 (relax_delay_slots): Likewise.
7677 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7679 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7681 PR tree-optimization/83965
7682 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7683 that grouped statements are part of a reduction chain. Return
7684 true if the statement is not marked as a reduction itself but
7686 (vect_recog_dot_prod_pattern): Don't check whether the statement
7687 is part of a group here.
7688 (vect_recog_sad_pattern): Likewise.
7689 (vect_recog_widen_sum_pattern): Likewise.
7691 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7694 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7696 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7699 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7701 (ix86_output_indirect_function_return): New prototype.
7702 (ix86_split_simple_return_pop_internal): Likewise.
7703 * config/i386/i386.c (indirect_return_via_cx): New.
7704 (indirect_return_via_cx_bnd): Likewise.
7705 (indirect_thunk_name): Handle return va CX_REG.
7706 (output_indirect_thunk_function): Create alias for
7707 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7708 (ix86_output_indirect_jmp): Remove the bool argument.
7709 (ix86_output_indirect_function_return): New function.
7710 (ix86_split_simple_return_pop_internal): Likewise.
7711 * config/i386/i386.md (*indirect_jump): Don't pass false
7712 to ix86_output_indirect_jmp.
7713 (*tablejump_1): Likewise.
7714 (simple_return_pop_internal): Change it to define_insn_and_split.
7715 Call ix86_split_simple_return_pop_internal to split it for
7717 (simple_return_indirect_internal): Call
7718 ix86_output_indirect_function_return instead of
7719 ix86_output_indirect_jmp.
7721 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7724 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7725 memset and value initialization afterwards.
7727 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7729 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7731 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7734 * common/config/aarch64/aarch64-common.c
7735 (aarch_option_optimization_table[]): Switch
7736 off fomit-frame-pointer
7738 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7739 Chung-Ju Wu <jasonwucj@gmail.com>
7741 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7743 (store_multiple): Ditto.
7745 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7747 * config.gcc: Add --with-cpu support for nds32 target.
7748 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7749 * config/nds32/nds32.opt: Add -mcpu= option.
7751 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7753 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7754 isel=yes): Warn for these deprecated options.
7756 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7758 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7759 ISA_2_5_MASKS_EMBEDDED.
7761 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7763 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7764 p->max as pointers rather than using iterative_hash_expr.
7766 2018-02-23 Carl Love <cel@us.ibm.com>
7768 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7769 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7771 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7772 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7773 P8V_BUILTIN_VEC_VUNSIGNED2.
7775 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7778 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7779 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7781 * lra-constraints.c (curr_insn_transform): Set up
7782 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7784 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7785 * lra-eliminations.c (spill_pseudos): Ditto.
7786 (process_insn_for_elimination): Ditto.
7787 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7788 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7789 LRA_NON_CLOBBERED_ALT.
7791 2018-02-22 Martin Sebor <msebor@redhat.com>
7793 PR tree-optimization/84480
7794 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7795 to maybe_diag_stxncpy_trunc. Call it.
7796 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7797 from gimple_fold_builtin_strcpy. Print inlining stack.
7798 (handle_builtin_stxncpy): Print inlining stack.
7799 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7801 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7804 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7805 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7806 and -fcheck-pointer-bounds are used together.
7807 (indirect_thunk_prefix): New enum.
7808 (indirect_thunk_need_prefix): New function.
7809 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7810 "_nt" instead of "_bnd" for NOTRACK prefix.
7811 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7812 (output_indirect_thunk_function): Likewise.
7814 (ix86_code_end): Update output_indirect_thunk_function calls.
7815 (ix86_output_indirect_branch_via_reg): Replace
7816 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7817 (ix86_output_indirect_branch_via_push): Likewise.
7818 (ix86_output_function_return): Likewise.
7819 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7820 incompatible with -fcf-protection=branch and
7821 -fcheck-pointer-bounds.
7823 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7826 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7827 Change gcc_assert call to output_operand_lossage.
7829 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7831 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7833 2018-02-22 DJ Delorie <dj@redhat.com>
7834 Sebastian Perta <sebastian.perta@renesas.com>
7835 Oleg Endo <olegendo@gcc.gnu.org>
7837 * config/rx/rx.c (rx_rtx_costs): New function.
7838 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7840 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7842 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7844 2018-02-22 Martin Liska <mliska@suse.cz>
7847 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7848 Add "native" as a possible value.
7850 2018-02-22 Martin Liska <mliska@suse.cz>
7853 * config/i386/i386.c (ix86_option_override_internal):
7854 Add "native" as a possible value for -march and -mtune.
7856 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7859 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7860 to all type variants.
7862 PR tree-optimization/84503
7863 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7864 width as info->bitpos + info->bitsize - start.
7865 (merged_store_group::merge_overlapping): Simplify width computation.
7866 (check_no_overlap): New function.
7867 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7868 start + width and last_order of the group, fail if check_no_overlap
7870 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7871 to group if check_no_overlap fails.
7873 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7875 * config/rs6000/altivec.md: Delete contraint arguments to
7876 define_expand, define_split, and define_peephole2, and in
7877 define_insn_and_split if always unused.
7878 * config/rs6000/darwin.md: Ditto.
7879 * config/rs6000/dfp.md: Ditto.
7880 * config/rs6000/rs6000.md: Ditto.
7881 * config/rs6000/sync.md: Ditto.
7882 * config/rs6000/vector.md: Ditto.
7883 * config/rs6000/vsx.md: Ditto.
7885 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7887 * config/rs6000/altivec.md: Write output control strings as braced
7888 blocks instead of double-quoted strings.
7889 * config/rs6000/darwin.md: Ditto.
7890 * config/rs6000/rs6000.md: Ditto.
7891 * config/rs6000/vector.md: Ditto.
7892 * config/rs6000/vsx.md: Ditto.
7894 2018-02-21 Jason Merrill <jason@redhat.com>
7896 PR c++/84314 - ICE with templates and fastcall attribute.
7897 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7899 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7901 * ipa-cp.c (determine_versionability): Fix comment typos.
7903 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7906 * ipa-cp.c (determine_versionability): Do not version functions caling
7909 2018-02-21 Martin Liska <mliska@suse.cz>
7912 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7913 Add "native" as a possible value.
7914 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7915 the macro when native cpu detection is available.
7917 2018-02-21 Martin Liska <mliska@suse.cz>
7920 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7921 Add "native" as a possible value.
7922 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7923 when native cpu detection is available.
7925 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7926 Martin Sebor <msebor@redhat.com>
7928 PR tree-optimization/84478
7929 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7931 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7932 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7933 support which is conservatively correct, for 2 only stay conservative
7934 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7935 argument to the 2 argument get_range_strlen, adjust 6 arg
7936 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7938 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7939 (gimple_fold_builtin_strlen): Pass true as last argument to
7942 2018-02-20 Martin Sebor <msebor@redhat.com>
7945 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7946 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7947 (builtin_memref::builtin_memref): Factor out parts into
7948 set_base_and_offset and call it.
7950 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7953 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7954 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7955 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7956 search at the associated MODE_INT.
7958 2018-02-20 Jeff Law <law@redhat.com>
7961 PR tree-optimization/81592
7963 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7964 for range data rather than using global data.
7965 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7966 range data rather than using global data.
7967 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7968 pass it to children as needed.
7969 (struct directive::fmtresult): Similarly.
7970 (struct directive::set_width): Similarly.
7971 (struct directive::set_precision): Similarly.
7972 (format_integer, format_directive, parse_directive): Similarly.
7973 (format_none): Accept unnamed vr_values parameter.
7974 (format_percent, format_floating, format_character): Similarly.
7975 (format_string, format_plain): Similarly.
7976 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7977 the EVRP range analyzer for range data rather than using global data.
7978 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7979 gimple-ssa-evrp-analyze.h
7980 (class sprintf_dom_walker): Add after_dom_children member function.
7981 Add evrp_range_analyzer member.
7982 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7983 range analyzer as needed.
7984 (sprintf_dom_walker::after_dom_children): New member function.
7985 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7987 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7988 (evrp_range_analyzer::pop_to_marker): Likewise.
7990 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7992 PR tree-optimization/84419
7993 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7994 with the required type if its current type is compatible but
7997 2018-02-20 Jakub Jelinek <jakub@redhat.com>
8000 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
8001 after vectorization.
8003 2018-02-20 Martin Liska <mliska@suse.cz>
8006 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
8007 possible values if we don't have a hint.
8009 2018-02-20 Martin Liska <mliska@suse.cz>
8013 * final.c (shorten_branches): Build align_tab array with one
8015 * opts.c (finish_options): Add alignment option limit check.
8016 (MAX_CODE_ALIGN): Likewise.
8017 (MAX_CODE_ALIGN_VALUE): Likewise.
8018 * doc/invoke.texi: Document maximum allowed option value for
8019 all -falign-* options.
8021 2018-02-19 Jakub Jelinek <jakub@redhat.com>
8024 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
8025 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
8026 * var-tracking.c (emit_note_insn_var_location): Remove all references
8027 to NOTE_INSN_CALL_ARG_LOCATION.
8028 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
8029 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
8030 Use copy_rtx_if_shared.
8031 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
8032 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
8033 (dwarf2out_var_location): Remove handling of
8034 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8036 * final.c (final_scan_insn): Remove all references to
8037 NOTE_INSN_CALL_ARG_LOCATION.
8038 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8039 before dumping final insns.
8040 * except.c (emit_note_eh_region_end): Remove all references to
8041 NOTE_INSN_CALL_ARG_LOCATION.
8042 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8043 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8044 * config/arc/arc.c (hwloop_optimize): Likewise.
8045 * config/arm/arm.c (create_fix_barrier): Likewise.
8046 * config/s390/s390.c (s390_chunkify_start): Likewise.
8047 * config/sh/sh.c (find_barrier): Likewise.
8048 * config/i386/i386.c (rest_of_insert_endbranch,
8049 ix86_seh_fixup_eh_fallthru): Likewise.
8050 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8051 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8052 * config/frv/frv.c (frv_function_prologue): Likewise.
8053 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8055 (note_outside_basic_block_p): Remove all references to
8056 NOTE_INSN_CALL_ARG_LOCATION.
8057 * gengtype.c (adjust_field_rtx_def): Likewise.
8058 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8060 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8061 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8064 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8067 PR tree-optimization/84452
8068 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8069 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8072 2018-02-19 Martin Liska <mliska@suse.cz>
8075 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8077 2018-02-19 Martin Liska <mliska@suse.cz>
8078 Richard Sandiford <richard.sandiford@linaro.org>
8080 PR tree-optimization/82491
8081 * gimple-fold.c (get_base_constructor): Make earlier bail out
8084 2018-02-19 Carl Love <cel@us.ibm.com>
8086 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8087 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8089 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8090 P8V_BUILTIN_VEC_NEG.
8092 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8094 * config/rl78/rl78.md (movdf): New define expand.
8096 2018-02-19 Martin Liska <mliska@suse.cz>
8099 * doc/invoke.texi: Fix typo.
8100 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8102 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8104 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8105 handle rs6000_single_float and rs6000_double_float specially for
8108 2018-02-16 Jeff Law <law@redhat.com>
8110 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8113 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8116 * ipa-inline.c (inline_small_functions): Fix a typo.
8118 2018-02-16 Nathan Sidwell <nathan@acm.org>
8120 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8122 2018-02-16 Carl Love <cel@us.ibm.com>
8124 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8125 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8126 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8127 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8128 expansion to P8V_BUILTIN_VEC_FLOAT2.
8130 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8132 PR rtl-optimization/70023
8133 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8134 src_regno into account.
8136 2018-02-16 Carl Love <cel@us.ibm.com>
8138 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8139 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8140 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8141 * config/rs6000/rs6000.c: Remove case statements for
8142 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8143 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8144 and P9V_BUILTIN_VEC_VINSERT4B.
8145 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8146 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8147 * config/rs6000/vsx.md:
8148 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8151 2018-02-16 Carl Love <cel@us.ibm.com>
8153 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8155 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8157 * config/rs6000/rs6000-c.c: Add the definitions for
8158 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8159 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8160 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8161 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8162 definition for insert4b and define insn *insert3b_internal.
8163 * doc/extend.texi: Add documentation for vec_extract4b.
8165 2018-02-16 Nathan Sidwell <nathan@acm.org>
8167 * doc/extend.texi (Backwards Compatibility): Mention friend
8168 injection. Note for-scope is deprecated.
8169 * doc/invoke.texi (-ffriend-injection): Deprecate.
8171 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8173 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8174 that moved to I2, also allow destinations that are a paradoxical
8175 subreg (instead of a normal reg).
8177 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8180 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8183 2018-02-16 Richard Biener <rguenther@suse.de>
8185 PR tree-optimization/84037
8186 PR tree-optimization/84016
8188 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8189 Adjust vec_construct for the fact we need additional higher latency
8190 128bit inserts for AVX256 and AVX512 vector builds.
8191 (ix86_add_stmt_cost): Scale vector construction cost for
8194 2018-02-16 Richard Biener <rguenther@suse.de>
8196 PR tree-optimization/84417
8197 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8198 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8199 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8201 2018-02-16 Martin Liska <mliska@suse.cz>
8204 * internal-fn.def (ASAN_CHECK): Set proper flags.
8205 (ASAN_MARK): Likewise.
8207 2018-02-16 Julia Koval <julia.koval@intel.com>
8209 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8210 from PTA_CANNONLAKE.
8212 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8215 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8216 Use ++iter rather than iter++ for std::list iterators.
8217 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8218 defer deleting them until all nodes in the forest are processed. Do
8219 free even leaf nodes. Change to_process into auto_vec.
8222 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8223 * vec.h (vec_default_construct): Use memset instead of placement new
8224 if BROKEN_VALUE_INITIALIZATION is defined.
8225 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8226 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8229 PR rtl-optimization/83723
8230 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8231 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8232 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8234 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8236 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8238 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8240 PR rtl-optimization/81443
8241 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8242 from inner REGs to paradoxical SUBREGs.
8244 2018-02-16 Richard Biener <rguenther@suse.de>
8246 PR tree-optimization/84399
8247 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8248 For operands we can analyze at their definition make sure we can
8249 analyze them at each use as well.
8251 2018-02-16 Richard Biener <rguenther@suse.de>
8253 PR tree-optimization/84190
8254 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8255 volatile accesses if the decl isn't volatile.
8257 2018-02-15 Jason Merrill <jason@redhat.com>
8259 PR c++/84314 - ICE with templates and fastcall attribute.
8260 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8261 TYPE_CANONICAL on an existing type.
8263 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8265 PR tree-optimization/84383
8266 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8267 dstoff nor call operand_equal_p if dstbase is NULL.
8269 PR tree-optimization/84334
8270 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8271 also a CONSTANT_CLASS_P, punt.
8273 2018-02-14 Jim Wilson <jimw@sifive.com>
8275 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8276 first SMALL_OPERAND check. New local min_second_step. Move assert
8277 to where locals are set. Add TARGET_RVC support.
8278 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8280 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8282 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8284 2018-02-14 Martin Sebor <msebor@redhat.com>
8286 PR tree-optimization/83698
8287 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8288 arrays constrain the offset range to their bounds.
8289 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8290 (builtin_access::overlap): Avoid setting the size of overlap if it's
8292 (maybe_diag_overlap): Also consider arrays when deciding what values
8293 of offsets to include in diagnostics.
8295 2018-02-14 Martin Sebor <msebor@redhat.com>
8298 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8299 that correspond to the kind of a declaration.
8301 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8304 * config/pa/pa.md: Load address of PIC label using the linkage table
8305 if the label is nonlocal.
8307 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8309 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8310 warning message if user requests -maltivec=be.
8311 * doc/invoke.texi: Document deprecation of -maltivec=be.
8313 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8316 * config/rs6000/rs6000-c.c: Update definitions for
8317 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8318 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8320 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8323 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8324 add _get_ssp intrinsics. Remove argument from
8325 __builtin_ia32_rdssp[d|q].
8326 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8327 * config/i386/i386-builtin.def: Remove argument from
8328 __builtin_ia32_rdssp[d|q].
8329 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8330 ix86_expand_special_args_builtin for _rdssp[d|q].
8331 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8332 Clear register before usage.
8333 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8334 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8336 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8338 PR tree-optimization/84357
8339 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8340 operand 1 of an ARRAY_REF too.
8342 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8345 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8346 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8348 (set_of_reg): New struct.
8349 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8350 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8351 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8353 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8354 Split into bitclr, bitset, bitinvert patterns if appropriate.
8355 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8356 use rx_fuse_in_memory_bitop.
8357 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8358 to named insn, correct maximum insn length.
8360 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8363 * machmode.def: Define a complex mode for PARTIAL_INT.
8364 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8366 * doc/rtl.texi: Document CSPImode.
8367 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8369 (msp430_hard_regno_nregs_with_padding): Likewise.
8371 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8374 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8376 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8378 PR rtl-optimization/84169
8379 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8380 we generated a parallel as new i3 and we split that to new i2 and i3
8381 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8382 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8383 those to i2, not i1. Partially rewrite this scan code.
8385 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8388 * stor-layout.c (place_field): For variable length fields, adjust
8389 offset_align afterwards not just based on the field's alignment,
8390 but also on the size.
8393 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8394 of exps and logs in the use_exp2 case.
8396 2018-02-13 Jeff Law <law@redhat.com>
8398 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8401 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8404 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8408 * common.opt (gas-loc-support, gas-locview-support): New.
8409 (ginline-points, ginternal-reset-location-views): New.
8410 * doc/invoke.texi: Document them. Use @itemx where intended.
8411 (gvariable-location-views): Adjust.
8412 * target.def (reset_location_view): New.
8413 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8414 (TARGET_RESET_LOCATION_VIEW): New.
8415 * doc/tm.texi: Rebuilt.
8416 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8417 (dwarf2out_default_as_locview_support): New.
8418 (output_asm_line_debug_info): Use option variables.
8419 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8420 (output_loc_list): Likewise.
8421 (add_high_low_attributes): Check option variables.
8422 Don't output entry view attribute in strict mode.
8423 (gen_inlined_subroutine_die): Check option variables.
8424 (dwarf2out_inline_entry): Likewise.
8425 (init_sections_and_labels): Likewise.
8426 (dwarf2out_early_finish): Likewise.
8427 (maybe_reset_location_view): New, from...
8428 (dwarf2out_var_location): ... here. Call it.
8429 * debug.h (dwarf2out_default_as_loc_support): Declare.
8430 (dwarf2out_default_as_locview_support): Declare.
8431 * hooks.c (hook_int_rtx_insn_0): New.
8432 * hooks.h (hook_int_rtx_insn_0): Declare.
8433 * toplev.c (process_options): Take -gas-loc-support and
8434 -gas-locview-support from dwarf2out. Enable
8435 -gvariable-location-views by default only with locview
8436 assembler support. Enable -ginternal-reset-location-views by
8437 default only if the target defines the corresponding hook.
8438 Enable -ginline-points by default if location views are
8439 enabled; force it disabled if statement frontiers are
8441 * tree-inline.c (expand_call_inline): Check option variables.
8442 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8444 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8446 PR tree-optimization/84321
8447 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8448 handling. Also check whether the anti-range contains any values
8449 that satisfy the mask; switch to a VR_RANGE if not.
8451 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8454 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8456 2018-02-13 Martin Jambor <mjambor@suse.cz>
8459 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8460 of call statements, also set location of a load to a temporary.
8462 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8464 * config/rl78/rl78.c (add_vector_labels): New function.
8465 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8466 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8467 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8468 which checks that no arguments are passed.
8469 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8470 * doc/extend.texi: Documentation for the new attribute.
8472 2018-02-13 Andreas Schwab <schwab@suse.de>
8474 * config/riscv/linux.h (CPP_SPEC): Define.
8476 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8479 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8480 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8481 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8482 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8483 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8484 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8485 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8486 temporarily for AES and PCLMUL builtins.
8488 PR tree-optimization/84339
8489 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8490 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8494 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8495 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8496 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8498 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8500 * omp-simd-clone.h: New file.
8501 * omp-simd-clone.c: Include omp-simd-clone.h.
8502 (expand_simd_clones): No longer static.
8503 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8504 cgraph.h and omp-simd-clone.h.
8505 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8506 (vect_recog_widen_shift_pattern): Formatting fix.
8507 (vect_pattern_recog_1): Don't check optab for calls.
8510 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8511 operands[2] into a REG before using gen_lowpart on it.
8513 2018-02-12 Jeff Law <law@redhat.com>
8516 * config/sh/sh.c (find_barrier): Consider a sibling call
8519 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8520 successfully back substituting a reg.
8522 2018-02-12 Richard Biener <rguenther@suse.de>
8524 PR tree-optimization/84037
8525 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8526 parameter, move visited init to caller.
8527 (vect_slp_analyze_operations): Separate cost from validity
8528 check, initialize visited once for all instances.
8529 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8531 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8533 (vect_model_promotion_demotion_cost): Likewise.
8534 (vectorizable_bswap): Guard cost modeling with !slp_node
8535 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8537 (vectorizable_call): Likewise.
8538 (vectorizable_conversion): Likewise.
8539 (vectorizable_assignment): Likewise.
8540 (vectorizable_shift): Likewise.
8541 (vectorizable_operation): Likewise.
8542 (vectorizable_store): Likewise.
8543 (vectorizable_load): Likewise.
8544 (vectorizable_condition): Likewise.
8545 (vectorizable_comparison): Likewise.
8547 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8550 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8551 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8552 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8554 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8557 * params.def (inline-min-speedup): Increase from 8 to 15.
8558 (max-inline-insns-auto): Decrease from 40 to 30.
8559 * ipa-split.c (consider_split): Add some buffer for function to
8560 be considered inlining candidate.
8561 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8564 2018-02-12 Richard Biener <rguenther@suse.de>
8566 PR tree-optimization/84037
8567 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8568 matched stmts if we cannot swap the non-matched ones.
8570 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8572 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8573 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8574 _mm_maskz_scalef_round_ss): New intrinsics.
8575 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8576 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8577 __builtin_ia32_scalefss_round): Remove.
8578 (__builtin_ia32_scalefsd_mask_round,
8579 __builtin_ia32_scalefss_mask_round): New intrinsics.
8580 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8581 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8582 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8583 "<round_constraint>")): Changed to ...
8584 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8585 "<round_scalar_constraint>")): ... this.
8586 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8587 %0, %1, %2<round_op3>}"): Changed to ...
8588 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8589 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8590 %2<round_scalar_mask_op3>}"): ... this.
8591 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8593 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8595 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8596 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8597 (_mm_maskz_sqrt_round_ss): New intrinsics.
8598 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8599 (__builtin_ia32_sqrtsd_mask_round)
8600 (__builtin_ia32_sqrtss_mask_round): New builtins.
8601 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8602 (__builtin_ia32_sqrtss_round): Remove.
8603 (__builtin_ia32_sqrtsd_mask_round)
8604 (__builtin_ia32_sqrtss_mask_round): New builtins.
8605 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8606 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8607 ((match_operand:VF_128 1 "vector_operand"
8608 "xBm,<round_constraint>")): Changed to ...
8609 ((match_operand:VF_128 1 "vector_operand"
8610 "xBm,<round_scalar_constraint>")): ... this.
8611 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8612 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8613 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8614 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8615 %<iptr>1<round_scalar_mask_op3>}): ... this.
8616 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8617 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8619 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8622 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8623 Cast vec_cmpeq result to correct type.
8624 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8625 Cast vec_cmpgt result to correct type.
8627 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8629 * final.c (final_scan_insn_1): Renamed from...
8630 (final_scan_insn): ... this. New wrapper, to recover
8631 seen from the outermost call in recursive ones.
8632 * config/sparc/sparc.c (output_return): Drop seen from call.
8633 (output_sibcall): Likewise.
8634 * config/visium/visium.c (output_branch): Likewise.
8636 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8638 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8641 2018-02-10 Alan Modra <amodra@gmail.com>
8644 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8645 Specify LR as an input.
8647 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8650 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8651 remove_member_access_dummy_vars): New functions.
8652 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8653 lower_omp_1, execute_lower_omp): Use them.
8655 PR rtl-optimization/84308
8656 * shrink-wrap.c (spread_components): Release todo vector.
8658 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8660 PR rtl-optimization/57193
8661 * ira-color.c (struct allocno_color_data): Add member
8662 conflict_allocno_hard_prefs.
8663 (update_conflict_allocno_hard_prefs): New.
8664 (bucket_allocno_compare_func): Add a preference based on
8665 conflict_allocno_hard_prefs.
8666 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8667 (color_allocnos): Remove a dead code. Initiate
8668 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8670 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8673 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8674 constraint from =wa to wa. Avoid a subreg on the output operand,
8675 instead use a pseudo and subreg it in a move.
8676 (p9_xxbrd_<mode>): Changed to ...
8677 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8678 (p9_xxbrd_v2df): New expander.
8679 (p9_xxbrw_<mode>): Changed to ...
8680 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8681 (p9_xxbrw_v4sf): New expander.
8683 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8685 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8687 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8690 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8691 multiply in 32-bit mode.
8692 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8693 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8696 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8698 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8699 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8700 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8701 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8703 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8706 * dwarf2out.c (is_trivial_indirect_ref): New function.
8707 (dwarf2out_late_global_decl): Do not generate a location
8708 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8709 and that are not defined in the current unit.
8711 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8713 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8714 instead of a libcall for UNORDERED.
8716 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8719 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8720 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8722 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8725 * config/s390/s390.c (s390_set_current_function): Invoke
8726 s390_indirect_branch_settings also if fndecl didn't change.
8728 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8730 * config/rs6000/rs6000.md (blockage): Set length to zero.
8732 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8734 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8736 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8739 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8740 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8744 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8745 PARALLEL incoming that failed vt_get_decl_and_offset check.
8748 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8749 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8751 (get_variable_section): For decls in named .bss* sections pass true as
8752 second argument to bss_initializer_p.
8754 2018-02-09 Marek Polacek <polacek@redhat.com>
8755 Jakub Jelinek <jakub@redhat.com>
8758 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8759 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8760 Sync some changes from cxx_fold_indirect_ref.
8762 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8764 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8766 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8767 (BLOCK_INLINE_ENTRY_LABEL): New.
8768 (dwarf2out_var_location): Disregard inline entry markers.
8769 (inline_entry_data): New struct.
8770 (inline_entry_data_hasher): New hashtable type.
8771 (inline_entry_data_hasher::hash): New.
8772 (inline_entry_data_hasher::equal): New.
8773 (inline_entry_data_table): New variable.
8774 (add_high_low_attributes): Add DW_AT_entry_pc and
8775 DW_AT_GNU_entry_view attributes if a pending entry is found
8776 in inline_entry_data_table. Add old entry_pc attribute only
8777 if debug nonbinding markers are disabled.
8778 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8779 markers are enabled.
8780 (block_within_block_p, dwarf2out_inline_entry): New.
8781 (dwarf2out_finish): Check that no entries remained in
8782 inline_entry_data_table.
8783 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8784 (final_scan_insn, notice_source_line): Likewise.
8785 (rest_of_clean_state): Skip inline entry markers.
8786 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8788 * gimple.c (gimple_build_debug_inline_entry): New.
8789 * gimple.h (enum gimple_debug_subcode): Add
8790 GIMPLE_DEBUG_INLINE_ENTRY.
8791 (gimple_build_debug_inline_entry): Declare.
8792 (gimple_debug_inline_entry_p): New.
8793 (gimple_debug_nonbind_marker_p): Adjust.
8794 * insn-notes.def (INLINE_ENTRY): New.
8795 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8796 inline entry marker notes.
8797 (print_insn): Likewise.
8798 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8799 (INSN_DEBUG_MARKER_KIND): Likewise.
8800 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8801 * tree-inline.c (expand_call_inline): Build and insert
8802 debug_inline_entry stmt.
8803 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8804 inline entry blocks early, if nonbind markers are enabled.
8805 (dump_scope_block): Dump fragment info.
8806 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8807 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8808 (gimple_build_debug_inline_entry): New.
8809 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8810 Enable/disable inline entry points too.
8811 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8812 (DEBUG_INSN): Describe inline entry markers.
8814 * common.opt (gvariable-location-views): New.
8815 (gvariable-location-views=incompat5): New.
8816 * config.in: Rebuilt.
8817 * configure: Rebuilt.
8818 * configure.ac: Test assembler for view support.
8819 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8820 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8821 * dwarf2out.c (var_loc_view): New typedef.
8822 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8823 (dwarf2out_locviews_in_attribute): New.
8824 (dwarf2out_locviews_in_loclist): New.
8825 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8826 (enum dw_line_info_opcode): Add LI_adv_address.
8827 (struct dw_line_info_table): Add view.
8828 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8829 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8830 (zero_view_p): New variable.
8831 (ZERO_VIEW_P): New macro.
8832 (output_asm_line_debug_info): New.
8833 (struct var_loc_node): Add view.
8834 (add_AT_view_list, AT_loc_list): New.
8835 (add_var_loc_to_decl): Add view param. Test it against last.
8836 (new_loc_list): Add view params. Record them.
8837 (AT_loc_list_ptr): Handle loc and view lists.
8838 (view_list_to_loc_list_val_node): New.
8839 (print_dw_val): Handle dw_val_class_view_list.
8840 (size_of_die): Likewise.
8841 (value_format): Likewise.
8842 (loc_list_has_views): New.
8843 (gen_llsym): Set vl_symbol too.
8844 (maybe_gen_llsym, skip_loc_list_entry): New.
8845 (dwarf2out_maybe_output_loclist_view_pair): New.
8846 (output_loc_list): Output view list or entries too.
8847 (output_view_list_offset): New.
8848 (output_die): Handle dw_val_class_view_list.
8849 (output_dwarf_version): New.
8850 (output_compilation_unit_header): Use it.
8851 (output_skeleton_debug_sections): Likewise.
8852 (output_rnglists, output_line_info): Likewise.
8853 (output_pubnames, output_aranges): Update version comments.
8854 (output_one_line_info_table): Output view numbers in asm comments.
8855 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8856 Call maybe_gen_llsym.
8857 (loc_list_from_tree_1): Adjust.
8858 (add_AT_location_description): Create view list attribute if
8859 needed, check it's absent otherwise.
8860 (convert_cfa_to_fb_loc_list): Adjust.
8861 (maybe_emit_file): Call output_asm_line_debug_info for test.
8862 (dwarf2out_var_location): Reset views as needed. Precompute
8863 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8864 attribute. Set view.
8865 (new_line_info_table): Reset next view.
8866 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8867 (dwarf2out_source_line): Likewise. Output view resets and labels to
8868 the assembler, or select appropriate line info opcodes.
8869 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8870 (optimize_string_length): Catch it. Adjust.
8871 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8872 dw_val_class_view_list, and remove it if no longer needed.
8873 (hash_loc_list): Hash view numbers.
8874 (loc_list_hasher::equal): Compare them.
8875 (optimize_location_lists): Check whether a view list symbol is
8876 needed, and whether the locview attribute is present, and
8877 whether they match. Remove the locview attribute if no longer
8879 (index_location_lists): Call skip_loc_list_entry for test.
8880 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8881 Use output_dwarf_version.
8882 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8883 (struct dw_val_node): Add val_view_list.
8884 * final.c (SEEN_NEXT_VIEW): New.
8885 (set_next_view_needed): New.
8886 (clear_next_view_needed): New.
8887 (maybe_output_next_view): New.
8888 (final_start_function): Rename to...
8889 (final_start_function_1): ... this. Take pointer to FIRST,
8890 add SEEN parameter. Emit param bindings in the initial view.
8891 (final_start_function): Reintroduce SEEN-less interface.
8892 (final): Rename to...
8893 (final_1): ... this. Take SEEN parameter. Output final pending
8894 next view at the end.
8895 (final): Reintroduce seen-less interface.
8896 (final_scan_insn): Output pending next view before switching
8897 sections or ending a block. Mark the next view as needed when
8898 outputting variable locations. Notify debug backend of section
8899 changes, and of location view changes.
8900 (rest_of_handle_final): Adjust.
8901 * toplev.c (process_options): Autodetect value for debug variable
8902 location views option. Warn on incompat5 without -gdwarf-5.
8903 * doc/invoke.texi (gvariable-location-views): New.
8904 (gvariable-location-views=incompat5): New.
8905 (gno-variable-location-views): New.
8907 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8909 PR tree-optimization/84136
8910 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8911 that the result of find_edge is non-NULL.
8913 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8916 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8917 storing integer register in SImode. Fix cost of 256 and 512
8918 byte aligned SSE register store.
8920 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8922 * config/i386/i386.c (ix86_multiplication_cost): Fix
8923 multiplication cost for TARGET_AVX512DQ.
8925 2018-02-08 Marek Polacek <polacek@redhat.com>
8927 PR tree-optimization/84238
8928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8931 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8933 PR tree-optimization/84265
8934 * tree-vect-stmts.c (vectorizable_store): Don't treat
8935 VMAT_CONTIGUOUS accesses as grouped.
8936 (vectorizable_load): Likewise.
8938 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8940 PR tree-optimization/81635
8941 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8942 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8943 (test_round_for_mask): New functions.
8944 (wide_int_cc_tests): Call test_round_for_mask.
8945 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8946 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8947 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8948 range returned by get_range_info.
8950 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8953 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8954 * symtab.c: Include builtins.h
8955 (symtab_node::output_to_lto_symbol_table_p): Move here
8956 from lto-streamer-out.c:output_symbol_p.
8957 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8958 (output_symbol_p): Move all logic to symtab.c
8959 (produce_symtab): Update.
8961 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8963 * config/s390/s390-opts.h (enum indirect_branch): Define.
8964 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8965 (s390_indirect_branch_via_thunk)
8966 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8967 (enum s390_indirect_branch_type): Define.
8968 * config/s390/s390.c (struct s390_frame_layout, struct
8969 machine_function): Remove.
8970 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8971 (indirect_branch_table_label_no, indirect_branch_table_name):
8973 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8974 (enum s390_indirect_branch_option): Define.
8975 (s390_return_addr_from_memory): New function.
8976 (s390_handle_string_attribute): New function.
8977 (s390_attribute_table): Add new attribute handler.
8978 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8979 (s390_indirect_branch_via_thunk): New function.
8980 (s390_indirect_branch_via_inline_thunk): New function.
8981 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8982 sibling call optimization for non z10 compiles.
8983 (s390_emit_call): Force indirect branch target to be a single
8984 register. Add r1 clobber for non-z10 compiles.
8985 (s390_emit_epilogue): Emit return jump via return_use expander.
8986 (s390_reorg): Handle JUMP_INSNs as execute targets.
8987 (s390_option_override_internal): Perform validity checks for the
8988 new command line options.
8989 (s390_indirect_branch_attrvalue): New function.
8990 (s390_indirect_branch_settings): New function.
8991 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8992 (s390_output_indirect_thunk_function): New function.
8993 (s390_code_end): Implement target hook.
8994 (s390_case_values_threshold): Implement target hook.
8995 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8997 * config/s390/s390.h (struct s390_frame_layout)
8998 (struct machine_function): Move here from s390.c.
8999 (TARGET_INDIRECT_BRANCH_NOBP_RET)
9000 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
9001 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
9002 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
9003 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
9004 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
9005 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
9006 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
9007 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
9008 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
9009 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
9010 (mnemonic attribute): Add values which aren't recognized
9012 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
9013 pattern for branch conversion. Fix mnemonic attribute.
9014 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
9015 indirect branch via thunk if requested.
9016 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
9017 ("*indirect_jump"): Disable for branch conversion using out of
9019 ("indirect_jump_via_thunk<mode>_z10")
9020 ("indirect_jump_via_thunk<mode>")
9021 ("indirect_jump_via_inlinethunk<mode>_z10")
9022 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
9023 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
9024 ("casesi_jump_via_inlinethunk<mode>_z10")
9025 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
9026 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
9027 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
9028 ("*indirect2_jump"): Disable for branch conversion.
9029 ("casesi_jump"): Turn into expander and expand patterns for branch
9031 ("return_use"): New expander.
9032 ("*return"): Emit return via thunk and rename it to ...
9033 ("*return<mode>"): ... this one.
9034 * config/s390/s390.opt: Add new options and and enum for the
9037 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9039 * lra-constraints.c (match_reload): Unconditionally use
9040 gen_lowpart_SUBREG, rather than selecting between that
9041 and equivalent gen_rtx_SUBREG code.
9043 2018-02-08 Richard Biener <rguenther@suse.de>
9045 PR tree-optimization/84233
9046 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9047 changed flag instead of boguously re-using phi_inserted.
9049 2018-02-08 Martin Jambor <mjambor@suse.cz>
9051 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9052 static local variables.
9054 2018-02-08 Richard Biener <rguenther@suse.de>
9056 PR tree-optimization/84278
9057 * tree-vect-stmts.c (vectorizable_store): When looking for
9058 smaller vector types to perform grouped strided loads/stores
9059 make sure the mode is supported by the target.
9060 (vectorizable_load): Likewise.
9062 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9064 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9065 Increase LDP/STP opportunities by adding adjacent callee-saves.
9067 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9069 PR rtl-optimization/84068
9070 PR rtl-optimization/83459
9071 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9073 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9075 PR tree-optimization/84224
9076 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9077 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9080 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9083 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9084 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9085 count, remove one USE.
9087 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9089 * doc/install.texi (Configuration): Document the
9090 --with-long-double-format={ibm,ieee} PowerPC configuration
9094 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9095 Convert from define_expand to be define_insn_and_split. Rework
9096 float/double/_Float128 conversions to QI/HI/SImode to work with
9097 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9098 conversions to QI/HImode types did a store and then a load to
9099 truncate the value. For conversions to VSX registers, don't split
9100 the insn, instead emit the code directly. Use the code iterator
9101 any_fix to combine signed and unsigned conversions.
9102 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9103 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9104 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9105 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9106 (fix_<mode>di2_hw): Likewise.
9107 (fixuns_<mode>di2_hw): Likewise.
9108 (fix_<mode>si2_hw): Likewise.
9109 (fixuns_<mode>si2_hw): Likewise.
9110 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9111 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9112 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9113 fix<uns>_trunc<SFDF:mode>si2_p8.
9114 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
9115 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9116 (fix<uns>_<mode>_mem): Likewise.
9117 (fctiw<u>z_<mode>_mem): Likewise.
9118 (fix<uns>_<mode>_mem): Likewise.
9119 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9120 the register allocator from doing a direct move to the GPRs to do
9121 a store, and instead use the ISA 3.0 store byte/half-word from
9122 vector register instruction. For IEEE 128-bit floating point,
9123 also optimize stores of 32-bit ints.
9124 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9126 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9128 * genextract.c (push_pathstr_operand): New function to support
9130 (walk_rtx): Call push_pathstr_operand.
9131 (print_path): Support [a-zA-Z].
9133 2018-02-07 Richard Biener <rguenther@suse.de>
9135 PR tree-optimization/84037
9136 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9137 (cse_and_gimplify_to_preheader): Declare.
9138 (vect_get_place_in_interleaving_chain): Likewise.
9139 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9141 (_loop_vec_info::~_loop_vec_info): Delete it.
9142 (cse_and_gimplify_to_preheader): New function.
9143 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9144 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9145 (vectorizable_load): Likewise. For grouped stores always base
9146 the IV on the first element.
9147 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9148 condition before gimplifying.
9150 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9152 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9153 *DIV_EXPR and *MOD_EXPR.
9155 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9158 * config/i386/i386.c (ix86_option_override_internal): Mask out
9159 the CF_SET bit when checking -fcf-protection.
9161 2018-02-07 Tom de Vries <tom@codesourcery.com>
9164 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9167 2018-02-07 Richard Biener <rguenther@suse.de>
9169 PR tree-optimization/84204
9170 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9173 PR tree-optimization/84205
9174 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9175 special-case isl_ast_op_zdiv_r.
9177 PR tree-optimization/84223
9178 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9179 Only add conditions from within the region.
9180 (gather_bbs::after_dom_children): Adjust.
9182 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9185 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9186 * config/avr/avr.md: Only post-reload split REG-REG moves if
9187 either register is GENERAL_REG_P.
9189 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9191 PR tree-optimization/84235
9192 * tree-ssa-scopedtables.c
9193 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9194 if the subtraction is performed in floating point type where NaNs are
9195 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9196 build 1. Formatting fix.
9198 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9201 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9202 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9203 and skip it regardless of bb boundaries. Use CALL_P macro,
9204 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9205 unnecessarily, formatting fix.
9207 2018-02-06 Michael Collison <michael.collison@arm.com>
9209 * config/arm/thumb2.md:
9210 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9211 (*thumb_mov_notscc): Ditto.
9213 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9216 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9217 unsigned_fix, not "s".
9219 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9221 * configure.ac (gcc_fn_eh_frame_ro): New function.
9222 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9223 correct .eh_frame permissions.
9224 * configure: Regenerate.
9226 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9228 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9231 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9233 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9234 Display warning message for -mno-speculate-indirect-jumps.
9236 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9238 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9240 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9242 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9244 PR tree-optimization/84225
9245 * tree-eh.c (find_trapping_overflow): Only call
9246 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9248 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9251 * config/i386/i386.c: Reimplement the check of possible options
9252 -mibt/-mshstk conbination. Change error messages.
9253 * doc/invoke.texi: Fix a typo: remove extra '='.
9255 2018-02-06 Marek Polacek <polacek@redhat.com>
9257 PR tree-optimization/84228
9258 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9260 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9263 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9264 emitted arch directives.
9265 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9266 __ARM_FEATURE_COPROC before changing architectures.
9268 2018-02-06 Richard Biener <rguenther@suse.de>
9270 * config/i386/i386.c (print_reg): Fix typo.
9271 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9273 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9275 * configure: Regenerate.
9277 2018-02-05 Martin Sebor <msebor@redhat.com>
9279 PR tree-optimization/83369
9280 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9283 2018-02-05 Martin Liska <mliska@suse.cz>
9285 * doc/invoke.texi: Cherry-pick upstream r323995.
9287 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9289 * ira.c (ira_init_register_move_cost): Adjust comment.
9291 2018-02-05 Martin Liska <mliska@suse.cz>
9293 PR gcov-profile/84137
9294 * doc/gcov.texi: Fix typo in documentation.
9296 2018-02-05 Martin Liska <mliska@suse.cz>
9298 PR gcov-profile/83879
9299 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9300 using dlopen functionality.
9302 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9304 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9305 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9306 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9307 _mm_maskz_range_round_ss): New intrinsics.
9308 (__builtin_ia32_rangesd128_round)
9309 (__builtin_ia32_rangess128_round): Remove.
9310 (__builtin_ia32_rangesd128_mask_round,
9311 __builtin_ia32_rangess128_mask_round): New builtins.
9312 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9313 __builtin_ia32_rangess128_round): Remove.
9314 (__builtin_ia32_rangesd128_mask_round,
9315 __builtin_ia32_rangess128_mask_round): New builtins.
9316 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9317 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9318 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9319 "<round_saeonly_constraint>")): Changed to ...
9320 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9321 "<round_saeonly_scalar_constraint>")): ... this.
9322 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9323 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9324 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9325 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9326 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9328 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9330 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9332 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9333 Remove all values except native, 8540 and 8548.
9335 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9337 * config/i386/i386.c (ix86_output_function_return): Pass
9338 INVALID_REGNUM, instead of -1, as invalid register number to
9339 indirect_thunk_name and output_indirect_thunk.
9341 2018-02-02 Julia Koval <julia.koval@intel.com>
9343 * config.gcc: Add -march=icelake.
9344 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9345 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9346 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9347 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9348 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9349 (processor_target_table): Add icelake.
9350 (ix86_option_override_internal): Handle new PTAs.
9351 (get_builtin_code_for_version): Handle icelake.
9352 (M_INTEL_COREI7_ICELAKE): New.
9353 (fold_builtin_cpu): Handle icelake.
9354 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9355 * doc/invoke.texi: Add -march=icelake.
9357 2018-02-02 Julia Koval <julia.koval@intel.com>
9359 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9360 to wide_int_bitmask.
9361 * wide-int-bitmask.h: New.
9363 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9366 * config/i386/i386.md: Replace Pmode with word_mode in
9367 builtin_setjmp_setup and builtin_longjmp to support x32.
9369 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9373 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9375 (rs6000_supported_cpu_names): New static variable.
9376 (linux_cpu_translation_table): Likewise.
9377 (elf_platform) <cpu>: Define new static variable and use it.
9378 Translate kernel AT_PLATFORM name to canonical name if needed.
9379 Error if platform name is unknown.
9381 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9384 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9386 2018-02-01 Jeff Law <law@redhat.com>
9389 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9390 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9391 the scratch if RELEASE_VIA_POP is false.
9392 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9393 If we have to save a temporary register, decrement SIZE appropriately.
9394 Pass new arguments to release_scratch_register_on_entry.
9395 (ix86_adjust_stack_and_probe): Likewise.
9396 (ix86_emit_probe_stack_range): Pass new arguments to
9397 release_scratch_register_on_entry.
9399 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9401 PR rtl-optimization/84157
9402 * combine.c (change_zero_ext): Use REG_P predicate in
9403 front of HARD_REGISTER_P predicate.
9405 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9407 * config/avr/avr.c (avr_option_override): Move disabling of
9408 -fdelete-null-pointer-checks to...
9409 * common/config/avr/avr-common.c (avr_option_optimization_table):
9412 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9414 PR tree-optimization/81635
9415 * tree-data-ref.c (split_constant_offset_1): For types that
9416 wrap on overflow, try to use range info to prove that wrapping
9419 2018-02-01 Renlin Li <renlin.li@arm.com>
9422 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9424 (aarch64_register_move_cost): Likewise.
9425 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9427 (REG_CLASS_NAMES): Likewise.
9428 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9429 TAILCALL_ADDR_REGS. Remove IP registers.
9430 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9432 2018-02-01 Richard Biener <rguenther@suse.de>
9434 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9435 for specifying RPO order and allow NULL for that.
9436 * domwalk.c (dom_walker::dom_walker): Likewise.
9437 (dom_walker::walk): Handle NULL RPO order.
9438 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9440 (rewrite_update_dom_walker): Likewise.
9441 (mark_def_dom_walker): Likewise.
9443 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9445 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9446 (aarch64_maybe_expand_sve_subreg_move): Declare.
9447 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9448 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9450 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9451 that are semantically a reverse operation.
9452 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9453 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9454 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9456 (aarch64_can_change_mode_class): For big-endian, forbid changes
9457 between two SVE modes if they have different element sizes.
9459 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9461 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9462 the TImode handling for big-endian targets.
9464 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9466 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9467 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9469 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9470 Remove BSWAP handing for big-endian targets and use the form of
9471 LD1RQ appropariate for the mode.
9473 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9475 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9476 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9479 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9482 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9483 check for operands that need to go through aarch64_sve_reload_be.
9485 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9487 PR tree-optimization/81661
9488 PR tree-optimization/84117
9489 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9490 * tree-eh.c: Include gimplify.h.
9491 (find_trapping_overflow, replace_trapping_overflow,
9492 rewrite_to_non_trapping_overflow): New functions.
9493 * tree-vect-loop.c: Include tree-eh.h.
9494 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9495 * tree-data-ref.c: Include tree-eh.h.
9496 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9498 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9500 PR rtl-optimization/84123
9501 * combine.c (change_zero_ext): Check if hard register satisfies
9502 can_change_dest_mode before calling gen_lowpart_SUBREG.
9504 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9507 * ira.c (ira_init_register_move_cost): Remove assert.
9509 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9511 PR rtl-optimization/84071
9512 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9513 * doc/tm.texi: Regenerate.
9515 2018-01-31 Richard Biener <rguenther@suse.de>
9517 PR tree-optimization/84132
9518 * tree-data-ref.c (analyze_miv_subscript): Properly
9519 check whether evolution_function_is_affine_multivariate_p
9520 before calling gcd_of_steps_may_divide_p.
9522 2018-01-31 Julia Koval <julia.koval@intel.com>
9525 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9526 * config/i386/i386.md (rdpid_rex64) New.
9527 (rdpid): Make 32bit only.
9529 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9532 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9533 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9535 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9538 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9540 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9542 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9544 PR rtl-optimization/84071
9545 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9546 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9548 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9550 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9551 (arc_attribute_table): Add 'aux' attribute.
9552 (arc_in_small_data_p): Consider aux like variables.
9553 (arc_is_aux_reg_p): New function.
9554 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9555 (arc_get_aux_arg): New function.
9556 (prepare_move_operands): Handle aux-register access.
9557 (arc_handle_aux_attribute): New function.
9558 * doc/extend.texi (ARC Variable attributes): Add subsection.
9560 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9562 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9563 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9564 (arc_attribute_table): Add 'uncached' attribute.
9565 (arc_print_operand): Print '.di' flag for uncached memory
9567 (arc_in_small_data_p): Do not consider for small data the uncached
9569 (arc_is_uncached_mem_p): New function.
9570 * config/arc/predicates.md (compact_store_memory_operand): Check
9571 for uncached memory accesses.
9572 (nonvol_nonimm_operand): Likewise.
9573 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9575 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9578 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9579 falign-loops=): Add Optimization flag.
9581 2018-01-30 Jeff Law <law@redhat.com>
9584 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9585 INT_REGISTERS_SAVED. Check it prior to calling
9586 get_scratch_register_on_entry.
9587 (ix86_adjust_stack_and_probe): Similarly.
9588 (ix86_emit_probe_stack_range): Similarly.
9589 (ix86_expand_prologue): Corresponding changes.
9591 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9594 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9595 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9597 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9600 * lra-constraints.c (curr_insn_transform): Process AND in the
9603 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9605 PR rtl-optimization/83986
9606 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9607 dependence against last_pending_memory_flush in addition to
9610 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9612 PR tree-optimization/81611
9613 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9616 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9619 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9622 2018-01-30 Richard Biener <rguenther@suse.de>
9623 Jakub Jelinek <jakub@redhat.com>
9625 PR tree-optimization/84111
9626 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9627 inner loops added during recursion, as they don't have up-to-date
9630 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9633 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9634 (can_inline_edge_by_limits_p): ... here.
9635 (can_early_inline_edge_p, check_callers,
9636 update_caller_keys, update_callee_keys, recursive_inlining,
9637 add_new_edges_to_heap, speculation_useful_p,
9638 inline_small_functions,
9639 inline_small_functions, flatten_function,
9640 inline_to_all_callers_1): Update.
9642 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9644 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9647 2018-01-30 Richard Biener <rguenther@suse.de>
9649 PR tree-optimization/83008
9650 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9651 invariant and constant vector uses in stmts when they need
9654 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9657 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9658 * configure: Regenerate.
9660 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9662 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9664 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9665 Use gen_rtx_REG rather than gen_lowpart.
9667 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9669 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9670 rather than 0 when creating partial subregs.
9672 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9674 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9677 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9680 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9681 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9682 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9683 flags. This restores the settings used before the 2017-07-24.
9684 Turning off pre increment/decrement/modify allows IVOPTS to
9685 optimize DF/SF loops where the index is an int.
9687 2018-01-29 Richard Biener <rguenther@suse.de>
9688 Kelvin Nilsen <kelvin@gcc.gnu.org>
9691 * tree-vect-stmts.c (vectorizable_call): Don't call
9692 targetm.vectorize_builtin_md_vectorized_function if callee is
9695 2018-01-22 Carl Love <cel@us.ibm.com>
9697 * doc/extend.tex: Fix typo in second arg in
9698 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9700 2018-01-29 Richard Biener <rguenther@suse.de>
9702 PR tree-optimization/84086
9703 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9704 (flush_ssaname_freelist): When SSA names were released reset
9705 the SCEV hash table.
9707 2018-01-29 Richard Biener <rguenther@suse.de>
9709 PR tree-optimization/84057
9710 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9711 removed paths when removing edges.
9713 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9715 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9716 -mfunction-return=@var{choice}.
9718 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9721 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9723 (layout::print_source_line): Likewise.
9724 (test_get_line_width_without_trailing_whitespace): Add test cases.
9726 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9729 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9732 2018-01-26 Jim Wilson <jimw@sifive.com>
9734 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9736 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9739 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9741 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9742 and CMP + SUB-immediate -> SUBS.
9744 2018-01-26 Martin Sebor <msebor@redhat.com>
9746 PR tree-optimization/83896
9747 * tree-ssa-strlen.c (get_string_len): Rename...
9748 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9749 Avoid assuming length is constant.
9750 (handle_char_store): Use HOST_WIDE_INT for string length.
9752 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9755 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9756 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9758 2018-01-26 Richard Biener <rguenther@suse.de>
9760 PR rtl-optimization/84003
9761 * dse.c (record_store): Only record redundant stores when
9762 the earlier store aliases at least all accesses the later one does.
9764 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9766 PR rtl-optimization/83985
9767 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9768 REG_CFA_RESTORE insns.
9769 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9770 REG_CFA_RESTORE insns here.
9773 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9774 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9776 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9778 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9779 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9780 (arc_init): Likewise.
9781 (arc_override_options): Likewise.
9782 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9784 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9786 * config/arc/arc.h (TARGET_DBNZ): Define.
9787 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9788 properly set the tune attribute.
9789 (dbnz): Use TARGET_DBNZ guard.
9790 * config/arc/arc.opt (mtune): Add core3 option.
9792 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9794 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9795 recognize new pic like addresses.
9796 (arc_delegitimize_address): Clean up.
9798 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9800 * config/arc/arc-arches.def: Option mrf16 valid for all
9802 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9803 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9804 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9805 * config/arc/arc-tables.opt: Regenerate.
9806 * config/arc/arc.c (arc_conditional_register_usage): Handle
9807 reduced register file case.
9808 (arc_file_start): Set must have build attributes.
9809 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9811 * config/arc/arc.opt (mrf16): Add new option.
9812 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9813 * config/arc/genmultilib.awk: Handle new mrf16 option.
9814 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9815 * config/arc/t-multilib: Regenerate.
9816 * doc/invoke.texi (ARC Options): Document mrf16 option.
9818 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9820 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9821 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9822 (arc_attribute_table): Add 'secure_call' attribute.
9823 (arc_print_operand): Print secure call operand.
9824 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9826 (arc_is_secure_call_p): New function. * config/arc/arc.md
9827 (call_i): Add support for sjli instruction.
9828 (call_value_i): Likewise.
9829 * config/arc/constraints.md (Csc): New constraint.
9831 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9832 John Eric Martin <John.Martin@emmicro-us.com>
9834 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9835 * config/arc/arc.c (_arc_jli_section): New struct.
9836 (arc_jli_section): New type.
9837 (rc_jli_sections): New static variable.
9838 (arc_handle_jli_attribute): New function.
9839 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9840 (arc_file_end): New function.
9841 (TARGET_ASM_FILE_END): Define.
9842 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9843 (arc_add_jli_section): New function.
9844 (jli_call_scan): Likewise.
9845 (arc_reorg): Call jli_call_scan.
9846 (arc_output_addsi): Remove 'S' from printing asm operand.
9847 (arc_is_jli_call_p): New function.
9848 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9850 (movhi_insn): Likewise.
9851 (movsi_insn): Likewise.
9852 (movsi_set_cc_insn): Likewise.
9853 (loadqi_update): Likewise.
9854 (load_zeroextendqisi_update): Likewise.
9855 (load_signextendqisi_update): Likewise.
9856 (loadhi_update): Likewise.
9857 (load_zeroextendhisi_update): Likewise.
9858 (load_signextendhisi_update): Likewise.
9859 (loadsi_update): Likewise.
9860 (loadsf_update): Likewise.
9861 (movsicc_insn): Likewise.
9862 (bset_insn): Likewise.
9863 (bxor_insn): Likewise.
9864 (bclr_insn): Likewise.
9865 (bmsk_insn): Likewise.
9866 (bicsi3_insn): Likewise.
9867 (cmpsi_cc_c_insn): Likewise.
9868 (movsi_ne): Likewise.
9869 (movsi_cond_exec): Likewise.
9870 (clrsbsi2): Likewise.
9881 (call_i): Remove 'S' asm letter, add jli instruction.
9882 (call_value_i): Likewise.
9883 * config/arc/arc.op (mjli-always): New option.
9884 * config/arc/constraints.md (Cji): New constraint.
9885 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9887 (subsf3_fpx): Likewise.
9888 (mulsf3_fpx): Likewise.
9889 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9891 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9893 * doc/invoke.texi (ARC): Document mjli-always option.
9895 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9897 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9898 avoid addition with 0 and use incw and decw where possible.
9900 2018-01-26 Richard Biener <rguenther@suse.de>
9902 PR tree-optimization/81082
9903 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9904 association if it requires casting to unsigned.
9905 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9906 from fold_plusminus_mult_expr to catch important cases late when
9907 range info is available.
9909 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9911 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9912 * configure.ac (hidden_linkonce): New test.
9913 * configure: Regenerate.
9914 * config.in: Regenerate.
9916 2018-01-26 Julia Koval <julia.koval@intel.com>
9918 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9919 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9920 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9921 _mm_mask_bitshuffle_epi64_mask): Fix type.
9922 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9923 USI_FTYPE_V4DI_V4DI_USI): Remove.
9924 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9925 __builtin_ia32_vpshufbitqmb256_mask,
9926 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9927 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9928 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9930 2018-01-26 Alan Modra <amodra@gmail.com>
9933 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9934 UNSPEC_VBPERMQ. Sort other unspecs.
9936 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9938 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9940 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9943 * predict.c (drop_profile): Do not push/pop cfun; update also
9945 (handle_missing_profiles): Fix logic looking for zero profiles.
9947 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9950 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9951 on functions with #pragma omp declare simd or functions with simd
9953 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9954 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9955 Remove trailing \n from warning_at calls.
9957 2018-01-25 Tom de Vries <tom@codesourcery.com>
9960 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9961 for neutered workers.
9963 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9966 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9967 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9969 2018-01-24 Jeff Law <law@redhat.com>
9972 * i386.c (get_probe_interval): Move to earlier point.
9973 (ix86_compute_frame_layout): If -fstack-clash-protection and
9974 the frame is larger than the probe interval, then use pushes
9975 to save registers rather than reg->mem moves.
9976 (ix86_expand_prologue): Remove conditional for int_registers_saved
9979 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9982 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9983 min/max for never referenced object.
9985 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9988 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9990 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9991 attributes from DECL_ATTRIBUTES (decl) without affecting
9992 DECL_ATTRIBUTES (current_function_decl).
9993 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9994 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9996 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9998 PR tree-optimization/83979
9999 * fold-const.c (fold_comparison): Use constant_boolean_node
10000 instead of boolean_{true,false}_node.
10002 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
10004 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
10007 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10009 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10010 Simplify the clause that sets the length attribute.
10011 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10012 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
10013 clause that sets the length attribute.
10014 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10016 2018-01-24 Tom de Vries <tom@codesourcery.com>
10019 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
10020 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
10021 Add strict parameter.
10022 (prevent_branch_around_nothing): Insert dummy insn between branch to
10023 label and label with no ptx insn inbetween.
10024 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
10026 2018-01-24 Tom de Vries <tom@codesourcery.com>
10029 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
10030 for neutered threads in warp.
10031 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
10033 2018-01-24 Richard Biener <rguenther@suse.de>
10035 PR tree-optimization/83176
10036 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10039 2018-01-24 Richard Biener <rguenther@suse.de>
10041 PR tree-optimization/82819
10042 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10043 code generating pluses that are no-ops in the target precision.
10045 2018-01-24 Richard Biener <rguenther@suse.de>
10047 PR middle-end/84000
10048 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10050 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10052 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10053 to merge probabilities.
10054 * predict.c (probably_never_executed): Also mark as cold functions
10055 with global 0 profile and guessed local profile.
10056 * profile-count.c (profile_probability::combine_with_count): New
10058 * profile-count.h (profile_probability::operator*,
10059 profile_probability::operator*=, profile_probability::operator/,
10060 profile_probability::operator/=): Reduce precision to adjusted
10061 and set value to guessed on contradictory divisions.
10062 (profile_probability::combine_with_freq): Remove.
10063 (profile_probability::combine_wiht_count): Declare.
10064 (profile_count::force_nonzero):: Set to adjusted.
10065 (profile_count::probability_in):: Set quality to adjusted.
10066 * tree-ssa-tail-merge.c (replace_block_by): Use
10067 combine_with_count.
10069 2018-01-23 Andrew Waterman <andrew@sifive.com>
10070 Jim Wilson <jimw@sifive.com>
10072 * config/riscv/riscv.c (riscv_stack_boundary): New.
10073 (riscv_option_override): Set riscv_stack_boundary. Handle
10074 riscv_preferred_stack_boundary_arg.
10075 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10076 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10077 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10078 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10079 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10080 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10082 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10085 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10086 of struct ix86_frame.
10087 (ix86_expand_epilogue): Likewise. Add a local variable for
10088 the reg_save_offset field in struct ix86_frame.
10090 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10092 PR tree-optimization/82604
10093 * tree-loop-distribution.c (enum partition_kind): New enum item
10094 PKIND_PARTIAL_MEMSET.
10095 (partition_builtin_p): Support above new enum item.
10096 (generate_code_for_partition): Ditto.
10097 (compute_access_range): Differentiate cases that equality can be
10098 proven at all loops, the innermost loops or no loops.
10099 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10100 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10101 (finalize_partitions, distribute_loop): Don't fuse partition of
10102 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10103 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10104 parloop is enabled.
10106 2018-01-23 Martin Liska <mliska@suse.cz>
10108 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10109 order to ignore the predictor.
10110 (PRED_POLYMORPHIC_CALL): Likewise.
10111 (PRED_RECURSIVE_CALL): Likewise.
10113 2018-01-23 Martin Liska <mliska@suse.cz>
10115 * tree-profile.c (tree_profiling): Print function header to
10116 aware reader which function we are working on.
10117 * value-prof.c (gimple_find_values_to_profile): Do not print
10118 not interesting value histograms.
10120 2018-01-23 Martin Liska <mliska@suse.cz>
10122 * profile-count.h (enum profile_quality): Add
10123 profile_uninitialized as the first value. Do not number values
10124 as they are zero based.
10125 (profile_count::verify): Update sanity check.
10126 (profile_probability::verify): Likewise.
10128 2018-01-23 Nathan Sidwell <nathan@acm.org>
10130 * doc/invoke.texi (ffor-scope): Deprecate.
10132 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10134 PR tree-optimization/83510
10135 * domwalk.c (set_all_edges_as_executable): New function.
10136 (dom_walker::dom_walker): Convert bool param
10137 "skip_unreachable_blocks" to enum reachability. Move setup of
10138 edge flags to set_all_edges_as_executable and only do it when
10139 reachability is REACHABLE_BLOCKS.
10140 * domwalk.h (enum dom_walker::reachability): New enum.
10141 (dom_walker::dom_walker): Convert bool param
10142 "skip_unreachable_blocks" to enum reachability.
10143 (set_all_edges_as_executable): New decl.
10144 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10145 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10147 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10148 but converting true to REACHABLE_BLOCKS.
10149 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10151 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10152 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10153 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10155 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10156 if check_all_array_refs will be called.
10158 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10160 * tree.c (selftest::test_location_wrappers): Add more test
10163 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10165 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10166 (selftest::test_bit_in_range): Likewise.
10168 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10171 * doc/sourcebuild.texi (vect_float): Say that the selector
10172 only describes the situation when -funsafe-math-optimizations is on.
10173 (vect_float_strict): Document.
10175 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10177 PR tree-optimization/83965
10178 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10179 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10180 instead of checking only for a reduction.
10181 (vect_recog_widen_sum_pattern): Likewise.
10183 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10185 * predict.c (probably_never_executed): Only use precise profile info.
10186 (compute_function_frequency): Skip after inlining hack since we now
10187 have quality checking.
10189 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10191 * profile-count.h (profile_probability::very_unlikely,
10192 profile_probability::unlikely, profile_probability::even): Set
10193 precision to guessed.
10195 2018-01-23 Richard Biener <rguenther@suse.de>
10197 PR tree-optimization/83963
10198 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10199 Properly terminate dominator walk when crossing the exit edge not
10200 when visiting its source block.
10202 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10205 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10206 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10208 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10210 PR tree-optimization/83957
10211 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10212 semicolon after for body surrounded by braces.
10214 PR tree-optimization/83081
10215 * profile-count.h (profile_probability::split): New method.
10216 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10217 Use profile_probability::split.
10218 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10219 when splitting a single conditional jump into 2.
10221 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10223 PR tree-optimization/69452
10224 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10227 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10229 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10230 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10231 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10233 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10235 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
10237 * config/rl78/rl78.md (movdi): New define_expand.
10238 * config/rl78/rl78.c (rl78_split_movdi): New function.
10240 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10243 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10245 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10246 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10247 128-bit to produce an UNSPEC move to get the double word with the
10248 signbit and then a shift directly to do signbit.
10249 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10250 implementation with a new version that just does either a direct
10251 move or a regular move. Move memory interface to separate insns.
10252 Move insns so they are next to the expander.
10253 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10254 with signbit move. Split big and little endian case.
10255 (signbit<mode>2_dm_mem_le): Likewise.
10256 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10257 (signbit<mode>2_dm2): Likewise.
10259 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10261 * config/rl78/rl78.md (anddi3): New define_expand.
10263 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10265 * config/rl78/rl78.md (umindi3): New define_expand.
10267 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10269 * config/rl78/rl78.md (smindi3): New define_expand.
10271 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10273 * config/rl78/rl78.md (smaxdi3): New define_expand.
10275 2018-01-22 Carl Love <cel@us.ibm.com>
10277 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10278 LVX_V1TI): Add macro expansion.
10279 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10280 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10281 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10282 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10283 Change check to determine if the instruction is a byte reversing
10284 entry. Fix typo in comment.
10285 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10286 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10287 Add def_builtin calls for new builtins.
10288 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10289 Add define_insn expansion.
10291 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10293 * config/rl78/rl78.md (umaxdi3): New define_expand.
10295 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10297 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10298 for non-QImode registers.
10300 2018-01-22 Richard Biener <rguenther@suse.de>
10302 PR tree-optimization/83963
10303 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10304 including the loop exit block.
10305 (scop_detection::merge_sese): Likewise.
10306 (scop_detection::add_scop): Do it here instead.
10308 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10310 * doc/sourcebuild.texi (arm_softfloat): Document.
10312 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10315 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10316 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10317 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10319 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10320 David Edelsohn <dje.gcc@gmail.com>
10323 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10324 Change "crset eq" to "crset 2".
10325 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10326 (*call_indirect_aix<mode>_nospec): Likewise.
10327 (*call_value_indirect_aix<mode>_nospec): Likewise.
10328 (*call_indirect_elfv2<mode>_nospec): Likewise.
10329 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10330 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10331 change assembly output from . to $.
10332 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10333 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10334 (*tablejump<mode>_internal1_nospec): Likewise.
10336 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10339 * config/sh/sh_optimize_sett_clrt.cc:
10340 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10342 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10344 PR tree-optimization/83940
10345 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10346 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10347 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10348 use it to pass back the definition type.
10349 (vect_check_store_rhs): Likewise rhs_dt_out.
10350 (vect_build_gather_load_calls): Add a mask_dt argument and use
10351 it instead of a call to vect_is_simple_use.
10352 (vectorizable_store): Update calls to vect_check_load_store_mask
10353 and vect_check_store_rhs. Use the dt returned by the latter instead
10354 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10355 instead of calls to vect_is_simple_use. Pass the scalar rather
10356 than the vector operand to vect_is_simple_use when handling
10357 second and subsequent copies of an rhs value.
10358 (vectorizable_load): Update calls to vect_check_load_store_mask
10359 and vect_build_gather_load_calls. Use the cached mask_dt and
10360 gs_info.offset_dt instead of calls to vect_is_simple_use.
10362 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10364 PR middle-end/83945
10365 * tree-emutls.c: Include gimplify.h.
10366 (lower_emutls_2): New function.
10367 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10368 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10369 it before further processing.
10372 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10373 UINTVAL (trueop1) instead of INTVAL (op1).
10375 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10379 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10380 INCOMING_FRAME_SP_OFFSET if not defined.
10381 (scan_trace): Add ENTRY argument. If true and
10382 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10383 emit a note to adjust the CFA offset.
10384 (create_cfi_notes): Adjust scan_trace callers.
10385 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10386 INCOMING_FRAME_SP_OFFSET in the CIE.
10387 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10388 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10390 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10391 * doc/tm.texi: Regenerated.
10393 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10395 PR rtl-optimization/83147
10396 * lra-constraints.c (remove_inheritance_pseudos): Use
10397 lra_substitute_pseudo_within_insn.
10399 2018-01-19 Tom de Vries <tom@codesourcery.com>
10400 Cesar Philippidis <cesar@codesourcery.com>
10403 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10405 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10408 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10409 spaces for function labels.
10411 2018-01-19 Martin Liska <mliska@suse.cz>
10413 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10414 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10415 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10416 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10417 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10418 (PRED_CONST_RETURN): Change from 69 to 65.
10419 (PRED_NULL_RETURN): Change from 91 to 71.
10420 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10421 (PRED_LOOP_GUARD): Change from 66 to 73.
10423 2018-01-19 Martin Liska <mliska@suse.cz>
10425 * predict.c (predict_insn_def): Add new assert.
10426 (struct branch_predictor): Change type to signed integer.
10427 (test_prediction_value_range): Amend test to cover
10428 PROB_UNINITIALIZED.
10429 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10430 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10431 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10432 (PRED_LOOP_IV_COMPARE): Likewise.
10433 * predict.h (PROB_UNINITIALIZED): Define new constant.
10435 2018-01-19 Martin Liska <mliska@suse.cz>
10437 * predict.c (dump_prediction): Add new format for
10438 analyze_brprob.py script which is enabled with -details
10440 * profile-count.h (precise_p): New function.
10442 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10444 PR tree-optimization/83922
10445 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10446 there are no statements that need masking.
10447 (vect_active_double_reduction_p): New function.
10448 (vect_analyze_loop_operations): Use it when handling phis that
10449 are not in the loop header.
10451 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10453 PR tree-optimization/83914
10454 * tree-vect-loop.c (vectorizable_induction): Don't convert
10455 init_expr or apply the peeling adjustment for inductions
10456 that are nested within the vectorized loop.
10458 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10460 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10463 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10467 * function.h (gimplify_parameters): Add gimple_seq * argument.
10468 * function.c: Include gimple.h and options.h.
10469 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10470 for the added local temporaries if needed.
10471 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10472 if there are any parameter cleanups, wrap whole body into a
10473 try/finally with the cleanups.
10475 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10478 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10479 Use GET_MODE_CLASS for scalar floating point.
10481 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10485 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10486 Fix call of call_cgraph_insertion_hooks.
10488 2018-01-18 Martin Sebor <msebor@redhat.com>
10490 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10492 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10495 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10498 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10501 * common.opt: (-ffile-prefix-map): New option.
10502 * opts.c (common_handle_option): Defer it.
10503 * opts-global.c (handle_common_deferred_options): Handle it.
10504 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10505 * file-prefix-map.h: New file.
10506 (remap_debug_filename, add_debug_prefix_map): ...here.
10507 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10508 * final.c (debug_prefix_map, add_debug_prefix_map
10509 remap_debug_filename): Move to...
10510 * file-prefix-map.c: New file.
10511 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10512 generalize, get rid of alloca(), use strrchr() instead of strchr().
10513 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10514 Implement in terms of add_prefix_map().
10515 (remap_macro_filename, remap_debug_filename): Implement in term of
10517 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10518 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10519 * dbxout.c: Include file-prefix-map.h.
10520 * varasm.c: Likewise.
10521 * vmsdbgout.c: Likewise.
10522 * xcoffout.c: Likewise.
10523 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10524 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10525 * doc/invoke.texi (-ffile-prefix-map): Document.
10526 (-fdebug-prefix-map): Update description.
10528 2018-01-18 Martin Liska <mliska@suse.cz>
10530 * config/i386/i386.c (indirect_thunk_name): Document that also
10532 (output_indirect_thunk): Document why both instructions
10533 (pause and lfence) are generated.
10535 2018-01-18 Richard Biener <rguenther@suse.de>
10537 PR tree-optimization/83887
10538 * graphite-scop-detection.c
10539 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10540 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10541 (scop_detection::merge_sese): Re-implement with a flood-fill
10542 algorithm that properly finds a SESE region if it exists.
10544 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10547 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10548 pointer_diff optimizations use view_convert instead of convert.
10550 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10552 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10553 Generate different code for -mno-speculate-indirect-jumps.
10554 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10555 (*call_indirect_aix<mode>): Disable for
10556 -mno-speculate-indirect-jumps.
10557 (*call_indirect_aix<mode>_nospec): New define_insn.
10558 (*call_value_indirect_aix<mode>): Disable for
10559 -mno-speculate-indirect-jumps.
10560 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10561 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10562 -mno-speculate-indirect-jumps.
10563 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10565 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10567 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10568 long double type, set the flags for noting the default long double
10569 type, even if we don't pass or return a long double type.
10571 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10574 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10577 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10579 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10580 support for merge[hl].
10581 (fold_mergehl_helper): New helper function.
10582 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10583 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10584 (altivec_vmrglw_direct): Add xxmrglw insn.
10586 2018-01-17 Andrew Waterman <andrew@sifive.com>
10588 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10589 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10591 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10594 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10595 call the lto_location_cache before reading the
10596 DECL_SOURCE_LOCATION of the types.
10598 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10599 Richard Sandiford <richard.sandiford@linaro.org>
10601 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10602 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10603 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10604 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10605 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10607 * config/aarch64/constraints.md (aarch64_movti_operand):
10609 * config/aarch64/predicates.md (Uti): Add new constraint.
10611 2018-01-17 Carl Love <cel@us.ibm.com>
10613 * config/rs6000/vsx.md (define_expand xl_len_r,
10614 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10615 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10617 (define_expand, define_insn): Move the shift left from the
10618 define_insn to the define_expand for lxvl and stxvl instructions.
10619 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10620 and XL_LEN_R definitions to PURE.
10622 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10624 * config/i386/i386.c (indirect_thunk_name): Declare regno
10625 as unsigned int. Compare regno with INVALID_REGNUM.
10626 (output_indirect_thunk): Ditto.
10627 (output_indirect_thunk_function): Ditto.
10628 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10629 in the call to output_indirect_thunk_function.
10631 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10633 PR middle-end/83884
10634 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10635 rather than the size of inner_type to determine the stack slot size
10636 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10638 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10641 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10644 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10646 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10647 endian Linux systems to optionally enable multilibs for selecting
10648 the long double type if the user configured an explicit type.
10649 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10650 have no long double multilibs if not defined.
10651 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10652 warn if the user used -mabi={ieee,ibm}longdouble and we built
10653 multilibs for long double.
10654 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10655 appropriate multilib option.
10656 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10658 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10659 for building long double multilibs.
10660 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10662 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10664 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10667 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10669 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10672 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10675 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10678 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10680 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10681 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10683 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10685 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10686 different rtl trees depending on TARGET_64BIT.
10687 (rs6000_gen_lvx): Likewise.
10689 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10691 * config/visium/visium.md (nop): Tweak comment.
10692 (hazard_nop): Likewise.
10694 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10696 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10697 -mspeculate-indirect-jumps.
10698 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10699 for -mno-speculate-indirect-jumps.
10700 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10701 (*call_value_indirect_elfv2<mode>): Disable for
10702 -mno-speculate-indirect-jumps.
10703 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10704 (indirect_jump): Emit different RTL for
10705 -mno-speculate-indirect-jumps.
10706 (*indirect_jump<mode>): Disable for
10707 -mno-speculate-indirect-jumps.
10708 (*indirect_jump<mode>_nospec): New define_insn.
10709 (tablejump): Emit different RTL for
10710 -mno-speculate-indirect-jumps.
10711 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10712 (tablejumpsi_nospec): New define_expand.
10713 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10714 (tablejumpdi_nospec): New define_expand.
10715 (*tablejump<mode>_internal1): Disable for
10716 -mno-speculate-indirect-jumps.
10717 (*tablejump<mode>_internal1_nospec): New define_insn.
10718 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10721 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10723 * caller-save.c (insert_save): Drop unnecessary parameter. All
10726 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10727 Richard Biener <rguenth@suse.de>
10730 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10731 return early, inline manually is_gimple_sizepos. Make sure if we
10732 call gimplify_expr we don't end up with a gimple constant.
10733 * tree.c (variably_modified_type_p): Don't return true for
10734 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10735 * gimplify.h (is_gimple_sizepos): Remove.
10737 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10739 PR tree-optimization/83857
10740 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10741 vectorizable_live_operation for pure SLP statements.
10742 (vectorizable_live_operation): Handle PHIs.
10744 2018-01-16 Richard Biener <rguenther@suse.de>
10746 PR tree-optimization/83867
10747 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10748 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10750 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10753 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10754 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10755 If off is not INTEGER_CST, issue a may not be aligned warning
10756 rather than isn't aligned. Use isn%'t rather than isn't.
10757 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10759 <case MULT_EXPR>: Improve the case when bottom and one of the
10760 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10761 operand, in that case check if the other operand is multiple of
10762 bottom divided by the INTEGER_CST operand.
10764 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10767 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10768 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10769 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10770 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10771 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10772 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10773 (pa_function_arg_size): New function.
10775 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10777 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10778 in a separate statement.
10780 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10782 PR tree-optimization/83847
10783 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10784 group gathers and scatters.
10786 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10788 PR rtl-optimization/86620
10789 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10791 PR rtl-optimization/83213
10792 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10793 to last if both are JUMP_INSNs.
10795 PR tree-optimization/83843
10796 * gimple-ssa-store-merging.c
10797 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10798 store_immediate_info for bswap/nop orig_stores.
10800 2018-01-15 Andrew Waterman <andrew@sifive.com>
10802 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10804 <UDIV>: Increase cost if !TARGET_DIV.
10806 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10808 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10809 (define_attr "cr_logical_3op"): New.
10810 (cceq_ior_compare): Adjust.
10811 (cceq_ior_compare_complement): Adjust.
10812 (*cceq_rev_compare): Adjust.
10813 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10814 (is_cracked_insn): Adjust.
10815 (insn_must_be_first_in_group): Adjust.
10816 * config/rs6000/40x.md: Adjust.
10817 * config/rs6000/440.md: Adjust.
10818 * config/rs6000/476.md: Adjust.
10819 * config/rs6000/601.md: Adjust.
10820 * config/rs6000/603.md: Adjust.
10821 * config/rs6000/6xx.md: Adjust.
10822 * config/rs6000/7450.md: Adjust.
10823 * config/rs6000/7xx.md: Adjust.
10824 * config/rs6000/8540.md: Adjust.
10825 * config/rs6000/cell.md: Adjust.
10826 * config/rs6000/e300c2c3.md: Adjust.
10827 * config/rs6000/e500mc.md: Adjust.
10828 * config/rs6000/e500mc64.md: Adjust.
10829 * config/rs6000/e5500.md: Adjust.
10830 * config/rs6000/e6500.md: Adjust.
10831 * config/rs6000/mpc.md: Adjust.
10832 * config/rs6000/power4.md: Adjust.
10833 * config/rs6000/power5.md: Adjust.
10834 * config/rs6000/power6.md: Adjust.
10835 * config/rs6000/power7.md: Adjust.
10836 * config/rs6000/power8.md: Adjust.
10837 * config/rs6000/power9.md: Adjust.
10838 * config/rs6000/rs64.md: Adjust.
10839 * config/rs6000/titan.md: Adjust.
10841 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10843 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10844 ix86_indirect_branch_register logic.
10846 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10848 * config/i386/constraints.md (Bs): Update
10849 ix86_indirect_branch_register check. Don't check
10850 ix86_indirect_branch_register with GOT_memory_operand.
10852 * config/i386/predicates.md (GOT_memory_operand): Don't check
10853 ix86_indirect_branch_register here.
10854 (GOT32_symbol_operand): Likewise.
10856 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10858 * config/i386/predicates.md (constant_call_address_operand):
10859 Rewrite ix86_indirect_branch_register logic.
10860 (sibcall_insn_operand): Likewise.
10862 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10864 * config/i386/constraints.md (Bs): Replace
10865 ix86_indirect_branch_thunk_register with
10866 ix86_indirect_branch_register.
10868 * config/i386/i386.md (indirect_jump): Likewise.
10869 (tablejump): Likewise.
10870 (*sibcall_memory): Likewise.
10871 (*sibcall_value_memory): Likewise.
10872 Peepholes of indirect call and jump via memory: Likewise.
10873 * config/i386/i386.opt: Likewise.
10874 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10875 (GOT_memory_operand): Likewise.
10876 (call_insn_operand): Likewise.
10877 (sibcall_insn_operand): Likewise.
10878 (GOT32_symbol_operand): Likewise.
10880 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10882 PR middle-end/83837
10883 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10884 type rather than type addr's type points to.
10885 (expand_omp_atomic_mutex): Likewise.
10886 (expand_omp_atomic): Likewise.
10888 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10891 * config/i386/i386.c (output_indirect_thunk_function): Use
10892 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10893 for __x86_return_thunk.
10895 2018-01-15 Richard Biener <rguenther@suse.de>
10897 PR middle-end/83850
10898 * expmed.c (extract_bit_field_1): Fix typo.
10900 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10903 * config/arm/iterators.md (VF): New mode iterator.
10904 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10905 Remove integer-related logic from pattern.
10906 (neon_vabd<mode>_3): Likewise.
10908 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10910 PR middle-end/82694
10911 * common.opt (fstrict-overflow): No longer an alias.
10912 (fwrapv-pointer): New option.
10913 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10914 also for pointer types based on flag_wrapv_pointer.
10915 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10916 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10917 opts->x_flag_wrapv got set.
10918 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10919 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10920 POINTER_TYPE_OVERFLOW_UNDEFINED.
10921 * match.pd: Likewise in address comparison pattern.
10922 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10924 2018-01-15 Richard Biener <rguenther@suse.de>
10927 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10928 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10929 Reset type names to their identifier if their TYPE_DECL doesn't
10930 have linkage (and thus is used for ODR and devirt).
10931 (save_debug_info_for_decl): Remove.
10932 (save_debug_info_for_type): Likewise.
10933 (add_tree_to_fld_list): Adjust.
10934 * tree-pretty-print.c (dump_generic_node): Make dumping of
10935 type names more robust.
10937 2018-01-15 Richard Biener <rguenther@suse.de>
10939 * BASE-VER: Bump to 8.0.1.
10941 2018-01-14 Martin Sebor <msebor@redhat.com>
10944 * builtins.c (check_access): Avoid warning when the no-warning bit
10947 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10949 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10950 * ira-color (allocno_hard_regs_compare): Likewise.
10952 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10955 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10956 Use .pushsection/.popsection.
10958 2018-01-14 Martin Sebor <msebor@redhat.com>
10961 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10963 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10965 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10966 entry from extra_headers.
10967 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10968 extra_headers, make the list bitwise identical to the i?86-*-* one.
10970 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10972 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10973 -mcmodel=large with -mindirect-branch=thunk,
10974 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10975 -mfunction-return=thunk-extern.
10976 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10977 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10978 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10980 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10982 * config/i386/i386.c (print_reg): Print the name of the full
10983 integer register without '%'.
10984 (ix86_print_operand): Handle 'V'.
10985 * doc/extend.texi: Document 'V' modifier.
10987 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10989 * config/i386/constraints.md (Bs): Disallow memory operand for
10990 -mindirect-branch-register.
10992 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10993 (GOT_memory_operand): Likewise.
10994 (call_insn_operand): Likewise.
10995 (sibcall_insn_operand): Likewise.
10996 (GOT32_symbol_operand): Likewise.
10997 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10998 for -mindirect-branch-register.
10999 (tablejump): Likewise.
11000 (*sibcall_memory): Likewise.
11001 (*sibcall_value_memory): Likewise.
11002 Disallow peepholes of indirect call and jump via memory for
11003 -mindirect-branch-register.
11004 (*call_pop): Replace m with Bw.
11005 (*call_value_pop): Likewise.
11006 (*sibcall_pop_memory): Replace m with Bs.
11007 * config/i386/i386.opt (mindirect-branch-register): New option.
11008 * doc/invoke.texi: Document -mindirect-branch-register option.
11010 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11012 * config/i386/i386-protos.h (ix86_output_function_return): New.
11013 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
11014 set function_return_type.
11015 (indirect_thunk_name): Add ret_p to indicate thunk for function
11017 (output_indirect_thunk_function): Pass false to
11018 indirect_thunk_name.
11019 (ix86_output_indirect_branch_via_reg): Likewise.
11020 (ix86_output_indirect_branch_via_push): Likewise.
11021 (output_indirect_thunk_function): Create alias for function
11022 return thunk if regno < 0.
11023 (ix86_output_function_return): New function.
11024 (ix86_handle_fndecl_attribute): Handle function_return.
11025 (ix86_attribute_table): Add function_return.
11026 * config/i386/i386.h (machine_function): Add
11027 function_return_type.
11028 * config/i386/i386.md (simple_return_internal): Use
11029 ix86_output_function_return.
11030 (simple_return_internal_long): Likewise.
11031 * config/i386/i386.opt (mfunction-return=): New option.
11032 (indirect_branch): Mention -mfunction-return=.
11033 * doc/extend.texi: Document function_return function attribute.
11034 * doc/invoke.texi: Document -mfunction-return= option.
11036 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11038 * config/i386/i386-opts.h (indirect_branch): New.
11039 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11040 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11041 with local indirect jump when converting indirect call and jump.
11042 (ix86_set_indirect_branch_type): New.
11043 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11044 (indirectlabelno): New.
11045 (indirect_thunk_needed): Likewise.
11046 (indirect_thunk_bnd_needed): Likewise.
11047 (indirect_thunks_used): Likewise.
11048 (indirect_thunks_bnd_used): Likewise.
11049 (INDIRECT_LABEL): Likewise.
11050 (indirect_thunk_name): Likewise.
11051 (output_indirect_thunk): Likewise.
11052 (output_indirect_thunk_function): Likewise.
11053 (ix86_output_indirect_branch_via_reg): Likewise.
11054 (ix86_output_indirect_branch_via_push): Likewise.
11055 (ix86_output_indirect_branch): Likewise.
11056 (ix86_output_indirect_jmp): Likewise.
11057 (ix86_code_end): Call output_indirect_thunk_function if needed.
11058 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11060 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11061 (ix86_attribute_table): Add indirect_branch.
11062 * config/i386/i386.h (machine_function): Add indirect_branch_type
11063 and has_local_indirect_jump.
11064 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11066 (tablejump): Likewise.
11067 (*indirect_jump): Use ix86_output_indirect_jmp.
11068 (*tablejump_1): Likewise.
11069 (simple_return_indirect_internal): Likewise.
11070 * config/i386/i386.opt (mindirect-branch=): New option.
11071 (indirect_branch): New.
11074 (thunk-inline): Likewise.
11075 (thunk-extern): Likewise.
11076 * doc/extend.texi: Document indirect_branch function attribute.
11077 * doc/invoke.texi: Document -mindirect-branch= option.
11079 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11082 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11084 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11086 * ipa-inline.c (want_inline_small_function_p): Return false if
11087 inlining has already failed with CIF_FINAL_ERROR.
11088 (update_caller_keys): Call want_inline_small_function_p before
11090 (update_callee_keys): Likewise.
11092 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11094 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11096 (rs6000_quadword_masked_address_p): Likewise.
11097 (quad_aligned_load_p): Likewise.
11098 (quad_aligned_store_p): Likewise.
11099 (const_load_sequence_p): Add comment to describe the outer-most loop.
11100 (mimic_memory_attributes_and_flags): New function.
11101 (rs6000_gen_stvx): Likewise.
11102 (replace_swapped_aligned_store): Likewise.
11103 (rs6000_gen_lvx): Likewise.
11104 (replace_swapped_aligned_load): Likewise.
11105 (replace_swapped_load_constant): Capitalize argument name in
11106 comment describing this function.
11107 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11108 and stores that access quad-word aligned addresses and replace
11109 with stvx or lvx instructions when appropriate.
11110 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11111 New function prototype.
11112 (rs6000_quadword_masked_address_p): Likewise.
11113 (rs6000_gen_lvx): Likewise.
11114 (rs6000_gen_stvx): Likewise.
11115 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11116 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11117 when memory address is aligned.
11118 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11119 this split to select lvx instruction when memory address is aligned.
11120 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11121 instruction when memory address is aligned.
11122 (*vsx_le_perm_load_v16qi): Likewise.
11123 (four unnamed splitters): Modify to select the stvx instruction
11124 when memory is aligned.
11126 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11128 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11129 which appears in the queue multiple times.
11131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11132 Alan Hayward <alan.hayward@arm.com>
11133 David Sherwood <david.sherwood@arm.com>
11135 * tree-vectorizer.h (vec_lower_bound): New structure.
11136 (_loop_vec_info): Add check_nonzero and lower_bounds.
11137 (LOOP_VINFO_CHECK_NONZERO): New macro.
11138 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11139 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11140 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11141 fields. Make seg_len the distance travelled, not including the
11143 (dr_direction_indicator): Declare.
11144 (dr_zero_step_indicator): Likewise.
11145 (dr_known_forward_stride_p): Likewise.
11146 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11148 (runtime_alias_check_p): Allow runtime alias checks with
11150 (operator ==): Compare access_size and align.
11151 (prune_runtime_alias_test_list): Rework for new distinction between
11152 the access_size and seg_len.
11153 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11155 (get_segment_min_max): New function.
11156 (create_intersect_range_checks): Use it.
11157 (dr_step_indicator): New function.
11158 (dr_direction_indicator): Likewise.
11159 (dr_zero_step_indicator): Likewise.
11160 (dr_known_forward_stride_p): Likewise.
11161 * tree-loop-distribution.c (data_ref_segment_size): Return
11162 DR_STEP * (niters - 1).
11163 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11165 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11166 (vect_preserves_scalar_order_p): New function, split out from...
11167 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11168 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11169 (vect_vfa_access_size): New function.
11170 (vect_vfa_align): Likewise.
11171 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11172 (dump_lower_bound): New function.
11173 (vect_check_lower_bound): Likewise.
11174 (vect_small_gap_p): Likewise.
11175 (vectorizable_with_step_bound_p): Likewise.
11176 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11177 depencies if the vectorization factor is 1. Convert the checks
11178 for nonzero steps into checks on the bounds of DR_STEP. Try using
11179 a bunds check for variable steps if the minimum required step is
11180 relatively small. Update calls to the dr_with_seg_len
11181 constructor and to vect_compile_time_alias.
11182 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11184 (vect_loop_versioning): Call it.
11185 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11187 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11190 Alan Hayward <alan.hayward@arm.com>
11191 David Sherwood <david.sherwood@arm.com>
11193 * doc/sourcebuild.texi (vect_scatter_store): Document.
11194 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11196 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11198 * genopinit.c (main): Add supports_vec_scatter_store and
11199 supports_vec_scatter_store_cached to target_optabs.
11200 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11201 IFN_MASK_SCATTER_STORE.
11202 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11204 * internal-fn.h (internal_store_fn_p): Declare.
11205 (internal_fn_stored_value_index): Likewise.
11206 * internal-fn.c (scatter_store_direct): New macro.
11207 (expand_scatter_store_optab_fn): New function.
11208 (direct_scatter_store_optab_supported_p): New macro.
11209 (internal_store_fn_p): New function.
11210 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11211 IFN_MASK_SCATTER_STORE.
11212 (internal_fn_mask_index): Likewise.
11213 (internal_fn_stored_value_index): New function.
11214 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11215 for scatter stores.
11216 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11217 * optabs-query.c (supports_vec_scatter_store_p): New function.
11218 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11219 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11220 true for scatter stores.
11221 (vect_gather_scatter_fn_p): Handle scatter stores too.
11222 (vect_check_gather_scatter): Consider using scatter stores if
11223 supports_vec_scatter_store_p.
11224 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11225 scatter stores too.
11226 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11227 internal_fn_stored_value_index.
11228 (check_load_store_masking): Handle scatter stores too.
11229 (vect_get_store_rhs): Make public.
11230 (vectorizable_call): Use internal_store_fn_p.
11231 (vectorizable_store): Handle scatter store internal functions.
11232 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11233 when deciding whether the end of the group has been reached.
11234 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11235 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11236 (mask_scatter_store<mode>): New insns.
11238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11239 Alan Hayward <alan.hayward@arm.com>
11240 David Sherwood <david.sherwood@arm.com>
11242 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11243 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11244 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11246 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11247 Use vect_truncate_gather_scatter_offset if we can't treat the
11248 operation as a normal gather load or scatter store.
11249 (get_group_load_store_type): Take the gather_scatter_info
11250 as argument. Try using a gather load or scatter store for
11251 single-element groups.
11252 (get_load_store_type): Update calls to get_group_load_store_type
11253 and vect_use_strided_gather_scatters_p.
11255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11256 Alan Hayward <alan.hayward@arm.com>
11257 David Sherwood <david.sherwood@arm.com>
11259 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11260 optional tree argument.
11261 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11263 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11264 but continue to use the current value as a fallback.
11265 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11266 to compare the updates.
11267 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11268 (get_load_store_type): Use it when handling a strided access.
11269 (vect_get_strided_load_store_ops): New function.
11270 (vect_get_data_ptr_increment): Likewise.
11271 (vectorizable_load): Handle strided gather loads. Always pass
11272 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11274 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11275 Alan Hayward <alan.hayward@arm.com>
11276 David Sherwood <david.sherwood@arm.com>
11278 * doc/md.texi (gather_load@var{m}): Document.
11279 (mask_gather_load@var{m}): Likewise.
11280 * genopinit.c (main): Add supports_vec_gather_load and
11281 supports_vec_gather_load_cached to target_optabs.
11282 * optabs-tree.c (init_tree_optimization_optabs): Use
11283 ggc_cleared_alloc to allocate target_optabs.
11284 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11285 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11287 * internal-fn.h (internal_load_fn_p): Declare.
11288 (internal_gather_scatter_fn_p): Likewise.
11289 (internal_fn_mask_index): Likewise.
11290 (internal_gather_scatter_fn_supported_p): Likewise.
11291 * internal-fn.c (gather_load_direct): New macro.
11292 (expand_gather_load_optab_fn): New function.
11293 (direct_gather_load_optab_supported_p): New macro.
11294 (direct_internal_fn_optab): New function.
11295 (internal_load_fn_p): Likewise.
11296 (internal_gather_scatter_fn_p): Likewise.
11297 (internal_fn_mask_index): Likewise.
11298 (internal_gather_scatter_fn_supported_p): Likewise.
11299 * optabs-query.c (supports_at_least_one_mode_p): New function.
11300 (supports_vec_gather_load_p): Likewise.
11301 * optabs-query.h (supports_vec_gather_load_p): Declare.
11302 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11303 and memory_type field.
11304 (NUM_PATTERNS): Bump to 15.
11305 * tree-vect-data-refs.c: Include internal-fn.h.
11306 (vect_gather_scatter_fn_p): New function.
11307 (vect_describe_gather_scatter_call): Likewise.
11308 (vect_check_gather_scatter): Try using internal functions for
11309 gather loads. Recognize existing calls to a gather load function.
11310 (vect_analyze_data_refs): Consider using gather loads if
11311 supports_vec_gather_load_p.
11312 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11313 (vect_get_gather_scatter_offset_type): Likewise.
11314 (vect_convert_mask_for_vectype): Likewise.
11315 (vect_add_conversion_to_patterm): Likewise.
11316 (vect_try_gather_scatter_pattern): Likewise.
11317 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11318 (vect_vect_recog_func_ptrs): Add it.
11319 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11320 internal_fn_mask_index and internal_gather_scatter_fn_p.
11321 (check_load_store_masking): Take the gather_scatter_info as an
11322 argument and handle gather loads.
11323 (vect_get_gather_scatter_ops): New function.
11324 (vectorizable_call): Check internal_load_fn_p.
11325 (vectorizable_load): Likewise. Handle gather load internal
11327 (vectorizable_store): Update call to check_load_store_masking.
11328 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11329 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11330 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11331 (aarch64_gather_scale_operand_d): New predicates.
11332 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11333 (mask_gather_load<mode>): New insns.
11335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11336 Alan Hayward <alan.hayward@arm.com>
11337 David Sherwood <david.sherwood@arm.com>
11339 * optabs.def (fold_left_plus_optab): New optab.
11340 * doc/md.texi (fold_left_plus_@var{m}): Document.
11341 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11342 * internal-fn.c (fold_left_direct): Define.
11343 (expand_fold_left_optab_fn): Likewise.
11344 (direct_fold_left_optab_supported_p): Likewise.
11345 * fold-const-call.c (fold_const_fold_left): New function.
11346 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11347 * tree-parloops.c (valid_reduction_p): New function.
11348 (gather_scalar_reductions): Use it.
11349 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11350 (vect_finish_replace_stmt): Declare.
11351 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11352 (needs_fold_left_reduction_p): New function, split out from...
11353 (vect_is_simple_reduction): ...here. Accept reductions that
11354 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11355 (vect_force_simple_reduction): Also store the reduction type in
11356 the assignment's STMT_VINFO_REDUC_TYPE.
11357 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11358 (merge_with_identity): New function.
11359 (vect_expand_fold_left): Likewise.
11360 (vectorize_fold_left_reduction): Likewise.
11361 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11362 scalar phi in place for it. Check for target support and reject
11363 cases that would reassociate the operation. Defer the transform
11364 phase to vectorize_fold_left_reduction.
11365 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11366 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11367 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11371 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11372 call to ifc_temp_var.
11374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * target.def (legitimize_address_displacement): Take the original
11379 offset as a poly_int.
11380 * targhooks.h (default_legitimize_address_displacement): Update
11382 * targhooks.c (default_legitimize_address_displacement): Likewise.
11383 * doc/tm.texi: Regenerate.
11384 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11385 as an argument, moving assert of ad->disp == ad->disp_term to...
11386 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11387 Try calling targetm.legitimize_address_displacement before expanding
11388 the address rather than afterwards, and adjust for the new interface.
11389 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11390 Match the new hook interface. Handle SVE addresses.
11391 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11392 new hook interface.
11394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11396 * Makefile.in (OBJS): Add early-remat.o.
11397 * target.def (select_early_remat_modes): New hook.
11398 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11399 * doc/tm.texi: Regenerate.
11400 * targhooks.h (default_select_early_remat_modes): Declare.
11401 * targhooks.c (default_select_early_remat_modes): New function.
11402 * timevar.def (TV_EARLY_REMAT): New timevar.
11403 * passes.def (pass_early_remat): New pass.
11404 * tree-pass.h (make_pass_early_remat): Declare.
11405 * early-remat.c: New file.
11406 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11408 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11410 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11411 Alan Hayward <alan.hayward@arm.com>
11412 David Sherwood <david.sherwood@arm.com>
11414 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11415 vfm1 with a bound_epilog parameter.
11416 (vect_do_peeling): Update calls accordingly, and move the prologue
11417 call earlier in the function. Treat the base bound_epilog as 0 for
11418 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11419 this base when peeling for gaps.
11420 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11421 with fully-masked loops.
11422 (vect_estimate_min_profitable_iters): Handle the single peeled
11423 iteration in that case.
11425 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11426 Alan Hayward <alan.hayward@arm.com>
11427 David Sherwood <david.sherwood@arm.com>
11429 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11430 single-element interleaving even if the size is not a power of 2.
11431 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11432 accesses for single-element interleaving if the group size is
11435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11436 Alan Hayward <alan.hayward@arm.com>
11437 David Sherwood <david.sherwood@arm.com>
11439 * doc/md.texi (fold_extract_last_@var{m}): Document.
11440 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11441 * optabs.def (fold_extract_last_optab): New optab.
11442 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11443 * internal-fn.c (fold_extract_direct): New macro.
11444 (expand_fold_extract_optab_fn): Likewise.
11445 (direct_fold_extract_optab_supported_p): Likewise.
11446 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11447 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11448 EXTRACT_LAST_REDUCTION.
11449 (get_initial_def_for_reduction): Do not create an initial vector
11450 for EXTRACT_LAST_REDUCTION reductions.
11451 (vectorizable_reduction): Leave the scalar phi in place for
11452 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11453 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11454 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11455 transform phase to vectorizable_condition.
11456 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11458 (vect_finish_stmt_generation): ...here.
11459 (vect_finish_replace_stmt): New function.
11460 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11461 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11463 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11466 Alan Hayward <alan.hayward@arm.com>
11467 David Sherwood <david.sherwood@arm.com>
11469 * doc/md.texi (extract_last_@var{m}): Document.
11470 * optabs.def (extract_last_optab): New optab.
11471 * internal-fn.def (EXTRACT_LAST): New internal function.
11472 * internal-fn.c (cond_unary_direct): New macro.
11473 (expand_cond_unary_optab_fn): Likewise.
11474 (direct_cond_unary_optab_supported_p): Likewise.
11475 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11476 loops using EXTRACT_LAST.
11477 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11478 (extract_last_<mode>): ...this optab.
11479 (vec_extract<mode><Vel>): Update accordingly.
11481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11482 Alan Hayward <alan.hayward@arm.com>
11483 David Sherwood <david.sherwood@arm.com>
11485 * target.def (empty_mask_is_expensive): New hook.
11486 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11487 * doc/tm.texi: Regenerate.
11488 * targhooks.h (default_empty_mask_is_expensive): Declare.
11489 * targhooks.c (default_empty_mask_is_expensive): New function.
11490 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11491 if the target says that empty masks are expensive.
11492 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11494 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11496 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11497 Alan Hayward <alan.hayward@arm.com>
11498 David Sherwood <david.sherwood@arm.com>
11500 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11501 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11502 (vect_use_loop_mask_for_alignment_p): New function.
11503 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11504 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11505 niters_skip argument. Make sure that the first niters_skip elements
11506 of the first iteration are inactive.
11507 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11508 Update call to vect_set_loop_masks_directly.
11509 (get_misalign_in_elems): New function, split out from...
11510 (vect_gen_prolog_loop_niters): ...here.
11511 (vect_update_init_of_dr): Take a code argument that specifies whether
11512 the adjustment should be added or subtracted.
11513 (vect_update_init_of_drs): Likewise.
11514 (vect_prepare_for_masked_peels): New function.
11515 (vect_do_peeling): Skip prologue peeling if we're using a mask
11516 instead. Update call to vect_update_inits_of_drs.
11517 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11519 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11520 alignment. Do not include the number of peeled iterations in
11521 the minimum threshold in that case.
11522 (vectorizable_induction): Adjust the start value down by
11523 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11524 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11525 Take the number of skipped iterations into account when calculating
11527 * tree-vect-stmts.c (vect_gen_while_not): New function.
11529 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11530 Alan Hayward <alan.hayward@arm.com>
11531 David Sherwood <david.sherwood@arm.com>
11533 * doc/sourcebuild.texi (vect_fully_masked): Document.
11534 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11535 default value to 0.
11536 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11538 (vect_analyze_loop_2): ...here. Don't check the vectorization
11539 factor against the number of loop iterations if the loop is
11542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11543 Alan Hayward <alan.hayward@arm.com>
11544 David Sherwood <david.sherwood@arm.com>
11546 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11547 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11548 (dump_groups): Update accordingly.
11549 (iv_use::mem_type): New member variable.
11550 (address_p): New function.
11551 (record_use): Add a mem_type argument and initialize the new
11553 (record_group_use): Add a mem_type argument. Use address_p.
11554 Remove obsolete null checks of base_object. Update call to record_use.
11555 (find_interesting_uses_op): Update call to record_group_use.
11556 (find_interesting_uses_cond): Likewise.
11557 (find_interesting_uses_address): Likewise.
11558 (get_mem_type_for_internal_fn): New function.
11559 (find_address_like_use): Likewise.
11560 (find_interesting_uses_stmt): Try find_address_like_use before
11561 calling find_interesting_uses_op.
11562 (addr_offset_valid_p): Use the iv mem_type field as the type
11563 of the addressed memory.
11564 (add_autoinc_candidates): Likewise.
11565 (get_address_cost): Likewise.
11566 (split_small_address_groups_p): Use address_p.
11567 (split_address_groups): Likewise.
11568 (add_iv_candidate_for_use): Likewise.
11569 (autoinc_possible_for_pair): Likewise.
11570 (rewrite_groups): Likewise.
11571 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11572 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11573 (get_alias_ptr_type_for_ptr_address): New function.
11574 (rewrite_use_address): Rewrite address uses in calls that were
11575 identified by find_address_like_use.
11577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11578 Alan Hayward <alan.hayward@arm.com>
11579 David Sherwood <david.sherwood@arm.com>
11581 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11583 * gimple-expr.h (is_gimple_addressable: Likewise.
11584 * gimple-expr.c (is_gimple_address): Likewise.
11585 * internal-fn.c (expand_call_mem_ref): New function.
11586 (expand_mask_load_optab_fn): Use it.
11587 (expand_mask_store_optab_fn): Likewise.
11589 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11590 Alan Hayward <alan.hayward@arm.com>
11591 David Sherwood <david.sherwood@arm.com>
11593 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11594 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11595 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11596 (cond_umax@var{mode}): Document.
11597 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11598 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11599 (cond_umin_optab, cond_umax_optab): New optabs.
11600 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11601 (COND_IOR, COND_XOR): New internal functions.
11602 * internal-fn.h (get_conditional_internal_fn): Declare.
11603 * internal-fn.c (cond_binary_direct): New macro.
11604 (expand_cond_binary_optab_fn): Likewise.
11605 (direct_cond_binary_optab_supported_p): Likewise.
11606 (get_conditional_internal_fn): New function.
11607 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11608 Cope with reduction statements that are vectorized as calls rather
11610 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11611 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11612 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11613 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11614 (UNSPEC_COND_EOR): New unspecs.
11615 (optab): Add mappings for them.
11616 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11617 (sve_int_op, sve_fp_op): New int attributes.
11619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11620 Alan Hayward <alan.hayward@arm.com>
11621 David Sherwood <david.sherwood@arm.com>
11623 * optabs.def (while_ult_optab): New optab.
11624 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11625 * internal-fn.def (WHILE_ULT): New internal function.
11626 * internal-fn.h (direct_internal_fn_supported_p): New override
11627 that takes two types as argument.
11628 * internal-fn.c (while_direct): New macro.
11629 (expand_while_optab_fn): New function.
11630 (convert_optab_supported_p): Likewise.
11631 (direct_while_optab_supported_p): New macro.
11632 * wide-int.h (wi::udiv_ceil): New function.
11633 * tree-vectorizer.h (rgroup_masks): New structure.
11634 (vec_loop_masks): New typedef.
11635 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11636 and fully_masked_p.
11637 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11638 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11639 (vect_max_vf): New function.
11640 (slpeel_make_loop_iterate_ntimes): Delete.
11641 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11642 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11643 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11644 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11645 internal-fn.h, stor-layout.h and optabs-query.h.
11646 (vect_set_loop_mask): New function.
11647 (add_preheader_seq): Likewise.
11648 (add_header_seq): Likewise.
11649 (interleave_supported_p): Likewise.
11650 (vect_maybe_permute_loop_masks): Likewise.
11651 (vect_set_loop_masks_directly): Likewise.
11652 (vect_set_loop_condition_masked): Likewise.
11653 (vect_set_loop_condition_unmasked): New function, split out from
11654 slpeel_make_loop_iterate_ntimes.
11655 (slpeel_make_loop_iterate_ntimes): Rename to..
11656 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11657 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11658 (vect_do_peeling): Update call accordingly.
11659 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11661 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11662 mask_compare_type, can_fully_mask_p and fully_masked_p.
11663 (release_vec_loop_masks): New function.
11664 (_loop_vec_info): Use it to free the loop masks.
11665 (can_produce_all_loop_masks_p): New function.
11666 (vect_get_max_nscalars_per_iter): Likewise.
11667 (vect_verify_full_masking): Likewise.
11668 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11669 retries, and free the mask rgroups before retrying. Check loop-wide
11670 reasons for disallowing fully-masked loops. Make the final decision
11671 about whether use a fully-masked loop or not.
11672 (vect_estimate_min_profitable_iters): Do not assume that peeling
11673 for the number of iterations will be needed for fully-masked loops.
11674 (vectorizable_reduction): Disable fully-masked loops.
11675 (vectorizable_live_operation): Likewise.
11676 (vect_halve_mask_nunits): New function.
11677 (vect_double_mask_nunits): Likewise.
11678 (vect_record_loop_mask): Likewise.
11679 (vect_get_loop_mask): Likewise.
11680 (vect_transform_loop): Handle the case in which the final loop
11681 iteration might handle a partial vector. Call vect_set_loop_condition
11682 instead of slpeel_make_loop_iterate_ntimes.
11683 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11684 (check_load_store_masking): New function.
11685 (prepare_load_store_mask): Likewise.
11686 (vectorizable_store): Handle fully-masked loops.
11687 (vectorizable_load): Likewise.
11688 (supportable_widening_operation): Use vect_halve_mask_nunits for
11690 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11691 (vect_gen_while): New function.
11692 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11693 (aarch64_uqdec<mode>): New insn.
11695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11696 Alan Hayward <alan.hayward@arm.com>
11697 David Sherwood <david.sherwood@arm.com>
11699 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11700 (reduc_xor_scal_optab): New optabs.
11701 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11702 (reduc_xor_scal_@var{m}): Document.
11703 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11704 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11705 internal functions.
11706 * fold-const-call.c (fold_const_call): Handle them.
11707 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11708 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11709 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11710 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11711 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11712 (UNSPEC_XORV): New unspecs.
11713 (optab): Add entries for them.
11714 (BITWISEV): New int iterator.
11715 (bit_reduc_op): New int attributes.
11717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11718 Alan Hayward <alan.hayward@arm.com>
11719 David Sherwood <david.sherwood@arm.com>
11721 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11722 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11723 * optabs.def (vec_shl_insert_optab): New optab.
11724 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11725 (duplicate_and_interleave): Likewise.
11726 * tree-vect-loop.c: Include internal-fn.h.
11727 (neutral_op_for_slp_reduction): New function, split out from
11728 get_initial_defs_for_reduction.
11729 (get_initial_def_for_reduction): Handle option 2 for variable-length
11730 vectors by loading the neutral value into a vector and then shifting
11731 the initial value into element 0.
11732 (get_initial_defs_for_reduction): Replace the code argument with
11733 the neutral value calculated by neutral_op_for_slp_reduction.
11734 Use gimple_build_vector for constant-length vectors.
11735 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11736 but the first group_size elements have a neutral value.
11737 Use duplicate_and_interleave otherwise.
11738 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11739 Update call to get_initial_defs_for_reduction. Handle SLP
11740 reductions for variable-length vectors by creating one vector
11741 result for each scalar result, with the elements associated
11742 with other scalar results stubbed out with the neutral value.
11743 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11744 Require IFN_VEC_SHL_INSERT for double reductions on
11745 variable-length vectors, or SLP reductions that have
11746 a neutral value. Require can_duplicate_and_interleave_p
11747 support for variable-length unchained SLP reductions if there
11748 is no neutral value, such as for MIN/MAX reductions. Also require
11749 the number of vector elements to be a multiple of the number of
11750 SLP statements when doing variable-length unchained SLP reductions.
11751 Update call to vect_create_epilog_for_reduction.
11752 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11753 and remove initial values.
11754 (duplicate_and_interleave): Make public.
11755 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11756 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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-slp.c: Include gimple-fold.h and internal-fn.h
11763 (can_duplicate_and_interleave_p): New function.
11764 (vect_get_and_check_slp_defs): Take the vector of statements
11765 rather than just the current one. Remove excess parentheses.
11766 Restriction rejectinon of vect_constant_def and vect_external_def
11767 for variable-length vectors to boolean types, or types for which
11768 can_duplicate_and_interleave_p is false.
11769 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11770 (duplicate_and_interleave): New function.
11771 (vect_get_constant_vectors): Use gimple_build_vector for
11772 constant-length vectors and suitable variable-length constant
11773 vectors. Use duplicate_and_interleave for other variable-length
11774 vectors. Don't defer the update when inserting new statements.
11776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11777 Alan Hayward <alan.hayward@arm.com>
11778 David Sherwood <david.sherwood@arm.com>
11780 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11781 min_profitable_iters doesn't go negative.
11783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11784 Alan Hayward <alan.hayward@arm.com>
11785 David Sherwood <david.sherwood@arm.com>
11787 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11788 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11789 * optabs.def (vec_mask_load_lanes_optab): New optab.
11790 (vec_mask_store_lanes_optab): Likewise.
11791 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11792 (MASK_STORE_LANES): Likewise.
11793 * internal-fn.c (mask_load_lanes_direct): New macro.
11794 (mask_store_lanes_direct): Likewise.
11795 (expand_mask_load_optab_fn): Handle masked operations.
11796 (expand_mask_load_lanes_optab_fn): New macro.
11797 (expand_mask_store_optab_fn): Handle masked operations.
11798 (expand_mask_store_lanes_optab_fn): New macro.
11799 (direct_mask_load_lanes_optab_supported_p): Likewise.
11800 (direct_mask_store_lanes_optab_supported_p): Likewise.
11801 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11803 (vect_load_lanes_supported): Likewise.
11804 * tree-vect-data-refs.c (strip_conversion): New function.
11805 (can_group_stmts_p): Likewise.
11806 (vect_analyze_data_ref_accesses): Use it instead of checking
11807 for a pair of assignments.
11808 (vect_store_lanes_supported): Take a masked_p parameter.
11809 (vect_load_lanes_supported): Likewise.
11810 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11811 vect_store_lanes_supported and vect_load_lanes_supported.
11812 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11813 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11814 parameter. Don't allow gaps for masked accesses.
11815 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11816 and vect_load_lanes_supported.
11817 (get_load_store_type): Take a masked_p parameter and update
11818 call to get_group_load_store_type.
11819 (vectorizable_store): Update call to get_load_store_type.
11820 Handle IFN_MASK_STORE_LANES.
11821 (vectorizable_load): Update call to get_load_store_type.
11822 Handle IFN_MASK_LOAD_LANES.
11824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11825 Alan Hayward <alan.hayward@arm.com>
11826 David Sherwood <david.sherwood@arm.com>
11828 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11830 * config/aarch64/aarch64-protos.h
11831 (aarch64_sve_struct_memory_operand_p): Declare.
11832 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11833 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11834 (VPRED, vpred): Handle SVE structure modes.
11835 * config/aarch64/constraints.md (Utx): New constraint.
11836 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11837 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11838 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11839 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11840 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11841 structure modes. Split into pieces after RA.
11842 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11843 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11845 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11846 SVE structure modes.
11847 (aarch64_classify_address): Likewise.
11848 (sizetochar): Move earlier in file.
11849 (aarch64_print_operand): Handle SVE register lists.
11850 (aarch64_array_mode): New function.
11851 (aarch64_sve_struct_memory_operand_p): Likewise.
11852 (TARGET_ARRAY_MODE): Redefine.
11854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11855 Alan Hayward <alan.hayward@arm.com>
11856 David Sherwood <david.sherwood@arm.com>
11858 * target.def (array_mode): New target hook.
11859 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11860 * doc/tm.texi: Regenerate.
11861 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11862 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11863 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11864 targetm.array_mode.
11865 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11869 Alan Hayward <alan.hayward@arm.com>
11870 David Sherwood <david.sherwood@arm.com>
11872 * fold-const.c (fold_binary_loc): Check the argument types
11873 rather than the result type when testing for a vector operation.
11875 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11877 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11878 * doc/tm.texi: Regenerate.
11880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11881 Alan Hayward <alan.hayward@arm.com>
11882 David Sherwood <david.sherwood@arm.com>
11884 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11885 (sve): Document new AArch64 extension.
11886 * doc/md.texi (w): Extend the description of the AArch64
11887 constraint to include SVE vectors.
11888 (Upl, Upa): Document new AArch64 predicate constraints.
11889 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11891 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11892 (msve-vector-bits=): New option.
11893 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11894 SVE when these are disabled.
11895 (sve): New extension.
11896 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11897 modes. Adjust their number of units based on aarch64_sve_vg.
11898 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11899 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11900 aarch64_addr_query_type.
11901 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11902 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11903 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11904 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11905 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11906 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11907 (aarch64_simd_imm_zero_p): Delete.
11908 (aarch64_check_zero_based_sve_index_immediate): Declare.
11909 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11910 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11911 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11912 (aarch64_sve_float_mul_immediate_p): Likewise.
11913 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11914 rather than an rtx.
11915 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11916 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11917 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11918 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11919 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11920 (aarch64_regmode_natural_size): Likewise.
11921 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11922 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11924 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11925 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11926 for VG and the SVE predicate registers.
11927 (V_ALIASES): Add a "z"-prefixed alias.
11928 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11929 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11930 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11931 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11932 (REG_CLASS_NAMES): Add entries for them.
11933 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11934 and the predicate registers.
11935 (aarch64_sve_vg): Declare.
11936 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11937 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11938 (REGMODE_NATURAL_SIZE): Define.
11939 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11941 * config/aarch64/aarch64.c: Include cfgrtl.h.
11942 (simd_immediate_info): Add a constructor for series vectors,
11943 and an associated step field.
11944 (aarch64_sve_vg): New variable.
11945 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11946 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11947 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11948 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11949 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11950 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11951 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11952 (aarch64_get_mask_mode): New functions.
11953 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11954 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11955 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11956 predicate modes and predicate registers. Explicitly restrict
11957 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11958 to store a vector mode if it is recognized by
11959 aarch64_classify_vector_mode.
11960 (aarch64_regmode_natural_size): New function.
11961 (aarch64_hard_regno_caller_save_mode): Return the original mode
11963 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11964 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11965 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11966 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11968 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11969 does not overlap dest if the function is frame-related. Handle
11971 (aarch64_split_add_offset): New function.
11972 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11973 them aarch64_add_offset.
11974 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11975 and update call to aarch64_sub_sp.
11976 (aarch64_add_cfa_expression): New function.
11977 (aarch64_expand_prologue): Pass extra temporary registers to the
11978 functions above. Handle the case in which we need to emit new
11979 DW_CFA_expressions for registers that were originally saved
11980 relative to the stack pointer, but now have to be expressed
11981 relative to the frame pointer.
11982 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11984 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11985 IP0 and IP1 values for SVE frames.
11986 (aarch64_expand_vec_series): New function.
11987 (aarch64_expand_sve_widened_duplicate): Likewise.
11988 (aarch64_expand_sve_const_vector): Likewise.
11989 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11990 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11991 into the register, rather than emitting a SET directly.
11992 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11993 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11994 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11995 (offset_9bit_signed_scaled_p): New functions.
11996 (aarch64_replicate_bitmask_imm): New function.
11997 (aarch64_bitmask_imm): Use it.
11998 (aarch64_cannot_force_const_mem): Reject expressions involving
11999 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
12000 (aarch64_classify_index): Handle SVE indices, by requiring
12001 a plain register index with a scale that matches the element size.
12002 (aarch64_classify_address): Handle SVE addresses. Assert that
12003 the mode of the address is VOIDmode or an integer mode.
12004 Update call to aarch64_classify_symbol.
12005 (aarch64_classify_symbolic_expression): Update call to
12006 aarch64_classify_symbol.
12007 (aarch64_const_vec_all_in_range_p): New function.
12008 (aarch64_print_vector_float_operand): Likewise.
12009 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
12010 "vN" for FP registers with SVE modes. Handle (const ...) vectors
12011 and the FP immediates 1.0 and 0.5.
12012 (aarch64_print_address_internal): Handle SVE addresses.
12013 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
12014 (aarch64_regno_regclass): Handle predicate registers.
12015 (aarch64_secondary_reload): Handle big-endian reloads of SVE
12017 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
12018 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
12019 (aarch64_convert_sve_vector_bits): New function.
12020 (aarch64_override_options): Use it to handle -msve-vector-bits=.
12021 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
12022 rather than an rtx.
12023 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
12024 Handle SVE vector and predicate modes. Accept VL-based constants
12025 that need only one temporary register, and VL offsets that require
12026 no temporary registers.
12027 (aarch64_conditional_register_usage): Mark the predicate registers
12028 as fixed if SVE isn't available.
12029 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
12030 Return true for SVE vector and predicate modes.
12031 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
12032 rather than an unsigned int. Handle SVE modes.
12033 (aarch64_preferred_simd_mode): Update call accordingly. Handle
12035 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12037 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12038 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12039 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12040 (aarch64_sve_float_mul_immediate_p): New functions.
12041 (aarch64_sve_valid_immediate): New function.
12042 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12043 Explicitly reject structure modes. Check for INDEX constants.
12044 Handle PTRUE and PFALSE constants.
12045 (aarch64_check_zero_based_sve_index_immediate): New function.
12046 (aarch64_simd_imm_zero_p): Delete.
12047 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12048 vector modes. Accept constants in the range of CNT[BHWD].
12049 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12050 ask for an Advanced SIMD mode.
12051 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12052 (aarch64_simd_vector_alignment): Handle SVE predicates.
12053 (aarch64_vectorize_preferred_vector_alignment): New function.
12054 (aarch64_simd_vector_alignment_reachable): Use it instead of
12056 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12057 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12059 (MAX_VECT_LEN): Delete.
12060 (expand_vec_perm_d): Add a vec_flags field.
12061 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12062 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12063 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12065 (aarch64_evpc_rev): Rename to...
12066 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12067 (aarch64_evpc_rev_global): New function.
12068 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12069 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12071 (aarch64_evpc_sve_tbl): New function.
12072 (aarch64_expand_vec_perm_const_1): Update after rename of
12073 aarch64_evpc_rev. Handle SVE permutes too, trying
12074 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12075 than aarch64_evpc_tbl.
12076 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12077 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12078 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12079 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12080 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12081 (aarch64_expand_sve_vcond): New functions.
12082 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12083 of aarch64_vector_mode_p.
12084 (aarch64_dwarf_poly_indeterminate_value): New function.
12085 (aarch64_compute_pressure_classes): Likewise.
12086 (aarch64_can_change_mode_class): Likewise.
12087 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12088 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12089 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12090 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12091 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12092 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12093 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12094 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12096 (Dn, Dl, Dr): Accept const as well as const_vector.
12097 (Dz): Likewise. Compare against CONST0_RTX.
12098 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12099 of "vector" where appropriate.
12100 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12101 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12102 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12103 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12104 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12105 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12106 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12107 (v_int_equiv): Extend to SVE modes.
12108 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12110 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12111 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12112 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12113 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12114 (SVE_COND_FP_CMP): New int iterators.
12115 (perm_hilo): Handle the new unpack unspecs.
12116 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12118 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12119 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12120 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12121 (aarch64_equality_operator, aarch64_constant_vector_operand)
12122 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12123 (aarch64_sve_nonimmediate_operand): Likewise.
12124 (aarch64_sve_general_operand): Likewise.
12125 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12126 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12127 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12128 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12129 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12130 (aarch64_sve_float_arith_immediate): Likewise.
12131 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12132 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12133 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12134 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12135 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12136 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12137 (aarch64_sve_float_arith_operand): Likewise.
12138 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12139 (aarch64_sve_float_mul_operand): Likewise.
12140 (aarch64_sve_vec_perm_operand): Likewise.
12141 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12142 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12143 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12144 as well as const_vector.
12145 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12146 in file. Use CONST0_RTX and CONSTM1_RTX.
12147 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12148 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12149 Use aarch64_simd_imm_zero.
12150 * config/aarch64/aarch64-sve.md: New file.
12151 * config/aarch64/aarch64.md: Include it.
12152 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12153 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12154 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12155 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12156 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12157 (sve): New attribute.
12158 (enabled): Disable instructions with the sve attribute unless
12160 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12161 aarch64_expand_mov_immediate.
12162 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12163 CNT[BHSD] immediates.
12164 (movti): Split CONST_POLY_INT moves into two halves.
12165 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12166 Split additions that need a temporary here if the destination
12167 is the stack pointer.
12168 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12169 (*add<mode>3_poly_1): New instruction.
12170 (set_clobber_cc): New expander.
12172 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12174 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12175 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12176 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12177 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12178 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12179 Change innermode from fixed_mode_size to machine_mode.
12180 (simplify_subreg): Update call accordingly. Handle a constant-sized
12181 subreg of a variable-length CONST_VECTOR.
12183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12184 Alan Hayward <alan.hayward@arm.com>
12185 David Sherwood <david.sherwood@arm.com>
12187 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12188 (add_offset_to_base): New function, split out from...
12189 (create_mem_ref): ...here. When handling a scale other than 1,
12190 check first whether the address is valid without the offset.
12191 Add it into the base if so, leaving the index and scale as-is.
12193 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12196 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12197 fold_for_warn before checking if arg2 is INTEGER_CST.
12199 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12201 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12202 (store_multiple_operation): Delete.
12203 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12204 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12205 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12206 guarded by TARGET_STRING.
12207 (rs6000_output_load_multiple): Delete.
12208 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12209 OPTION_MASK_STRING / TARGET_STRING handling.
12210 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12211 (const rs6000_opt_masks) <"string">: Change mask to 0.
12212 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12213 (MASK_STRING): Delete.
12214 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12216 (load_multiple): Delete.
12223 (store_multiple): Delete.
12230 (movmemsi_8reg): Delete.
12231 (corresponding unnamed define_insn): Delete.
12232 (movmemsi_6reg): Delete.
12233 (corresponding unnamed define_insn): Delete.
12234 (movmemsi_4reg): Delete.
12235 (corresponding unnamed define_insn): Delete.
12236 (movmemsi_2reg): Delete.
12237 (corresponding unnamed define_insn): Delete.
12238 (movmemsi_1reg): Delete.
12239 (corresponding unnamed define_insn): Delete.
12240 * config/rs6000/rs6000.opt (mno-string): New.
12241 (mstring): Replace by deprecation warning stub.
12242 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12244 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12246 * regrename.c (regrename_do_replace): If replacing the same
12247 reg multiple times, try to reuse last created gen_raw_REG.
12250 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12251 main to workaround a bug in GDB.
12253 2018-01-12 Tom de Vries <tom@codesourcery.com>
12256 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12258 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12260 PR rtl-optimization/80481
12261 * ira-color.c (get_cap_member): New function.
12262 (allocnos_conflict_by_live_ranges_p): Use it.
12263 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12264 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12266 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12269 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12270 (*saddl_se_1): Ditto.
12271 (*ssubsi_1): Ditto.
12272 (*ssubl_se_1): Ditto.
12274 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12276 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12277 rather than wi::to_widest for DR_INITs.
12278 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12279 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12280 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12282 (vect_analyze_group_access_1): Note that here.
12284 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12286 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12287 polynomial type sizes.
12289 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12291 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12292 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12293 (gimple_add_tmp_var): Likewise.
12295 2018-01-12 Martin Liska <mliska@suse.cz>
12297 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12298 (gimple_alloc_sizes): Likewise.
12299 (dump_gimple_statistics): Use PRIu64 in printf format.
12300 * gimple.h: Change uint64_t to int.
12302 2018-01-12 Martin Liska <mliska@suse.cz>
12304 * tree-core.h: Use uint64_t instead of int.
12305 * tree.c (tree_node_counts): Likewise.
12306 (tree_node_sizes): Likewise.
12307 (dump_tree_statistics): Use PRIu64 in printf format.
12309 2018-01-12 Martin Liska <mliska@suse.cz>
12311 * Makefile.in: As qsort_chk is implemented in vec.c, add
12312 vec.o to linkage of gencfn-macros.
12313 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12314 passing the info to record_node_allocation_statistics.
12315 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12317 * ggc-common.c (struct ggc_usage): Add operator== and use
12318 it in operator< and compare function.
12319 * mem-stats.h (struct mem_usage): Likewise.
12320 * vec.c (struct vec_usage): Remove operator< and compare
12321 function. Can be simply inherited.
12323 2018-01-12 Martin Jambor <mjambor@suse.cz>
12326 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12327 * tree-ssa-math-opts.c: Include domwalk.h.
12328 (convert_mult_to_fma_1): New function.
12329 (fma_transformation_info): New type.
12330 (fma_deferring_state): Likewise.
12331 (cancel_fma_deferring): New function.
12332 (result_of_phi): Likewise.
12333 (last_fma_candidate_feeds_initial_phi): Likewise.
12334 (convert_mult_to_fma): Added deferring logic, split actual
12335 transformation to convert_mult_to_fma_1.
12336 (math_opts_dom_walker): New type.
12337 (math_opts_dom_walker::after_dom_children): New method, body moved
12338 here from pass_optimize_widening_mul::execute, added deferring logic
12340 (pass_optimize_widening_mul::execute): Moved most of code to
12341 math_opts_dom_walker::after_dom_children.
12342 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12343 * config/i386/i386.c (ix86_option_override_internal): Added
12344 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12346 2018-01-12 Richard Biener <rguenther@suse.de>
12349 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12350 inline instance vars.
12352 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12355 * config/rx/rx.c (rx_is_restricted_memory_address):
12356 Handle SUBREG case.
12358 2018-01-12 Richard Biener <rguenther@suse.de>
12360 PR tree-optimization/80846
12361 * target.def (split_reduction): New target hook.
12362 * targhooks.c (default_split_reduction): New function.
12363 * targhooks.h (default_split_reduction): Declare.
12364 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12365 target requests first reduce vectors by combining low and high
12367 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12368 (get_vectype_for_scalar_type_and_size): Export.
12369 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12370 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12371 * doc/tm.texi: Regenerate.
12372 * config/i386/i386.c (ix86_split_reduction): Implement
12373 TARGET_VECTORIZE_SPLIT_REDUCTION.
12375 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12378 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12379 in PIC mode except for TARGET_VXWORKS_RTP.
12380 * config/sparc/sparc.c: Include cfgrtl.h.
12381 (TARGET_INIT_PIC_REG): Define.
12382 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12383 (sparc_pic_register_p): New predicate.
12384 (sparc_legitimate_address_p): Use it.
12385 (sparc_legitimize_pic_address): Likewise.
12386 (sparc_delegitimize_address): Likewise.
12387 (sparc_mode_dependent_address_p): Likewise.
12388 (gen_load_pcrel_sym): Remove 4th parameter.
12389 (load_got_register): Adjust call to above. Remove obsolete stuff.
12390 (sparc_expand_prologue): Do not call load_got_register here.
12391 (sparc_flat_expand_prologue): Likewise.
12392 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12393 (sparc_use_pseudo_pic_reg): New function.
12394 (sparc_init_pic_reg): Likewise.
12395 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12396 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12398 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12400 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12401 Add item for branch_cost.
12403 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12405 PR rtl-optimization/83565
12406 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12407 not extend the result to a larger mode for rotate operations.
12408 (num_sign_bit_copies1): Likewise.
12410 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12413 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12415 Use values-Xc.o for -pedantic.
12416 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12418 2018-01-12 Martin Liska <mliska@suse.cz>
12421 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12423 (possible_polymorphic_call_targets): Use it.
12424 (ipa_devirt): Likewise.
12426 2018-01-12 Martin Liska <mliska@suse.cz>
12428 * profile-count.h (enum profile_quality): Use 0 as invalid
12429 enum value of profile_quality.
12431 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12433 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12434 -mext-string options.
12436 2018-01-12 Richard Biener <rguenther@suse.de>
12438 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12439 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12440 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12442 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12444 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12446 * configure.ac (--with-long-double-format): Add support for the
12447 configuration option to change the default long double format on
12449 * config.gcc (powerpc*-linux*-*): Likewise.
12450 * configure: Regenerate.
12451 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12452 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12453 used without modification.
12455 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12457 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12458 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12459 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12460 MISC_BUILTIN_SPEC_BARRIER.
12461 (rs6000_init_builtins): Likewise.
12462 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12464 (speculation_barrier): New define_insn.
12465 * doc/extend.texi: Document __builtin_speculation_barrier.
12467 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12470 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12471 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12472 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12474 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12475 integral modes instead of "ss" and "sd".
12476 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12477 vectors with 32-bit and 64-bit elements.
12478 (vecdupssescalarmodesuffix): New mode attribute.
12479 (vec_dup<mode>): Use it.
12481 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12484 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12485 frame if argument is passed on stack.
12487 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12490 * ree.c (combine_reaching_defs): Optimize also
12491 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12492 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12494 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12496 PR middle-end/83189
12497 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12499 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12501 PR middle-end/83718
12502 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12503 after they are computed.
12505 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12507 PR tree-optimization/83695
12508 * gimple-loop-linterchange.cc
12509 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12510 reset cached scev information after interchange.
12511 (pass_linterchange::execute): Remove call to scev_reset_htab.
12513 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12515 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12516 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12517 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12518 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12519 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12520 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12521 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12522 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12523 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12524 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12525 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12526 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12527 (V_lane_reg): Likewise.
12528 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12530 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12531 (vfmal_lane_low<mode>_intrinsic,
12532 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12533 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12534 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12535 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12536 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12537 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12539 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12541 * config/arm/arm-cpus.in (fp16fml): New feature.
12542 (ALL_SIMD): Add fp16fml.
12543 (armv8.2-a): Add fp16fml as an option.
12544 (armv8.3-a): Likewise.
12545 (armv8.4-a): Add fp16fml as part of fp16.
12546 * config/arm/arm.h (TARGET_FP16FML): Define.
12547 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12549 * config/arm/arm-modes.def (V2HF): Define.
12550 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12551 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12552 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12553 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12554 vfmsl_low, vfmsl_high): New set of builtins.
12555 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12556 (vfml_op): New code attribute.
12557 (VFMLHALVES): New int iterator.
12558 (VFML, VFMLSEL): New mode attributes.
12559 (V_reg): Define mapping for V2HF.
12560 (V_hi, V_lo): New mode attributes.
12561 (VF_constraint): Likewise.
12562 (vfml_half, vfml_half_selector): New int attributes.
12563 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12565 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12566 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12568 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12569 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12570 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12571 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12573 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12574 Document new effective target and option set.
12576 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12578 * config/arm/arm-cpus.in (armv8_4): New feature.
12579 (ARMv8_4a): New fgroup.
12580 (armv8.4-a): New arch.
12581 * config/arm/arm-tables.opt: Regenerate.
12582 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12583 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12584 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12585 Add matching rules for -march=armv8.4-a and extensions.
12586 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12588 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12591 * config/rx/rx.md (BW): New mode attribute.
12592 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12594 2018-01-11 Richard Biener <rguenther@suse.de>
12596 PR tree-optimization/83435
12597 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12598 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12599 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12601 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12602 Alan Hayward <alan.hayward@arm.com>
12603 David Sherwood <david.sherwood@arm.com>
12605 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12607 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12608 (aarch64_print_address_internal): Use it to check for a zero offset.
12610 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12611 Alan Hayward <alan.hayward@arm.com>
12612 David Sherwood <david.sherwood@arm.com>
12614 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12615 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12616 Return a poly_int64 rather than a HOST_WIDE_INT.
12617 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12618 rather than a HOST_WIDE_INT.
12619 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12620 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12621 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12622 final_offset from HOST_WIDE_INT to poly_int64.
12623 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12624 to_constant when getting the number of units in an Advanced SIMD
12626 (aarch64_builtin_vectorized_function): Check for a constant number
12628 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12630 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12631 attribute instead of GET_MODE_NUNITS.
12632 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12633 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12634 GET_MODE_SIZE for fixed-size registers.
12635 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12636 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12637 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12638 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12639 (aarch64_print_operand, aarch64_print_address_internal)
12640 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12641 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12642 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12643 Handle polynomial GET_MODE_SIZE.
12644 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12645 wider than SImode without modification.
12646 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12647 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12648 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12649 passing and returning SVE modes.
12650 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12651 rather than GEN_INT.
12652 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12653 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12654 (aarch64_allocate_and_probe_stack_space): Likewise.
12655 (aarch64_layout_frame): Cope with polynomial offsets.
12656 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12657 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12658 polynomial offsets.
12659 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12660 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12661 poly_int64 rather than a HOST_WIDE_INT.
12662 (aarch64_get_separate_components, aarch64_process_components)
12663 (aarch64_expand_prologue, aarch64_expand_epilogue)
12664 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12665 (aarch64_anchor_offset): New function, split out from...
12666 (aarch64_legitimize_address): ...here.
12667 (aarch64_builtin_vectorization_cost): Handle polynomial
12668 TYPE_VECTOR_SUBPARTS.
12669 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12671 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12672 number of elements from the PARALLEL rather than the mode.
12673 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12674 rather than GET_MODE_BITSIZE.
12675 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12676 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12677 (aarch64_expand_vec_perm_const_1): Handle polynomial
12678 d->perm.length () and d->perm elements.
12679 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12680 Apply to_constant to d->perm elements.
12681 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12682 polynomial CONST_VECTOR_NUNITS.
12683 (aarch64_move_pointer): Take amount as a poly_int64 rather
12685 (aarch64_progress_pointer): Avoid temporary variable.
12686 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12687 the mode attribute instead of GET_MODE.
12689 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12690 Alan Hayward <alan.hayward@arm.com>
12691 David Sherwood <david.sherwood@arm.com>
12693 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12694 x exists before using it.
12695 (aarch64_add_constant_internal): Rename to...
12696 (aarch64_add_offset_1): ...this. Replace regnum with separate
12697 src and dest rtxes. Handle the case in which they're different,
12698 including when the offset is zero. Replace scratchreg with an rtx.
12699 Use 2 additions if there is no spare register into which we can
12700 move a 16-bit constant.
12701 (aarch64_add_constant): Delete.
12702 (aarch64_add_offset): Replace reg with separate src and dest
12703 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12704 Use aarch64_add_offset_1.
12705 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12706 an rtx rather than an int. Take the delta as a poly_int64
12707 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12708 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12709 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12710 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12711 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12712 and aarch64_add_sp.
12713 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12714 aarch64_add_constant.
12716 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12718 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12719 Use scalar_float_mode.
12721 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12723 * config/aarch64/aarch64-simd.md
12724 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12725 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12726 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12727 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12728 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12729 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12730 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12731 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12732 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12733 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12735 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12738 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12739 targ_options->x_arm_arch_string is non NULL.
12741 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12743 * config/aarch64/aarch64.h
12744 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12746 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12749 * expmed.c (emit_store_flag_force): Swap if const op0
12750 and change VOIDmode to mode of op0.
12752 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12754 PR rtl-optimization/83761
12755 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12756 than bytes to mode_for_size.
12758 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12760 PR middle-end/83189
12761 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12762 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12765 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12767 PR middle-end/83575
12768 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12769 when in layout mode.
12770 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12771 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12774 2018-01-10 Michael Collison <michael.collison@arm.com>
12776 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12777 * config/aarch64/aarch64-option-extension.def: Add
12778 AARCH64_OPT_EXTENSION of 'fp16fml'.
12779 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12780 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12781 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12782 * config/aarch64/constraints.md (Ui7): New constraint.
12783 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12784 (VFMLA_SEL_W): Ditto.
12787 (VFMLA16_LOW): New int iterator.
12788 (VFMLA16_HIGH): Ditto.
12789 (UNSPEC_FMLAL): New unspec.
12790 (UNSPEC_FMLSL): Ditto.
12791 (UNSPEC_FMLAL2): Ditto.
12792 (UNSPEC_FMLSL2): Ditto.
12793 (f16mac): New code attribute.
12794 * config/aarch64/aarch64-simd-builtins.def
12795 (aarch64_fmlal_lowv2sf): Ditto.
12796 (aarch64_fmlsl_lowv2sf): Ditto.
12797 (aarch64_fmlalq_lowv4sf): Ditto.
12798 (aarch64_fmlslq_lowv4sf): Ditto.
12799 (aarch64_fmlal_highv2sf): Ditto.
12800 (aarch64_fmlsl_highv2sf): Ditto.
12801 (aarch64_fmlalq_highv4sf): Ditto.
12802 (aarch64_fmlslq_highv4sf): Ditto.
12803 (aarch64_fmlal_lane_lowv2sf): Ditto.
12804 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12805 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12806 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12807 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12808 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12809 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12810 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12811 (aarch64_fmlal_lane_highv2sf): Ditto.
12812 (aarch64_fmlsl_lane_highv2sf): Ditto.
12813 (aarch64_fmlal_laneq_highv2sf): Ditto.
12814 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12815 (aarch64_fmlalq_lane_highv4sf): Ditto.
12816 (aarch64_fmlsl_lane_highv4sf): Ditto.
12817 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12818 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12819 * config/aarch64/aarch64-simd.md:
12820 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12821 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12822 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12823 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12824 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12825 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12826 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12827 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12828 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12829 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12830 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12831 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12832 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12833 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12834 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12835 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12836 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12837 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12838 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12839 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12840 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12841 (vfmlsl_low_u32): Ditto.
12842 (vfmlalq_low_u32): Ditto.
12843 (vfmlslq_low_u32): Ditto.
12844 (vfmlal_high_u32): Ditto.
12845 (vfmlsl_high_u32): Ditto.
12846 (vfmlalq_high_u32): Ditto.
12847 (vfmlslq_high_u32): Ditto.
12848 (vfmlal_lane_low_u32): Ditto.
12849 (vfmlsl_lane_low_u32): Ditto.
12850 (vfmlal_laneq_low_u32): Ditto.
12851 (vfmlsl_laneq_low_u32): Ditto.
12852 (vfmlalq_lane_low_u32): Ditto.
12853 (vfmlslq_lane_low_u32): Ditto.
12854 (vfmlalq_laneq_low_u32): Ditto.
12855 (vfmlslq_laneq_low_u32): Ditto.
12856 (vfmlal_lane_high_u32): Ditto.
12857 (vfmlsl_lane_high_u32): Ditto.
12858 (vfmlal_laneq_high_u32): Ditto.
12859 (vfmlsl_laneq_high_u32): Ditto.
12860 (vfmlalq_lane_high_u32): Ditto.
12861 (vfmlslq_lane_high_u32): Ditto.
12862 (vfmlalq_laneq_high_u32): Ditto.
12863 (vfmlslq_laneq_high_u32): Ditto.
12864 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12865 (AARCH64_FL_FOR_ARCH8_4): New.
12866 (AARCH64_ISA_F16FML): New ISA flag.
12867 (TARGET_F16FML): New feature flag for fp16fml.
12868 (doc/invoke.texi): Document new fp16fml option.
12870 2018-01-10 Michael Collison <michael.collison@arm.com>
12872 * config/aarch64/aarch64-builtins.c:
12873 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12874 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12875 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12876 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12877 (AARCH64_ISA_SHA3): New ISA flag.
12878 (TARGET_SHA3): New feature flag for sha3.
12879 * config/aarch64/iterators.md (sha512_op): New int attribute.
12880 (CRYPTO_SHA512): New int iterator.
12881 (UNSPEC_SHA512H): New unspec.
12882 (UNSPEC_SHA512H2): Ditto.
12883 (UNSPEC_SHA512SU0): Ditto.
12884 (UNSPEC_SHA512SU1): Ditto.
12885 * config/aarch64/aarch64-simd-builtins.def
12886 (aarch64_crypto_sha512hqv2di): New builtin.
12887 (aarch64_crypto_sha512h2qv2di): Ditto.
12888 (aarch64_crypto_sha512su0qv2di): Ditto.
12889 (aarch64_crypto_sha512su1qv2di): Ditto.
12890 (aarch64_eor3qv8hi): Ditto.
12891 (aarch64_rax1qv2di): Ditto.
12892 (aarch64_xarqv2di): Ditto.
12893 (aarch64_bcaxqv8hi): Ditto.
12894 * config/aarch64/aarch64-simd.md:
12895 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12896 (aarch64_crypto_sha512su0qv2di): Ditto.
12897 (aarch64_crypto_sha512su1qv2di): Ditto.
12898 (aarch64_eor3qv8hi): Ditto.
12899 (aarch64_rax1qv2di): Ditto.
12900 (aarch64_xarqv2di): Ditto.
12901 (aarch64_bcaxqv8hi): Ditto.
12902 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12903 (vsha512h2q_u64): Ditto.
12904 (vsha512su0q_u64): Ditto.
12905 (vsha512su1q_u64): Ditto.
12906 (veor3q_u16): Ditto.
12907 (vrax1q_u64): Ditto.
12908 (vxarq_u64): Ditto.
12909 (vbcaxq_u16): Ditto.
12910 * config/arm/types.md (crypto_sha512): New type attribute.
12911 (crypto_sha3): Ditto.
12912 (doc/invoke.texi): Document new sha3 option.
12914 2018-01-10 Michael Collison <michael.collison@arm.com>
12916 * config/aarch64/aarch64-builtins.c:
12917 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12918 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12919 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12920 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12921 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12922 (AARCH64_ISA_SM4): New ISA flag.
12923 (TARGET_SM4): New feature flag for sm4.
12924 * config/aarch64/aarch64-simd-builtins.def
12925 (aarch64_sm3ss1qv4si): Ditto.
12926 (aarch64_sm3tt1aq4si): Ditto.
12927 (aarch64_sm3tt1bq4si): Ditto.
12928 (aarch64_sm3tt2aq4si): Ditto.
12929 (aarch64_sm3tt2bq4si): Ditto.
12930 (aarch64_sm3partw1qv4si): Ditto.
12931 (aarch64_sm3partw2qv4si): Ditto.
12932 (aarch64_sm4eqv4si): Ditto.
12933 (aarch64_sm4ekeyqv4si): Ditto.
12934 * config/aarch64/aarch64-simd.md:
12935 (aarch64_sm3ss1qv4si): Ditto.
12936 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12937 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12938 (aarch64_sm4eqv4si): Ditto.
12939 (aarch64_sm4ekeyqv4si): Ditto.
12940 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12941 (sm3part_op): Ditto.
12942 (CRYPTO_SM3TT): Ditto.
12943 (CRYPTO_SM3PART): Ditto.
12944 (UNSPEC_SM3SS1): New unspec.
12945 (UNSPEC_SM3TT1A): Ditto.
12946 (UNSPEC_SM3TT1B): Ditto.
12947 (UNSPEC_SM3TT2A): Ditto.
12948 (UNSPEC_SM3TT2B): Ditto.
12949 (UNSPEC_SM3PARTW1): Ditto.
12950 (UNSPEC_SM3PARTW2): Ditto.
12951 (UNSPEC_SM4E): Ditto.
12952 (UNSPEC_SM4EKEY): Ditto.
12953 * config/aarch64/constraints.md (Ui2): New constraint.
12954 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12955 * config/arm/types.md (crypto_sm3): New type attribute.
12956 (crypto_sm4): Ditto.
12957 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12958 (vsm3tt1aq_u32): Ditto.
12959 (vsm3tt1bq_u32): Ditto.
12960 (vsm3tt2aq_u32): Ditto.
12961 (vsm3tt2bq_u32): Ditto.
12962 (vsm3partw1q_u32): Ditto.
12963 (vsm3partw2q_u32): Ditto.
12964 (vsm4eq_u32): Ditto.
12965 (vsm4ekeyq_u32): Ditto.
12966 (doc/invoke.texi): Document new sm4 option.
12968 2018-01-10 Michael Collison <michael.collison@arm.com>
12970 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12971 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12972 (AARCH64_FL_FOR_ARCH8_4): New.
12973 (AARCH64_FL_V8_4): New flag.
12974 (doc/invoke.texi): Document new armv8.4-a option.
12976 2018-01-10 Michael Collison <michael.collison@arm.com>
12978 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12979 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12980 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12981 * config/aarch64/aarch64-option-extension.def: Add
12982 AARCH64_OPT_EXTENSION of 'sha2'.
12983 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12984 (crypto): Disable sha2 and aes if crypto disabled.
12985 (crypto): Enable aes and sha2 if enabled.
12986 (simd): Disable sha2 and aes if simd disabled.
12987 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12989 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12990 (TARGET_SHA2): New feature flag for sha2.
12991 (TARGET_AES): New feature flag for aes.
12992 * config/aarch64/aarch64-simd.md:
12993 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12994 conditional on TARGET_AES.
12995 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12996 (aarch64_crypto_sha1hsi): Make pattern conditional
12998 (aarch64_crypto_sha1hv4si): Ditto.
12999 (aarch64_be_crypto_sha1hv4si): Ditto.
13000 (aarch64_crypto_sha1su1v4si): Ditto.
13001 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
13002 (aarch64_crypto_sha1su0v4si): Ditto.
13003 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
13004 (aarch64_crypto_sha256su0v4si): Ditto.
13005 (aarch64_crypto_sha256su1v4si): Ditto.
13006 (doc/invoke.texi): Document new aes and sha2 options.
13008 2018-01-10 Martin Sebor <msebor@redhat.com>
13010 PR tree-optimization/83781
13011 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
13014 2018-01-11 Martin Sebor <msebor@gmail.com>
13015 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13017 PR tree-optimization/83501
13018 PR tree-optimization/81703
13020 * tree-ssa-strlen.c (get_string_cst): Rename...
13021 (get_string_len): ...to this. Handle global constants.
13022 (handle_char_store): Adjust.
13024 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
13025 Jim Wilson <jimw@sifive.com>
13027 * config/riscv/riscv-protos.h (riscv_output_return): New.
13028 * config/riscv/riscv.c (struct machine_function): New naked_p field.
13029 (riscv_attribute_table, riscv_output_return),
13030 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
13031 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
13032 (riscv_compute_frame_info): Only compute frame->mask if not a naked
13034 (riscv_expand_prologue): Add early return for naked function.
13035 (riscv_expand_epilogue): Likewise.
13036 (riscv_function_ok_for_sibcall): Return false for naked function.
13037 (riscv_set_current_function): New.
13038 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13039 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13040 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13041 * doc/extend.texi (RISC-V Function Attributes): New.
13043 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13045 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13046 check for 128-bit long double before checking TCmode.
13047 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13048 128-bit long doubles before checking TFmode or TCmode.
13049 (FLOAT128_IBM_P): Likewise.
13051 2018-01-10 Martin Sebor <msebor@redhat.com>
13053 PR tree-optimization/83671
13054 * builtins.c (c_strlen): Unconditionally return zero for the empty
13056 Use -Warray-bounds for warnings.
13057 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13058 for non-constant array indices with COMPONENT_REF, arrays of
13059 arrays, and pointers to arrays.
13060 (gimple_fold_builtin_strlen): Determine and set length range for
13061 non-constant character arrays.
13063 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13065 PR middle-end/81897
13066 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13069 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13071 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13073 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13076 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13077 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13078 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13079 indexed_or_indirect_operand predicate.
13080 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13081 (*vsx_le_perm_load_v8hi): Likewise.
13082 (*vsx_le_perm_load_v16qi): Likewise.
13083 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13084 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13085 (*vsx_le_perm_store_v8hi): Likewise.
13086 (*vsx_le_perm_store_v16qi): Likewise.
13087 (eight unnamed splitters): Likewise.
13089 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13091 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13092 * config/rs6000/emmintrin.h: Likewise.
13093 * config/rs6000/mmintrin.h: Likewise.
13094 * config/rs6000/xmmintrin.h: Likewise.
13096 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13099 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13101 * tree.c (tree_nop_conversion): Return true for location wrapper
13103 (maybe_wrap_with_location): New function.
13104 (selftest::check_strip_nops): New function.
13105 (selftest::test_location_wrappers): New function.
13106 (selftest::tree_c_tests): Call it.
13107 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13108 (maybe_wrap_with_location): New decl.
13109 (EXPR_LOCATION_WRAPPER_P): New macro.
13110 (location_wrapper_p): New inline function.
13111 (tree_strip_any_location_wrapper): New inline function.
13113 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13116 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13117 stack_realign_offset for the largest alignment of stack slot
13119 (ix86_find_max_used_stack_alignment): New function.
13120 (ix86_finalize_stack_frame_flags): Use it. Set
13121 max_used_stack_alignment if we don't realign stack.
13122 * config/i386/i386.h (machine_function): Add
13123 max_used_stack_alignment.
13125 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13127 * config/arm/arm.opt (-mbranch-cost): New option.
13128 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13131 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13134 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13135 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13137 2018-01-10 Richard Biener <rguenther@suse.de>
13140 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13141 early out so it also covers the case where we have a non-NULL
13144 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13146 PR tree-optimization/83753
13147 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13148 for non-strided grouped accesses if the number of elements is 1.
13150 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13153 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13154 * i386.h (TARGET_USE_GATHER): Define.
13155 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13157 2018-01-10 Martin Liska <mliska@suse.cz>
13160 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13161 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13163 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13164 CLEANUP_NO_PARTITIONING is not set.
13166 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13168 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13169 for vectors, as a partial revert of r254296.
13170 * rtl.h (const_vec_p): Delete.
13171 (const_vec_duplicate_p): Don't test for vector CONSTs.
13172 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13173 * expmed.c (make_tree): Likewise.
13176 * common.md (E, F): Use CONSTANT_P instead of checking for
13178 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13179 checking for CONST_VECTOR.
13181 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13183 PR middle-end/83575
13184 * predict.c (force_edge_cold): Handle in more sane way edges
13185 with no prediction.
13187 2018-01-09 Carl Love <cel@us.ibm.com>
13189 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13191 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13192 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13193 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13194 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13195 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13196 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13197 * config/rs6000/rs6000-protos.h: Add extern defition for
13198 rs6000_generate_float2_double_code.
13199 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13201 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13202 (float2_v2df): Add define_expand.
13204 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13207 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13208 op_mode in the force_to_mode call.
13210 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13212 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13213 instead of checking each element individually.
13214 (aarch64_evpc_uzp): Likewise.
13215 (aarch64_evpc_zip): Likewise.
13216 (aarch64_evpc_ext): Likewise.
13217 (aarch64_evpc_rev): Likewise.
13218 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13219 instead of checking each element individually. Return true without
13221 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13222 whether all selected elements come from the same input, instead of
13223 checking each element individually. Remove calls to gen_rtx_REG,
13224 start_sequence and end_sequence and instead assert that no rtl is
13227 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13229 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13230 order of HIGH and CONST checks.
13232 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13234 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13235 if the destination isn't an SSA_NAME.
13237 2018-01-09 Richard Biener <rguenther@suse.de>
13239 PR tree-optimization/83668
13240 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13242 (canonicalize_loop_form): ... here, renamed from ...
13243 (canonicalize_loop_closed_ssa_form): ... this and amended to
13244 swap successor edges for loop exit blocks to make us use
13245 the RPO order we need for initial schedule generation.
13247 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13249 PR tree-optimization/64811
13250 * match.pd: When optimizing comparisons with Inf, avoid
13251 introducing or losing exceptions from comparisons with NaN.
13253 2018-01-09 Martin Liska <mliska@suse.cz>
13256 * asan.c (shadow_mem_size): Add gcc_assert.
13258 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13260 Don't save registers in main().
13263 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13264 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13265 * config/avr/avr.c (avr_set_current_function): Don't error if
13266 naked, OS_task or OS_main are specified at the same time.
13267 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13269 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13271 * common/config/avr/avr-common.c (avr_option_optimization_table):
13272 Switch on -mmain-is-OS_task for optimizing compilations.
13274 2018-01-09 Richard Biener <rguenther@suse.de>
13276 PR tree-optimization/83572
13277 * graphite.c: Include cfganal.h.
13278 (graphite_transform_loops): Connect infinite loops to exit
13279 and remove fake edges at the end.
13281 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13283 * ipa-inline.c (edge_badness): Revert accidental checkin.
13285 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13288 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13289 symbols; not inline clones.
13291 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13294 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13295 hard registers. Formatting fixes.
13297 PR preprocessor/83722
13298 * gcc.c (try_generate_repro): Pass
13299 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13300 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13303 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13304 Kito Cheng <kito.cheng@gmail.com>
13306 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13307 (riscv_leaf_function_p): Delete.
13308 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13310 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13312 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13314 (do_ifelse): New function.
13315 (do_isel): New function.
13316 (do_sub3): New function.
13317 (do_add3): New function.
13318 (do_load_mask_compare): New function.
13319 (do_overlap_load_compare): New function.
13320 (expand_compare_loop): New function.
13321 (expand_block_compare): Call expand_compare_loop() when appropriate.
13322 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13323 option description.
13324 (-mblock-compare-inline-loop-limit): New option.
13326 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13329 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13330 Reverse order of second and third operands in first alternative.
13331 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13332 of first and second elements in UNSPEC_VPERMR vector.
13333 (altivec_expand_vec_perm_le): Likewise.
13335 2018-01-08 Jeff Law <law@redhat.com>
13337 PR rtl-optimizatin/81308
13338 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13339 (process_switch): If group_case_labels makes a change, then set
13341 (pass_convert_switch::execute): If a switch is converted, then
13342 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13344 PR rtl-optimization/81308
13345 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13348 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13350 PR target/83663 - Revert r255946
13351 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13352 generation for cases where splatting a value is not useful.
13353 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13354 across a vec_duplicate and a paradoxical subreg forming a vector
13355 mode to a vec_concat.
13357 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13359 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13360 -march=armv8.3-a variants.
13361 * config/arm/t-multilib: Likewise.
13362 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13364 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13366 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13368 (cceq_ior_compare_complement): Give it a name so I can use it, and
13369 change boolean_or_operator predicate to boolean_operator so it can
13370 be used to generate a crand.
13371 (eqne): New code iterator.
13372 (bd/bd_neg): New code_attrs.
13373 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13374 a single define_insn.
13375 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13376 decrement (bdnzt/bdnzf/bdzt/bdzf).
13377 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13378 with the new names of the branch decrement patterns, and added the
13379 names of the branch decrement conditional patterns.
13381 2018-01-08 Richard Biener <rguenther@suse.de>
13383 PR tree-optimization/83563
13384 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13387 2018-01-08 Richard Biener <rguenther@suse.de>
13389 PR middle-end/83713
13390 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13392 2018-01-08 Richard Biener <rguenther@suse.de>
13394 PR tree-optimization/83685
13395 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13396 references to abnormals.
13398 2018-01-08 Richard Biener <rguenther@suse.de>
13401 * dwarf2out.c (output_indirect_strings): Handle empty
13402 skeleton_debug_str_hash.
13403 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13405 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13407 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13408 (emit_store_direct): Likewise.
13409 (arc_trampoline_adjust_address): Likewise.
13410 (arc_asm_trampoline_template): New function.
13411 (arc_initialize_trampoline): Use asm_trampoline_template.
13412 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13413 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13414 * config/arc/arc.md (flush_icache): Delete pattern.
13416 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13418 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13419 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13422 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13425 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13426 by not USED_FOR_TARGET.
13427 (make_pass_resolve_sw_modes): Likewise.
13429 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13431 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13434 2018-01-08 Richard Biener <rguenther@suse.de>
13436 PR middle-end/83580
13437 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13439 2018-01-08 Richard Biener <rguenther@suse.de>
13441 PR middle-end/83517
13442 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13444 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13446 PR middle-end/81897
13447 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13448 basic blocks with a small number of successors.
13449 (convert_control_dep_chain_into_preds): Improve handling of
13451 (dump_predicates): Split apart into...
13452 (dump_pred_chain): ...here...
13453 (dump_pred_info): ...and here.
13454 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13455 (can_chain_union_be_invalidated_p): Improve check for invalidation
13457 (uninit_uses_cannot_happen): Avoid unnecessary if
13458 convert_control_dep_chain_into_preds yielded nothing.
13460 2018-01-06 Martin Sebor <msebor@redhat.com>
13462 PR tree-optimization/83640
13463 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13464 subtracting negative offset from size.
13465 (builtin_access::overlap): Adjust offset bounds of the access to fall
13466 within the size of the object if possible.
13468 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13470 PR rtl-optimization/83699
13471 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13472 extract_bit_field_as_subreg to cases in which the extracted
13473 value is also a vector.
13475 * lra-constraints.c (process_alt_operands): Test for the equivalence
13476 substitutions when detecting a possible reload cycle.
13478 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13481 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13482 by default if flag_selective_schedling{,2}. Formatting fixes.
13484 PR rtl-optimization/83682
13485 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13486 if it has non-VECTOR_MODE element mode.
13487 (vec_duplicate_p): Likewise.
13489 PR middle-end/83694
13490 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13491 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13493 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13496 * config/i386/i386-builtin.def
13497 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13498 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13499 Require also OPTION_MASK_ISA_AVX512F in addition to
13500 OPTION_MASK_ISA_GFNI.
13501 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13502 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13503 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13504 to OPTION_MASK_ISA_GFNI.
13505 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13506 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13507 OPTION_MASK_ISA_AVX512BW.
13508 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13509 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13510 addition to OPTION_MASK_ISA_GFNI.
13511 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13512 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13513 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13514 to OPTION_MASK_ISA_GFNI.
13515 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13516 a requirement for all ISAs rather than any of them with a few
13518 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13520 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13521 bitmasks to be enabled with 3 exceptions, instead of requiring any
13522 enabled ISA with lots of exceptions.
13523 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13524 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13525 Change avx512bw in isa attribute to avx512f.
13526 * config/i386/sgxintrin.h: Add license boilerplate.
13527 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13528 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13529 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13530 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13532 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13533 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13534 temporarily sse2 rather than sse if not enabled already.
13537 * config/i386/sse.md (VI248_VLBW): Rename to ...
13538 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13539 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13540 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13541 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13542 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13543 mode iterator instead of VI248_VLBW.
13545 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13547 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13548 (record_modified): Skip clobbers; add debug output.
13549 (param_change_prob): Use sreal frequencies.
13551 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13553 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13554 punt for user-aligned variables.
13556 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13558 * tree-chrec.c (chrec_contains_symbols): Return true for
13561 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13564 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13565 of (x|y) == x for BICS pattern.
13567 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13569 PR tree-optimization/83605
13570 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13571 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13574 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13576 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13577 * config/epiphany/rtems.h: New file.
13579 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13580 Uros Bizjak <ubizjak@gmail.com>
13583 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13584 QIreg_operand instead of register_operand predicate.
13585 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13586 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13587 comments instead of -fmitigate[-_]rop.
13589 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13592 * cgraphunit.c (symbol_table::compile): Switch to text_section
13593 before calling assembly_start debug hook.
13594 * run-rtl-passes.c (run_rtl_passes): Likewise.
13597 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13599 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13600 range_int_cst_p rather than !symbolic_range_p before calling
13601 extract_range_from_multiplicative_op_1.
13603 2018-01-04 Jeff Law <law@redhat.com>
13605 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13606 redundant test in assertion.
13608 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13610 * doc/rtl.texi: Document machine_mode wrapper classes.
13612 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13614 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13615 using tree_to_uhwi.
13617 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13619 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13620 the VEC_PERM_EXPR fold to fail.
13622 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13625 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13626 to switched_sections.
13628 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13631 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13632 test for d.testing.
13634 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13637 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13638 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13640 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13643 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13644 is BLKmode and bitpos not zero or mode change is needed.
13646 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13649 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13652 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13655 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13656 instead of MULT rtx. Update all corresponding splitters.
13657 (*saddl_se): Ditto.
13658 (*ssub<modesuffix>): Ditto.
13659 (*ssubl_se): Ditto.
13660 (*cmp_sadd_di): Update split patterns.
13661 (*cmp_sadd_si): Ditto.
13662 (*cmp_sadd_sidi): Ditto.
13663 (*cmp_ssub_di): Ditto.
13664 (*cmp_ssub_si): Ditto.
13665 (*cmp_ssub_sidi): Ditto.
13666 * config/alpha/predicates.md (const23_operand): New predicate.
13667 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13668 Look for ASHIFT, not MULT inner operand.
13669 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13671 2018-01-04 Martin Liska <mliska@suse.cz>
13673 PR gcov-profile/83669
13674 * gcov.c (output_intermediate_file): Add version to intermediate
13676 * doc/gcov.texi: Document new field 'version' in intermediate
13677 file format. Fix location of '-k' option of gcov command.
13679 2018-01-04 Martin Liska <mliska@suse.cz>
13682 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13684 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13686 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13688 2018-01-03 Martin Sebor <msebor@redhat.com>
13690 PR tree-optimization/83655
13691 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13692 checking calls with invalid arguments.
13694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13696 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13697 (vectorizable_mask_load_store): Delete.
13698 (vectorizable_call): Return false for masked loads and stores.
13699 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13700 instead of gimple_assign_rhs1.
13701 (vectorizable_load): Handle IFN_MASK_LOAD.
13702 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13706 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13708 (vectorizable_mask_load_store): ...here.
13709 (vectorizable_load): ...and here.
13711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13713 * tree-vect-stmts.c (vect_build_all_ones_mask)
13714 (vect_build_zero_merge_argument): New functions, split out from...
13715 (vectorizable_load): ...here.
13717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13719 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13721 (vectorizable_mask_load_store): ...here.
13722 (vectorizable_store): ...and here.
13724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13726 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13728 (vectorizable_mask_load_store): ...here.
13730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13732 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13733 (vect_model_store_cost): Take a vec_load_store_type instead of a
13735 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13736 (vect_model_store_cost): Take a vec_load_store_type instead of a
13738 (vectorizable_mask_load_store): Update accordingly.
13739 (vectorizable_store): Likewise.
13740 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13744 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13745 IFN_MASK_LOAD calls here rather than...
13746 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13748 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13749 Alan Hayward <alan.hayward@arm.com>
13750 David Sherwood <david.sherwood@arm.com>
13752 * expmed.c (extract_bit_field_1): For vector extracts,
13753 fall back to extract_bit_field_as_subreg if vec_extract
13756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13757 Alan Hayward <alan.hayward@arm.com>
13758 David Sherwood <david.sherwood@arm.com>
13760 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13761 they are variable or constant sized.
13762 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13763 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13770 handling COND_EXPRs with boolean comparisons, try to find a better
13771 basis for the mask type than the boolean itself.
13773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13775 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13776 is calculated and how it can be overridden.
13777 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13778 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13780 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13784 Alan Hayward <alan.hayward@arm.com>
13785 David Sherwood <david.sherwood@arm.com>
13787 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13788 Remove the mode argument.
13789 (aarch64_simd_valid_immediate): Remove the mode and inverse
13791 * config/aarch64/iterators.md (bitsize): New iterator.
13792 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13793 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13794 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13795 aarch64_simd_valid_immediate.
13796 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13797 (aarch64_reg_or_bic_imm): Likewise.
13798 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13799 with an insn_type enum and msl with a modifier_type enum.
13800 Replace element_width with a scalar_mode. Change the shift
13801 to unsigned int. Add constructors for scalar_float_mode and
13802 scalar_int_mode elements.
13803 (aarch64_vect_float_const_representable_p): Delete.
13804 (aarch64_can_const_movi_rtx_p)
13805 (aarch64_simd_scalar_immediate_valid_for_move)
13806 (aarch64_simd_make_constant): Update call to
13807 aarch64_simd_valid_immediate.
13808 (aarch64_advsimd_valid_immediate_hs): New function.
13809 (aarch64_advsimd_valid_immediate): Likewise.
13810 (aarch64_simd_valid_immediate): Remove mode and inverse
13811 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13812 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13813 and aarch64_float_const_representable_p on the result.
13814 (aarch64_output_simd_mov_immediate): Remove mode argument.
13815 Update call to aarch64_simd_valid_immediate and use of
13816 simd_immediate_info.
13817 (aarch64_output_scalar_simd_mov_immediate): Update call
13820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13821 Alan Hayward <alan.hayward@arm.com>
13822 David Sherwood <david.sherwood@arm.com>
13824 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13825 (mode_nunits): Likewise CONST_MODE_NUNITS.
13826 * machmode.def (ADJUST_NUNITS): Document.
13827 * genmodes.c (mode_data::need_nunits_adj): New field.
13828 (blank_mode): Update accordingly.
13829 (adj_nunits): New variable.
13830 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13832 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13833 listed in adj_nunits.
13834 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13835 listed in adj_nunits. Don't emit case statements for such modes.
13836 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13837 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13838 nothing if adj_nunits is nonnull.
13839 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13840 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13841 (emit_mode_fbit): Update use of print_maybe_const_decl.
13842 (emit_move_size): Likewise. Treat the array as non-const
13844 (emit_mode_adjustments): Handle adj_nunits.
13846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13848 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13849 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13850 (VECTOR_MODES): Use it.
13851 (make_vector_modes): Take the prefix as an argument.
13853 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13854 Alan Hayward <alan.hayward@arm.com>
13855 David Sherwood <david.sherwood@arm.com>
13857 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13858 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13859 for MODE_VECTOR_BOOL.
13860 * machmode.def (VECTOR_BOOL_MODE): Document.
13861 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13862 (make_vector_bool_mode): New function.
13863 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13865 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13866 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13868 * stor-layout.c (int_mode_for_mode): Likewise.
13869 * tree.c (build_vector_type_for_mode): Likewise.
13870 * varasm.c (output_constant_pool_2): Likewise.
13871 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13872 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13873 for MODE_VECTOR_BOOL.
13874 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13875 of mode class checks.
13876 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13877 instead of a list of mode class checks.
13878 (expand_vector_scalar_condition): Likewise.
13879 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13882 Alan Hayward <alan.hayward@arm.com>
13883 David Sherwood <david.sherwood@arm.com>
13885 * machmode.h (mode_size): Change from unsigned short to
13887 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13888 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13889 or if measurement_type is not polynomial.
13890 (fixed_size_mode::includes_p): Check for constant-sized modes.
13891 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13892 return a poly_uint16 rather than an unsigned short.
13893 (emit_mode_size): Change the type of mode_size from unsigned short
13894 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13895 (emit_mode_adjustments): Cope with polynomial vector sizes.
13896 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13898 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13900 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13901 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13902 * caller-save.c (setup_save_areas): Likewise.
13903 (replace_reg_with_saved_mem): Likewise.
13904 * calls.c (emit_library_call_value_1): Likewise.
13905 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13906 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13907 (gen_lowpart_for_combine): Likewise.
13908 * convert.c (convert_to_integer_1): Likewise.
13909 * cse.c (equiv_constant, cse_insn): Likewise.
13910 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13911 (cselib_subst_to_values): Likewise.
13912 * dce.c (word_dce_process_block): Likewise.
13913 * df-problems.c (df_word_lr_mark_ref): Likewise.
13914 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13915 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13916 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13917 (rtl_for_decl_location): Likewise.
13918 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13919 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13920 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13921 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13922 (expand_expr_real_1): Likewise.
13923 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13924 (pad_below): Likewise.
13925 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13926 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13927 * ira.c (get_subreg_tracking_sizes): Likewise.
13928 * ira-build.c (ira_create_allocno_objects): Likewise.
13929 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13930 (ira_sort_regnos_for_alter_reg): Likewise.
13931 * ira-costs.c (record_operand_costs): Likewise.
13932 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13933 (resolve_simple_move): Likewise.
13934 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13935 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13936 (lra_constraints): Likewise.
13937 (CONST_POOL_OK_P): Reject variable-sized modes.
13938 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13939 (add_pseudo_to_slot, lra_spill): Likewise.
13940 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13941 * optabs-query.c (get_best_extraction_insn): Likewise.
13942 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13943 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13944 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13945 * recog.c (offsettable_address_addr_space_p): Likewise.
13946 * regcprop.c (maybe_mode_change): Likewise.
13947 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13948 * regrename.c (build_def_use): Likewise.
13949 * regstat.c (dump_reg_info): Likewise.
13950 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13951 (find_reloads, find_reloads_subreg_address): Likewise.
13952 * reload1.c (eliminate_regs_1): Likewise.
13953 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13954 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13955 (simplify_binary_operation_1, simplify_subreg): Likewise.
13956 * targhooks.c (default_function_arg_padding): Likewise.
13957 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13958 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13959 (verify_gimple_assign_ternary): Likewise.
13960 * tree-inline.c (estimate_move_cost): Likewise.
13961 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13962 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13963 (get_address_cost_ainc): Likewise.
13964 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13965 (vect_supportable_dr_alignment): Likewise.
13966 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13967 (vectorizable_reduction): Likewise.
13968 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13969 (vectorizable_operation, vectorizable_load): Likewise.
13970 * tree.c (build_same_sized_truth_vector_type): Likewise.
13971 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13972 * var-tracking.c (emit_note_insn_var_location): Likewise.
13973 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13974 (ADDR_VEC_ALIGN): Likewise.
13976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13977 Alan Hayward <alan.hayward@arm.com>
13978 David Sherwood <david.sherwood@arm.com>
13980 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13982 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13983 or if measurement_type is polynomial.
13984 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13985 * combine.c (make_extraction): Likewise.
13986 * dse.c (find_shift_sequence): Likewise.
13987 * dwarf2out.c (mem_loc_descriptor): Likewise.
13988 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13989 (extract_bit_field, extract_low_bits): Likewise.
13990 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13991 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13992 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13993 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13994 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13995 * reload.c (find_reloads): Likewise.
13996 * reload1.c (alter_reg): Likewise.
13997 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13998 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13999 * tree-if-conv.c (predicate_mem_writes): Likewise.
14000 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
14001 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
14002 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
14003 * valtrack.c (dead_debug_insert_temp): Likewise.
14004 * varasm.c (mergeable_constant_section): Likewise.
14005 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
14007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14008 Alan Hayward <alan.hayward@arm.com>
14009 David Sherwood <david.sherwood@arm.com>
14011 * expr.c (expand_assignment): Cope with polynomial mode sizes
14012 when assigning to a CONCAT.
14014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14015 Alan Hayward <alan.hayward@arm.com>
14016 David Sherwood <david.sherwood@arm.com>
14018 * machmode.h (mode_precision): Change from unsigned short to
14020 (mode_to_precision): Return a poly_uint16 rather than an unsigned
14022 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
14023 or if measurement_type is not polynomial.
14024 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
14025 in which the mode is already known to be a scalar_int_mode.
14026 * genmodes.c (emit_mode_precision): Change the type of mode_precision
14027 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
14029 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14030 for GET_MODE_PRECISION.
14031 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14032 for GET_MODE_PRECISION.
14033 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
14035 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14036 (expand_field_assignment, make_extraction): Likewise.
14037 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14038 (get_last_value): Likewise.
14039 * convert.c (convert_to_integer_1): Likewise.
14040 * cse.c (cse_insn): Likewise.
14041 * expr.c (expand_expr_real_1): Likewise.
14042 * lra-constraints.c (simplify_operand_subreg): Likewise.
14043 * optabs-query.c (can_atomic_load_p): Likewise.
14044 * optabs.c (expand_atomic_load): Likewise.
14045 (expand_atomic_store): Likewise.
14046 * ree.c (combine_reaching_defs): Likewise.
14047 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14048 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14049 * tree.h (type_has_mode_precision_p): Likewise.
14050 * ubsan.c (instrument_si_overflow): Likewise.
14052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14053 Alan Hayward <alan.hayward@arm.com>
14054 David Sherwood <david.sherwood@arm.com>
14056 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14057 polynomial numbers of units.
14058 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14059 (valid_vector_subparts_p): New function.
14060 (build_vector_type): Remove temporary shim and take the number
14061 of units as a poly_uint64 rather than an int.
14062 (build_opaque_vector_type): Take the number of units as a
14063 poly_uint64 rather than an int.
14064 * tree.c (build_vector_from_ctor): Handle polynomial
14065 TYPE_VECTOR_SUBPARTS.
14066 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14067 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14068 (build_vector_from_val): If the number of units is variable,
14069 use build_vec_duplicate_cst for constant operands and
14070 VEC_DUPLICATE_EXPR otherwise.
14071 (make_vector_type): Remove temporary is_constant ().
14072 (build_vector_type, build_opaque_vector_type): Take the number of
14073 units as a poly_uint64 rather than an int.
14074 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14076 * cfgexpand.c (expand_debug_expr): Likewise.
14077 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14078 (store_constructor, expand_expr_real_1): Likewise.
14079 (const_scalar_mask_from_tree): Likewise.
14080 * fold-const-call.c (fold_const_reduction): Likewise.
14081 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14082 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14083 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14084 (fold_relational_const): Likewise.
14085 (native_interpret_vector): Likewise. Change the size from an
14086 int to an unsigned int.
14087 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14088 TYPE_VECTOR_SUBPARTS.
14089 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14090 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14091 duplicating a non-constant operand into a variable-length vector.
14092 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14093 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14094 * ipa-icf.c (sem_variable::equals): Likewise.
14095 * match.pd: Likewise.
14096 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14097 * print-tree.c (print_node): Likewise.
14098 * stor-layout.c (layout_type): Likewise.
14099 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14100 * tree-cfg.c (verify_gimple_comparison): Likewise.
14101 (verify_gimple_assign_binary): Likewise.
14102 (verify_gimple_assign_ternary): Likewise.
14103 (verify_gimple_assign_single): Likewise.
14104 * tree-pretty-print.c (dump_generic_node): Likewise.
14105 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14106 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14107 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14108 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14109 (vect_shift_permute_load_chain): Likewise.
14110 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14111 (expand_vector_condition, optimize_vector_constructor): Likewise.
14112 (lower_vec_perm, get_compute_type): Likewise.
14113 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14114 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14115 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14116 (vect_recog_mask_conversion_pattern): Likewise.
14117 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14118 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14119 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14120 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14121 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14122 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14123 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14124 (supportable_widening_operation): Likewise.
14125 (supportable_narrowing_operation): Likewise.
14126 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14128 * varasm.c (output_constant): Likewise.
14130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14131 Alan Hayward <alan.hayward@arm.com>
14132 David Sherwood <david.sherwood@arm.com>
14134 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14135 so that both the length == 3 and length != 3 cases set up their
14136 own permute vectors. Add comments explaining why we know the
14137 number of elements is constant.
14138 (vect_permute_load_chain): Likewise.
14140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14141 Alan Hayward <alan.hayward@arm.com>
14142 David Sherwood <david.sherwood@arm.com>
14144 * machmode.h (mode_nunits): Change from unsigned char to
14146 (ONLY_FIXED_SIZE_MODES): New macro.
14147 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14148 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14149 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14151 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14152 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14153 or if measurement_type is not polynomial.
14154 * genmodes.c (ZERO_COEFFS): New macro.
14155 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14157 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14158 Use ZERO_COEFFS when emitting initializers.
14159 * data-streamer.h (bp_pack_poly_value): New function.
14160 (bp_unpack_poly_value): Likewise.
14161 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14162 for GET_MODE_NUNITS.
14163 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14164 for GET_MODE_NUNITS.
14165 * tree.c (make_vector_type): Remove temporary shim and make
14166 the real function take the number of units as a poly_uint64
14167 rather than an int.
14168 (build_vector_type_for_mode): Handle polynomial nunits.
14169 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14170 * emit-rtl.c (const_vec_series_p_1): Likewise.
14171 (gen_rtx_CONST_VECTOR): Likewise.
14172 * fold-const.c (test_vec_duplicate_folding): Likewise.
14173 * genrecog.c (validate_pattern): Likewise.
14174 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14175 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14176 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14177 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14178 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14179 * rtlanal.c (subreg_get_info): Likewise.
14180 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14181 (vect_grouped_load_supported): Likewise.
14182 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14183 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14184 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14185 (simplify_const_unary_operation, simplify_binary_operation_1)
14186 (simplify_const_binary_operation, simplify_ternary_operation)
14187 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14188 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14189 instead of CONST_VECTOR_NUNITS.
14190 * varasm.c (output_constant_pool_2): Likewise.
14191 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14192 explicit-encoded elements in the XVEC for variable-length vectors.
14194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14196 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14198 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14199 Alan Hayward <alan.hayward@arm.com>
14200 David Sherwood <david.sherwood@arm.com>
14202 * coretypes.h (fixed_size_mode): Declare.
14203 (fixed_size_mode_pod): New typedef.
14204 * builtins.h (target_builtins::x_apply_args_mode)
14205 (target_builtins::x_apply_result_mode): Change type to
14206 fixed_size_mode_pod.
14207 * builtins.c (apply_args_size, apply_result_size, result_vector)
14208 (expand_builtin_apply_args_1, expand_builtin_apply)
14209 (expand_builtin_return): Update accordingly.
14211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14213 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14214 * cselib.c (cselib_hash_rtx): Likewise.
14215 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14216 CONST_VECTOR encoding.
14218 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14219 Jeff Law <law@redhat.com>
14222 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14223 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14224 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14225 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14228 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14229 explicitly probe *sp in a noreturn function if there were any callee
14230 register saves or frame pointer is needed.
14232 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14235 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14236 BLKmode for ternary, binary or unary expressions.
14239 * var-tracking.c (delete_vta_debug_insn): New inline function.
14240 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14241 insns from get_insns () to NULL instead of each bb separately.
14242 Use delete_vta_debug_insn. No longer static.
14243 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14244 delete_vta_debug_insns callers.
14245 * rtl.h (delete_vta_debug_insns): Declare.
14246 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14247 instead of variable_tracking_main.
14249 2018-01-03 Martin Sebor <msebor@redhat.com>
14251 PR tree-optimization/83603
14252 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14253 arguments past the endof the argument list in functions declared
14254 without a prototype.
14255 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14256 Avoid checking when arguments are null.
14258 2018-01-03 Martin Sebor <msebor@redhat.com>
14261 * doc/extend.texi (attribute const): Fix a typo.
14262 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14263 issuing -Wsuggest-attribute for void functions.
14265 2018-01-03 Martin Sebor <msebor@redhat.com>
14267 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14268 offset_int::from instead of wide_int::to_shwi.
14269 (maybe_diag_overlap): Remove assertion.
14270 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14271 * gimple-ssa-sprintf.c (format_directive): Same.
14272 (parse_directive): Same.
14273 (sprintf_dom_walker::compute_format_length): Same.
14274 (try_substitute_return_value): Same.
14276 2018-01-03 Jeff Law <law@redhat.com>
14278 PR middle-end/83654
14279 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14280 non-constant residual for zero at runtime and avoid probing in
14281 that case. Reorganize code for trailing problem to mirror handling
14284 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14286 PR tree-optimization/83501
14287 * tree-ssa-strlen.c (get_string_cst): New.
14288 (handle_char_store): Call get_string_cst.
14290 2018-01-03 Martin Liska <mliska@suse.cz>
14292 PR tree-optimization/83593
14293 * tree-ssa-strlen.c: Include tree-cfg.h.
14294 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14295 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14296 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14298 (strlen_dom_walker::before_dom_children): Call
14299 gimple_purge_dead_eh_edges. Dump tranformation with details
14301 (strlen_dom_walker::before_dom_children): Update call by adding
14302 new argument cleanup_eh.
14303 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14305 2018-01-03 Martin Liska <mliska@suse.cz>
14308 * cif-code.def (VARIADIC_THUNK): New enum value.
14309 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14312 2018-01-03 Jan Beulich <jbeulich@suse.com>
14314 * sse.md (mov<mode>_internal): Tighten condition for when to use
14315 vmovdqu<ssescalarsize> for TI and OI modes.
14317 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14319 Update copyright years.
14321 2018-01-03 Martin Liska <mliska@suse.cz>
14324 * ipa-visibility.c (function_and_variable_visibility): Skip
14325 functions with noipa attribure.
14327 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14329 * gcc.c (process_command): Update copyright notice dates.
14330 * gcov-dump.c (print_version): Ditto.
14331 * gcov.c (print_version): Ditto.
14332 * gcov-tool.c (print_version): Ditto.
14333 * gengtype.c (create_file): Ditto.
14334 * doc/cpp.texi: Bump @copying's copyright year.
14335 * doc/cppinternals.texi: Ditto.
14336 * doc/gcc.texi: Ditto.
14337 * doc/gccint.texi: Ditto.
14338 * doc/gcov.texi: Ditto.
14339 * doc/install.texi: Ditto.
14340 * doc/invoke.texi: Ditto.
14342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14344 * vector-builder.h (vector_builder::m_full_nelts): Change from
14345 unsigned int to poly_uint64.
14346 (vector_builder::full_nelts): Update prototype accordingly.
14347 (vector_builder::new_vector): Likewise.
14348 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14349 (vector_builder::operator ==): Likewise.
14350 (vector_builder::finalize): Likewise.
14351 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14352 Take the number of elements as a poly_uint64 rather than an
14354 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14355 from unsigned int to poly_uint64.
14356 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14357 (vec_perm_indices::new_vector): Likewise.
14358 (vec_perm_indices::length): Likewise.
14359 (vec_perm_indices::nelts_per_input): Likewise.
14360 (vec_perm_indices::input_nelts): Likewise.
14361 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14362 number of elements per input as a poly_uint64 rather than an
14363 unsigned int. Use the original encoding for variable-length
14364 vectors, rather than clamping each individual element.
14365 For the second and subsequent elements in each pattern,
14366 clamp the step and base before clamping their sum.
14367 (vec_perm_indices::series_p): Handle polynomial element counts.
14368 (vec_perm_indices::all_in_range_p): Likewise.
14369 (vec_perm_indices_to_tree): Likewise.
14370 (vec_perm_indices_to_rtx): Likewise.
14371 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14372 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14373 (tree_vector_builder::new_binary_operation): Handle polynomial
14374 element counts. Return false if we need to know the number
14375 of elements at compile time.
14376 * fold-const.c (fold_vec_perm): Punt if the number of elements
14377 isn't known at compile time.
14379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14381 * vec-perm-indices.h (vec_perm_builder): Change element type
14382 from HOST_WIDE_INT to poly_int64.
14383 (vec_perm_indices::element_type): Update accordingly.
14384 (vec_perm_indices::clamp): Handle polynomial element_types.
14385 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14386 (vec_perm_indices::all_in_range_p): Likewise.
14387 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14389 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14390 polynomial vec_perm_indices element types.
14391 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14392 * fold-const.c (fold_vec_perm): Likewise.
14393 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14394 * tree-vect-generic.c (lower_vec_perm): Likewise.
14395 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14396 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14397 element type to HOST_WIDE_INT.
14399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14400 Alan Hayward <alan.hayward@arm.com>
14401 David Sherwood <david.sherwood@arm.com>
14403 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14404 rather than an int. Use plus_constant.
14405 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14406 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14413 a HOST_WIDE_INT to a poly_int64.
14415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14416 Alan Hayward <alan.hayward@arm.com>
14417 David Sherwood <david.sherwood@arm.com>
14419 * calls.c (load_register_parameters): Cope with polynomial
14420 mode sizes. Require a constant size for BLKmode parameters
14421 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14422 forces a parameter to be padded at the lsb end in order to
14423 fill a complete number of words, require the parameter size
14424 to be ordered wrt UNITS_PER_WORD.
14426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14427 Alan Hayward <alan.hayward@arm.com>
14428 David Sherwood <david.sherwood@arm.com>
14430 * reload1.c (spill_stack_slot_width): Change element type
14431 from unsigned int to poly_uint64_pod.
14432 (alter_reg): Treat mode sizes as polynomial.
14434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14435 Alan Hayward <alan.hayward@arm.com>
14436 David Sherwood <david.sherwood@arm.com>
14438 * reload.c (complex_word_subreg_p): New function.
14439 (reload_inner_reg_of_subreg, push_reload): Use it.
14441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14442 Alan Hayward <alan.hayward@arm.com>
14443 David Sherwood <david.sherwood@arm.com>
14445 * lra-constraints.c (process_alt_operands): Reject matched
14446 operands whose sizes aren't ordered.
14447 (match_reload): Refer to this check here.
14449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14450 Alan Hayward <alan.hayward@arm.com>
14451 David Sherwood <david.sherwood@arm.com>
14453 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14454 that the mode size is in the set {1, 2, 4, 8, 16}.
14456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14457 Alan Hayward <alan.hayward@arm.com>
14458 David Sherwood <david.sherwood@arm.com>
14460 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14461 Use plus_constant instead of gen_rtx_PLUS.
14463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14464 Alan Hayward <alan.hayward@arm.com>
14465 David Sherwood <david.sherwood@arm.com>
14467 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14468 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14469 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14470 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14471 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14472 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14473 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14474 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14475 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14476 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14478 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14479 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14480 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14481 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14482 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14483 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14484 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14485 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14486 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14487 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14489 * expr.c (emit_move_resolve_push): Treat the input and result
14490 of PUSH_ROUNDING as a poly_int64.
14491 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14492 (emit_push_insn): Likewise.
14493 * lra-eliminations.c (mark_not_eliminable): Likewise.
14494 * recog.c (push_operand): Likewise.
14495 * reload1.c (elimination_effects): Likewise.
14496 * rtlanal.c (nonzero_bits1): Likewise.
14497 * calls.c (store_one_arg): Likewise. Require the padding to be
14498 known at compile time.
14500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14501 Alan Hayward <alan.hayward@arm.com>
14502 David Sherwood <david.sherwood@arm.com>
14504 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14505 Use plus_constant instead of gen_rtx_PLUS.
14507 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14508 Alan Hayward <alan.hayward@arm.com>
14509 David Sherwood <david.sherwood@arm.com>
14511 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14512 rather than an int.
14514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14515 Alan Hayward <alan.hayward@arm.com>
14516 David Sherwood <david.sherwood@arm.com>
14518 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14519 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14520 via stack temporaries. Treat the mode size as polynomial too.
14522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14523 Alan Hayward <alan.hayward@arm.com>
14524 David Sherwood <david.sherwood@arm.com>
14526 * expr.c (expand_expr_real_2): When handling conversions involving
14527 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14528 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14529 as a poly_uint64 too.
14531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14532 Alan Hayward <alan.hayward@arm.com>
14533 David Sherwood <david.sherwood@arm.com>
14535 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14538 Alan Hayward <alan.hayward@arm.com>
14539 David Sherwood <david.sherwood@arm.com>
14541 * combine.c (can_change_dest_mode): Handle polynomial
14542 REGMODE_NATURAL_SIZE.
14543 * expmed.c (store_bit_field_1): Likewise.
14544 * expr.c (store_constructor): Likewise.
14545 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14546 and polynomial REGMODE_NATURAL_SIZE.
14547 (gen_lowpart_common): Likewise.
14548 * reginfo.c (record_subregs_of_mode): Likewise.
14549 * rtlanal.c (read_modify_subreg_p): Likewise.
14551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14552 Alan Hayward <alan.hayward@arm.com>
14553 David Sherwood <david.sherwood@arm.com>
14555 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14556 numbers of elements.
14558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14559 Alan Hayward <alan.hayward@arm.com>
14560 David Sherwood <david.sherwood@arm.com>
14562 * match.pd: Cope with polynomial numbers of vector elements.
14564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14565 Alan Hayward <alan.hayward@arm.com>
14566 David Sherwood <david.sherwood@arm.com>
14568 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14569 in a POINTER_PLUS_EXPR.
14571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14572 Alan Hayward <alan.hayward@arm.com>
14573 David Sherwood <david.sherwood@arm.com>
14575 * omp-simd-clone.c (simd_clone_subparts): New function.
14576 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14577 (ipa_simd_modify_function_body): Likewise.
14579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14580 Alan Hayward <alan.hayward@arm.com>
14581 David Sherwood <david.sherwood@arm.com>
14583 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14584 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14585 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14586 (expand_vector_condition, vector_element): Likewise.
14587 (subparts_gt): New function.
14588 (get_compute_type): Use subparts_gt.
14589 (count_type_subparts): Delete.
14590 (expand_vector_operations_1): Use subparts_gt instead of
14591 count_type_subparts.
14593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14594 Alan Hayward <alan.hayward@arm.com>
14595 David Sherwood <david.sherwood@arm.com>
14597 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14598 (vect_compile_time_alias): ...this new function. Do the calculation
14599 on poly_ints rather than trees.
14600 (vect_prune_runtime_alias_test_list): Update call accordingly.
14602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14603 Alan Hayward <alan.hayward@arm.com>
14604 David Sherwood <david.sherwood@arm.com>
14606 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14608 (vect_schedule_slp_instance): Likewise.
14610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14611 Alan Hayward <alan.hayward@arm.com>
14612 David Sherwood <david.sherwood@arm.com>
14614 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14615 constant and extern definitions for variable-length vectors.
14616 (vect_get_constant_vectors): Note that the number of units
14617 is known to be constant.
14619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14620 Alan Hayward <alan.hayward@arm.com>
14621 David Sherwood <david.sherwood@arm.com>
14623 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14624 of units as polynomial. Choose between WIDE and NARROW based
14627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14628 Alan Hayward <alan.hayward@arm.com>
14629 David Sherwood <david.sherwood@arm.com>
14631 * tree-vect-stmts.c (simd_clone_subparts): New function.
14632 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14635 Alan Hayward <alan.hayward@arm.com>
14636 David Sherwood <david.sherwood@arm.com>
14638 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14639 vectors as polynomial. Use build_index_vector for
14640 IFN_GOMP_SIMD_LANE.
14642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14643 Alan Hayward <alan.hayward@arm.com>
14644 David Sherwood <david.sherwood@arm.com>
14646 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14647 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14648 for variable-length vectors.
14649 (vectorizable_mask_load_store): Treat the number of units as
14650 polynomial, asserting that it is constant if the condition has
14651 already been enforced.
14652 (vectorizable_store, vectorizable_load): Likewise.
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-vect-loop.c (vectorizable_live_operation): Treat the number
14659 of units as polynomial. Punt if we can't tell at compile time
14660 which vector contains the final result.
14662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14663 Alan Hayward <alan.hayward@arm.com>
14664 David Sherwood <david.sherwood@arm.com>
14666 * tree-vect-loop.c (vectorizable_induction): Treat the number
14667 of units as polynomial. Punt on SLP inductions. Use an integer
14668 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14669 cast of such a series for variable-length floating-point
14672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14673 Alan Hayward <alan.hayward@arm.com>
14674 David Sherwood <david.sherwood@arm.com>
14676 * tree.h (build_index_vector): Declare.
14677 * tree.c (build_index_vector): New function.
14678 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14679 of units as polynomial, forcibly converting it to a constant if
14680 vectorizable_reduction has already enforced the condition.
14681 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14682 to create a {1,2,3,...} vector.
14683 (vectorizable_reduction): Treat the number of units as polynomial.
14684 Choose vectype_in based on the largest scalar element size rather
14685 than the smallest number of units. Enforce the restrictions
14688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14689 Alan Hayward <alan.hayward@arm.com>
14690 David Sherwood <david.sherwood@arm.com>
14692 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14693 number of units as polynomial.
14695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14696 Alan Hayward <alan.hayward@arm.com>
14697 David Sherwood <david.sherwood@arm.com>
14699 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14700 * target.def (autovectorize_vector_sizes): Return the vector sizes
14701 by pointer, using vector_sizes rather than a bitmask.
14702 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14703 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14704 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14706 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14707 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14708 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14709 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14710 * omp-general.c (omp_max_vf): Likewise.
14711 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14712 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14713 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14714 * tree-vect-slp.c (vect_slp_bb): Likewise.
14715 * doc/tm.texi: Regenerate.
14716 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14718 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14719 the vector size as a poly_uint64 rather than an unsigned int.
14720 (current_vector_size): Change from an unsigned int to a poly_uint64.
14721 (get_vectype_for_scalar_type): Update accordingly.
14722 * tree.h (build_truth_vector_type): Take the size and number of
14723 units as a poly_uint64 rather than an unsigned int.
14724 (build_vector_type): Add a temporary overload that takes
14725 the number of units as a poly_uint64 rather than an unsigned int.
14726 * tree.c (make_vector_type): Likewise.
14727 (build_truth_vector_type): Take the number of units as a poly_uint64
14728 rather than an unsigned int.
14730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14731 Alan Hayward <alan.hayward@arm.com>
14732 David Sherwood <david.sherwood@arm.com>
14734 * target.def (get_mask_mode): Take the number of units and length
14735 as poly_uint64s rather than unsigned ints.
14736 * targhooks.h (default_get_mask_mode): Update accordingly.
14737 * targhooks.c (default_get_mask_mode): Likewise.
14738 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14739 * doc/tm.texi: Regenerate.
14741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14742 Alan Hayward <alan.hayward@arm.com>
14743 David Sherwood <david.sherwood@arm.com>
14745 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14746 * omp-general.c (omp_max_vf): Likewise.
14747 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14748 (expand_omp_simd): Handle polynomial safelen.
14749 * omp-low.c (omplow_simd_context): Add a default constructor.
14750 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14751 (lower_rec_simd_input_clauses): Update accordingly.
14752 (lower_rec_input_clauses): Likewise.
14754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14755 Alan Hayward <alan.hayward@arm.com>
14756 David Sherwood <david.sherwood@arm.com>
14758 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14759 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14760 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14761 (vect_analyze_slp_cost): Likewise.
14762 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14763 (vect_model_load_cost): Likewise.
14765 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14766 Alan Hayward <alan.hayward@arm.com>
14767 David Sherwood <david.sherwood@arm.com>
14769 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14770 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14771 from an unsigned int * to a poly_uint64_pod *.
14772 (calculate_unrolling_factor): New function.
14773 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14776 Alan Hayward <alan.hayward@arm.com>
14777 David Sherwood <david.sherwood@arm.com>
14779 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14780 from an unsigned int to a poly_uint64.
14781 (_loop_vec_info::slp_unrolling_factor): Likewise.
14782 (_loop_vec_info::vectorization_factor): Change from an int
14784 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14785 (vect_get_num_vectors): New function.
14786 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14787 (vect_get_num_copies): Use vect_get_num_vectors.
14788 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14789 to an unsigned int *.
14790 (vect_analyze_data_refs): Change min_vf from an int * to a
14792 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14793 than an unsigned HOST_WIDE_INT.
14794 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14795 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14796 to an unsigned int *.
14797 (vect_analyze_data_ref_dependences): Likewise.
14798 (vect_compute_data_ref_alignment): Handle polynomial vf.
14799 (vect_enhance_data_refs_alignment): Likewise.
14800 (vect_prune_runtime_alias_test_list): Likewise.
14801 (vect_shift_permute_load_chain): Likewise.
14802 (vect_supportable_dr_alignment): Likewise.
14803 (dependence_distance_ge_vf): Take the vectorization factor as a
14804 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14805 (vect_analyze_data_refs): Change min_vf from an int * to a
14807 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14808 vfm1 as a poly_uint64 rather than an int. Make the same change
14809 for the returned bound_scalar.
14810 (vect_gen_vector_loop_niters): Handle polynomial vf.
14811 (vect_do_peeling): Likewise. Update call to
14812 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14813 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14815 * tree-vect-loop.c (vect_determine_vectorization_factor)
14816 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14817 (vect_get_known_peeling_cost): Likewise.
14818 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14819 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14820 (vect_transform_loop): Likewise. Use the lowest possible VF when
14821 updating the upper bounds of the loop.
14822 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14823 rather than an int.
14824 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14825 polynomial unroll factors.
14826 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14827 (vect_make_slp_decision): Likewise.
14828 (vect_supported_load_permutation_p): Likewise, and polynomial
14830 (vect_analyze_slp_cost): Handle polynomial vf.
14831 (vect_slp_analyze_node_operations): Likewise.
14832 (vect_slp_analyze_bb_1): Likewise.
14833 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14834 than an unsigned HOST_WIDE_INT.
14835 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14836 (vectorizable_load): Handle polynomial vf.
14837 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14839 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14842 Alan Hayward <alan.hayward@arm.com>
14843 David Sherwood <david.sherwood@arm.com>
14845 * match.pd: Handle bit operations involving three constants
14846 and try to fold one pair.
14848 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14850 * tree-vect-loop-manip.c: Include gimple-fold.h.
14851 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14852 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14853 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14854 Add a path that uses a step of VF instead of 1, but disable it
14856 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14857 and niters_no_overflow parameters. Update calls to
14858 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14859 Create a new SSA name if the latter choses to use a ste other
14860 than zero, and return it via niters_vector_mult_vf_var.
14861 * tree-vect-loop.c (vect_transform_loop): Update calls to
14862 vect_do_peeling, vect_gen_vector_loop_niters and
14863 slpeel_make_loop_iterate_ntimes.
14864 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14865 (vect_gen_vector_loop_niters): Update declarations after above changes.
14867 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14869 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14870 128-bit round to integer instructions.
14871 (ceil<mode>2): Likewise.
14872 (btrunc<mode>2): Likewise.
14873 (round<mode>2): Likewise.
14875 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14877 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14878 unaligned VSX load/store on P8/P9.
14879 (expand_block_clear): Allow the use of unaligned VSX
14880 load/store on P8/P9.
14882 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14884 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14886 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14887 swap associated with both a load and a store.
14889 2018-01-02 Andrew Waterman <andrew@sifive.com>
14891 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14892 * config/riscv/riscv.md (clear_cache): Use it.
14894 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14896 * web.c: Remove out-of-date comment.
14898 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14900 * expr.c (fixup_args_size_notes): Check that any existing
14901 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14902 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14903 (emit_single_push_insn): ...here.
14905 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14907 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14908 (const_vector_encoded_nelts): New function.
14909 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14910 (const_vector_int_elt, const_vector_elt): Declare.
14911 * emit-rtl.c (const_vector_int_elt_1): New function.
14912 (const_vector_elt): Likewise.
14913 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14914 of CONST_VECTOR_ELT.
14916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14918 * expr.c: Include rtx-vector-builder.h.
14919 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14920 directly on the tree encoding.
14921 (const_vector_from_tree): Likewise.
14922 * optabs.c: Include rtx-vector-builder.h.
14923 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14924 sequence of "u" values.
14925 * vec-perm-indices.c: Include rtx-vector-builder.h.
14926 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14927 directly on the vec_perm_indices encoding.
14929 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14931 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14932 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14933 * rtx-vector-builder.h: New file.
14934 * rtx-vector-builder.c: Likewise.
14935 * rtl.h (rtx_def::u2): Add a const_vector field.
14936 (CONST_VECTOR_NPATTERNS): New macro.
14937 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14938 (CONST_VECTOR_DUPLICATE_P): Likewise.
14939 (CONST_VECTOR_STEPPED_P): Likewise.
14940 (CONST_VECTOR_ENCODED_ELT): Likewise.
14941 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14942 (unwrap_const_vec_duplicate): Likewise.
14943 (const_vec_series_p): Check for a non-duplicated vector encoding.
14944 Say that the function only returns true for integer vectors.
14945 * emit-rtl.c: Include rtx-vector-builder.h.
14946 (gen_const_vec_duplicate_1): Delete.
14947 (gen_const_vector): Call gen_const_vec_duplicate instead of
14948 gen_const_vec_duplicate_1.
14949 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14950 (gen_const_vec_duplicate): Use rtx_vector_builder.
14951 (gen_const_vec_series): Likewise.
14952 (gen_rtx_CONST_VECTOR): Likewise.
14953 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14954 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14955 Build a new vector rather than modifying a CONST_VECTOR in-place.
14956 (handle_special_swappables): Update call accordingly.
14957 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14958 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14959 Build a new vector rather than modifying a CONST_VECTOR in-place.
14960 (handle_special_swappables): Update call accordingly.
14962 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14964 * simplify-rtx.c (simplify_const_binary_operation): Use
14965 CONST_VECTOR_ELT instead of XVECEXP.
14967 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14969 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14970 the selector elements to be different from the data elements
14971 if the selector is a VECTOR_CST.
14972 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14973 ssizetype for the selector.
14975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14977 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14978 before testing each element individually.
14979 * tree-vect-generic.c (lower_vec_perm): Likewise.
14981 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14983 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14984 * selftest-run-tests.c (selftest::run_tests): Call it.
14985 * vector-builder.h (vector_builder::operator ==): New function.
14986 (vector_builder::operator !=): Likewise.
14987 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14988 (vec_perm_indices::all_from_input_p): New function.
14989 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14990 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14991 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14992 instead of reading the VECTOR_CST directly. Detect whether both
14993 vector inputs are the same before constructing the vec_perm_indices,
14994 and update the number of inputs argument accordingly. Use the
14995 utility functions added above. Only construct sel2 if we need to.
14997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14999 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
15000 the broadcast of the low byte.
15001 (expand_mult_highpart): Use an explicit encoding for the permutes.
15002 * optabs-query.c (can_mult_highpart_p): Likewise.
15003 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
15004 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15005 (vectorizable_bswap): Likewise.
15006 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
15007 explicit encoding for the power-of-2 permutes.
15008 (vect_permute_store_chain): Likewise.
15009 (vect_grouped_load_supported): Likewise.
15010 (vect_permute_load_chain): Likewise.
15012 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15014 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
15015 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
15016 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
15017 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15018 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15019 (vect_gen_perm_mask_any): Likewise.
15021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15023 * int-vector-builder.h: New file.
15024 * vec-perm-indices.h: Include int-vector-builder.h.
15025 (vec_perm_indices): Redefine as an int_vector_builder.
15026 (auto_vec_perm_indices): Delete.
15027 (vec_perm_builder): Redefine as a stand-alone class.
15028 (vec_perm_indices::vec_perm_indices): New function.
15029 (vec_perm_indices::clamp): Likewise.
15030 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
15031 (vec_perm_indices::new_vector): New function.
15032 (vec_perm_indices::new_expanded_vector): Update for new
15033 vec_perm_indices class.
15034 (vec_perm_indices::rotate_inputs): New function.
15035 (vec_perm_indices::all_in_range_p): Operate directly on the
15036 encoded form, without computing elided elements.
15037 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15038 encoding. Update for new vec_perm_indices class.
15039 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15040 the given vec_perm_builder.
15041 (expand_vec_perm_var): Update vec_perm_builder constructor.
15042 (expand_mult_highpart): Use vec_perm_builder instead of
15043 auto_vec_perm_indices.
15044 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15045 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15046 or double series encoding as appropriate.
15047 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15048 vec_perm_indices instead of auto_vec_perm_indices.
15049 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15050 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15051 (vect_permute_store_chain): Likewise.
15052 (vect_grouped_load_supported): Likewise.
15053 (vect_permute_load_chain): Likewise.
15054 (vect_shift_permute_load_chain): Likewise.
15055 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15056 (vect_transform_slp_perm_load): Likewise.
15057 (vect_schedule_slp_instance): Likewise.
15058 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15059 (vectorizable_mask_load_store): Likewise.
15060 (vectorizable_bswap): Likewise.
15061 (vectorizable_store): Likewise.
15062 (vectorizable_load): Likewise.
15063 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15064 vec_perm_indices instead of auto_vec_perm_indices. Use
15065 tree_to_vec_perm_builder to read the vector from a tree.
15066 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15067 vec_perm_builder instead of a vec_perm_indices.
15068 (have_whole_vector_shift): Use vec_perm_builder and
15069 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15070 truncation to calc_vec_perm_mask_for_shift.
15071 (vect_create_epilog_for_reduction): Likewise.
15072 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15073 from auto_vec_perm_indices to vec_perm_indices.
15074 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15075 instead of changing individual elements.
15076 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15077 the vector in d.perm.
15078 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15079 from auto_vec_perm_indices to vec_perm_indices.
15080 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15081 instead of changing individual elements.
15082 (arm_vectorize_vec_perm_const): Use new_vector to install
15083 the vector in d.perm.
15084 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15085 Update vec_perm_builder constructor.
15086 (rs6000_expand_interleave): Likewise.
15087 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15088 (rs6000_expand_interleave): Likewise.
15090 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15092 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15093 to qimode could truncate the indices.
15094 * optabs.c (expand_vec_perm_var): Likewise.
15096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15098 * Makefile.in (OBJS): Add vec-perm-indices.o.
15099 * vec-perm-indices.h: New file.
15100 * vec-perm-indices.c: Likewise.
15101 * target.h (vec_perm_indices): Replace with a forward class
15103 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15104 * optabs.h: Include vec-perm-indices.h.
15105 (expand_vec_perm): Delete.
15106 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15107 (expand_vec_perm_const): Declare.
15108 * target.def (vec_perm_const_ok): Replace with...
15109 (vec_perm_const): ...this new hook.
15110 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15111 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15112 * doc/tm.texi: Regenerate.
15113 * optabs.def (vec_perm_const): Delete.
15114 * doc/md.texi (vec_perm_const): Likewise.
15115 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15116 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15117 expand_vec_perm for constant permutation vectors. Assert that
15118 the mode of variable permutation vectors is the integer equivalent
15119 of the mode that is being permuted.
15120 * optabs-query.h (selector_fits_mode_p): Declare.
15121 * optabs-query.c: Include vec-perm-indices.h.
15122 (selector_fits_mode_p): New function.
15123 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15124 is defined, instead of checking whether the vec_perm_const_optab
15125 exists. Use targetm.vectorize.vec_perm_const instead of
15126 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15127 fit in the vector mode before using a variable permute.
15128 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15129 vec_perm_indices instead of an rtx.
15130 (expand_vec_perm): Replace with...
15131 (expand_vec_perm_const): ...this new function. Take the selector
15132 as a vec_perm_indices rather than an rtx. Also take the mode of
15133 the selector. Update call to shift_amt_for_vec_perm_mask.
15134 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15135 Use vec_perm_indices::new_expanded_vector to expand the original
15136 selector into bytes. Check whether the indices fit in the vector
15137 mode before using a variable permute.
15138 (expand_vec_perm_var): Make global.
15139 (expand_mult_highpart): Use expand_vec_perm_const.
15140 * fold-const.c: Includes vec-perm-indices.h.
15141 * tree-ssa-forwprop.c: Likewise.
15142 * tree-vect-data-refs.c: Likewise.
15143 * tree-vect-generic.c: Likewise.
15144 * tree-vect-loop.c: Likewise.
15145 * tree-vect-slp.c: Likewise.
15146 * tree-vect-stmts.c: Likewise.
15147 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15149 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15150 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15151 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15152 (aarch64_vectorize_vec_perm_const): ...this new function.
15153 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15154 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15155 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15156 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15157 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15158 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15159 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15161 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15162 check for NEON modes.
15163 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15164 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15165 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15166 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15168 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15169 the old VEC_PERM_CONST conditions.
15170 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15171 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15172 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15173 (ia64_vectorize_vec_perm_const_ok): Merge into...
15174 (ia64_vectorize_vec_perm_const): ...this new function.
15175 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15176 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15177 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15178 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15179 * config/mips/mips.c (mips_expand_vec_perm_const)
15180 (mips_vectorize_vec_perm_const_ok): Merge into...
15181 (mips_vectorize_vec_perm_const): ...this new function.
15182 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15183 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15184 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15185 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15186 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15187 (rs6000_expand_vec_perm_const): Delete.
15188 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
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.
15198 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15199 an element count and rtx array.
15200 (rs6000_expand_extract_even): Update call accordingly.
15201 (rs6000_expand_interleave): Likewise.
15202 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15203 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15204 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15205 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15206 (rs6000_expand_vec_perm_const): Delete.
15207 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15208 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15209 (altivec_expand_vec_perm_const_le): Take each operand individually.
15210 Operate on constant selectors rather than rtxes.
15211 (altivec_expand_vec_perm_const): Likewise. Update call to
15212 altivec_expand_vec_perm_const_le.
15213 (rs6000_expand_vec_perm_const): Delete.
15214 (rs6000_vectorize_vec_perm_const_ok): Delete.
15215 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15216 reference to the SPE evmerge intructions.
15217 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15218 an element count and rtx array.
15219 (rs6000_expand_extract_even): Update call accordingly.
15220 (rs6000_expand_interleave): Likewise.
15221 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15222 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15224 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15228 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15229 vector mode and that that mode matches the mode of the data
15231 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15232 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15233 directly using expand_vec_perm_1 when forcing selectors into
15235 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15239 * optabs-query.h (can_vec_perm_p): Delete.
15240 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15241 * optabs-query.c (can_vec_perm_p): Split into...
15242 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15243 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15244 particular selector is valid.
15245 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15246 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15247 (vect_grouped_load_supported): Likewise.
15248 (vect_shift_permute_load_chain): Likewise.
15249 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15250 (vect_transform_slp_perm_load): Likewise.
15251 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15252 (vectorizable_bswap): Likewise.
15253 (vect_gen_perm_mask_checked): Likewise.
15254 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15255 implementations of variable permutation vectors into account
15256 when deciding which selector to use.
15257 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15258 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15259 with a false third argument.
15260 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15261 to test whether the constant selector is valid and can_vec_perm_var_p
15262 to test whether a variable selector is valid.
15264 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15266 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15267 * optabs-query.c (can_vec_perm_p): Likewise.
15268 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15269 instead of vec_perm_indices.
15270 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15271 (vect_gen_perm_mask_checked): Likewise,
15272 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15273 (vect_gen_perm_mask_checked): Likewise,
15275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15277 * optabs-query.h (qimode_for_vec_perm): Declare.
15278 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15279 (qimode_for_vec_perm): ...this new function.
15280 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15282 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15284 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15285 does not have a conditional at the top.
15287 2018-01-02 Richard Biener <rguenther@suse.de>
15289 * ipa-inline.c (big_speedup_p): Fix expression.
15291 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15294 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15297 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15301 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15302 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15303 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15304 cond_taken_branch_cost 3->4.
15306 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15308 PR tree-optimization/83581
15309 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15310 TODO_cleanup_cfg if any changes have been made.
15312 PR middle-end/83608
15313 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15314 convert_modes if target mode has the right side, but different mode
15317 PR middle-end/83609
15318 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15319 last argument when extracting from CONCAT. If either from_real or
15320 from_imag is NULL, use expansion through memory. If result is not
15321 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15322 the parts directly to inner mode, if even that fails, use expansion
15325 PR middle-end/83623
15326 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15327 check for bswap in mode rather than HImode and use that in expand_unop
15330 Copyright (C) 2018 Free Software Foundation, Inc.
15332 Copying and distribution of this file, with or without modification,
15333 are permitted in any medium without royalty provided the copyright
15334 notice and this notice are preserved.