1 2018-06-05 Richard Biener <rguenther@suse.de>
3 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4 removal pretend DOM info isn't available so we do not update
5 it and only remove edges, not dominated blocks. Actually free
6 DOM info in case we removed something. Remove unreachable blocks.
7 (mfb_keep_latches): Work with either DOM info or marked backedges.
8 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
9 first. Mark backedges if DOM info isn't available.
10 (Re-)compute DOM info after cleanup_control_flow_pre.
12 2018-06-05 Richard Biener <rguenther@suse.de>
14 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
15 (locus_discrim_hasher::hash): Adjust.
16 (locus_discrim_hasher::equal): Likewise.
17 (next_discriminator_for_locus): Work on line directly.
18 (same_line_p): Pass in expanded locus1 as well.
19 (assign_discriminators): Avoid redundant location expansions.
21 2018-06-05 Richard Biener <rguenther@suse.de>
23 PR tree-optimization/86046
24 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
25 if required after clearing TREE_ADDRESSABLE.
27 2018-06-05 Richard Biener <rguenther@suse.de>
29 PR tree-optimization/86047
30 * tree-ssa-loop.c (for_each_index): Glob handling of all
31 decls and constants and really handle all of them.
33 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
36 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
37 qualifier_void_pointer and qualifier_const_void_pointer.
38 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
39 (arm_init_builtins): Handle the above.
40 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
41 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
42 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
45 2018-06-05 Martin Liska <mliska@suse.cz>
47 * auto-profile.c (read_autofdo_file): Do not use
48 gcov_ctr_summary struct.
49 (afdo_callsite_hot_enough_for_early_inline): Likewise.
50 * coverage.c (struct counts_entry): Likewise.
51 (read_counts_file): Read just single summary entry.
52 (get_coverage_counts): Use gcov_summary struct.
53 * coverage.h (get_coverage_counts): Likewise.
54 * gcov-dump.c (dump_working_sets): Likewise.
55 (tag_summary): Dump just single summary.
56 * gcov-io.c (gcov_write_summary): Write just histogram
58 (gcov_read_summary): Read just single summary.
59 (compute_working_sets): Use gcov_summary struct.
60 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
61 of GCOV_COUNTERS_SUMMABLE.
62 (GCOV_COUNTERS_SUMMABLE): Remove.
63 (GCOV_FIRST_VALUE_COUNTER): Replace with
64 GCOV_COUNTER_V_INTERVAL.
65 (struct gcov_ctr_summary): Remove.
66 (struct gcov_summary): Directly use fields of former
68 (compute_working_sets): Use gcov_summary struct.
69 * gcov.c (read_count_file): Do not use ctrs fields.
70 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
72 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
74 * profile.c: Likewise.
75 * profile.h: Likewise.
77 2018-06-05 Martin Liska <mliska@suse.cz>
80 * gcov.c (output_lines): Print working directory only
81 in intermediate format.
83 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
85 * config/s390/s390-builtin-types.def: Add void function type.
86 * config/s390/s390-builtins.def: Use the function type for the
89 2018-06-04 Jim Wilson <jimw@sifive.com>
91 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
93 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
94 and maybe_eh_return. Change regno to unsigned int. Use new args to
95 handle EH_RETURN_DATA_REGNO registers properly.
96 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
97 (riscv_expand_epilogue): Update comment. Change argument name and
98 type. Update code to use new name and type. Pass new args to
99 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
101 * config/riscv/riscv.md (NORMAL_RETURN): New.
102 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
103 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
104 (eh_return): Call gen_eh_return_internal and emit barrier.
105 (eh_return_internal): Call riscv_expand_epilogue.
107 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
109 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
110 bit_insertion field and declare can_be_merged_into method.
111 (merged_store_group::can_be_merged_into): New method.
112 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
113 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
114 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
116 2018-06-04 Richard Biener <rguenther@suse.de>
118 PR tree-optimization/85955
119 * builtins.c (fold_builtin_sincos): Convert pointers to
120 destination to appropriate type before dereferencing.
122 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
124 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
126 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
128 * expr.c (expand_expr_real_1): Force the operand into memory if
129 its TYPE_MODE is BLKmode and if there is no integer mode for
130 the number of bits being extracted.
132 2018-06-04 Jakub Jelinek <jakub@redhat.com>
136 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
137 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
139 2018-06-04 Richard Biener <rguenther@suse.de>
141 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
142 (cleanup_tree_cfg_noloop): ... single caller. Do
143 start_recording_case_labels later.
145 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
147 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
148 to _IMMINTRIN_H_INCLUDED.
149 * config/i386/pconfigintrin.h: Ditto.
150 * config/i386/waitpkgintrin.h: Ditto.
151 * config/i386/immintrin.h: Add includes for sgxintrin.h,
152 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
153 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
154 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
155 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
156 waitpkgintrin.h and cldemoteintrin.h.
158 2018-06-04 Richard Biener <rguenther@suse.de>
160 PR tree-optimization/86038
161 * tracer.c (find_best_successor): Check probability for
162 being initialized, bail out if not.
164 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
167 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
168 of bits to ignore when comparing architectures.
170 2018-06-04 Jakub Jelinek <jakub@redhat.com>
172 PR tree-optimization/69615
173 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
174 maximum or minimum of the type, try to merge it also as if
175 range1 is + [-, x - 1] or + [x + 1, -].
178 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
180 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
182 PR tree-optimization/86034
183 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
184 the unsigned bitfield type in a bit insertion sequence if it does not
185 have a larger precision than the bitfield size.
186 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
188 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
190 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
192 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
194 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
195 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
196 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
197 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
199 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
201 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
202 Disable -fdelete-null-pointer-checks for ELF toolchain.
204 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
205 Kito Cheng <kito.cheng@gmail.com>
207 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
208 (nds32le-*-*, nds32be-*-*): Integrate checking process.
209 (nds32*-*-*): Add glibc and uclibc conditions.
210 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
211 (TARGET_EXCEPT_UNWIND_INFO): Define.
212 * config/nds32/elf.h: New file.
213 * config/nds32/linux.h: New file.
214 * config/nds32/nds32-elf.opt: New file.
215 * config/nds32/nds32-linux.opt: New file.
216 * config/nds32/nds32-fp-as-gp.c
217 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
218 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
220 (nds32_asm_file_end): Ditto.
221 (nds32_print_operand): Ditto.
222 (nds32_insert_attributes): Ditto.
223 (nds32_init_libfuncs): New function.
224 (TARGET_HAVE_TLS): Define.
225 (TARGET_INIT_LIBFUNCS): Define.
226 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
228 (TARGET_ELF): Apply different mcmodel setting.
229 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
230 been migrated into elf.h and linux.h files.
231 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
232 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
233 (mcmodel): The content has been migrated into nds32-elf.opt and
234 nds32-linux.opt files.
235 * config/nds32/t-elf: New file.
236 * config/nds32/t-linux: New file.
238 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
239 Shiva Chen <shiva0217@gmail.com>
241 * config/nds32/constants.md (unspec_volatile_element): Add
242 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
243 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
245 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
246 (make_pass_nds32_fp_as_gp): Declare.
247 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
249 (nds32_asm_function_end_prologue): Remove unused asm output.
250 (nds32_asm_function_begin_epilogue): Remove unused asm output.
251 (nds32_asm_file_start): Output necessary fp_as_gp information.
252 (nds32_option_override): Adjust register usage.
253 (nds32_expand_prologue): Consider fp_as_gp situation.
254 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
255 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
258 (simple_return): Ditto.
259 (omit_fp_begin): Output special directive for fp_as_gp.
260 (omit_fp_end): Output special directive for fp_as_gp.
261 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
262 mforbid-fp-as-gp): New options.
264 2018-06-01 Mark Wielaard <mark@klomp.org>
266 * dwarf2out.c (dwarf2out_finish): Remove generation of
269 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
271 * gimple-ssa-store-merging.c: Include gimple-fold.h.
272 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
273 (struct merged_store_group): Add bit_insertion field.
274 (dump_char_array): Use standard hexadecimal format.
275 (merged_store_group::merged_store_group): Set bit_insertion to false.
276 (merged_store_group::apply_stores): Use optimal buffer size. Deal
277 with BIT_INSERT_EXPR stores. Move up code updating the mask and
278 also print the mask in the dump file.
279 (pass_store_merging::gate): Minor tweak.
280 (imm_store_chain_info::coalesce_immediate): Fix wrong association
281 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
282 stores with INTEGER_CST stores.
283 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
284 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
285 and use it throughout. Generate bit insertion sequences if need be.
286 (pass_store_merging::process_store): Remove redundant condition.
287 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
289 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
291 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
292 the 128-bit floating point types. Fix function comment.
294 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
296 * config/aarch64/aarch64-simd.md
297 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
299 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
302 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
304 PR tree-optimization/85989
305 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
307 (backprop::intersect_uses): Check it when deciding whether this
308 is a backedge reference.
309 (backprop::process_block): Add each phi to m_visited_phis
310 after visiting it, then clear it at the end.
312 2018-06-01 Richard Biener <rguenther@suse.de>
314 * tree-vectorizer.h (vect_dr_stmt): New function.
315 (vect_get_load_cost): Adjust.
316 (vect_get_store_cost): Likewise.
317 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
318 Use vect_dr_stmt instead of DR_SMTT.
319 (vect_record_base_alignments): Likewise.
320 (vect_calculate_target_alignment): Likewise.
321 (vect_compute_data_ref_alignment): Likewise and make static.
322 (vect_update_misalignment_for_peel): Likewise.
323 (vect_verify_datarefs_alignment): Likewise.
324 (vector_alignment_reachable_p): Likewise.
325 (vect_get_data_access_cost): Likewise. Pass down
326 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
327 (vect_get_peeling_costs_all_drs): Likewise.
328 (vect_peeling_hash_get_lowest_cost): Likewise.
329 (vect_enhance_data_refs_alignment): Likewise.
330 (vect_find_same_alignment_drs): Likewise.
331 (vect_analyze_data_refs_alignment): Likewise.
332 (vect_analyze_group_access_1): Likewise.
333 (vect_analyze_group_access): Likewise.
334 (vect_analyze_data_ref_access): Likewise.
335 (vect_analyze_data_ref_accesses): Likewise.
336 (vect_vfa_segment_size): Likewise.
337 (vect_small_gap_p): Likewise.
338 (vectorizable_with_step_bound_p): Likewise.
339 (vect_prune_runtime_alias_test_list): Likewise.
340 (vect_analyze_data_refs): Likewise.
341 (vect_supportable_dr_alignment): Likewise.
342 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
343 (vect_gen_prolog_loop_niters): Likewise.
344 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
345 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
347 (vect_recog_mask_conversion_pattern): Likewise.
348 (vect_try_gather_scatter_pattern): Likewise.
349 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
350 to vect_get_store_cost.
351 (vect_get_store_cost): Get stmt_info instead of DR.
352 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
353 (vect_get_load_cost): Get stmt_info instead of DR.
355 2018-06-01 Richard Biener <rguenther@suse.de>
358 * gimple-fold.c (var_decl_component_p): Also allow offsetted
359 vars wrapped in MEM_REFs.
361 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
363 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
364 Fix subreg tests so that we only return a choice between
365 GENERAL_REGS and FP_REGS if the original classes included both.
367 2018-06-01 Richard Biener <rguenther@suse.de>
370 * tree-ssa-structalias.c (get_function_part_constraint):
371 Handle NULL fi->decl.
372 (find_func_aliases_for_call): Properly handle indirect
374 (find_func_clobbers): Likewise.
375 (ipa_pta_execute): Likewise.
376 (create_variable_info_for): For functions that are ifunc_resolver
377 resolve to a varinfo that contains the result of the resolver
379 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
382 2018-05-31 Michael Collison <michael.collison@arm.com>
384 * config/aarch64/aarch64.md:
385 (*fix_to_zero_extenddfdi2): New pattern.
386 * gcc.target/aarch64/fix_extend1.c: New testcase.
388 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
392 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
393 and BUILT_IN_STRNCMP_EQ.
394 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
396 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
397 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
398 (gimple_fold_builtin): Likewise.
399 * tree-ssa-strlen.c (compute_string_length): New function.
400 (determine_min_obsize): New function.
401 (handle_builtin_string_cmp): New function to handle calls to
402 string compare functions.
403 (strlen_optimize_stmt): Add handling to builtin string compare
405 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
406 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
407 * tree.c (build_common_builtin_nodes): Add new defines of
408 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
410 2018-05-31 Jakub Jelinek <jakub@redhat.com>
413 * bb-reorder.c (pass_partition_blocks::gate): Return false for
414 functions with naked attribute.
416 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
418 * config/i386/sse.md (avx_vec_concat<mode>):
419 Substitute concat_tg_mode mode attribute with xtg_mode.
420 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
421 (concat_tg_mode): Remove mode attribute.
423 2018-05-31 Martin Sebor <msebor@redhat.com>
426 * calls.c (alloc_max_size): Correct a logic error/typo.
427 Treat excessive arguments as infinite. Warn for invalid arguments.
428 * doc/invoke.texi (-Walloc-size-larger-than): Update.
430 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
433 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
434 and movx for Haswell.
436 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
437 Cesar Philippidis <cesar@codesourcery.com>
440 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
441 when emitting error on private/firstprivate reductions.
442 * omp-low.c (lower_omp_target): Avoid reference-type processing
443 on pointers for firstprivate clause.
445 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
447 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
450 * config/aarch64/aarch64-simd.md
451 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
452 (aarch64_ld1_x3_<mode>): Likewise
453 (aarch64_st1x2<VALLDIF:mode>): Likewise
454 (aarch64_st1_x2_<mode>): Likewise
455 (aarch64_st1x3<VALLDIF:mode>): Likewise
456 (aarch64_st1_x3_<mode>): Likewise
457 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
458 (vld1_s8_x3): Likewise.
459 (vld1_u16_x3): Likewise.
460 (vld1_s16_x3): Likewise.
461 (vld1_u32_x3): Likewise.
462 (vld1_s32_x3): Likewise.
463 (vld1_u64_x3): Likewise.
464 (vld1_s64_x3): Likewise.
465 (vld1_f16_x3): Likewise.
466 (vld1_f32_x3): Likewise.
467 (vld1_f64_x3): Likewise.
468 (vld1_p8_x3): Likewise.
469 (vld1_p16_x3): Likewise.
470 (vld1_p64_x3): Likewise.
471 (vld1q_u8_x3): Likewise.
472 (vld1q_s8_x3): Likewise.
473 (vld1q_u16_x3): Likewise.
474 (vld1q_s16_x3): Likewise.
475 (vld1q_u32_x3): Likewise.
476 (vld1q_s32_x3): Likewise.
477 (vld1q_u64_x3): Likewise.
478 (vld1q_s64_x3): Likewise.
479 (vld1q_f16_x3): Likewise.
480 (vld1q_f32_x3): Likewise.
481 (vld1q_f64_x3): Likewise.
482 (vld1q_p8_x3): Likewise.
483 (vld1q_p16_x3): Likewise.
484 (vld1q_p64_x3): Likewise.
485 (vst1_s64_x2): Likewise.
486 (vst1_u64_x2): Likewise.
487 (vst1_f64_x2): Likewise.
488 (vst1_s8_x2): Likewise.
489 (vst1_p8_x2): Likewise.
490 (vst1_s16_x2): Likewise.
491 (vst1_p16_x2): Likewise.
492 (vst1_s32_x2): Likewise.
493 (vst1_u8_x2): Likewise.
494 (vst1_u16_x2): Likewise.
495 (vst1_u32_x2): Likewise.
496 (vst1_f16_x2): Likewise.
497 (vst1_f32_x2): Likewise.
498 (vst1_p64_x2): Likewise.
499 (vst1q_s8_x2): Likewise.
500 (vst1q_p8_x2): Likewise.
501 (vst1q_s16_x2): Likewise.
502 (vst1q_p16_x2): Likewise.
503 (vst1q_s32_x2): Likewise.
504 (vst1q_s64_x2): Likewise.
505 (vst1q_u8_x2): Likewise.
506 (vst1q_u16_x2): Likewise.
507 (vst1q_u32_x2): Likewise.
508 (vst1q_u64_x2): Likewise.
509 (vst1q_f16_x2): Likewise.
510 (vst1q_f32_x2): Likewise.
511 (vst1q_f64_x2): Likewise.
512 (vst1q_p64_x2): Likewise.
513 (vst1_s64_x3): Likewise.
514 (vst1_u64_x3): Likewise.
515 (vst1_f64_x3): Likewise.
516 (vst1_s8_x3): Likewise.
517 (vst1_p8_x3): Likewise.
518 (vst1_s16_x3): Likewise.
519 (vst1_p16_x3): Likewise.
520 (vst1_s32_x3): Likewise.
521 (vst1_u8_x3): Likewise.
522 (vst1_u16_x3): Likewise.
523 (vst1_u32_x3): Likewise.
524 (vst1_f16_x3): Likewise.
525 (vst1_f32_x3): Likewise.
526 (vst1_p64_x3): Likewise.
527 (vst1q_s8_x3): Likewise.
528 (vst1q_p8_x3): Likewise.
529 (vst1q_s16_x3): Likewise.
530 (vst1q_p16_x3): Likewise.
531 (vst1q_s32_x3): Likewise.
532 (vst1q_s64_x3): Likewise.
533 (vst1q_u8_x3): Likewise.
534 (vst1q_u16_x3): Likewise.
535 (vst1q_u32_x3): Likewise.
536 (vst1q_u64_x3): Likewise.
537 (vst1q_f16_x3): Likewise.
538 (vst1q_f32_x3): Likewise.
539 (vst1q_f64_x3): Likewise.
540 (vst1q_p64_x3): Likewise.
542 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
544 * config/msp430/msp430.c (msp430_output_labelref): Prepend
545 user_label_prefix to name.
547 * tree-core.h: Update comment about the format of NAME string
548 passed to handler in attribute_spec.
550 * config/msp430/msp430.md: Remove erroneous subreg expression from
551 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
552 zero_extend{q,h}isi2.
554 2018-05-30 Borislav Petkov <bp@suse.de>
556 * doc/extend.texi: Document some architecture specific
557 constraints and sort entries.
559 2018-05-30 Martin Sebor <msebor@redhat.com>
562 * builtins.c (expand_builtin_stpcpy_1): New function.
563 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
564 only if the former succeeds.
566 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
568 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
571 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
573 * doc/invoke.texi (-flinker-output): Document
575 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
577 * passes.c (ipa_write_summaries): Only modify statements if body
579 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
580 incrementally linking.
581 (ipa_passes): Likewise.
582 * lto-cgraph.c (lto_output_node): When incrementally linking do not
583 pass down resolution info.
584 * common.opt (flag_incremental_link): Update info.
585 * gcc.c (plugin specs): Turn flinker-output=* to
586 -plugin-opt=-linker-output-known
587 * toplev.c (compile_file): Also cut compilation when doing incremental
589 * flag-types. (enum lto_partition_model): Add
590 LTO_LINKER_OUTPUT_NOLTOREL.
591 (invoke.texi): Add -flinker-output docs.
592 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
593 link same way as WPA; do not stream in dead initializers.
595 * dwarf2out.c (dwarf2out_die_ref_for_decl,
596 darf2out_register_external_decl): Support incremental link.
598 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
600 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
602 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
604 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
605 it down to simple_object_copy_lto_debug_sections.
606 (run_gcc): Determine incremental LTO link time and configure
607 lto1 into non-wpa mode, disable renaming of debug sections.
609 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
611 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
612 descriptions of various incorrectly documented functions.
614 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
617 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
618 address check not strict.
620 2018-05-30 Richard Biener <rguenther@suse.de>
622 PR tree-optimization/85964
623 * tracer.c (better_p): Drop initialized count check, we only
624 call the function with initialized counts now.
625 (find_best_successor): Do find a best edge if one
626 has uninitialized count.
627 (find_best_predecessor): Likewise. Do BB frequency check only
628 if count is initialized.
630 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
632 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
633 (aarch64_ldrstr_offset_compare): New.
634 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
635 load/store orderings.
636 (aarch64_gen_adjusted_ldpstp): Likewise.
638 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
640 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
641 Check for subset of GENERAL_REGS and FP_REGS.
642 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
645 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
647 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
648 and wi::to_poly_offset. Add the current offset and then check
649 whether the sum fits, rather than using an unchecked addition of
650 a checked term. Check for a shwi rather than a uhwi.
651 * expr.c (get_bit_range): Use tree_to_poly_uint64.
652 (store_constructor): Use poly_int_tree_p.
653 (expand_expr_real_1): Likewise.
654 * function.c (assign_temp): Likewise.
655 * fold-const.c (const_binop): Use poly_int_tree_p and
657 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
659 * ipa-icf-gimple.c (func_checker::compare_operand): Use
660 to_poly_offset for MEM offsets.
661 * ipa-icf.c (sem_variable::equals): Likewise.
662 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
663 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
664 wi::to_poly_offset for BIT_FIELD_REF offsets.
665 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
667 * var-tracking.c (emit_note_insn_var_location): Use
670 2018-05-29 Jim Wilson <jimw@sifive.com>
672 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
674 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
677 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
678 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
680 (sse4_1_round<mode>2): Use nonimmediate_operand
681 for operand 1 predicate.
683 2018-05-29 Martin Sebor <msebor@redhat.com>
684 Richard Biener <rguenther@suse.de>
687 * calls.c (get_size_range): Call determine_value_range instead
689 * tree-vrp.h (determine_value_range): Declared new function.
690 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
692 2018-05-29 Richard Biener <rguenther@suse.de>
694 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
695 sure to use non-pattern stmts for get_earlier_stmt arguments.
696 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
697 called on pattern stmts.
698 (get_later_stmt): Likewise.
700 2018-05-29 Martin Liska <mliska@suse.cz>
702 PR gcov-profile/85759
703 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
706 2018-05-29 Jakub Jelinek <jakub@redhat.com>
708 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
710 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
714 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
715 VEC_PACK_FLOAT_EXPR): New tree codes.
716 * tree-pretty-print.c (op_code_prio): Handle
717 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
718 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
719 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
720 * tree-inline.c (estimate_operator_cost): Likewise.
721 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
722 * fold-const.c (const_binop): Likewise.
723 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
724 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
725 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
726 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
727 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
728 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
729 * expr.c (expand_expr_real_2): Likewise.
730 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
731 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
732 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
734 * optabs.c (expand_widen_pattern_expr): For
735 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
736 sign from result type rather than operand's type.
737 (expand_binop_directly): For vec_packu_float_optab and
738 vec_packs_float_optab allow result type to be different from operand's
740 * optabs-tree.c (optab_for_tree_code): Handle
741 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
742 VEC_PACK_FLOAT_EXPR. Formatting fixes.
743 * tree-vect-generic.c (expand_vector_operations_1): Handle
744 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
746 * tree-vect-stmts.c (supportable_widening_operation): Handle
748 (supportable_narrowing_operation): Handle FLOAT_EXPR.
749 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
750 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
751 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
752 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
754 (vec_pack<floatprefix>_float_<mode>): New expander.
755 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
757 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
758 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
759 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
760 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
761 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
763 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
764 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
765 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
766 VEC_PACK_FLOAT_EXPR): Document.
768 2018-05-29 Richard Biener <rguenther@suse.de>
770 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
772 (stmt_vec_info_vec): Make pointer.
773 (init_stmt_vec_info_vec): Remove.
774 (free_stmt_vec_info_vec): Likewise.
775 (set_stmt_vec_info_vec): New function.
776 (free_stmt_vec_infos): Likewise.
777 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
778 (set_vinfo_for_stmt): Likewise.
779 (get_earlier_stmt): Likewise.
780 (get_later_stmt): Likewise.
781 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
782 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
783 (vec_info::~vec_info): Free stmt_vec_infos.
784 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
785 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
786 (pass_slp_vectorize::execute): Likewise.
787 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
788 (free_stmt_vec_info_vec): Likewise.
789 (set_stmt_vec_info_vec): New function.
790 (free_stmt_vec_infos): Likewise.
791 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
792 the global stmt_vec_info_vec.
793 * tree-parloops.c (gather_scalar_reductions): Use
794 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
797 2018-05-29 Richard Biener <rguenther@suse.de>
799 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
801 2018-05-29 Martin Liska <mliska@suse.cz>
802 David Malcolm <dmalcolm@redhat.com>
804 * vec.c (test_reverse): New.
805 (vec_c_tests): Add new test.
806 * vec.h (vl_ptr>::reverse): New function.
808 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
810 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
812 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
815 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
817 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
819 2018-05-28 Richard Biener <rguenther@suse.de>
821 PR tree-optimization/85933
822 * tree-vect-data-refs.c (vect_record_base_alignments): Only
823 look at stmts marked as vectorizable.
825 2018-05-28 Richard Biener <rguenther@suse.de>
827 PR tree-optimization/85934
828 * tree-vect-generic.c (expand_vector_operations_1): Hoist
829 vector boolean check before scalar optimization.
831 2018-05-28 Jakub Jelinek <jakub@redhat.com>
833 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
836 2018-05-28 Mark Wielaard <mark@klomp.org>
838 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
839 if it is an expression containing a minus sign.
841 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
843 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
845 2018-05-27 Paul Koning <ni1d@arrl.net>
847 * config/pdp11/pdp11.md (truncsihi2): Remove.
849 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
850 Chung-Ju Wu <jasonwucj@gmail.com>
852 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
854 (unaligned_store_dw): Ditto.
855 * config/nds32/nds32-memory-manipulation.c
856 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
857 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
858 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
859 (emit_setmem_word_loop): Rename to ...
860 (emit_setmem_doubleword_loop): ... this.
861 (nds32_gen_dup_4_byte_to_word_value): New function.
862 (nds32_gen_dup_8_byte_to_double_word_value): New function.
863 (nds32_expand_setmem_loop): Refine implementation.
864 (nds32_expand_setmem_loop_v3m): Ditto.
865 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
868 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
870 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
872 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
874 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
875 (nds32_init_machine_status): Initialize machine->attr_naked_p and
876 machine->attr_no_prologue_p.
877 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
878 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
879 (nds32_expand_epilogue): Consider attr_naked_p.
880 (nds32_expand_epilogue_v3pop): Likewise.
881 (nds32_can_use_return_insn): Likewise.
882 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
883 attr_no_prologue_p fields.
884 * config/nds32/nds32.opt (mret-in-naked-func): New option.
886 2018-05-27 Jakub Jelinek <jakub@redhat.com>
889 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
892 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
894 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
896 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
898 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
900 (*<floatsuffix>floatv2div2sf2): Rename to ...
901 (*float<floatunssuffix>v2div2sf2): ... this.
902 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
903 (float<floatunssuffix>v2div2sf2_mask): ... this.
904 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
905 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
906 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
908 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
910 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
912 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
914 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
916 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
918 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
919 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
920 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
921 gen_ufix_truncv8dfv8si2.
922 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
923 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
924 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
925 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
926 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
927 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
928 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
929 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
931 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
935 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
937 2018-05-25 Jim Wilson <jimw@sifive.com>
939 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
940 * config/riscv/riscv.c (struct machine_function): Add
941 interrupt_handler_p and attribute_checked_p fields.
942 (riscv_attribute_table): Add interrupt.
943 (riscv_interrupt_type_p): New.
944 (riscv_save_reg_p): Save extra regs for interrupt handler.
945 (riscv_use_save_libcall): Return false for interrupt handler.
946 (riscv_first_stack_step): Add forward declaration.
947 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
948 for interrupt handler with large frame. Use it for saved reg list.
949 (riscv_expand_prologue): Move flag_stack_usage_info support to
950 eliminate duplication.
951 (riscv_expand_epilogue): Generate mret for interrupt handler.
952 (riscv_epilogue_uses): New.
953 (riscv_can_use_return_insn): Return false for interrupt handler.
954 (riscv_function_ok_for_sibcall): Likewise.
955 (riscv_set_current_function): Add interrupt handler support.
956 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
957 * config/riscv/riscv.md (UNSPECV_MRET): New.
959 (riscv_frflags, riscv_fsflags): Use tab after opcode.
961 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
963 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
965 PR tree-optimization/85712
966 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
967 this candidate has already been replaced in-situ by a copy.
969 2018-05-25 Jason Merrill <jason@redhat.com>
971 PR c++/80485 - inline function non-zero address.
972 * symtab.c (nonzero_address): Check DECL_COMDAT.
974 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
977 * config/alpha/alpha.md (ashlsi3): New insn pattern.
978 (*ashlsi_se): Rename from *ashldi_se. Define as sign
979 extension of SImode operation. Use const123_operand predicate.
981 (*saddl_se_1): Ditto.
983 (*ssubl_se_1): Ditto.
984 * config/alpha/predicates.md (const123_operand): New predicate.
985 * config/alpha/constraints.md (P): Use IN_RANGE.
987 2018-05-25 Richard Biener <rguenther@suse.de>
989 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
991 (ref_maybe_used_by_stmt_p): Likewise.
992 (stmt_may_clobber_ref_p): Likewise.
993 (stmt_may_clobber_ref_p_1): Likewise.
994 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
996 (ref_maybe_used_by_stmt_p): Likewise.
997 (stmt_may_clobber_ref_p): Likewise.
998 (stmt_may_clobber_ref_p_1): Likewise.
999 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1000 the alias oracle to disambiguate DRs with stmts DR analysis
1002 (vect_analyze_data_refs): Do not give up on not analyzable
1003 DRs for BB vectorization. Remove code truncating the dataref
1006 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1009 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1010 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1011 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1013 2018-05-25 Richard Biener <rguenther@suse.de>
1015 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1016 function, combining stmt data ref gathering and fatal analysis
1018 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1019 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1020 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1021 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1022 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1024 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1026 PR tree-optimization/85720
1027 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1028 SCC if all partitions are builtins.
1029 (version_loop_by_alias_check): New parameter. Generate cancelable
1030 runtime alias check if all partitions are builtins.
1031 (distribute_loop): Update call to above function.
1033 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1035 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1036 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1037 (parm_default_def_partition_arg): Ditto.
1038 (set_parm_default_def_partition): Ditto.
1039 (get_parm_default_def_partitions): Ditto and make it static.
1040 (get_undefined_value_partitions): Ditto and make it static.
1041 (remove_ssa_form): Refactor call to init_var_map here.
1042 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1043 computation for loop region.
1044 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1045 (register_default_def): Delete.
1046 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1047 (parm_default_def_partition_arg): Ditto.
1048 (set_parm_default_def_partition): Ditto.
1049 (get_parm_default_def_partitions): Ditto and make it static.
1050 (get_undefined_value_partitions): Ditto and make it static.
1051 (coalesce_with_default, coalesce_with_default): Update comment.
1052 (create_coalesce_list_for_region): New func factored out from
1053 create_outofssa_var_map.
1054 (populate_coalesce_list_for_outofssa): New func factored out from
1055 create_outofssa_var_map and coalesce_ssa_name.
1056 (create_outofssa_var_map): Delete.
1057 (coalesce_ssa_name): Refactor to support live range computation.
1058 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1059 (get_parm_default_def_partitions): Delete.
1060 (get_undefined_value_partitions): Ditto.
1061 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1062 computation for loop region.
1063 (new_tree_live_info, loe_visit_block): Ditto.
1064 (live_worklist, set_var_live_on_entry): Ditto.
1065 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1066 * tree-ssa-live.h (struct _var_map): New fields.
1067 (init_var_map): Change decl.
1068 (region_contains_p): New.
1070 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1072 * tree-ssa-live.h (live_merge_and_clear): Delete.
1074 2018-05-25 Richard Biener <rguenther@suse.de>
1077 * tree-dump.c (dequeue_and_dump): Remove access to removed
1078 operand 2 of a SWITCH_EXPR.
1080 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1082 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1083 multiplication and division.
1084 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1085 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1086 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1087 (cond_udiv_optab, cond_umod_optab): New optabs.
1088 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1089 (IFN_COND_RDIV): New internal functions.
1090 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1091 TRUNC_MOD_EXPR and RDIV_EXPR.
1092 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1093 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1095 (SVE_INT_BINARY): Include mult.
1096 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1097 (optab, sve_int_op): Handle mult.
1098 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1100 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1101 for SVE_INT_BINARY_SD.
1103 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1105 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1106 (optab, sve_int_op): Handle div and udiv.
1107 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1108 for SVE_INT_BINARY_SD.
1109 (*<optab><mode>3): New insn for the same.
1111 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1113 * tree-vect-patterns.c: Include predict.h.
1114 (vect_recog_divmod_pattern): Restrict check for division support
1115 to when optimizing for size.
1117 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1119 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1120 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1121 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1122 (gimple_match_op::set_op): Likewise.
1123 (gimple_resimplify4): Declare.
1124 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1125 (expr::gen_transform): Likewise.
1126 (decision_tree::gen): Generate a simplification routine for 4 operands.
1127 * gimple-match-head.c (gimple_simplify): Add an overload for
1128 4 operands. In the top-level function, handle up to 4 call
1129 arguments and call gimple_resimplify4.
1130 (gimple_resimplify4): New function.
1131 (build_call_internal): Pass a fourth operand.
1132 (maybe_push_to_seq): Likewise.
1133 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1134 Fold VEC_COND_EXPRs of an operation and a default value into
1135 an IFN_COND_* function if possible.
1136 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1138 (SVE_COND_FP_BINARY): Include them.
1139 (optab, sve_fp_op): Handle them.
1140 (SVE_INT_BINARY_REV): New code iterator.
1141 (SVE_COND_FP_BINARY_REV): New int iterator.
1142 (commutative): New int attribute.
1143 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1145 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1147 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1148 (*cond_<optab><mode>): New patterns for reversed operands.
1150 2018-05-25 Richard Biener <rguenther@suse.de>
1152 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1153 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1154 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1155 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1156 * tree-vect-data-refs.c (everywhere): Adjust users.
1157 * tree-vect-loop.c (everywhere): Likewise.
1158 * tree-vect-slp.c (everywhere): Likewise.
1159 * tree-vect-stmts.c (everywhere): Likewise.
1160 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1162 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1164 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1166 (gcc_cv_as_section_exclude): ... this.
1167 Try Solaris as #exclude syntax.
1168 * configure: Regenerate.
1169 * config.in: Regenerate.
1170 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1172 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1173 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1175 * varasm.c (default_elf_asm_named_section): Don't check if
1176 HAVE_GAS_SECTION_EXCLUDE is defined.
1178 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1180 * doc/md.texi: Update the documentation of the cond_* optabs
1181 to mention the new final operand. Fix GET_MODE_NUNITS call.
1182 Describe the scalar case too.
1183 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1184 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1186 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1187 (get_conditional_internal_fn): Update comment.
1188 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1189 accumulator value as a final argument to conditional functions.
1190 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1191 a define_expand and add an "else" operand. Assert for now that
1192 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1193 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1194 (*cond_<optab><mode>): New patterns.
1195 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1196 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1197 (UNSPEC_COND_EOR): Delete.
1198 (optab): Remove associated mappings.
1199 (SVE_INT_BINARY): New code iterator.
1200 (sve_int_op): Remove int attribute and add "minus" to the code
1202 (SVE_COND_INT_OP): Delete.
1203 (SVE_COND_FP_OP): Rename to...
1204 (SVE_COND_FP_BINARY): ...this.
1206 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1208 * optabs.c (can_reuse_operands_p): New function.
1209 (maybe_legitimize_operands): Try to reuse the results for
1212 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1214 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1215 Add {q} suffix to insn mnemonic.
1217 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1219 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1220 (msp430_warn_func_return): New.
1222 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1224 * fold-const.c (tree_nonzero_bits): New function.
1225 * fold-const.h (tree_nonzero_bits): Likewise.
1226 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1227 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1229 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1233 * varasm.c (assemble_alias): Check ifunc_resolver only on
1236 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1239 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1240 when memory input operand is handled.
1242 2018-05-24 Luis Machado <luis.machado@linaro.org>
1244 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1246 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1248 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1250 * match.pd: Delay FMA folds until after vectorization.
1252 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1255 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1256 address check not strict.
1258 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1260 * gimple-match.h (gimple_match_op): New class.
1261 (mprts_hook): Replace parameters with a gimple_match_op *.
1262 (maybe_build_generic_op): Likewise.
1263 (gimple_simplified_result_is_gimple_val): Replace parameters with
1264 a const gimple_match_op *.
1265 (gimple_simplify): Replace code_helper * and tree * parameters with
1266 a gimple_match_op * parameter.
1267 (gimple_resimplify1): Replace code_helper *, tree and tree *
1268 parameters with a gimple_match_op * parameter.
1269 (gimple_resimplify2): Likewise.
1270 (gimple_resimplify3): Likewise.
1271 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1272 parameters with a gimple_match_op * parameter.
1273 * gimple-match-head.c (gimple_simplify): Change prototypes of
1274 auto-generated functions to take a gimple_match_op * instead of
1275 separate code_helper * and tree * parameters. Make the same
1276 change in the top-level overload and update calls to the
1277 gimple_resimplify routines. Update calls to the auto-generated
1278 functions and to maybe_push_res_to_seq in the publicly-facing
1279 operation-specific gimple_simplify overloads.
1280 (gimple_match_op::MAX_NUM_OPS): Define.
1281 (gimple_resimplify1): Replace rcode and ops with a single res_op
1282 parameter. Update call to gimple_simplify.
1283 (gimple_resimplify2): Likewise.
1284 (gimple_resimplify3): Likewise.
1285 (mprts_hook): Replace parameters with a gimple_match_op *.
1286 (maybe_build_generic_op): Likewise.
1287 (build_call_internal): Replace type, nargs and ops with
1288 a gimple_match_op *.
1289 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1290 with a single gimple_match_op *. Update calls to mprts_hook,
1291 build_call_internal and gimple_simplified_result_is_gimple_val.
1292 Factor out code that is common to the tree_code and combined_fn cases.
1293 * genmatch.c (expr::gen_transform): Replace tem_code and
1294 tem_ops with a gimple_match_op called tem_op. Update calls
1295 to the gimple_resimplify functions and maybe_push_res_to_seq.
1296 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1297 res_ops. Update call to the gimple_resimplify functions.
1298 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1299 (decision_tree::gen): Make the functions take a gimple_match_op *
1300 called res_op instead of separate res_code and res_ops parameters.
1301 Update call accordingly.
1302 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1303 and ops with a single res_op parameter. Update calls to
1304 maybe_build_generic_op and maybe_push_res_to_seq.
1305 (fold_stmt_1): Update calls to gimple_simplify and
1306 replace_stmt_with_simplification.
1307 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1308 and gimple_simplified_result_is_gimple_val.
1309 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1311 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1312 with a gimple_match_op *.
1313 (vn_nary_build_or_lookup): Likewise. Update call to
1314 vn_nary_build_or_lookup_1.
1315 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1316 gimple_match_op *. Update calls to the gimple_resimplify routines
1317 and to gimple_simplified_result_is_gimple_val.
1318 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1319 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1320 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1321 (visit_nary_op): Likewise.
1322 (visit_reference_op_load): Likewise.
1324 2018-05-23 Luis Machado <luis.machado@linaro.org>
1326 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1327 modifier for printing the step amount.
1329 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1332 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1335 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1337 * doc/sourcebuild.texi (Endianness): New subsubsection.
1339 2018-05-23 Luis Machado <luis.machado@linaro.org>
1341 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1342 <prefetch_dynamic_strides>: New const bool field.
1343 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1344 prefetch_dynamic_strides.
1345 (exynosm1_prefetch_tune): Likewise.
1346 (thunderxt88_prefetch_tune): Likewise.
1347 (thunderx_prefetch_tune): Likewise.
1348 (thunderx2t99_prefetch_tune): Likewise.
1349 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1351 (aarch64_override_options_internal): Update to set
1352 PARAM_PREFETCH_DYNAMIC_STRIDES.
1353 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1354 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1355 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1356 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1357 prefetch-dynamic-strides setting.
1359 2018-05-23 Luis Machado <luis.machado@linaro.org>
1361 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1362 <minimum_stride>: New const int field.
1363 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1364 minimum_stride field defaulting to -1.
1365 (exynosm1_prefetch_tune): Likewise.
1366 (thunderxt88_prefetch_tune): Likewise.
1367 (thunderx_prefetch_tune): Likewise.
1368 (thunderx2t99_prefetch_tune): Likewise.
1369 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1370 <default_opt_level>: Set to 3.
1371 (aarch64_override_options_internal): Update to set
1372 PARAM_PREFETCH_MINIMUM_STRIDE.
1373 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1374 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1375 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1376 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1377 stride is constant and is below the minimum stride threshold.
1379 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1381 * config/arm/arm-cpus.in (mode26): Delete.
1382 (armv4): Delete mode26 reference.
1383 * config/arm/arm.c (arm_configure_build_target): Delete use of
1386 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1388 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1390 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1391 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1393 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1394 (floatunsdidf2): Ditto.
1396 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1398 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1399 (fixuns_trunc<mode>si2_avx512f): Ditto.
1400 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1401 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1402 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1404 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1406 PR rtl-optimization/79985
1407 * df-scan.c (df_insn_refs_collect): Remove special case for
1408 global registers and asm statements.
1410 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1412 * extend.texi (Global Register Variables): Rewrite the bullet list.
1413 Note that the register is available for allocation. Note that access
1414 via inline asm must use constraints. Add note about async-signal
1415 handlers. Remove paragraph about automagic register selection.
1417 2018-05-23 Richard Biener <rguenther@suse.de>
1419 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1420 of fixed offset from memset VN.
1422 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1424 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1426 (alloc_cand_and_find_basis): Initialize first_interp field.
1427 (slsr_process_mul): Modify first_interp field.
1428 (slsr_process_add): Likewise.
1429 (slsr_process_cast): Modify first_interp field for each new
1431 (slsr_process_copy): Likewise.
1432 (dump_candidate): Dump first_interp field.
1433 (replace_mult_candidate): Process all interpretations, not just
1435 (replace_rhs_if_not_dup): Likewise.
1436 (replace_one_candidate): Likewise.
1438 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1440 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1442 (aarch64_needs_frame_chain): New function.
1443 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1445 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1448 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1449 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1450 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1451 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1452 as true for strict-align.
1453 (aarch64_can_inline_p): Perform checks even when callee has no
1454 attributes to check for strict alignment.
1455 * doc/extend.texi (AArch64 Function Attributes): Document
1457 * doc/invoke.texi: (AArch64 Options): Likewise.
1459 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1461 PR tree-optimization/85853
1462 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1463 the handling of the root of the node to...
1464 (vect_slp_analyze_node_operations_1): ...this new function,
1465 and run the whole thing with the child nodes' def types
1466 set according to their SLP node's def type.
1468 2018-05-23 Richard Biener <rguenther@suse.de>
1471 * tree-data-ref.c (create_runtime_alias_checks): Defer
1472 and ignore overflow warnings.
1474 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1476 PR tree-optimization/85822
1477 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1480 2018-05-23 Richard Biener <rguenther@suse.de>
1482 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1483 memset constants via native_interpret_expr.
1485 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1488 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1490 (cgraph_node::create_alias): Likewise.
1491 (cgraph_node::get_availability): Check ifunc_resolver instead
1492 of looking up ifunc attribute.
1493 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1494 * varasm.c (do_assemble_alias): Likewise.
1495 (assemble_alias): Likewise.
1496 (default_binds_local_p_3): Likewise.
1497 * cgraph.h (cgraph_node): Add ifunc_resolver.
1498 (cgraph_node::only_called_directly_or_aliased_p): Return false
1500 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1502 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1503 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1504 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1505 instead of looking up ifunc attribute.
1507 2018-05-22 Luis Machado <luis.machado@linaro.org>
1509 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1511 2018-05-22 Martin Sebor <msebor@redhat.com>
1514 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1515 only when expasion succeeds.
1516 (expand_builtin_strcmp): Same.
1517 (expand_builtin_strncmp): Same.
1519 2018-05-22 Martin Sebor <msebor@redhat.com>
1521 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1523 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1524 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1526 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1527 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1529 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1530 Add check for legitimate_address.
1531 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1532 (aarch64_swap_ldrstr_operands): New.
1533 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1536 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1537 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1539 * config/aarch64/aarch64.md: New patterns to generate stp
1541 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1542 single words and double words.
1543 (load_pair_sw, load_pair_dw): Likewise.
1544 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1546 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1548 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1549 for different mode ldpstp and add peephole for merged zero stores.
1551 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1553 (aarch64_gen_store_pair): Rename calls to match new patterns.
1554 (aarch64_gen_load_pair): Rename calls to match new patterns.
1555 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1556 (load_pair<DREG:mode><DREG2:mode>): ... This.
1557 (store_pair<mode>): Rename to...
1558 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1559 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1561 (V_INT_EQUIV): Handle SImode.
1562 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1565 2018-05-22 Martin Sebor <msebor@redhat.com>
1568 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1569 or ajust the presumed bound on an operation to avoid unnecessary
1572 2018-05-22 Martin Sebor <msebor@redhat.com>
1574 PR tree-optimization/85826
1575 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1576 assuming that a DECL necesarily has a constant size.
1578 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1581 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1583 2018-05-22 Richard Biener <rguenther@suse.de>
1585 PR tree-optimization/85834
1586 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1587 non-constant and non-zero memset arguments.
1589 2018-05-22 Martin Liska <mliska@suse.cz>
1592 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1594 2018-05-22 Richard Biener <rguenther@suse.de>
1596 PR tree-optimization/85863
1597 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1598 comparisons when vectype is specified.
1599 (vectorizable_condition): Do not specify vectype for
1600 vect_is_simple_cond when SLP vectorizing.
1602 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1605 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1606 define __ibm128 as long double.
1607 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1608 as a distinct type when IEEE 128-bit support is enabled.
1609 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1610 128-bit types to use the correct functions.
1611 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1612 convert between 128-bit floating point types that have different
1613 modes but the same representation, instead of using gen_lowpart to
1615 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1617 (IFKF_reg): New attributes to give the register constraints for
1619 (extend<mode>tf2_internal): New insns to mark an explicit
1620 conversion between 128-bit floating point types that have a
1621 different mode but share the same representation.
1623 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1625 PR tree-optimization/85814
1626 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1627 a null return from get_strinfo when unsharing the next
1628 strinfo in the chain.
1630 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1633 * varasm.c (weak_finish): Clean up weak_decls.
1635 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1637 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1638 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1639 UNSPEC_UADALP values.
1640 * config/aarch64/iterators.md (ABAL): New int iterator.
1643 (sur): Add mappings for the above.
1644 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1646 (aarch64_<sur>abal<mode>_4): Likewise.
1647 (aarch64_<sur>adalp<mode>_3): Likewise.
1648 (<sur>sadv16qi): New define_expand.
1650 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1652 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1653 (*movdf_internal): Ditto.
1654 (*rcpsf2_sse): Ditto.
1655 (*rsqrtsf2_sse): Ditto.
1656 (*sqrt<mode>2_sse): Ditto.
1658 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1660 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1662 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1663 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1664 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1665 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1667 * config/aarch64/arm_neon.h: Likewise.
1668 * config/aarch64/iterators.md (VQ_I): New.
1670 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1672 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1674 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1677 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1679 * config/nds32/constraints.md (S): New constraint.
1680 * config/nds32/nds32.md (call_internal): Use constraint S.
1681 (call_value_internal): Likewise.
1682 (sibcall_internal): Likewise.
1683 (sibcall_value_internal): Likewise.
1685 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1686 Chung-Ju Wu <jasonwucj@gmail.com>
1688 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1691 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1692 Chung-Ju Wu <jasonwucj@gmail.com>
1694 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1695 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1696 (nds32_rtx_costs_impl): Simplify.
1697 (nds32_address_cost_impl): Simplify.
1698 (nds32_init_rtx_costs): New function.
1699 (nds32_rtx_costs_speed_prefer): Likewise.
1700 (nds32_rtx_costs_size_prefer): Likewise.
1701 (nds32_address_cost_speed_prefer): Likewise.
1702 (nds32_address_cost_speed_fwprop): Likewise.
1703 (nds32_address_cost_size_prefer): Likewise.
1704 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1705 * config/nds32/nds32.c (nds32_option_override): Use
1706 nds32_init_rtx_costs function.
1708 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1710 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1711 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1712 (TARGET_PIPELINE_N8): Likewise.
1713 (TARGET_PIPELINE_N10): Likewise.
1714 (TARGET_PIPELINE_N13): Likewise.
1715 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1717 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1719 * config/nds32/nds32-fpu.md: Update copyright year.
1721 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1725 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1727 * config/nds32/nds32.c
1728 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1729 * config/nds32/nds32.opt (minline-asm-r15): New option.
1731 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1733 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1735 * config/nds32/nds32.md (abssi2): New pattern.
1737 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1739 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1740 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1741 Merge insn pattern from sse_cvtsi2ss<round_name> and
1742 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1743 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1744 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1745 using SWI48 mode iterator.
1746 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1747 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1748 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1749 pattern from sse_cvttss2si<round_saeonly_name>
1750 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1751 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1752 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1753 using SWI48 mode iterator.
1754 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1755 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1756 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1757 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1758 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1759 using SWI48 mode iterator.
1760 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1761 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1762 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1763 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1764 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1765 SWI48 mode iterator.
1766 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1767 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1768 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1769 pattern from sse_cvttsd2si<round_saeonly_name>
1770 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1772 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1774 * config/nds32/nds32-md-auxiliary.c
1775 (nds32_valid_smw_lwm_base_p): Refine.
1776 (nds32_output_smw_single_word): Refine.
1777 (nds32_output_smw_double_word): New.
1778 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1780 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1782 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1783 (nds32_output_stack_pop): Refine.
1784 (nds32_expand_unaligned_load): Refine.
1785 (nds32_expand_unaligned_store): Refine.
1787 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1788 Chung-Ju Wu <jasonwucj@gmail.com>
1790 * config/nds32/constants.md: Add TP_REGNUM constant.
1791 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1792 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1794 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1795 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1796 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1797 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1798 and PIC code generation.
1799 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1801 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1803 * config/nds32/nds32.md: Support TLS and PIC.
1804 * config/nds32/nds32.c: Support TLS and PIC.
1805 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1806 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1809 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1811 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1812 mode with E_ prefix.
1814 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1815 Chung-Ju Wu <jasonwucj@gmail.com>
1817 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1818 * config/nds32/nds32-md-auxiliary.c
1819 (symbolic_reference_mentioned_p): New.
1820 (nds32_legitimize_ict_address): New.
1821 (nds32_expand_ict_move): New.
1822 (nds32_indirect_call_referenced_p): New.
1823 (nds32_symbol_binds_local_p): Delete.
1824 (nds32_long_call_p): Modify.
1825 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1826 * config/nds32/nds32-protos.h
1827 (symbolic_reference_mentioned_p): Declare.
1828 (nds32_legitimize_ict_address): Declare.
1829 (nds32_expand_ict_move): Declare.
1830 (nds32_indirect_call_referenced_p): Declare.
1831 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1832 (nds32_relax_group): Use nds32_ict_const_p as condition.
1833 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1834 (nds32_asm_file_start): Output ict_model directive in asm code.
1835 (nds32_legitimate_address_p): Consider indirect call.
1836 (nds32_print_operand): Consider indirect call.
1837 (nds32_print_operand_address): Consider indirect call.
1838 (nds32_insert_attributes): Handle "indirect_call" attribute.
1839 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1840 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1841 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1842 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1843 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1844 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1845 (TARGET_ICT_MODEL_SMALL): Define.
1846 (TARGET_ICT_MODEL_LARGE): Define.
1847 * config/nds32/nds32.md (movsi): Consider ict model.
1848 (call, call_value): Consider ict model.
1849 (sibcall, sibcall_value): Consider ict model.
1850 * config/nds32/nds32.opt (mict-model): New option.
1851 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1854 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1855 Monk Chiang <sh.chiang04@gmail.com>
1856 Jim Wilson <jimw@sifive.com>
1858 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1859 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1860 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1861 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1862 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1863 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1864 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1865 compute save_libcall_adjustment properly.
1866 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1867 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1868 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1869 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1870 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1871 (ABI_SPEC): Handle mabi=ilp32e.
1872 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1873 (RVE): Add RVE mask.
1874 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1875 <-march>: Add rv32e as an example.
1877 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1880 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1881 (intra_create_variable_infos): Handle C++ constructors.
1883 2018-05-18 Martin Liska <mliska@suse.cz>
1885 * passes.def: Remove a redundant pass.
1887 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1890 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1892 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1894 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1896 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1898 (armv2, armv2a, armv3, armv3m): Delete architectures.
1899 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1900 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1901 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1903 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1904 (*mulsidi3adddi): Likewise.
1905 (mulsidi3): Likewise.
1906 (*mulsidi3_nov6): Likewise.
1907 (umulsidi3): Likewise.
1908 (umulsidi3_nov6): Likewise.
1909 (umaddsidi4): Likewise.
1910 (*umulsidi3adddi): Likewise.
1911 (smulsi3_highpart): Likewise.
1912 (*smulsi3_highpart_nov6): Likewise.
1913 (umulsi3_highpart): Likewise.
1914 (*umulsi3_highpart_nov6): Likewise.
1915 * config/arm/arm.h (arm_arch3m): Delete.
1916 * config/arm/arm.c (arm_arch3m): Delete.
1917 (arm_option_override_internal): Update armv3-related comment.
1918 (arm_configure_build_target): Delete use of isa_bit_mode32.
1919 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1920 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1921 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1923 (mulusa3): Likewise.
1924 * config/arm/arm-protos.h (arm_arch3m): Delete.
1925 * config/arm/arm-tables.opt: Regenerate.
1926 * config/arm/arm-tune.md: Likewise.
1927 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1928 deleted architectures.
1930 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1932 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1933 (armv5t, armv5te): New features.
1934 (ARMv5, ARMv5e): Delete fgroups.
1935 (ARMv5t, ARMv5te): Adjust for above changes.
1937 (armv5, armv5e): Delete arches.
1938 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1940 (*call_reg_arm): Likewise.
1941 (*call_value_reg_armv5): Likewise.
1942 (*call_value_reg_arm): Likewise.
1943 (*call_symbol): Likewise.
1944 (*call_value_symbol): Likewise.
1945 (*sibcall_insn): Likewise.
1946 (*sibcall_value_insn): Likewise.
1948 (prefetch): Likewise.
1949 (define_split and define_peephole2 dependent on arm_arch5):
1951 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1953 (TARGET_ARM_QBIT): Likewise.
1954 (TARGET_DSP_MULTIPLY): Likewise.
1955 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1956 (arm_arch5, arm_arch5e): Delete.
1957 (arm_arch5t, arm_arch5te): Declare.
1958 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1959 (arm_arch5t): Declare.
1960 (arm_option_reconfigure_globals): Update for the above.
1961 (arm_options_perform_arch_sanity_checks): Update comment, replace
1962 use of arm_arch5 with arm_arch5t.
1963 (use_return_insn): Likewise.
1964 (arm_emit_call_insn): Likewise.
1965 (output_return_instruction): Likewise.
1966 (arm_final_prescan_insn): Likewise.
1967 (arm_coproc_builtin_available): Likewise.
1968 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1969 arm_arch5e with arm_arch5t and arm_arch5te.
1970 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1971 (arm_arch5t, arm_arch5te): Declare.
1972 * config/arm/arm-tables.opt: Regenerate.
1973 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1974 * config/arm/t-multilib: Likewise.
1975 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1976 instead of arm_arch5.
1977 (*call_reg_thumb1): Likewise.
1978 (*call_value_reg_thumb1_v5): Likewise.
1979 (*call_value_reg_thumb1): Likewise.
1980 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1982 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1984 2018-05-18 Martin Liska <mliska@suse.cz>
1986 PR gcov-profile/84846
1987 * doc/gcov.texi: Document -t option of gcov tool.
1989 2018-05-18 Martin Liska <mliska@suse.cz>
1991 PR gcov-profile/84846
1992 * gcov.c (print_usage): Add new -t option.
1993 (process_args): Handle the option.
1994 (generate_results): Use stdout as output when requested by
1997 2018-05-18 Martin Liska <mliska@suse.cz>
1999 PR gcov-profile/84846
2000 * coverage.c (coverage_init): Write PWD to .gcno file.
2001 * doc/gcov.texi: Document how working directory is printed.
2002 * gcov-dump.c (dump_gcov_file): Print PWD.
2003 * gcov.c (output_intermediate_file): Likewise.
2004 (read_graph_file): Read PWD string.
2005 (output_lines): Print PWD.
2007 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2010 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2011 for retval and return false if all args to phi are zero.
2013 2018-05-18 Richard Biener <rguenther@suse.de>
2015 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2017 (evrp_dom_walker::before_dom_children): Call it.
2019 2018-05-18 Richard Biener <rguenther@suse.de>
2021 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2022 results when processing array refs with variable index.
2024 2018-05-18 Toon Moene <toon@moene.org>
2026 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2027 directly after that of -floop-interchange. Indicate that both
2028 options are enabled by default when specifying -O3.
2030 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2032 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2033 iterator. Delete separate integer-mode vec_set<mode> expander.
2034 (aarch64_simd_vec_setv2di): Delete.
2035 (vec_setv2di): Delete.
2036 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2037 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2038 the "w, r" alternative.
2040 2018-05-18 Martin Liska <mliska@suse.cz>
2042 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2043 * tree-pass.h (make_pass_lower_switch_O0): New function.
2044 * tree-switch-conversion.c (node_has_low_bound): Remove.
2045 (node_has_high_bound): Likewise.
2046 (node_is_bounded): Likewise.
2047 (class pass_lower_switch): Make it a template type and create
2049 (pass_lower_switch::execute): Add template argument.
2050 (make_pass_lower_switch): New function.
2051 (make_pass_lower_switch_O0): New function.
2052 (do_jump_if_equal): Remove.
2053 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2054 all the hard work to tree optimization passes.
2056 2018-05-18 Martin Liska <mliska@suse.cz>
2058 * dbgcnt.c (limit_low): Renamed from limit.
2059 (limit_high): New variable.
2060 (dbg_cnt_is_enabled): Check for upper limit.
2061 (dbg_cnt): Adjust dumping.
2062 (dbg_cnt_set_limit_by_index): Add new argument for high
2064 (dbg_cnt_set_limit_by_name): Likewise.
2065 (dbg_cnt_process_single_pair): Parse new format.
2066 (dbg_cnt_process_opt): Use strtok.
2067 (dbg_cnt_list_all_counters): Remove 'value' and add
2069 * doc/invoke.texi: Document changes.
2071 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2073 * doc/sourcebuild.texi (scalar_all_fma): Document.
2074 * tree.def (FMA_EXPR): Delete.
2075 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2076 * internal-fn.c (ternary_direct): New macro.
2077 (expand_ternary_optab_fn): Likewise.
2078 (direct_ternary_optab_supported_p): Likewise.
2079 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2080 * builtins.c (fold_builtin_fma): Delete.
2081 (fold_builtin_3): Don't call it.
2082 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2083 * expr.c (expand_expr_real_2): Likewise.
2084 * fold-const.c (operand_equal_p): Likewise.
2085 (fold_ternary_loc): Likewise.
2086 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2087 * gimple.c (DEFTREECODE): Likewise.
2088 * gimplify.c (gimplify_expr): Likewise.
2089 * optabs-tree.c (optab_for_tree_code): Likewise.
2090 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2091 * tree-eh.c (operation_could_trap_p): Likewise.
2092 (stmt_could_throw_1_p): Likewise.
2093 * tree-inline.c (estimate_operator_cost): Likewise.
2094 * tree-pretty-print.c (dump_generic_node): Likewise.
2095 (op_code_prio): Likewise.
2096 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2097 * tree-ssa-operands.c (get_expr_operands): Likewise.
2098 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2099 * fold-const-call.h (fold_fma): Delete.
2100 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2101 CFN_FNMA and CFN_FNMS.
2103 * genmatch.c (combined_fn): New enum.
2104 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2105 (commutative_op): New function.
2106 (commutate): Use it. Handle more than 2 operands.
2107 (dt_operand::gen_gimple_expr): Use commutative_op.
2108 (parser::parse_expr): Allow :c to be used with non-binary
2109 operators if the commutative operand is known.
2110 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2111 CFN_FMS, CFN_FNMA and CFN_FNMS.
2112 (backprop::process_assign_use): Remove FMA_EXPR handling.
2113 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2114 (gen_hsa_fma): New function.
2115 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2116 IFN_FNMA and IFN_FNMS.
2117 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2118 * gimple-fold.h (follow_all_ssa_edges): Declare.
2119 * gimple-fold.c (follow_all_ssa_edges): New function.
2120 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2121 gimple_build interface and use follow_all_ssa_edges to fold the result.
2122 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2123 instead of checking for optabs directly.
2124 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2125 rather than FMA_EXPRs.
2126 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2127 call to IFN_FMA instead of an FMA_EXPR.
2129 2018-05-17 Jim Wilson <jimw@sifive.com>
2131 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2132 sign extended promoted subreg, and the range does not have the sign bit
2133 set, then do a sign extend.
2135 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2136 test, check for sign extended subreg and/or constant operands, and
2137 do a sign extend in that case.
2139 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2141 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2142 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2144 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2145 Change logics_shift_reg to logics_shift_imm.
2146 (thunderx2t99_fp_loadpair_basic): Delete.
2147 (thunderx2t99_fp_storepair_basic): Delete.
2148 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2149 (thunderx2t99_asimd_polynomial): Delete.
2150 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2151 and neon_fp_mul_d_scalar_q.
2152 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2153 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2154 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2155 (thunderx2t99_asimd_lut): Add missing tbl types.
2156 (thunderx2t99_asimd_ext): Delete.
2157 (thunderx2t99_asimd_load1_1_mult): Delete.
2158 (thunderx2t99_asimd_load1_2_mult): Delete.
2159 (thunderx2t99_asimd_load1_ldp): New.
2160 (thunderx2t99_asimd_load1): New.
2161 (thunderx2t99_asimd_load2): Add missing *load2* types.
2162 (thunderx2t99_asimd_load3): New.
2163 (thunderx2t99_asimd_load4): New.
2164 (thunderx2t99_asimd_store1_1_mult): Delete.
2165 (thunderx2t99_asimd_store1_2_mult): Delete.
2166 (thunderx2t99_asimd_store2_mult): Delete.
2167 (thunderx2t99_asimd_store2_onelane): Delete.
2168 (thunderx2t99_asimd_store_stp): New.
2169 (thunderx2t99_asimd_store1): New.
2170 (thunderx2t99_asimd_store2): New.
2171 (thunderx2t99_asimd_store3): New.
2172 (thunderx2t99_asimd_store4): New.
2174 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2176 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2177 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2179 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2180 Segher Boessenkool <segher@kernel.crashing.org>
2183 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2186 2018-05-17 Richard Biener <rguenther@suse.de>
2188 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2189 for pruning loop and prune defs feeding only already visited PHIs.
2191 2018-05-17 Richard Biener <rguenther@suse.de>
2193 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2195 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2196 Richard Biener <rguenther@suse.de>
2198 PR tree-optimization/85793
2199 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2200 for VMAT_ELEMENTWISE.
2202 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2204 * internal-fn.h (lookup_internal_fn): Declare
2205 * internal-fn.c (lookup_internal_fn): New function.
2206 * gimple.c (gimple_build_call_from_tree): Handle calls to
2208 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2209 internal function names.
2210 * tree-pretty-print.c (dump_generic_node): Likewise.
2211 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2213 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2215 * gimple-fold.h (gimple_build): Make the function forms take
2216 combined_fn rather than built_in_function.
2217 (gimple_simplify): Likewise.
2218 * gimple-match-head.c (gimple_simplify): Likewise.
2219 * gimple-fold.c (gimple_build): Likewise.
2220 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2221 rather than gimple_build_call_internal.
2222 (get_initial_defs_for_reduction): Likewise.
2223 (vect_create_epilog_for_reduction): Likewise.
2224 (vectorizable_live_operation): Likewise.
2226 2018-05-17 Martin Liska <mliska@suse.cz>
2228 * gimple-ssa-sprintf.c (format_directive): Do not use
2229 space in between 'G_' and '('.
2231 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2234 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2235 even if the mask is not all ones.
2238 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2240 (ix86_gimple_fold_builtin): Likewise.
2243 * config/i386/i386.c: Include tree-vector-builder.h.
2244 (ix86_vector_shift_count): New function.
2245 (ix86_fold_builtin): Fold shift builtins by scalar count.
2246 (ix86_gimple_fold_builtin): Likewise.
2248 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2249 _mm512_setzero): New intrinsics.
2251 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2252 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2254 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2255 code generation for cases where splatting a value is not useful.
2256 * simplify-rtx.c (simplify_ternary_operation): Simplify
2257 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2258 mode to a vec_concat.
2260 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2262 * config.gcc: Support "goldmont-plus".
2263 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2265 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2266 PROCESSOR_GOLDMONT_PLUS.
2267 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2268 (processor_target_table): Add "goldmont-plus".
2269 (PTA_GOLDMONT_PLUS): Define.
2270 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2271 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2272 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2273 (fold_builtin_cpu): Add "goldmont-plus".
2274 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2275 (ix86_option_override_internal): Add "goldmont-plus".
2276 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2277 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2278 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2279 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2281 2018-05-17 Richard Biener <rguenther@suse.de>
2283 PR tree-optimization/85757
2284 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2285 remove defs that only feed that PHI from further processing.
2287 2018-05-16 Jim Wilson <jimw@sifive.com>
2289 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2291 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2293 2018-05-16 Mark Wielaard <mark@klomp.org>
2295 * dwarf2out.c (count_index_strings): New function.
2296 (output_indirect_strings): Call count_index_strings and generate
2297 header for dwarf_version >= 5.
2299 2018-05-16 Mark Wielaard <mark@klomp.org>
2301 * dwarf2out.c (dwarf_FORM): New function.
2302 (set_indirect_string): Use dwarf_FORM.
2303 (reset_indirect_string): Likewise.
2304 (size_of_die): Likewise.
2305 (value_format): Likewise.
2306 (output_die): Likewise.
2307 (add_skeleton_AT_string): Likewise.
2308 (output_macinfo_op): Likewise.
2309 (index_string): Likewise.
2310 (output_index_string_offset): Likewise.
2311 (output_index_string): Likewise.
2312 (count_index_strings): Likewise.
2314 2018-05-16 Carl Love <cel@us.ibm.com>
2316 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2317 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2319 2018-05-16 Martin Jambor <mjambor@suse.cz>
2321 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2322 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2324 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2326 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2327 (fnma<mode>4): Likewise.
2328 (fms<mode>4): Likewise.
2329 (fnms<mode>4): Likewise.
2330 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2331 (aarch64_fnma<mode>4): Likewise.
2332 (aarch64_fms<mode>4): Likewise.
2333 (aarch64_fnms<mode>4): Likewise.
2334 (aarch64_fnmadd<mode>4): Likewise.
2336 2018-05-16 Jason Merrill <jason@redhat.com>
2338 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2340 2018-05-16 Richard Biener <rguenther@suse.de>
2342 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2343 (dump_stmt_cost): Declare.
2344 (add_stmt_cost): Dump cost we add.
2345 (add_stmt_costs): New function.
2346 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2348 (vect_analyze_stmt): Adjust prototype.
2349 (vectorizable_condition): Likewise.
2350 (vectorizable_live_operation): Likewise.
2351 (vectorizable_reduction): Likewise.
2352 (vectorizable_induction): Likewise.
2353 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2354 cost vector to pass to vectorizable_ and record afterwards.
2355 (vect_model_reduction_cost): Take cost vector argument and adjust.
2356 (vect_model_induction_cost): Likewise.
2357 (vectorizable_reduction): Likewise.
2358 (vectorizable_induction): Likewise.
2359 (vectorizable_live_operation): Likewise.
2360 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2361 SLP_TREE_NUMBER_OF_VEC_STMTS.
2362 (vect_analyze_slp_cost_1): Remove.
2363 (vect_analyze_slp_cost): Likewise.
2364 (vect_slp_analyze_node_operations): Take visited args and
2365 a target cost vector. Avoid processing already visited stmt sets.
2366 (vect_slp_analyze_operations): Use a local cost vector to gather
2367 costs and register those of non-discarded instances.
2368 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2369 (vect_schedule_slp_instance): Remove copying of
2370 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2372 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2373 adding cost. Record cost entry location.
2374 (vect_prologue_cost_for_slp_op): Function to compute cost of
2375 a constant or invariant generated for SLP vect in the prologue,
2376 split out from vect_analyze_slp_cost_1.
2377 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2378 (vect_model_promotion_demotion_cost): Likewise.
2379 (vect_model_store_cost): Likewise, make static.
2380 (vect_model_load_cost): Likewise.
2381 (vectorizable_bswap): Add cost vector arg and adjust.
2382 (vectorizable_call): Likewise.
2383 (vectorizable_simd_clone_call): Likewise.
2384 (vectorizable_conversion): Likewise.
2385 (vectorizable_assignment): Likewise.
2386 (vectorizable_shift): Likewise.
2387 (vectorizable_operation): Likewise.
2388 (vectorizable_store): Likewise.
2389 (vectorizable_load): Likewise.
2390 (vectorizable_condition): Likewise.
2391 (vectorizable_comparison): Likewise.
2392 (can_vectorize_live_stmts): Likewise.
2393 (vect_analyze_stmt): Likewise.
2394 (vect_transform_stmt): Adjust calls to vectorizable_*.
2395 * tree-vectorizer.c: Include gimple-pretty-print.h.
2396 (dump_stmt_cost): New function.
2398 2018-05-16 Richard Biener <rguenther@suse.de>
2400 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2401 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2402 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2403 (check_name): New callback.
2404 (dse_classify_store): Track cycles via a visited bitmap of PHI
2405 defs and simplify handling of in-loop and across loop dead stores
2406 and properly fail for loop-variant refs. Handle byte-tracking with
2407 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2410 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2412 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2413 (vect_get_mask_type_for_stmt): Likewise.
2414 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2416 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2417 to determine the statement's vector type and the vector type that
2418 should be used for calculating nunits. Deal with cases in which
2419 the type has to be deferred.
2420 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2421 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2422 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2423 (vect_determine_vf_for_stmt): New functions, split out from...
2424 (vect_determine_vectorization_factor): ...here.
2425 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2426 (vect_get_mask_type_for_stmt): New functions, split out from
2427 vect_determine_vectorization_factor.
2429 2018-05-16 Richard Biener <rguenther@suse.de>
2431 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2432 verify the [VEC_]COND_EXPR embedded comparison.
2434 2018-05-15 Martin Sebor <msebor@redhat.com>
2436 PR tree-optimization/85753
2437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2438 RECORD_TYPE in addition to ARRAY_TYPE.
2440 2018-05-15 Martin Sebor <msebor@redhat.com>
2443 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2445 2018-05-15 Richard Biener <rguenther@suse.de>
2447 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2448 add by_clobber_p one. Change algorithm to collect all defs
2449 representing uses we need to walk and try reducing them to
2450 a single one before failing.
2451 (dse_dom_walker::dse_optimize_stmt): Adjust.
2453 2018-05-13 Mark Wielaard <mark@klomp.org>
2455 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2456 (size_of_loc_descr): Likewise.
2457 (output_loc_operands): Likewise.
2458 (output_loc_operands_raw): Likewise.
2459 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2460 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2461 (hash_loc_operands): Likewise.
2462 (compare_loc_operands): Likewise.
2464 2018-05-14 Mark Wielaard <mark@klomp.org>
2466 * dwarf2out.c (count_index_addrs): New function.
2467 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2469 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2471 PR tree-optimization/83648
2472 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2473 return value as malloc candidate.
2475 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2478 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2479 as true in call to suggest_attribute.
2481 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2483 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2484 -mreadonly-in-sdata.
2486 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2488 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2490 (aarch64_crypto_aesd_fused): Likewise.
2492 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2494 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2495 (movsi_aarch64): Likewise.
2496 (load_pairsi): Likewise.
2497 (load_pairdi): Likewise.
2498 (store_pairsi): Likewise.
2499 (store_pairdi): Likewise.
2500 (load_pairsf): Likewise.
2501 (load_pairdf): Likewise.
2502 (store_pairsf): Likewise.
2503 (store_pairdf): Likewise.
2504 (zero_extend): Likewise.
2505 (trunc): Swap alternatives.
2506 (fcvt_target): Add '?' to prefer w over r.
2508 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2511 * config/i386/i386.md: Disallow non-commutative arithmetics in
2512 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2513 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2514 in the peephole2 before it.
2516 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2518 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2519 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2520 (ix86_handle_option): Handle -mcldemote.
2521 * config.gcc: New header.
2522 * config/i386/cldemoteintrin.h: New file.
2523 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2524 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2526 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2527 OPTION_MASK_ISA_CLDEMOTE.
2528 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2529 (ix86_valid_target_attribute_inner_p): Ditto.
2530 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2531 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2532 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2533 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2534 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2536 * config/i386/i386.opt: Add -mcldemote.
2537 * config/i386/x86intrin.h: New header.
2538 * doc/invoke.texi: Add -mcldemote.
2540 2018-05-14 Richard Biener <rguenther@suse.de>
2542 * doc/match-and-simplify.texi: Adjust :s documentation.
2544 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2546 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2547 intended memcpy size.
2548 (REORDER_45): Likewise.
2550 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2552 * sort.cc: New file.
2553 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2554 * vec.c (qsort_chk): Use gcc_qsort.
2555 * Makefile.in (OBJS-libcommon): Add sort.o.
2556 (build/sort.o): New target. Use it...
2557 (BUILD_RTL): ... here, and...
2558 (build/gencfn-macros): ... here, and...
2559 (build/genmatch): ... here.
2561 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2562 Chung-Ju Wu <jasonwucj@gmail.com>
2564 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2565 * config/nds32/nds32-graywolf.md: New file.
2566 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2567 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2569 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2570 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2571 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2572 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2573 * config/nds32/pipelines.md: Include n15 settings.
2575 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2576 Chung-Ju Wu <jasonwucj@gmail.com>
2578 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2579 * config/nds32/nds32-n13.md: New file.
2580 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2581 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2583 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2584 * config/nds32/nds32.md (pipeline_model): Add n13.
2585 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2586 * config/nds32/pipelines.md: Include n13 settings.
2588 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2589 Chung-Ju Wu <jasonwucj@gmail.com>
2591 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2592 * config/nds32/nds32-n10.md: New file.
2593 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2594 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2596 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2597 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2598 * config/nds32/nds32.md (pipeline_model): Add n10.
2599 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2600 * config/nds32/pipelines.md: Include n10 settings.
2602 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2603 Kito Cheng <kito.cheng@gmail.com>
2604 Chung-Ju Wu <jasonwucj@gmail.com>
2606 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2607 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2608 Add enum values for DSP extension instructions.
2609 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2611 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2612 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2614 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2615 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2616 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2617 * config/nds32/nds32-intrinsic.md: Likewise.
2618 * config/nds32/nds32_intrinsic.h: Likewise.
2619 * config/nds32/nds32-md-auxiliary.c: Likewise.
2620 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2621 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2622 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2623 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2624 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2625 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2626 TYPE_DMAC in switch statement.
2627 * config/nds32/nds32.c: New checking and implementation for DSP
2628 extension instructions.
2629 * config/nds32/nds32.h: Likewise.
2630 * config/nds32/nds32.md: Likewise.
2631 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2632 * config/nds32/predicates.md: Implement new predicates for DSP
2635 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2637 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2638 Reformat alternatives and attributes so it is easier to identify
2639 which constraints/attributes go with which instruction.
2640 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2641 (mov<mode>_softfloat32, FMOVE64): Likewise.
2642 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2643 (mov<mode>_softfloat64, FMOVE64): Likewise.
2645 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2647 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2649 (Basic PowerPC Built-in Functions): The new name of the
2650 subsection previously known as "PowerPC Built-in Functions".
2651 (Basic PowerPC Built-in Functions Available on all Configurations):
2653 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2654 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2655 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2656 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2658 2018-05-11 Martin Jambor <mjambor@suse.cz>
2661 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2664 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2667 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2669 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2671 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2672 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2673 (ix86_handle_option): Handle -mwaitpkg.
2674 * config.gcc: New header.
2675 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2676 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2677 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2679 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2680 OPTION_MASK_ISA_WAITPKG.
2681 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2682 (ix86_option_override_internal): Add PTA_WAITPKG.
2683 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2684 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2685 IX86_BUILTIN_TPAUSE.
2686 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2687 __builtin_ia32_umwait and __builtin_ia32_tpause.
2688 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2689 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2690 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2691 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2692 UNSPECV_TPAUSE): New.
2693 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2694 * config/i386/i386.opt: Add -mwaitpkg.
2695 * config/i386/waitpkgintrin.h: New file.
2696 * config/i386/x86intrin.h: New header.
2697 * doc/invoke.texi: Add -mwaitpkg.
2699 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2702 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2704 (cortex-m0): Use armv6s-m isa.
2705 (cortex-m0plus): Likewise.
2706 (cortex-m1): Likewise.
2707 (cortex-m0.small-multiply): Likewise.
2708 (cortex-m0plus.small-multiply): Likewise.
2709 (cortex-m1.small-multiply): Likewise.
2711 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2712 Jakub Jelinek <jakub@redhat.com>
2714 PR tree-optimization/85692
2715 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2716 source permute as well.
2718 2018-05-11 Martin Liska <mliska@suse.cz>
2721 * doc/extend.texi: Document LLVM style format for no_sanitize
2724 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2726 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2727 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2728 (mode_supports_vsx_dform_quad): Likewise.
2729 (mode_supports_vmx_dform): Move these functions to be next to the
2730 other mode_supports functions.
2731 (mode_supports_dq_form): Likewise.
2732 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2733 mode_supports_dq_form.
2734 (reg_offset_addressing_ok_p): Likewise.
2735 (offsettable_ok_by_alignment): Likewise.
2736 (rs6000_legitimate_offset_address_p): Likewise.
2737 (legitimate_lo_sum_address_p): Likewise.
2738 (rs6000_legitimize_address): Likewise.
2739 (rs6000_legitimize_reload_address): Likewise.
2740 (rs6000_secondary_reload_inner): Likewise.
2741 (rs6000_preferred_reload_class): Likewise.
2742 (rs6000_output_move_128bit): Likewise.
2744 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2746 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2747 Generate SImode target register for null target.
2748 <case IX86_BUILTIN_XGETBV>: Ditto.
2749 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2750 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2752 2018-05-10 Carl Love <cel@us.ibm.com>
2754 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2755 dcbtt and dcbtstt if operands[2] is 0.
2757 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2760 * config/i386/sse.md (usadv64qi): New expander.
2762 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2764 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2765 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2766 -maltivec=be support.
2767 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2768 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2769 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2770 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2771 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2772 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2773 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2774 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2775 altivec_vsumsws): Adjust.
2776 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2777 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2778 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2780 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2781 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2782 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2783 (altivec_lve<VI_char>x): Delete expand.
2784 (*altivec_lve<VI_char>x_internal): Rename to...
2785 (altivec_lve<VI_char>x): ... this.
2786 (altivec_lvxl_<mode>): Delete expand.
2787 (*altivec_lvxl_<mode>_internal): Rename to ...
2788 (altivec_lvxl_<mode>): ... this.
2789 (altivec_stvxl_<mode>): Delete expand.
2790 (*altivec_stvxl_<mode>_internal): Rename to ...
2791 (altivec_stvxl_<mode>): ... this.
2792 (altivec_stve<VI_char>x): Delete expand.
2793 (*altivec_stve<VI_char>x_internal): Rename to ...
2794 (altivec_stve<VI_char>x): ... this.
2795 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2796 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2797 reduc_plus_scal_<mode>): Adjust.
2798 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2800 (rs6000_cpu_cpp_builtins): Adjust.
2801 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2802 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2803 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2804 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2805 -maltivec=be support.
2806 (rs6000_split_vec_extract_var): Adjust.
2807 (rs6000_split_v4si_init): Adjust.
2808 (swap_selector_for_mode): Delete.
2809 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2810 altivec_expand_stvex_be): Delete.
2811 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2812 -maltivec=be support.
2813 (rs6000_gimple_fold_builtin): Ditto.
2814 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2816 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2817 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2818 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2819 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2820 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2821 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2822 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2823 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2824 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2825 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2826 anonymous split): Adjust.
2827 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2828 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2830 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2832 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2833 when --with-gxx-include-dir is also specified.
2834 * configure: Regenerate.
2836 2018-05-09 Jim Wilson <jimw@sifive.com>
2839 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2840 * config/riscv/t-withmultilib: New.
2841 * config/riscv/withmultilib.h: New.
2842 * doc/install.texi: Document RISC-V --with-multilib-list support.
2844 2018-05-09 Richard Biener <rguenther@suse.de>
2846 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2848 (vect_bb_vectorization_profitable_p): Adjust. Compute
2849 actual scalar cost using the cost vector and the add_stmt_cost
2852 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2854 PR rtl-optimization/85645
2855 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2856 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2858 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2860 PR rtl-optimization/85645
2861 * shrink-wrap.c (spread_components): Return a boolean saying if
2862 anything was changed.
2863 (try_shrink_wrapping_separate): Iterate spread_components until
2864 nothing changes anymore.
2866 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2868 PR rtl-optimization/85645
2869 * regrename.c (build_def_use): Also kill the chains that include the
2870 destination of a REG_CFA_REGISTER note.
2872 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2874 PR rtl-optimization/85645
2875 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2876 insn that has a REG_CFA_REGISTER note.
2878 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2880 * cfgexpand.c (expand_clobber): New function.
2881 (expand_gimple_stmt_1): Use it.
2882 * tree-vect-stmts.c (vect_clobber_variable): New function,
2884 (vectorizable_simd_clone_call): ...here.
2885 (vectorizable_store): Emit a clobber either side of an
2886 IFN_STORE_LANES sequence.
2887 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2889 2018-05-09 Tom de Vries <tom@codesourcery.com>
2892 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2893 (define_insn "trap_if_false"): Add exit after trap.
2895 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2897 PR rtl-optimization/85638
2898 * bb-reorder.c: Include common/common-target.h.
2899 (create_forwarder_block): New function extracted from...
2900 (fix_up_crossing_landing_pad): ...here. Rename into...
2901 (dw2_fix_up_crossing_landing_pad): ...this.
2902 (sjlj_fix_up_crossing_landing_pad): New function.
2903 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2904 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2905 from both partitions and exit the loop after one iteration.
2907 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2910 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2912 (Basic PowerPC Built-in Functions): The new name of the
2913 subsection previously known as "PowerPC Built-in Functions".
2914 (Basic PowerPC Built-in Functions Available on all Configurations):
2916 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2918 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2920 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2922 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2925 2018-05-08 Jim Wilson <jimw@sifive.com>
2927 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2928 (LD_EMUL_SUFFIX): New.
2929 (LINK_SPEC): Use it.
2931 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2933 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2935 (Basic PowerPC Built-in Functions): The new name of the
2936 subsection previously known as "PowerPC Built-in Functions".
2937 (Basic PowerPC Built-in Functions Available on all Configurations):
2939 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2941 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2943 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2945 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2948 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2951 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2952 after cmpelim optimization.
2954 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2956 * config.gcc: Support "goldmont".
2957 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2958 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2960 * config/i386/i386.c (m_GOLDMONT): Define.
2961 (processor_target_table): Add "goldmont".
2962 (PTA_GOLDMONT): Define.
2963 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2964 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2965 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2966 (fold_builtin_cpu): Add "goldmont".
2967 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2968 (ix86_option_override_internal): Add "goldmont".
2969 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2970 (processor_type): Add PROCESSOR_GOLDMONT.
2971 * config/i386/i386.md: Add CPU "glm".
2972 * config/i386/glm.md: New file.
2973 * config/i386/x86-tune.def: Add m_GOLDMONT.
2974 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2976 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2979 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2981 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2982 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2983 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2986 * config/i386/i386.c (ix86_fold_builtin): Handle
2987 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2990 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2991 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2993 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2996 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2997 (check_arch): Likewise.
2998 (check_fpu): Return the result rather than printing it.
2999 (end arch): Fix operator precedence.
3000 (end cpu): Likewise.
3001 (END): Print the result from check_fpu.
3003 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3004 Alan Hayward <alan.hayward@arm.com>
3005 David Sherwood <david.sherwood@arm.com>
3007 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3008 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3009 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3010 (*fcmuo<mode>_and): New patterns.
3012 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3014 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3015 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3016 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3017 (cmp_op, sve_imm_con): New code attributes.
3018 (SVE_COND_INT_CMP, imm_con): Delete.
3019 (cmp_op): Remove above unspecs from int attribute.
3020 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3022 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3023 comparison-specific unspecs.
3024 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3025 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3026 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3027 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3028 (*vec_fcm<cmp_op><mode>): Rename to...
3029 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3030 (*vec_fcmuo<mode>): Rename to...
3031 (*fcmuo<mode>): ...this and adjust likewise.
3032 (*pred_fcm<cmp_op><mode>): New pattern.
3033 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3034 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3036 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3038 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3039 (aarch64_emit_sve_predicated_cond): New function.
3040 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3041 (aarch64_emit_unspec_cond_or): Replace with...
3042 (aarch64_emit_sve_or_conds): ...this new function. Use
3043 aarch64_emit_sve_ptrue_op for the individual comparisons and
3044 aarch64_emit_binop to OR them together.
3045 (aarch64_emit_inverted_unspec_cond): Replace with...
3046 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3047 aarch64_emit_sve_ptrue_op for the comparison and
3048 aarch64_emit_unop to invert the result.
3049 (aarch64_expand_sve_vec_cmp_float): Update after the above
3050 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3052 2018-05-07 Nathan Sidwell <nathan@acm.org>
3054 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3055 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3056 (Backwards Compatibility): Likewise.
3058 2018-05-07 Luis Machado <luis.machado@linaro.org>
3062 2018-05-07 Luis Machado <luis.machado@linaro.org>
3064 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3065 <prefetch_dynamic_strides>: New const bool field.
3066 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3067 prefetch_dynamic_strides.
3068 (exynosm1_prefetch_tune): Likewise.
3069 (thunderxt88_prefetch_tune): Likewise.
3070 (thunderx_prefetch_tune): Likewise.
3071 (thunderx2t99_prefetch_tune): Likewise.
3072 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3073 (aarch64_override_options_internal): Update to set
3074 PARAM_PREFETCH_DYNAMIC_STRIDES.
3075 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3076 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3077 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3078 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3079 prefetch-dynamic-strides setting.
3081 2018-05-07 Luis Machado <luis.machado@linaro.org>
3083 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3084 <minimum_stride>: New const int field.
3085 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3086 minimum_stride field.
3087 (exynosm1_prefetch_tune): Likewise.
3088 (thunderxt88_prefetch_tune): Likewise.
3089 (thunderx_prefetch_tune): Likewise.
3090 (thunderx2t99_prefetch_tune): Likewise.
3091 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3092 (aarch64_override_options_internal): Update to set
3093 PARAM_PREFETCH_MINIMUM_STRIDE.
3094 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3095 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3096 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3097 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3098 stride is constant and is below the minimum stride threshold.
3100 2018-05-07 Luis Machado <luis.machado@linaro.org>
3102 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3105 2018-05-07 Luis Machado <luis.machado@linaro.org>
3107 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3108 <prefetch_dynamic_strides>: New const bool field.
3109 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3110 prefetch_dynamic_strides.
3111 (exynosm1_prefetch_tune): Likewise.
3112 (thunderxt88_prefetch_tune): Likewise.
3113 (thunderx_prefetch_tune): Likewise.
3114 (thunderx2t99_prefetch_tune): Likewise.
3115 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3116 (aarch64_override_options_internal): Update to set
3117 PARAM_PREFETCH_DYNAMIC_STRIDES.
3118 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3119 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3120 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3121 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3122 prefetch-dynamic-strides setting.
3124 2018-05-07 Luis Machado <luis.machado@linaro.org>
3126 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3127 <minimum_stride>: New const int field.
3128 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3129 minimum_stride field.
3130 (exynosm1_prefetch_tune): Likewise.
3131 (thunderxt88_prefetch_tune): Likewise.
3132 (thunderx_prefetch_tune): Likewise.
3133 (thunderx2t99_prefetch_tune): Likewise.
3134 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3135 (aarch64_override_options_internal): Update to set
3136 PARAM_PREFETCH_MINIMUM_STRIDE.
3137 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3138 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3139 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3140 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3141 stride is constant and is below the minimum stride threshold.
3143 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3146 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3147 the type is addressable. Don't force op into register if it has
3150 2018-05-05 Roland McGrath <mcgrathr@google.com>
3153 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3154 any section for which we don't know a specific type it should have,
3155 regardless of name. Previously this was done only for the exact
3156 names ".init_array", ".fini_array", and ".preinit_array".
3157 (default_elf_asm_named_section): Add comment about
3158 relationship with default_section_type_flags and SECTION_NOTYPE.
3159 (get_section): Don't consider it a type conflict if one side has
3160 SECTION_NOTYPE and the other doesn't, as long as neither has the
3161 SECTION_BSS et al used in the default_section_type_flags logic.
3163 2018-05-05 Tom de Vries <tom@codesourcery.com>
3166 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3167 (workaround_barsyncs): New function.
3168 (nvptx_reorg): Use workaround_barsyncs.
3169 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3170 (define_expand "nvptx_membar_cta"): New define_expand.
3171 (define_insn "*nvptx_membar_cta"): New insn.
3173 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3175 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3176 To improve optimization opportunities.
3177 * builtin-types.def: The new needed builtin types for the above.
3179 2018-05-04 Richard Biener <rguenther@suse.de>
3181 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3182 * gimple-ssa-store-merging.c
3183 (imm_store_chain_info::output_merged_store): Remove redundant create,
3184 release split_store vector contents on failure.
3185 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3186 scalar stmt vector on cache hit.
3188 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3190 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3192 * config.gcc (powerpc-xilinx-eabi*): Remove.
3193 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3195 (fusion_addis_mem_combo_load): Ditto.
3196 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3198 (rs6000_cpu_cpp_builtins): Ditto.
3199 * config/rs6000/rs6000-linux.c
3200 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3201 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3202 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3204 (rs6000_setup_reg_addr_masks): Ditto.
3205 (rs6000_init_hard_regno_mode_ok): Ditto.
3206 (rs6000_option_override_internal): Ditto.
3207 (legitimate_lo_sum_address_p): Ditto.
3208 (rs6000_legitimize_address): Ditto.
3209 (rs6000_legitimize_reload_address): Ditto.
3210 (rs6000_legitimate_address_p): Ditto.
3211 (abi_v4_pass_in_fpr): Ditto.
3212 (setup_incoming_varargs): Ditto.
3213 (rs6000_gimplify_va_arg): Ditto.
3214 (rs6000_split_multireg_move): Ditto.
3215 (rs6000_savres_strategy): Ditto.
3216 (rs6000_emit_prologue_components): Ditto.
3217 (rs6000_emit_epilogue_components): Ditto.
3218 (rs6000_emit_prologue): Ditto.
3219 (rs6000_emit_epilogue): Ditto.
3220 (rs6000_elf_file_end): Ditto.
3221 (rs6000_function_value): Ditto.
3222 (rs6000_libcall_value): Ditto.
3223 * config/rs6000/rs6000.h: Ditto.
3224 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3225 (TARGET_MINMAX): ... this. New.
3226 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3227 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3228 (*movsi_internal1_single): Delete.
3229 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3230 mfpu=, mxilinx-fpu): Delete.
3231 * config/rs6000/singlefp.h: Delete.
3232 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3233 * config/rs6000/t-rs6000: Ditto.
3234 * config/rs6000/t-xilinx: Delete.
3235 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3236 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3237 (VStype_simple): Delete.
3238 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3239 * config/rs6000/xfpu.h: Delete.
3240 * config/rs6000/xfpu.md: Delete.
3241 * config/rs6000/xilinx.h: Delete.
3242 * config/rs6000/xilinx.opt: Delete.
3243 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3244 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3246 2018-05-04 Tom de Vries <tom@codesourcery.com>
3249 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3252 2018-05-04 Richard Biener <rguenther@suse.de>
3255 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3256 (expand_complex_div_wide): Likewise.
3257 (expand_complex_operations_1): Likewise.
3258 (expand_complex_libcall): Preserve EH info of the original stmt.
3259 (tree_lower_complex): Handle removed blocks.
3260 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3261 on complex multiplication and division libcall builtins.
3263 2018-05-04 Richard Biener <rguenther@suse.de>
3266 * fold-const.c (negate_expr_p): Restrict negation of operand
3267 zero of a division to when we know that can happen without
3269 (fold_negate_expr_1): Likewise.
3271 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3274 * real.h (real_nextafter): Declare.
3275 * real.c (real_nextafter): New function.
3276 * fold-const-call.c (fold_const_nextafter): New function.
3277 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3278 CASE_CFN_NEXTTOWARD.
3279 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3280 even when arg1_mode is different from arg0_mode.
3282 2018-05-03 Nathan Sidwell <nathan@acm.org>
3284 * doc/extend.texi (Deprecated Features): Remove
3286 (Backwards Compatibility): Likewise.
3287 * doc/invoke.texi (C++ Language Options): Likewise.
3288 (C++ Dialect Options): Likewise.
3290 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3293 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3294 _mm512_mask_mullox_epi64): New intrinsics.
3296 2018-05-03 Tom de Vries <tom@codesourcery.com>
3299 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3300 dump files): Add offload-tree.
3302 2018-05-03 Richard Biener <rguenther@suse.de>
3304 PR tree-optimization/85615
3305 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3306 to loops not nested in BBs loop father to avoid creating multi-entry
3309 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3311 PR tree-optimization/70291
3312 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3313 arguments. Change return type to tree. Emit libcall as a new
3314 statement rather than replacing existing one when inplace_p is true.
3315 (expand_complex_multiplication_components): New function.
3316 (expand_complex_multiplication): Expand floating-point complex
3317 multiplication using the above.
3318 (expand_complex_division): Rename inner_type parameter to type.
3319 Update expand_complex_libcall call-site.
3320 (expand_complex_operations_1): Update expand_complex_multiplication
3321 and expand_complex_division call-sites.
3323 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3326 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3327 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3328 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3329 the highest significant bit of the shift count mask is clear. In
3330 check whether and[sq]i3 is needed verify that all significant bits
3331 of the shift count other than the highest are set.
3333 2018-05-02 Tom de Vries <tom@codesourcery.com>
3336 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3337 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3338 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3339 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3340 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3341 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3342 __builtin_goacc_parlevel_size.
3344 2018-05-02 Richard Biener <rguenther@suse.de>
3346 PR tree-optimization/85597
3347 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3348 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3351 2018-05-02 Tom de Vries <tom@codesourcery.com>
3354 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3355 dump files): Add ltrans-tree.
3357 2018-05-02 Tom de Vries <tom@codesourcery.com>
3360 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3361 dump files): Add wpa-ipa.
3363 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3365 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3366 powerpc*-*-linux*paired* target.
3367 * config/rs6000/750cl.h: Delete.
3368 * config/rs6000/paired.h: Delete.
3369 * config/rs6000/paired.md: Delete.
3370 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3372 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3373 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3374 comment. Remove paired float support.
3375 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3376 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3378 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3379 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3381 * config/rs6000/rs6000.c: Remove paired float support.
3382 (paired_expand_vector_init, paired_expand_vector_move,
3383 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3384 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3385 paired_expand_builtin, paired_expand_predicate_builtin,
3386 paired_init_builtins): Delete.
3387 * config/rs6000/rs6000.h: Remove paired float support.
3388 * config/rs6000/rs6000.md: Remove paired float support.
3389 (move_from_CR_ov_bit): Delete.
3390 * config/rs6000/rs6000.opt (mpaired): Delete.
3391 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3392 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3394 2018-05-02 Richard Biener <rguenther@suse.de>
3397 * gimplify.c (gimplify_save_expr): When in SSA form allow
3398 SAVE_EXPRs to compute to SSA vars.
3400 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3403 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3404 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3405 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3406 clobber operands[2], instead use a new pseudo. Formatting fixes.
3408 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3410 PR tree-optimization/85586
3411 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3412 exit early for statements in the same group if the accesses are
3415 2018-05-02 Tom de Vries <tom@codesourcery.com>
3418 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3421 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3423 PR tree-optimization/85143
3424 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3426 2018-05-01 Tom de Vries <tom@codesourcery.com>
3429 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3430 not found" error message.
3432 2018-05-01 Tom de Vries <tom@codesourcery.com>
3435 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3436 * vec.c (test_ordered_remove_if): New function.
3437 (vec_c_tests): Call test_ordered_remove_if.
3438 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3439 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3440 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3441 VEC_ORDERED_REMOVE_IF.
3443 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3445 PR tree-optimization/82665
3446 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3447 pointer subtraction where arguments come from a memchr call.
3449 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3451 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3452 --push-state --as-needed and --pop-state instead of --as-needed and
3453 --no-as-needed if ld supports it.
3454 * configure: Regenerated.
3457 * doc/install.texi2html: Replace _002d with - and _002a with * in
3458 generated html files using sed.
3460 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3463 * gcc-rich-location.c (blank_line_before_p): New function.
3464 (use_new_line): New function.
3465 (gcc_rich_location::add_fixit_insert_formatted): New function.
3466 * gcc-rich-location.h
3467 (gcc_rich_location::add_fixit_insert_formatted): New function.
3469 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3471 * selftest.c (assert_streq): Rename "expected" and "actual" to
3472 "val1" and "val2". Extend NULL-handling to cover both inputs
3473 symmetrically, while still requiring both to be non-NULL for a pass.
3474 * selftest.h (assert_streq): Rename "expected" and "actual" to
3476 (ASSERT_EQ): Likewise.
3477 (ASSERT_EQ_AT): Likewise.
3478 (ASSERT_KNOWN_EQ): Likewise.
3479 (ASSERT_KNOWN_EQ_AT): Likewise.
3480 (ASSERT_NE): Likewise.
3481 (ASSERT_MAYBE_NE): Likewise.
3482 (ASSERT_MAYBE_NE_AT): Likewise.
3483 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3484 the assertion to pass.
3485 (ASSERT_STREQ_AT): Likewise.
3487 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3489 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3490 interaction with -pie.
3492 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3494 * selftest.h: Fix alphabetization of per-source-file selftest
3497 2018-04-30 Jason Merrill <jason@redhat.com>
3499 PR c++/61982 - dead stores to destroyed objects.
3500 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3503 2018-04-30 Jason Merrill <jason@redhat.com>
3505 * tree.c (build_clobber): New.
3506 * tree.h: Declare it.
3507 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3509 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3511 * diagnostic-show-locus.c (layout::layout): Update for
3512 location_get_source_line returning a char_span.
3513 (struct char_span): Move to input.h.
3514 (struct correction): Update for fields in char_span becoming
3516 (struct source_line): Update for location_get_source_line
3517 returning a char_span.
3518 (layout::print_line): Likewise.
3519 * edit-context.c (edited_file::print_content): Likewise.
3520 (edited_file::print_diff_hunk): Likewise.
3521 (edited_file::print_run_of_changed_lines): Likewise.
3522 (edited_file::get_num_lines): Likewise.
3523 (edited_line::edited_line): Likewise.
3524 * final.c (asm_show_source): Likewise.
3525 * input.c (location_get_source_line): Convert return type
3526 from const char * to char_span, losing the final "line_len"
3528 (dump_location_info): Update for the above.
3529 (get_substring_ranges_for_loc): Likewise. Use a char_span
3530 when handling the literal within the line.
3531 (test_reading_source_line): Update for location_get_source_line
3532 returning a char_span.
3533 * input.h (class char_span): Move here from
3534 diagnostic-show-locus.c, converting from a struct to a class.
3535 Make data members private.
3536 (char_span::operator bool): New.
3537 (char_span::length): New.
3538 (char_span::get_buffer): New.
3539 (char_span::operator[]): New.
3540 (char_span::subspan): Make const.
3541 (char_span::xstrdup): New.
3542 (location_get_source_line): Convert return type from const char *
3543 to char_span, losing the final "line_size" param.
3545 2018-04-30 Jan Hubicka <jh@suse.cz>
3547 * lto-wrapper.c (ltrans_priorities): New static var.
3548 (cmp_priority): New.
3549 (run_gcc): Read priorities and if doing parallel build order
3550 the Makefile by them.
3552 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3554 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3556 2018-04-30 Richard Biener <rguenther@suse.de>
3558 * tree-cfg.c (verify_address): Remove base argument, add
3559 flag whether to check TREE_ADDRESSABLE and do that.
3560 (verify_expr): Remove.
3561 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3562 (verify_gimple_assign_single): Likewise.
3563 (verify_gimple_switch): Likewise.
3564 (verify_expr_location_1): Dereference tp once. Add (disabled)
3565 piece from verify_expr.
3566 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3568 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3570 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3572 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3574 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3575 (small_data_pattern): Likewise.
3576 (arc_rewrite_small_data): Likewise.
3577 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3578 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3579 (get_symbol_alignment): New function.
3580 (legitimate_small_data_address_p): Likewise.
3581 (legitimate_scaled_address): Update, call
3582 legitimate_small_data_address_p.
3583 (output_sdata): New static variable.
3584 (arc_print_operand): Update how we handle small data operands.
3585 (arc_print_operand_address): Likewise.
3586 (arc_legitimate_address_p): Update, use
3587 legitimate_small_data_address_p.
3588 (arc_rewrite_small_data_p): Remove.
3589 (arc_rewrite_small_data_1): Likewise.
3590 (arc_rewrite_small_data): Likewise.
3591 (small_data_pattern): Likewise.
3592 (compact_sda_memory_operand): Update to use
3593 legitimate_small_data_address_p and get_symbol_alignment.
3594 (prepare_move_operands): Don't rewite sdata pattern.
3595 (prepare_extend_operands): Remove.
3596 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3598 (zero_extendqisi2): Likewise.
3599 (zero_extendhisi2): Likewise.
3600 (extendqihi2): Likewise.
3601 (extendqisi2): Likewise.
3602 (extendhisi2): Likewise.
3606 * config/arc/constraints.md (Usd): Change it to memory constraint.
3608 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3610 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3611 as source of std instructions.
3612 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3613 allow 6-bit constants as source for store instructions.
3614 (movdi_insn): Update instruction pattern to allow 6-bit constants
3615 as source for store instructions.
3617 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3619 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3621 2018-04-30 Nathan Sidwell <nathan@acm.org>
3622 Sandra Loosemore <sandra@codesourcery.com>
3624 * dumpfile.c (dump_open): Allow '-' for stdout.
3625 * doc/invoke.texi (Developer Options): Document dump filename
3626 determination early. Document stdin/stdout selection.
3628 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3630 Microblaze Target: PIC data text relative
3632 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3633 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3635 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3636 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3637 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3638 New addressing mode for data-text relative position indepenedent code.
3639 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3640 'ADDRESS_SYMBOLIC_TXT_REL'.
3641 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3642 (microblaze_legitimate_pic_operand): Exclude function calls from
3643 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3644 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3646 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3647 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3648 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3649 for 'address + offset'.
3650 (microblaze_expand_prologue): Add new function prologue call for
3652 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3653 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3654 table in case of TARGET_PIC_DATA_TEXT_REL.
3655 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3656 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3657 Add new macros 'UNSPEC_TEXT',
3658 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3659 + exclude function calls from 'UNSPEC_PLT' in case of data text
3661 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3662 new target hook for generating address diff vector tables in case of
3664 * doc/tm.texi : Regenerate.
3665 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3666 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3667 of addr diff vector generation.
3668 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3669 target hook definition.
3670 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3671 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3672 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3673 option for data text relative.
3675 2018-04-30 Richard Biener <rguenther@suse.de>
3677 * tree-chrec.h (evolution_function_is_constant_p): Remove
3679 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3681 2018-04-30 Richard Biener <rguenther@suse.de>
3684 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3686 2018-04-30 Richard Biener <rguenther@suse.de>
3688 PR tree-optimization/28364
3689 PR tree-optimization/85275
3690 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3691 copying first exit test.
3693 2018-04-28 Mark Wielaard <mark@klomp.org>
3695 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3697 (dwarf_AT): Handle DW_AT_addr_base.
3698 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3700 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3703 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3704 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3705 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3706 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3708 2018-04-28 Richard Biener <rguenther@suse.de>
3710 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3711 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3712 to reflect use. Only add interesting stmts.
3714 2018-04-27 Martin Jambor <mjambor@suse.cz>
3717 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3718 the jump function allows for passing through aggregate values.
3720 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3722 * input.h (in_system_header_at): Convert from macro to inline
3724 (from_macro_expansion_at): Likewise.
3725 (from_macro_definition_at): Likewise.
3727 2018-04-27 Jeff Law <law@redhat.com>
3729 * config.gcc: Mark tile* targets as deprecated/obsolete.
3731 2018-04-27 Richard Biener <rguenther@suse.de>
3733 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3736 2018-04-27 Richard Biener <rguenther@suse.de>
3738 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3740 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3742 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3743 with Yd constraint. Set "preferred_for_speed" attribute from
3744 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3746 (*movdi_internal): Ditto.
3747 (movti_interunit splitters): Remove
3748 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3749 (movdi_interunit splitters): Ditto.
3750 * config/i386/constraints.md (Ye): Remove.
3751 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3753 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3756 * config/aarch64/constraints.md (Usg): Limit to 31.
3759 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3761 PR tree-optimization/85529
3762 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3763 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3764 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3765 zero extension or masking of the MSB bit.
3766 (optimize_range_tests): Add FIRST_BB argument, pass it through
3767 to optimize_range_tests_var_bound.
3768 (maybe_optimize_range_tests, reassociate_bb): Adjust
3769 optimize_range_tests callers.
3771 2018-04-26 Richard Biener <rguenther@suse.de>
3772 Jakub Jelinek <jakub@redhat.com>
3774 * cgraph.h (symbol_table): Just declare debug method here.
3775 * symtab.c (symbol_table::debug): Define.
3777 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3779 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3781 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3783 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3784 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3785 (*movdi_internal): Substitute Yi and Yj constraint with x
3786 and Ym and Yn constraint with y constraint. Update "isa"
3787 attribute and set "preferred_for_speed" attribute from
3788 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3789 (*movsi_internal): Ditto.
3790 (*movdf_internal): Ditto.
3791 (*movsf_internal): Ditto.
3792 (*zero_extendsidi2): Ditto.
3793 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3794 (sse2_loadld): Ditto.
3795 (*vec_extract<ssevecmodelower>_0): Ditto.
3796 (*vec_extractv4si_0_zext_sse4): Ditto.
3797 (vec_concatv2di): Ditto.
3798 (*vec_dup<mode>): Ditto.
3799 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3800 * config/i386/constraints.md (Yi): Remove.
3805 2018-04-26 Nathan Sidwell <nathan@acm.org>
3807 * dumpfile.c (dump_open): New.
3808 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3809 (dump_finish): Detect stdio/stderr by value not name.
3811 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3813 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3815 2018-04-26 Tom de Vries <tom@codesourcery.com>
3818 * config/nvptx/nvptx.c (verify_neutering_jumps)
3819 (verify_neutering_labels): New function
3820 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3822 2018-04-26 Tom de Vries <tom@codesourcery.com>
3825 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3826 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3829 2018-04-26 Richard Biener <rguenther@suse.de>
3830 Tom de Vries <tom@codesourcery.com>
3833 * lto-streamer-out.c (output_function): Fixup loops if required to match
3834 discovery done in the reader.
3836 2018-04-26 Richard Biener <rguenther@suse.de>
3838 PR tree-optimization/85116
3839 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3840 have a loop exit from the single latch predecessor. Remove
3841 case of header with just condition.
3842 (ch_base::copy_headers): Exclude infinite loops from any
3844 (pass_ch::execute): Record exits.
3846 2018-04-26 Richard Biener <rguenther@suse.de>
3848 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3849 prologue cost vector and pass it to vect_get_load_cost.
3850 (vect_get_peeling_costs_all_drs): Likewise.
3851 (vect_peeling_hash_get_lowest_cost): Likewise.
3852 (vect_enhance_data_refs_alignment): Likewise.
3854 2018-04-26 Richard Biener <rguenther@suse.de>
3857 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3858 checking of integer<->pointer conversions.
3859 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3860 sign-/zero-extending pointer types.
3861 (expand_omp_for_static_chunk): Likewise.
3863 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3864 Jean Lee <xiaoyur347@gmail.com>
3866 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3867 (TARGET_ASAN_SHADOW_OFFSET): Define.
3868 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3869 true for -fsanitize=address.
3871 2018-04-25 Mark Wielaard <mark@klomp.org>
3873 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3876 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3878 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3879 than "alu", remove explicit "memory" and "imm_disp" attributes.
3880 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3883 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3884 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3885 gen_lowpart_no_emit.
3887 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3890 * config/i386/i386.c (ix86_expand_builtin): Change memory
3891 operand to XI, extend p0 to Pmode.
3892 * config/i386/i386.md: Change unspec volatile and operand
3893 1 mode to XI, change operand 0 mode to P.
3895 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3897 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3898 GET_MODE_MASK before any checking.
3899 (nds32_can_use_bset_p): Likewise.
3900 (nds32_can_use_btgl_p): Likewise.
3902 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3904 * config/nds32/nds32-doubleword.md: New define_split pattern for
3905 illegal register number.
3907 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3909 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3911 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3913 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3915 2018-04-25 Richard Biener <rguenther@suse.de>
3917 * lto-streamer.h (LTO_major_version): Bump to 8.
3919 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3921 * BASE-VER: Set to 9.0.0.
3923 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3925 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3926 in __abskf2 and __powikf2.
3928 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3931 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3932 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3933 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3934 Use the above on operand 2. Reindent.
3935 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3937 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3940 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3942 * config/i386/i386.opt (mcet): Removed.
3943 * doc/install.texi: Remove -mcet documentation.
3944 * doc/invoke.texi: Likewise.
3946 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3949 * doc/install.texi: Remove -mcet from bootstrap-cet.
3951 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3954 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3955 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3959 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3960 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3961 containing a CONST_VECTOR.
3963 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3965 * doc/install.texi: Update newlib dependency for nvptx.
3967 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3970 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3971 instead of INTVAL when shifting x left.
3973 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3975 PR tree-optimization/85478
3976 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3977 vect_grouped_store_supported for single element vectors.
3979 2018-04-24 Richard Biener <rguenther@suse.de>
3982 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3983 load cost increase to the case of non-constant step.
3985 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3988 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3989 destination if any_malformed_asm.
3991 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3994 * expr.c (store_field): In the bitfield case, if the value comes from
3995 a function call and is returned in registers by means of a PARALLEL,
3996 do not change the mode of the temporary unless BLKmode and VOIDmode.
3998 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4000 PR rtl-optimization/85423
4001 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4002 dependencies to debug insns when the previous insn is non-debug.
4004 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4006 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4007 enums into a single definition.
4008 (fls): Fix predicates and printing.
4011 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4013 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4014 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4015 and short u6 immediate.
4016 (check_if_valid_sleep_operand): Remove.
4017 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4019 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4021 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4022 flag_always_save_lp condition.
4023 * config/nds32/nds32.opt (malways-save-lp): New option.
4025 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4027 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4028 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4029 * config/nds32/nds32.h
4030 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4031 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4033 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4035 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4036 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4038 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4039 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4043 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4044 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4046 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4048 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4050 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4052 * config/nds32/nds32-protos.h (nds32_data_alignment,
4053 nds32_local_alignment): Declare.
4054 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4055 nds32_local_alignment): New functions.
4056 (TARGET_CONSTANT_ALIGNMENT): Define.
4057 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4059 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4061 * config/nds32/nds32.c
4062 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4063 (TARGET_MODES_TIEABLE_P): Likewise.
4065 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4067 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4070 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4071 Chung-Ju Wu <jasonwucj@gmail.com>
4073 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4074 for unaligned access.
4075 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4077 * config/nds32/nds32-intrinsic.md: Likewise.
4078 * config/nds32/nds32_intrinsic.h: Likewise.
4079 * config/nds32/nds32.h (nds32_builtins): Likewise.
4080 * config/nds32/nds32.opt (munaligned-access): New option.
4081 * config/nds32/nds32.c (nds32_asm_file_start): Display
4082 flag_unaligned_access status.
4084 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4086 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4087 -mno-relax is present.
4088 * config/riscv/linux.h (LINK_SPEC): Ditto.
4090 2018-04-20 Martin Sebor <msebor@redhat.com>
4093 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4095 (gimple_fold_builtin_stxcpy_chk): Same.
4096 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4098 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4101 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4102 __powikf2 when long double is IEEE 128-bit.
4104 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4106 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4107 step to make sure stack always aligned.
4109 2018-04-20 Carl Love <cel@us.ibm.com>
4112 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4113 size check for arg0.
4115 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4116 Tom de Vries <tom@codesourcery.com>
4119 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4120 Emit insns for calls too.
4121 (nvptx_find_par): Always look for worker-level predecessor insn.
4122 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4124 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4125 (nvptx_process_pars): Propagate frames for calls.
4127 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4130 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4132 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4133 (ix86_handle_option): Don't handle OPT_mibt.
4134 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4136 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4138 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4139 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4140 (ix86_target_macros): Define __CET__ with flag_cf_protection
4141 for -fcf-protection.
4142 * config/i386/i386.c (isa2_opts): Remove -mibt.
4143 * config/i386/i386.h (TARGET_IBT): Removed.
4144 (TARGET_IBT_P): Likewise.
4145 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4146 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4147 * config/i386/i386.opt (mcet): Update help message.
4150 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4151 -mcet as an alias for -mshstk.
4153 2018-04-20 Richard Biener <rguenther@suse.de>
4156 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4157 complexity by forcing a single use of the multiply operand.
4159 2018-04-20 Martin Jambor <mjambor@suse.cz>
4162 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4163 recursion dependency to only apply to non-clones.
4165 2018-04-20 Martin Jambor <mjambor@suse.cz>
4168 * ipa-cp.c (create_specialized_node): Check that clones of
4169 self-recursive edges exist during IPA-CP.
4171 2018-04-19 Toon Moene <toon@moene.org>
4173 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4176 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4178 PR tree-optimization/85467
4179 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4180 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4181 VECTOR_CST element to type.
4183 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4186 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4187 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4188 (builtin_longjmp): Likewise.
4189 (save_stack_nonlocal): New pattern.
4190 (restore_stack_nonlocal): Likewise.
4192 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4195 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4196 Replace ASM_OUTPUT_LABEL with fprintf.
4198 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4201 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4202 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4203 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4204 define __IBT__ and __SHSTK__ for -fcf-protection.
4205 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4207 (ix86_trampoline_init): Likewise.
4208 (x86_output_mi_thunk): Likewise.
4209 (ix86_notrack_prefixed_insn_p): Likewise.
4210 (ix86_option_override_internal): Don't disallow -fcf-protection.
4211 * config/i386/i386.md (rdssp<mode>): Also enable for
4213 (incssp<mode>): Likewise.
4214 (nop_endbr): Likewise.
4215 * config/i386/i386.opt (mcet): Change help message to built-in
4219 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4220 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4221 enable CET built-in functions.
4223 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4225 * common/config/i386/i386-common.c
4226 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4227 OPTION_MASK_ISA_MOVDIRI_UNSET,
4228 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4229 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4230 * config.gcc (movdirintrin.h): New header.
4231 * config/i386/cpuid.h (bit_MOVDIRI,
4232 bit_MOVDIR64B): New bits.
4233 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4235 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4236 (VOID, PVOID, PCVOID)): New function types.
4237 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4238 __builtin_ia32_directstoreu_u64,
4239 __builtin_ia32_movdir64b): New builtins.
4240 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4241 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4243 (ix86_valid_target_attribute_inner_p): Ditto.
4244 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4245 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4246 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4247 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4248 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4249 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4250 (movdiri<mode>, movdir64b_<mode>): New.
4251 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4252 * config/i386/immintrin.h: Include movdirintrin.h.
4253 * config/i386/movdirintrin.h: New file.
4254 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4256 2018-04-19 Richard Biener <rguenther@suse.de>
4259 * cfg.c (clear_bb_flags): When loop state says we have
4260 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4262 2018-04-19 Richard Biener <rguenther@suse.de>
4264 PR tree-optimization/84737
4265 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4266 copying restrict info.
4267 (vect_setup_realignment): Use it.
4268 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4269 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4270 the first DR to all generated stores.
4271 (vectorizable_load): Likewise for loads.
4273 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4275 PR tree-optimization/85446
4276 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4277 the integral and pointer types to have the same precision.
4279 * doc/install.texi: Document --disable-cet being the default and
4282 2018-04-18 Martin Liska <mliska@suse.cz>
4284 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4287 2018-04-18 Martin Liska <mliska@suse.cz>
4290 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4293 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4294 arguments if they are comparable.
4296 2018-04-18 Martin Liska <mliska@suse.cz>
4299 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4302 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4305 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4308 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4309 ENDBR after calling __morestack.
4311 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4314 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4315 by using gcc_base_ver to generate a gcc_driver_version, and use
4316 it when generating GCC_DRIVER_NAME.
4317 * configure: Regenerate.
4319 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4322 * config.gcc: Obsolete powerpc*-*-*spe*.
4324 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4327 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4328 (stabstr_D): Change type of unum from unsigned int to
4329 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4332 2018-04-17 Jim Wilson <jimw@sifive.com>
4335 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4336 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4337 Set arg_pointer_offset after using pretend_args_size.
4339 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4341 PR rtl-optimization/85431
4342 * dse.c (record_store): Ignore zero width stores.
4345 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4346 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4347 __builtin_stack_restore rather than after it.
4348 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4349 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4350 argument instead of virtual_dynamic_stack_rtx.
4352 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4354 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4356 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4357 Add note to error message to explain internal mapping of overloaded
4358 built-in function name to non-overloaded built-in function name.
4359 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4362 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4365 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4366 where the inputs overlap with the output.
4368 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4370 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4371 (=v, v) alternative and explicit "memory" attribute.
4372 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4373 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4375 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4376 "sselog1" type instead of "sselog".
4377 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4378 "sselog". Remove explicit "memory" attribute.
4379 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4380 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4382 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4383 "sselog1" type instead of "sselog". Remove explicit "memory"
4385 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4386 use "sselog1" type instead of "sselog". Remove explicit "memory"
4388 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4389 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4391 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4392 "sselog1" type instead of "sselog". Remove explicit "memory"
4394 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4395 use "sselog1" type instead of "sselog". Remove explicit "memory"
4399 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4402 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4405 2018-04-17 Martin Jambor <mjambor@suse.cz>
4408 * ipa-cp.c (create_specialized_node): Call
4409 expand_all_artificial_thunks if necessary.
4411 2018-04-17 Martin Liska <mliska@suse.cz>
4414 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4415 in message, remote space in between '_G' and '('.
4417 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4420 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4421 avx512f_vmcmp<mode>3<round_saeonly_name>,
4422 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4423 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4424 avx512f_rndscale<mode><round_saeonly_name>,
4425 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4426 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4427 Use %<iptr>2 instead of %2 for -masm=intel.
4428 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4429 avx512f_vcvttss2usi<round_saeonly_name>,
4430 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4432 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4433 avx512f_vcvttsd2usi<round_saeonly_name>,
4434 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4435 Use %q1 instead of %1 for -masm=intel.
4436 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4437 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4438 of %3 for -masm=intel.
4439 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4441 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4443 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4445 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4446 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4448 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4450 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4451 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4452 %g1 and one with %0 and %1.
4453 (avx512er_vmrcp28<mode><round_saeonly_name>,
4454 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4456 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4457 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4458 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4459 of %0 and %{%4%} for -masm=intel.
4460 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4461 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4462 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4463 order of %0 and %{%5%}%{z%} for -masm=intel.
4465 2018-04-17 Jan Hubicka <jh@suse.cz>
4468 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4470 2018-04-17 Martin Liska <mliska@suse.cz>
4473 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4474 for target_clone error message. Make default implementation
4475 clone to be a local declaration.
4476 (separate_attrs): Add new argument and check for an empty
4478 (expand_target_clones): Handle it.
4479 (ipa_target_clone): Make redirection just for target_clones
4482 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4483 Tom de Vries <tom@codesourcery.com>
4486 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4487 tiled basic blocks without omp continue statements.
4489 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4492 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4493 vec_extract expression as having side effects to make sure it gets
4496 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4499 * config/i386/i386.c (get_builtin_code_for_version): Check
4502 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4505 * gcc/config.gcc: Support "skylake".
4506 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4508 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4509 (processor_target_table): Add "skylake".
4510 (ix86_option_override_internal): Add "skylake".
4511 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4512 PROCESSOR_CANNONLAKE.
4513 (get_builtin_code_for_version): Fix priority for
4514 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4515 PROCESSOR_SKYLAKE-AVX512.
4516 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4517 (processor_type): Add PROCESSOR_SKYLAKE.
4519 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4520 Jason Merrill <jason@redhat.com>
4523 * convert.c (convert_to_integer_1): Use direct recursion for
4524 enumeral types and types with a precision less than the number
4525 of bits in their mode.
4527 2018-04-16 Julia Koval <julia.koval@intel.com>
4530 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4531 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4533 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4536 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4537 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4538 and -mno-direct-move.
4540 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4543 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4544 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4545 Ensure negative shifts result in {0}.
4547 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4549 PR rtl-optimization/79916
4550 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4551 regs (if any) to define how to gnerate SD moves when LRA is in
4554 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4556 PR rtl-optimization/85393
4557 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4558 * except.c (expand_dw2_landing_pad_for_region): Make static.
4559 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4560 a label and unconditional jump to old_bb, rather than
4561 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4564 PR rtl-optimization/85376
4565 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4566 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4567 instead of a specific value.
4569 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4570 Bin Cheng <bin.cheng@arm.com>
4572 PR tree-optimization/82965
4573 PR tree-optimization/83991
4574 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4575 by_profile_only parameter.
4576 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4577 information if the loop was predicted to iterate too many times.
4578 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4580 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4583 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4586 2018-04-13 Martin Liska <mliska@suse.cz>
4587 Jakub Jelinek <jakub@redhat.com>
4590 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4591 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4592 * builtins.c (expand_builtin_memory_copy_args): Use
4593 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4594 handle dest_addr == pc_rtx.
4596 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4599 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4600 asked to not generate direct moves.
4601 (fix_trunc<mode>si2_stfiwx): Similar.
4602 (fix_trunc<mode>si2_internal): Similar.
4604 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4607 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4608 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4609 lookup if dest in some wider mode is known to be const0_rtx and
4610 if so, record permanent equivalence for it to be ZERO_EXTEND of
4611 the narrower mode destination.
4613 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4615 * lto-streamer-out.c (output_function): Revert 259346.
4616 * omp-expand.c (expand_oacc_for): Likewise.
4618 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4620 PR rtl-optimization/85354
4621 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4622 * sel-sched.c (sel_global_init): ... here.
4624 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4627 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4628 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4629 mode for PE-COFF targets.
4630 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4631 (i386_pe_asm_lto_end): Likewise.
4632 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4633 (TARGET_ASM_LTO_END): Likewise.
4634 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4635 (i386_pe_asm_lto_start): New function.
4636 (i386_pe_asm_lto_end): Likewise.
4638 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4639 Richard Biener <rguenther@suse.de>
4642 * lto-streamer-out.c (output_function): Fix CFG loop state before
4644 * omp-expand.c (expand_oacc_for): Handle calls to internal
4645 functions like regular functions.
4647 2018-04-12 Richard Biener <rguenther@suse.de>
4650 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4651 for the early LTO debug to properly generate references to it
4652 during DIE emission. Do not re-use that for the skeleton for
4654 (dwarf2out_early_finish): Likewise.
4656 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4659 * config/i386/sse.md
4660 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4661 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4662 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4663 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4664 and output is a reg, avoid creating invalid lowpart subreg, but
4665 instead split into a 512-bit move. Don't split if not AVX512VL,
4666 input is xmm16+ reg and output is a mem.
4667 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4668 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4669 xmm16+ reg and output is a mem.
4671 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4673 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4674 also for flag_dwarf2_cfi_asm.
4676 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4678 PR rtl-optimization/85342
4679 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4680 a bool scalar var inside of the loop instead. Don't try to update
4681 recog_data.operand after failed apply_change_group.
4683 2018-04-12 Tom de Vries <tom@codesourcery.com>
4686 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4687 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4688 array with flexible array member as array without given dimension.
4689 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4690 argument for undefined param to true.
4692 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4695 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4696 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4697 from PowerPC section.
4698 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4699 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4700 help text that is too long.
4701 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4702 help text that is too long.
4703 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4704 help text that is too long.
4706 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4708 * config/alpha/alpha.md (stack_probe_internal): Rename
4709 from "probe_stack". Update all callers.
4711 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4713 PR rtl-optimization/84566
4714 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4715 sched_macro_fuse_insns.
4717 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4720 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4721 (compute_block_dependences): ... from here.
4723 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4725 PR tree-optimization/85331
4726 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4727 from int to HOST_WIDE_INT.
4729 2018-04-11 Martin Jambor <mjambor@suse.cz>
4732 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4733 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4734 not the same as the source val.
4735 (cgraph_edge_brings_value_p): New parameter.
4736 (gather_edges_for_value): Pass destination value to
4737 cgraph_edge_brings_value_p.
4738 (perhaps_add_new_callers): Likewise.
4739 (get_info_about_necessary_edges): Likewise and exclude values brought
4740 only by self-recursive edges.
4741 (create_specialized_node): Redirect only clones of self-calling edges.
4742 (+self_recursive_pass_through_p): New function.
4743 (find_more_scalar_values_for_callers_subset): Use it.
4744 (find_aggregate_values_for_callers_subset): Likewise.
4745 (known_aggs_to_agg_replacement_list): Removed.
4746 (decide_whether_version_node): Re-calculate known constants for all
4747 remaining context clones.
4749 2018-04-11 Richard Biener <rguenther@suse.de>
4752 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4753 from early DWARF output.
4754 (dwarf2out_early_finish): Output line info unconditionally into
4755 early DWARF and add reference to it.
4757 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4760 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4761 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4762 other than V2DFmode using iptr mode attribute.
4763 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4765 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4767 PR rtl-optimization/84659
4768 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4770 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4773 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4775 (output_loc_list): Pass address of a dummy size variable even in the
4776 locview handling loop.
4777 (index_location_lists): Add comment on why skip_loc_list_entry can't
4780 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4783 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4786 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4789 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4790 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4791 and -mstring-compare-inline-limit.
4793 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4796 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4797 for stack clash protection in a register whenever we need it to be in
4800 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4802 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4803 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4805 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4808 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4810 (mlong-double-): Ditto.
4811 * config/rs6000/sysv4.opt (msdata=): Ditto.
4812 (mtls-size=): Ditto.
4814 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4816 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4817 erroneous entries for
4818 "vector int vec_ldl (int, long int *)", and
4819 "vector unsigned int vec_ldl (int, unsigned long int *)".
4820 Add comments and entries for
4821 "vector bool char vec_ldl (int, bool char *)",
4822 "vector bool short vec_ldl (int, bool short *)",
4823 "vector bool int vec_ldl (int, bool int *)",
4824 "vector bool long long vec_ldl (int, bool long long *)",
4825 "vector pixel vec_ldl (int, pixel *)",
4826 "vector long long vec_ldl (int, long long *)",
4827 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4828 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4829 type tree bool_long_long_type_node and correct definition of
4830 bool_V2DI_type_node to make reference to this new type tree.
4831 (rs6000_mangle_type): Replace erroneous reference to
4832 bool_long_type_node with bool_long_long_type_node.
4833 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4834 comments to emphasize sign distinctions for char and int types and
4835 replace RS6000_BTI_bool_long constant with
4836 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4837 use of RS6000_BTI_pixel.
4838 (bool_long_type_node): Remove this macro definition.
4839 (bool_long_long_type_node): New macro definition
4841 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4843 PR rtl-optimization/85300
4844 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4845 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4846 simplify_unary_operation fails.
4848 2018-04-10 Martin Liska <mliska@suse.cz>
4850 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4851 cgraph_edge and ipa_ref.
4853 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4857 * config/i386/sse.md
4858 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4859 computation of the VEC_MERGE selector from mask.
4860 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4861 Fix decoding of the VEC_MERGE selector into mask.
4863 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4865 PR tree-optimization/85286
4866 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4868 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4870 * final.c (final_1): Set insn_last_address as well as
4871 insn_current_address.
4873 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4876 * explow.c (emit_stack_probe): Call validize_mem on memory location
4877 before passing it to gen_probe_stack. Create address operand and
4878 legitimize it for the probe_stack_address case.
4880 2018-04-09 Jan Hubicka <jh@suse.cz>
4883 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4884 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4885 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4887 (free_lang_data): Rebuild type inheritance graph.
4889 2018-04-09 Martin Sebor <msebor@redhat.com>
4891 * invoke.texi (-finline-small-functions): Mention other optimization
4893 (-findirect-inlining, -fpartial-inlining): Same.
4894 (-finline-functions-called-once): Same.
4895 (-freorder-blocks-and-partition): Same.
4897 2018-04-09 Jan Hubicka <jh@suse.cz>
4900 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4901 jumps; choose last target that matches the criteria (i.e.
4902 no partition changes for non-crossing jumps).
4903 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4904 support for redirecting crossing jumps to non-crossing.
4906 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4908 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4909 also for naked functions.
4911 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4913 * config/arc/arc.md (add_shift): New pattern.
4914 (add_shift2): Likewise.
4915 (sub_shift): Likewise.
4916 (sub_shift_cmp0_noout): Likewise.
4917 (compare_si_ashiftsi): Likewise.
4918 (xbfu_cmp0_noout): New combine pattern.
4919 (xbfu_cmp0"): Likewise.
4920 (movsi_set_cc_insn): Place the predicable variant first.
4921 (commutative_binary_cmp0_noout): Remove clobber.
4922 (commutative_binary_cmp0): New pattern.
4923 (noncommutative_binary_cmp0): Likewise.
4924 (noncommutative_binary_cmp0_noout): Likewise.
4925 (noncommutative_binary_comparison_result_used): Removed.
4926 (rsub_cmp0): New pattern.
4927 (rsub_cmp0_noout): Likewise.
4928 (extzvsi): Changed, keep only meaningful variants.
4929 (SQH, SEZ): New iterators.
4930 (SQH_postfix): New mode attribute.
4931 (SEZ_prefix): New code attribute.
4932 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4933 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4934 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4936 (noncommutative_operator): Check the availability of barrel
4939 2018-04-09 Richard Biener <rguenther@suse.de>
4941 PR tree-optimization/85284
4942 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4943 Only use the niter constraining form of simple_iv when the exit
4946 2018-04-09 Tom de Vries <tom@codesourcery.com>
4949 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4950 (define_expand "*memory_barrier"): New define_expand.
4951 (define_insn "memory_barrier"): New insn.
4953 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4955 PR rtl-optimization/80463
4956 PR rtl-optimization/83972
4957 PR rtl-optimization/83480
4959 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4960 correct producer for the insn.
4961 (tidy_control_flow): Fixup seqnos in case of debug insns.
4963 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4965 PR rtl-optimization/83913
4967 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4968 different sched-times when merging exprs.
4970 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4972 PR rtl-optimization/83962
4974 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4975 tidy_fallthru_edge and tidy_control_flow.
4977 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4979 PR rtl-optimization/83530
4981 * sel-sched.c (force_next_insn): New global variable.
4982 (remove_insn_for_debug): When force_next_insn is true, also leave only
4983 next insn in the ready list.
4984 (sel_sched_region): When the region wasn't scheduled, make another pass
4985 over it with force_next_insn set to 1.
4987 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4989 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4991 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4992 for interrupt control.
4993 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4994 functions for interrupt control.
4995 * config/nds32/nds32-intrinsic.md: Likewise.
4996 * config/nds32/nds32_intrinsic.h: Likewise.
4997 * config/nds32/nds32.h (nds32_builtins): Likewise.
4999 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5001 * config/nds32/nds32.c (nds32_init_machine_status,
5002 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5003 strict_aligned_p field.
5004 (nds32_expand_to_rtl_hook): New function.
5005 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5006 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5008 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5009 Chung-Ju Wu <jasonwucj@gmail.com>
5011 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5012 * config/nds32/nds32-n7.md: New file.
5013 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5014 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5016 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5017 * config/nds32/nds32.md (pipeline_model): Add n7.
5018 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5019 * config/nds32/pipelines.md: Include n7 settings.
5021 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5022 Chung-Ju Wu <jasonwucj@gmail.com>
5024 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5025 * config/nds32/nds32-e8.md: New file.
5026 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5027 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5029 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5030 * config/nds32/nds32.md (pipeline_model): Add e8.
5031 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5032 * config/nds32/pipelines.md: Include e8 settings.
5034 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5035 Chung-Ju Wu <jasonwucj@gmail.com>
5037 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5038 * config/nds32/nds32-n8.md: New file.
5039 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5040 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5042 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5043 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5044 * config/nds32/nds32.md (pipeline_model): Add n8.
5045 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5046 * config/nds32/pipelines.md: Include n8 settings.
5048 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5049 Chung-Ju Wu <jasonwucj@gmail.com>
5051 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5052 * config/nds32/nds32-n9-2r1w.md: New file.
5053 * config/nds32/nds32-n9-3r2w.md: New file.
5054 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5055 nds32_register_ports): New or modify for cpu n9.
5056 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5058 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5059 * config/nds32/nds32-utils.c: New file.
5060 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5061 TARGET_MUL_SLOW): Define.
5062 * config/nds32/nds32.md (pipeline_model): New attribute.
5063 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5064 New options that support cpu n9.
5065 * config/nds32/pipelines.md: Include n9 settings.
5066 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5068 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5070 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5071 information if necessary.
5072 (output_cond_branch_compare_zero): Likewise.
5073 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5074 (nds32_target_alignment): Refine for alignment.
5075 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5076 (FUNCTION_BOUNDARY): Modify.
5077 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5079 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5081 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5083 * config/nds32/constants.md (unspec_volatile_element): Add values for
5084 TLB operation and data prefetch.
5085 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5086 functions for TLB operation and data prefetch.
5087 * config/nds32/nds32-intrinsic.md: Likewise.
5088 * config/nds32/nds32_intrinsic.h: Likewise.
5089 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5090 (nds32_print_operand): Likewise.
5091 * config/nds32/nds32.h (nds32_builtins): Likewise.
5093 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5094 Andrew Pinski <pinsika@gcc.gnu.org>
5097 * match.pd: Use constant_boolean_node of correct type instead of
5098 boolean_true_node or boolean_false_node for simplifying
5099 pointer comparisons to zero.
5101 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5103 PR tree-optimization/80021
5104 * tree.c (verify_type_variant): Make error call in verify_variant_match
5105 translatable and remove final full stop.
5107 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5109 * config/nds32/constants.md (unspec_volatile_element): Add
5110 UNSPEC_VOLATILE_EH_RETURN.
5111 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5112 nds32_output_stack_pop): Support dwarf exception handling process.
5113 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5114 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5115 exception handling process.
5116 (nds32_compute_stack_frame): Likewise.
5117 (nds32_return_addr_rtx): Likewise.
5118 (nds32_initial_elimination_offset): Likewise.
5119 (nds32_expand_prologue): Likewise.
5120 (nds32_expand_epilogue): Likewise.
5121 (nds32_dynamic_chain_address): New function.
5122 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5124 (DYNAMIC_CHAIN_ADDRESS): Define.
5125 (EH_RETURN_DATA_REGNO): Define.
5126 (EH_RETURN_STACKADJ_RTX): Define.
5127 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5128 patterns for dwarf exception handling.
5130 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5132 * config/nds32/nds32.h: Clean up obsolete macros.
5134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5136 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5137 Add enum values for particular instructions.
5138 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5139 particular intrinsic functions.
5140 * config/nds32/nds32-intrinsic.md: Likewise.
5141 * config/nds32/nds32_intrinsic.h: Likewise.
5142 * config/nds32/nds32.h (nds32_builtins): Likewise.
5143 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5144 (btst, ave): New patterns for particular instructions.
5146 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5148 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5149 Add enum values for atomic load/store and memory sync.
5150 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5152 * config/nds32/nds32-intrinsic.md: Likewise.
5153 * config/nds32/nds32_intrinsic.h: Likewise.
5154 * config/nds32/nds32.h (nds32_builtins): Likewise.
5156 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5158 PR tree-optimization/85257
5159 * fold-const.c (native_encode_vector): If not all elts could fit
5160 and off is -1, return 0 rather than offset.
5161 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5162 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5163 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5164 adjust buffer in native_interpret_expr call.
5166 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5168 * config/nds32/constants.md (unspec_volatile_element): Add cache
5169 control enum values.
5170 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5171 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5172 * config/nds32/nds32.c (nds32_cctl_names): New.
5173 (nds32_print_operand): Handle cache control register names.
5174 * config/nds32/nds32.h (nds32_builtins): New enum values.
5175 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5177 * config/nds32/nds32.md (type): Add mmu.
5178 * config/nds32/pipelines.md (simple_insn): Add mmu.
5180 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5182 * config/nds32/nds32.md (type): Remove call.
5183 * config/nds32/pipelines.md (simple_insn): Likewise.
5185 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5187 * config/nds32/constants.md (unspec_volatile_element): Add
5188 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5189 UNSPEC_VOLATILE_FMFCFG.
5190 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5191 description for fmfcfg and fmfcsr.
5192 (bdesc_1arg): Add fmtcsr.
5193 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5194 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5195 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5196 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5197 unspec_fmfcfg): New patterns.
5198 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5199 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5200 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5201 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5202 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5203 __nds32__fmfcfg): Define.
5205 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5207 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5208 intrinsic register names.
5209 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5210 intrinsic register enum values and macros.
5212 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5214 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5215 for load/store addressing form.
5216 (nds32_print_operand_address): Likewise.
5218 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5221 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5222 based on LABEL_REF. Remove useless assertion.
5223 (pic_address_needs_scratch): Fix formatting.
5224 (sparc_legitimize_pic_address): Minor tweaks.
5225 (sparc_delegitimize_address): Adjust assertion accordingly.
5226 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5227 into symbolic_operand.
5228 (movsi_high_pic_label_ref): Likewise.
5229 (movsi_lo_sum_pic_label_ref): Likewise.
5230 (movdi_pic_label_ref): Likewise.
5231 (movdi_high_pic_label_ref): Likewise.
5232 (movdi_lo_sum_pic_label_ref): Likewise.
5234 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5236 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5237 custom LIB_SPEC setup.
5239 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5240 Kito Cheng <kito.cheng@gmail.com>
5242 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5243 * config/riscv/freebsd.h: New.
5245 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5247 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5248 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5251 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5252 Kito Cheng <kito.cheng@gmail.com>
5254 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5255 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5256 * config/nds32/nds32-protos.h (nds32_output_call,
5257 nds32_output_return): Declare.
5258 * config/nds32/nds32.md: Refine all the call and return patterns.
5260 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5263 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5264 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5266 PR rtl-optimization/84872
5267 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5268 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5271 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5273 * expr.c (copy_blkmode_to_reg): Revert 254862.
5274 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5276 2018-04-06 Richard Biener <rguenther@suse.de>
5279 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5280 after seeing a component reference with an adjacent field. Treat
5281 refs to arrays at struct end of external decls similar to
5282 refs to unconstrained commons.
5284 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5287 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5288 look through SAVE_EXPRs with non-side-effects argument. Adjust
5290 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5293 2018-04-06 Richard Biener <rguenther@suse.de>
5296 * alias.c (find_base_term): New wrapper around find_base_term
5297 unwinding CSELIB_VAL_PTR changes.
5298 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5301 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5303 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5305 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5306 constant definitions.
5307 ("nop"): lr 0,0 -> nopr r0
5308 ("nop_lr0", "nop_lr1"): New insn definitions.
5310 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5312 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5313 NDS32_V3PUSH_AVAILABLE_P macro.
5315 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5316 Chung-Ju Wu <jasonwucj@gmail.com>
5318 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5319 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5320 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5321 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5322 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5323 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5324 * config/nds32/constraints.md: New constraints and checking for hard
5325 float configuration.
5326 * config/nds32/iterators.md: New mode iterator and attribute for hard
5327 float configuration.
5328 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5330 * config/nds32/nds32-fpu.md: New file.
5331 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5332 deal with hard float code generation.
5333 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5335 (abi_type, float_reg_number): New enum type.
5336 * config/nds32/nds32-predicates.c: New predicates for hard float.
5337 * config/nds32/nds32-protos.h: Declare functions for hard float.
5338 * config/nds32/nds32.c: Implementation for hard float configuration.
5339 * config/nds32/nds32.h: Definitions for hard float configuration.
5340 * config/nds32/nds32.md: Include hard float machine description and
5341 modify patterns for hard float configuration.
5342 * config/nds32/nds32.opt: New options for hard float configuration.
5343 * config/nds32/predicates.md: New predicates for hard float
5346 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5348 * common/config/nds32/nds32-common.c
5349 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5351 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5354 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5355 CONSTRUCTOR_ELT (ctor, ...)->value.
5357 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5360 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5362 2018-04-05 Tom de Vries <tom@codesourcery.com>
5365 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5368 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5369 Kito Cheng <kito.cheng@gmail.com>
5371 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5372 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5373 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5374 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5376 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5377 Kito Cheng <kito.cheng@gmail.com>
5379 * config/nds32/constraints.md (Ufe): New memory constraint.
5380 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5381 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5382 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5384 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5385 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5387 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5389 * config/nds32/nds32.md: Use optimize_size in the condition for
5390 alu-shift instructions.
5392 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5394 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5396 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5398 * config/nds32/nds32.md (negsi2): Refine pattern.
5400 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5401 Chung-Ju Wu <jasonwucj@gmail.com>
5403 * config/nds32/iterators.md (shift_rotate): New code iterator.
5404 (shift): New code attribute.
5405 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5406 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5407 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5408 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5409 bit-wise operations.
5410 (andsi3, *andsi3): Ditto.
5411 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5412 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5413 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5414 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5415 nds32_ior_operand, nds32_xor_operand): New predicates.
5417 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5419 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5420 (addsi3, subsi3): ... this.
5422 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5424 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5426 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5428 * config/nds32/nds32.md: Adjust indention.
5430 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5432 * config/nds32/nds32.md (feature): New attribute.
5434 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config/nds32/nds32.md (subtype): New attribute.
5438 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5441 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5442 expansion to perform a bitwise AND of the argument followed by a
5443 boolean negation of the result.
5445 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5447 PR rtl-optimization/84878
5448 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5449 the basic block. Assert the use reference is not artificial and that
5450 it has an associated insn.
5452 2018-04-04 Michael Matz <matz@suse.de>
5454 * builtins.c (compute_objsize): Pass correct operand
5455 to array_at_struct_end_p.
5457 2018-04-04 Richard Biener <rguenther@suse.de>
5460 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5461 from contexts for DINFO_LEVEL_TERSE and below.
5463 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5465 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5466 resiter_operand condition.
5467 * config/nds32/nds32.md (*move<mode>): Ditto.
5469 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5470 Monk Chiang <sh.chiang04@gmail.com>
5472 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5474 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5476 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5478 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5479 Kito Cheng <kito.cheng@gmail.com>
5481 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5482 nds32_cond_code_str, output_cond_branch,
5483 output_cond_branch_compare_zero, nds32_expand_cbranch,
5484 nds32_expand_cstore, nds32_expand_movcc,
5485 nds32_output_cbranchsi4_equality_zero,
5486 nds32_output_cbranchsi4_equality_reg,
5487 nds32_output_cbranchsi4_equality_reg_or_const_int,
5488 nds32_output_cbranchsi4_greater_less_zero: New functions.
5489 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5490 nds32_expand_cstore, nds32_expand_movcc,
5491 nds32_output_cbranchsi4_equality_zero,
5492 nds32_output_cbranchsi4_equality_reg,
5493 nds32_output_cbranchsi4_equality_reg_or_const_int,
5494 nds32_output_cbranchsi4_greater_less_zero): Declare.
5495 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5496 nds32_rimm11s_operand): New predicates.
5497 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5498 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5501 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5503 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5504 * config/nds32/nds32.md: Ditto.
5505 * config/nds32/pipelines.md: Ditto.
5507 2018-04-04 Richard Biener <rguenther@suse.de>
5509 PR tree-optimization/85168
5510 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5511 propagating abnormals.
5513 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5515 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5517 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5518 Kito Cheng <kito.cheng@gmail.com>
5520 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5521 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5522 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5523 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5524 * config/nds32/nds32.md (sibcall_internal): New.
5525 (sibcall_register): Remove.
5526 (sibcall_immediate): Remove.
5527 (sibcall_value_internal): New.
5528 (sibcall_value_register): Remove.
5529 (sibcall_value_immediate): Remove.
5530 * config/nds32/predicates.md (nds32_general_register_operand): New.
5531 (nds32_call_address_operand): New.
5533 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5535 PR rtl-optimization/85167
5536 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5537 bb_defs if *split_p, instead preinitialize it to NULL.
5539 PR tree-optimization/85156
5540 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5541 evaluating the argument multiple times.
5543 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5545 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5547 (_mm_cvtpd_ps): Likewise.
5548 (_mm_cvttpd_epi32): Likewise.
5549 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5550 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5551 vector, pixel, and bool following altivec.h include.
5553 2018-04-03 Martin Sebor <msebor@redhat.com>
5555 * doc/extend.texi (Common Function Attributes): Clarify.
5556 (const attribute): Likewise.
5557 (pure attribute): Likewise.
5559 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5562 * config/i386/i386.c (ix86_expand_vector_set): Use
5563 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5565 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5567 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5568 instructions when changing rounding bits to preserve precision bits
5569 in the x87 control word.
5571 2018-04-03 Martin Liska <mliska@suse.cz>
5573 PR tree-optimization/82491
5574 * rtl.h (strip_offset_and_add): Replace += suboffset with
5575 poly_uint64 () + suboffset.
5577 2018-03-29 Martin Liska <mliska@suse.cz>
5578 Martin Jambor <mjambor@suse.cz>
5581 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5582 param_type is not an integral or pointer type.
5584 2018-04-03 Richard Biener <rguenther@suse.de>
5586 * sese.h (recompute_all_dominators): Remove.
5588 2018-04-02 Martin Sebor <msebor@redhat.com>
5590 * doc/invoke.texi (-Wrestrict): Fix typos.
5592 2018-04-02 Jim Wilson <jimw@sifive.com>
5594 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5595 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5596 (<optab>di3, <optab>si3_extend): Likewise.
5597 (<optab>si3_mask, <optab>si3_mask_1): New.
5598 (<optab>di3_mask, <optab>di3_mask_1): New.
5599 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5600 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5601 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5603 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5605 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5608 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5610 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5611 (nds32_canonicalize_comparison): New function.
5613 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5614 Kito Cheng <kito.cheng@gmail.com>
5615 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5617 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5618 * config/nds32/constants.md (unspec_volatile_element): Add
5619 UNSPEC_VOLATILE_RELAX_GROUP.
5620 * config/nds32/nds32-relax-opt.c: New file.
5621 * config/nds32/nds32-predicates.c
5622 (nds32_symbol_load_store_p): New function.
5623 * config/nds32/nds32-protos.h
5624 (nds32_symbol_load_store_p): Declare function.
5625 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5626 * config/nds32/nds32.c
5627 (nds32_register_pass): New function to register pass.
5628 (nds32_register_passes): New function to register passes.
5629 * config/nds32/nds32.md (relax_group): New pattern.
5630 * config/nds32/nds32.opt (mrelax-hint): New option.
5631 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5633 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5635 * config/nds32/t-nds32: Modify files dependency.
5637 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5639 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5640 (PROFILE_HOOK): Define its implementation.
5642 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5644 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5645 type and 32-bit size.
5647 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5650 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5651 (V_128_256): New mode iterator.
5652 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5653 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5654 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5656 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5657 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5659 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5662 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5663 NaN inputs correctly.
5665 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5668 * config/rs6000/vsx.md (??r): New mode attribute.
5669 (*vsx_mov<mode>_64bit): Use it.
5670 (*vsx_mov<mode>_32bit): Likewise.
5672 2018-03-30 Martin Sebor <msebor@redhat.com>
5674 PR tree-optimization/84818
5675 * builtins.c (check_access): Use warning_n.
5677 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5680 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5682 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5685 2018-03-30 Julia Koval <julia.koval@intel.com>
5688 * x86-tune.def (movx, partial_reg_dependency): Enable for
5691 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5694 * lra-constraints.c (process_alt_operands): Move setting
5695 this_alternative_matches below.
5697 2018-03-29 Martin Liska <mliska@suse.cz>
5700 * doc/invoke.texi: Document how LTO works with debug info.
5701 Describe auto-load support of binutils. Mention 'x86-64'
5702 as valid option value of -march option.
5704 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5706 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5709 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5710 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5711 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5714 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5717 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5718 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5719 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5720 for RS6000_BTM_POWERPC64.
5721 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5722 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5723 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5728 2018-03-28 Carl Love <cel@us.ibm.com>
5731 2017-09-27 Carl Love <cel@us.ibm.com>
5733 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5734 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5735 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5738 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5740 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5741 instead of __vector bool.
5742 (_mm_max_pu8): Likewise.
5743 (_mm_min_pi16): Likewise.
5745 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5748 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5749 (DIVWEUO): Likewise.
5751 (DIVDEUO): Likewise.
5752 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5753 DIVWEUO and DIVDEUO.
5754 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5755 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5756 (div_extend): Likewise.
5757 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5759 (__builtin_divweuo): Likewise.
5760 (__builtin_divdeo): Likewise.
5761 (__builtin_divdeuo): Likewise.
5763 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5766 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5767 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5769 PR tree-optimization/82004
5770 * gimple-match-head.c (optimize_pow_to_exp): New function.
5771 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5772 Don't fold to exp if optimize_pow_to_exp is false.
5774 2018-03-28 Martin Liska <mliska@suse.cz>
5777 * calls.c (initialize_argument_information): Fix trailing space.
5778 * common.opt: Fix typo and provide better explanation for
5779 -fsanitize-coverage option.
5780 * config/i386/i386.opt: Fix typo.
5782 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5783 Martin Liska <mliska@suse.cz>
5786 * gimplify.c (asan_poison_variable): Don't do the check for
5787 gimplify_omp_ctxp here.
5788 (gimplify_decl_expr): Do it here.
5789 (gimplify_target_expr): Likewise.
5791 2018-03-28 Martin Liska <mliska@suse.cz>
5794 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5795 chkp_type_bounds_count if MPX is not enabled.
5797 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5799 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5801 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5804 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5805 function to create the function decl for complex long double
5806 multiply and divide for -mabi=ieeelongdouble.
5807 (init_float128_ieee): Call it.
5809 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5812 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5813 -fcf-protection=branch -mibt.
5814 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5816 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5819 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5821 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5824 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5825 extern array declarations.
5827 2018-03-27 Richard Biener <rguenther@suse.de>
5830 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5831 explicit single_use checks.
5833 2018-03-27 Richard Biener <rguenther@suse.de>
5835 PR tree-optimization/85082
5836 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5839 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5841 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5842 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5843 Turn on fasynchronous-unwind-tables and funwind-tables.
5845 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5848 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5849 (*bmi_blsr_<mode>_ccz): Ditto.
5851 2018-03-26 Tom de Vries <tom@codesourcery.com>
5853 PR tree-optimization/85063
5854 * omp-general.c (offloading_function_p): New function. Factor out
5856 * omp-offload.c (pass_omp_target_link::gate): ... here.
5857 * omp-general.h (offloading_function_p): Declare.
5858 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5859 with attribute omp declare target for offloading functions.
5861 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5863 PR tree-optimization/84005
5864 * tree-data-ref.h (get_base_for_alignment): Declare.
5865 * tree-data-ref.c (get_base_for_alignment_1): New function.
5866 (get_base_for_alignment): Likewise.
5867 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5868 get_base_for_alignment to find a suitable base object, instead
5869 of always using drb->base_address.
5871 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5874 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5875 known size by default.
5877 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5880 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5881 and non BLKmode operands.
5883 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5886 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5887 Clean up attributes.
5889 2018-03-23 Richard Biener <rguenther@suse.de>
5892 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5893 we are going to emit early debug for LTO.
5895 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5898 * function.c (match_asm_constraints_1): Don't optimize if input
5899 doesn't satisfy general_operand predicate for output's mode.
5902 * alias.c (write_dependence_p): Don't require for x_canonicalized
5903 non-VOIDmode if x has VOIDmode.
5906 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5907 just don't try to optimize it rather than assert it never happens.
5909 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5911 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5912 macro expansions for definition of ST_INTERNAL_<mode> and
5913 LD_INTERNAL_<mode> builtins.
5914 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5916 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5918 (altivec_expand_st_builtin): Likewise.
5919 (altivec_expand_builtin): Remove calls to deleted functions.
5920 (rs6000_address_for_altivec): Delete this function.
5921 * config/rs6000/vector.md: Remove expands for
5922 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5924 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5927 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5928 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5929 re-computing once computed.
5930 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5931 (arm_init_machine_status): Initialize
5932 machine->static_chain_stack_bytes.
5934 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5937 * doc/extend.texi: Add four new prototypes for vec_ld.
5938 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5939 definitions for more logical presentation.
5940 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5941 entries for V1TI variants of __builtin_altivec_ld builtin.
5942 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5943 handling of V1TI variant of LVX icode pattern.
5944 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5945 (rs6000_gimple_fold_builtin): Likewise.
5946 (altivec_init_builtins): Add code to define
5947 __builtin_altivec_lvx_v1ti function.
5949 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5952 * function.c (match_asm_constraints_1): Don't do the optimization
5953 if input isn't a REG, SUBREG, MEM or constant.
5955 2018-03-22 Tom de Vries <tom@codesourcery.com>
5957 PR tree-optimization/84956
5958 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5959 bb_has_abnormal_pred.
5961 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5964 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5965 DECL_INITIAL (decl) to decl at the end.
5966 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5969 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5971 * doc/extend.texi (__builtin_tgmath): Document when complex
5972 integer types are treated as _Complex _Float64.
5974 2018-03-21 Tom de Vries <tom@codesourcery.com>
5976 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5978 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5980 PR tree-optimization/84960
5981 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5982 if it is ENTRY block, move them into single succ of ENTRY in that case.
5984 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5986 PR tree-optimization/84811
5987 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5988 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5989 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5990 (known_subrange_p): Update accordingly. Cast each value involved
5991 in the size comparison, rather than casting the result of the
5994 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5996 PR tree-optimization/84982
5997 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5998 by flipping the least significant bit rather than all bits from
5999 bitpos to bitpos + bitsize - 1.
6001 2018-03-21 Nathan Sidwell <nathan@acm.org>
6003 * doc/extend.texi (Deprecated Features): Remove mention of
6004 long-deleted deprecations.
6006 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6009 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6010 * configure: Regenerate.
6012 2018-03-21 Tom de Vries <tom@codesourcery.com>
6014 PR tree-optimization/83126
6015 * tree-parloops.c (num_phis): New function.
6016 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6018 2018-03-21 Nathan Sidwell <nathan@acm.org>
6020 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6021 mention anon-struct/union members and trailing attributes.
6023 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6025 PR tree-optimization/84969
6026 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6027 builtin memset partitions if they set different rhs values.
6029 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6031 PR rtl-optimization/84989
6032 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6033 VEC_DUPLICATE with scalar result mode.
6035 2018-03-21 Martin Liska <mliska@suse.cz>
6038 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6039 not intended return statement.
6041 2018-03-21 Martin Liska <mliska@suse.cz>
6044 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6045 (chkp_find_bound_slots_1): Limit number of iterations.
6047 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6050 * Minor grammar fixes for x86 options.
6052 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6055 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6056 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6058 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6061 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6062 (altivec_lvx_<mode>_1op): Likewise.
6063 (altivec_stvx_<mode>_2op): Likewise.
6064 (altivec_stvx_<mode>_1op): Likewise.
6065 (altivec_lvx_<VM2:mode>): New define_expand.
6066 (altivec_stvx_<VM2:mode>): Likewise.
6067 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6068 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6069 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6070 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6071 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6072 (rs6000_gen_lvx): Likewise.
6073 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6074 (altivec_expand_stv_builtin): Likewise.
6075 (altivec_expand_builtin): Likewise.
6076 * config/rs6000/vector.md: Likewise.
6078 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6081 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6084 2018-03-20 Richard Biener <rguenther@suse.de>
6087 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6088 sign-conversions as zero, fall back to standard scalar_stmt
6091 2018-03-20 Martin Liska <mliska@suse.cz>
6094 * predict.c (rebuild_frequencies): Handle case when we have
6095 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6097 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6100 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6101 flag_section_anchors.
6102 * varasm.c (use_blocks_for_decl_p): Remove hack for
6103 dw2_force_const_mem.
6106 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6108 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6109 be created, use lowpart_subreg of operands[0] rather than operands[0]
6111 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6112 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6113 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6114 and n constraint instead of aarch64_shift_imm_di and Usd.
6115 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6116 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6118 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6121 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6122 to favor GPR over NEON registers.
6123 (<shift>di3_neon): Likewise.
6125 2018-03-20 Tom de Vries <tom@codesourcery.com>
6128 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6129 (nvptx_process_pars): Emit bar.sync asap and alap.
6131 2018-03-20 Tom de Vries <tom@codesourcery.com>
6134 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6135 seen_label if seen_label is already set.
6137 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6140 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6141 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6142 Use 1U instead of 1. Formatting fixes.
6145 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6146 instead of TREE_TYPE (s1) for the return value.
6148 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6150 PR tree-optimization/84946
6151 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6152 bitsize + bitsize in poly_uint64 rather than poly_int64.
6155 * dwarf2asm.c: Include fold-const.c.
6156 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6157 of decl rather than decl itself.
6159 PR rtl-optimization/84643
6160 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6162 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6165 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6166 calling assemble_variable.
6168 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6171 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6172 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6174 2018-03-19 Jim Wilson <jimw@sifive.com>
6177 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6178 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6179 (riscv_first_stack_step): Likewise.
6180 (riscv_option_override): Use STACK_BOUNDARY instead of
6182 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6184 (BIGGEST_ALIGNMENT): Set to 128.
6185 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6186 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6189 2018-03-19 Richard Biener <rguenther@suse.de>
6191 PR tree-optimization/84933
6192 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6193 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6195 2018-03-19 Richard Biener <rguenther@suse.de>
6197 PR tree-optimization/84859
6198 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6199 (cond_if_else_store_replacement): Perform sinking operation on
6200 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6201 Generalize what a BB with a single eligible store is.
6203 2018-03-19 Richard Biener <rguenther@suse.de>
6205 PR tree-optimization/84929
6206 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6207 chrec_is_positive against non-chrec arg.
6209 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6212 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6214 2018-03-18 Martin Liska <mliska@suse.cz>
6216 PR rtl-optimization/84635
6217 * regrename.c (build_def_use): Use matches_mode only when
6220 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6222 PR tree-optimization/84913
6223 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6224 vectorize chains of COND_EXPRs.
6226 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6228 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6230 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6232 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6234 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6236 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6238 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6239 Kito Cheng <kito.cheng@gmail.com>
6241 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6242 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6243 (nds32_adjust_reg_alloc_order): New function.
6244 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6246 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6248 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6249 nds32_print_operand, nds32_print_operand_address): Use
6250 HOST_WIDE_INT_PRINT_DEC instead.
6252 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6254 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6256 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6259 * config/i386/i386.c (initial_ix86_tune_features,
6260 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6262 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6263 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6264 rather than 1u << ix86_tune. Formatting fix.
6265 (ix86_option_override_internal): Change ix86_arch_mask from
6266 unsigned int to unsigned HOST_WIDE_INT, initialize to
6267 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6268 (ix86_function_specific_restore): Likewise.
6270 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6273 * postreload.c (reload_combine_recognize_pattern): Perform
6274 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6275 truncate_int_for_mode the result for the destination's mode.
6278 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6279 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6281 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6284 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6287 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6288 regno_allocno_class_array and sorted_pseudos.
6289 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6290 insns where regno is used.
6292 2018-03-16 Martin Liska <mliska@suse.cz>
6295 * multiple_target.c (create_dispatcher_calls): Redirect
6296 reference in the symbol table.
6298 2018-03-16 Martin Liska <mliska@suse.cz>
6301 * multiple_target.c (create_dispatcher_calls): Redirect also
6304 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6308 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6309 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6310 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6312 2018-03-16 Julia Koval <julia.koval@intel.com>
6314 * doc/invoke.texi (Skylake Server): Add CLWB.
6315 Cannonlake): Remove CLWB.
6317 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6319 PR tree-optimization/84841
6320 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6322 (FLOAT_ONE_CONST_TYPE): Define.
6323 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6324 (sort_by_operand_rank): Put entries with higher constant_type last
6325 rather than first to match comments.
6327 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6329 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6332 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6335 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6336 check and use address of target always.
6338 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6341 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6342 (indirect_thunk_bnd_needed): Likewise.
6343 (indirect_thunks_used): Likewise.
6344 (indirect_thunks_bnd_used): Likewise.
6345 (indirect_return_needed): New.
6346 (indirect_return_bnd_needed): Likewise.
6347 (output_indirect_thunk_function): Add a bool argument for
6349 (output_indirect_thunk_function): Don't generate alias for
6350 function return thunk.
6351 (ix86_code_end): Call output_indirect_thunk_function to generate
6352 function return thunks.
6353 (ix86_output_function_return): Set indirect_return_bnd_needed
6354 and indirect_return_needed instead of indirect_thunk_bnd_needed
6355 and indirect_thunk_needed.
6357 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6359 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6360 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6361 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6363 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6364 Paul Hua <paul.hua.gm@gmail.com>
6367 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6369 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6371 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6372 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6375 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6378 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6379 instead of GET_MODE_SIZE when comparing Units.
6381 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6384 * varasm.c (hash_section): Return an unchangeble hash value
6385 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6386 Return !aarch64_can_use_per_function_literal_pools_p ().
6388 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6391 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6392 rather than address of cmode as last argument to prepare_cmp_insn.
6394 2018-03-15 Julia Koval <julia.koval@intel.com>
6396 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6397 F_AVX512VNNI, F_AVX512BITALG): New.
6399 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6402 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6403 insn for floating-point loads and stores.
6405 2018-03-14 Carl Love <cel@us.ibm.com>
6407 * config/rs6000/rs6000-c.c: Add macro definitions for
6408 ALTIVEC_BUILTIN_VEC_PERMXOR.
6409 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6410 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6411 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6412 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6414 * config/doc/extend.texi: Add prototypes for vec_permxor.
6416 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6419 * diagnostic-show-locus.c (class layout_point): Convert m_line
6420 from int to linenum_type.
6421 (line_span::comparator): Use linenum "compare" function when
6422 comparing line numbers.
6423 (test_line_span): New function.
6424 (layout_range::contains_point): Convert param "row" from int to
6426 (layout_range::intersects_line_p): Likewise.
6427 (layout::will_show_line_p): Likewise.
6428 (layout::print_source_line): Likewise.
6429 (layout::should_print_annotation_line_p): Likewise.
6430 (layout::print_annotation_line): Likewise.
6431 (layout::print_leading_fixits): Likewise.
6432 (layout::annotation_line_showed_range_p): Likewise.
6433 (struct line_corrections): Likewise for field m_row.
6434 (line_corrections::line_corrections): Likewise for param "row".
6435 (layout::print_trailing_fixits): Likewise.
6436 (layout::get_state_at_point): Likewise.
6437 (layout::get_x_bound_for_row): Likewise.
6438 (layout::print_line): Likewise.
6439 (diagnostic_show_locus): Likewise for locals "last_line" and
6441 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6442 * input.c (selftest::test_linenum_comparisons): New function.
6443 (selftest::input_c_tests): Call it.
6444 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6445 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6446 * selftest.h (ASSERT_GT): New macro.
6447 (ASSERT_GT_AT): New macro.
6448 (ASSERT_LT): New macro.
6449 (ASSERT_LT_AT): New macro.
6451 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6453 PR rtl-optimization/84780
6454 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6456 2018-03-14 Martin Liska <mliska@suse.cz>
6458 * tree.c (record_node_allocation_statistics): Use
6459 get_stats_node_kind.
6460 (get_stats_node_kind): New function extracted from
6461 record_node_allocation_statistics.
6462 (free_node): Use get_stats_node_kind.
6464 2018-03-14 Richard Biener <rguenther@suse.de>
6466 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6467 that the value-set of ANTIC_IN doesn't grow.
6470 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6472 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6473 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6475 2018-03-14 Julia Koval <julia.koval@intel.com>
6477 * config.gcc (icelake-client, icelake-server): New.
6479 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6480 (initial_ix86_arch_features): Ditto.
6481 (PTA_SKYLAKE): Add SGX.
6482 (PTA_ICELAKE): Remove.
6483 (PTA_ICELAKE_CLIENT): New.
6484 (PTA_ICELAKE_SERVER): New.
6485 (ix86_option_override_internal): Split up icelake on icelake client and
6487 (get_builtin_code_for_version): Ditto.
6488 (fold_builtin_cpu): Ditto.
6489 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6490 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6491 * config/i386/i386.h (processor_type): Ditto.
6492 * doc/invoke.texi: Ditto.
6494 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6497 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6498 INTEGER_CST offset, add it together with bitpos / 8 and
6499 sign extend based on POINTER_SIZE.
6503 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6506 * config/i386/constraints.md (Yc): New register constraint.
6507 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6508 Use Yc constraint for alternative 2 of operand 0. Remove
6509 preferred_for_speed attribute.
6511 2018-03-14 Richard Biener <rguenther@suse.de>
6513 PR tree-optimization/84830
6514 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6515 with the old one to avoid oscillations.
6517 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6520 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6522 (assign_by_spills): Return a flag of reload assignment failure.
6523 Do not process the reload assignment failures. Do not spill other
6524 reload pseudos if they has the same reg class. Update n if
6526 (lra_assign): Add a return arg. Set up from the result of
6527 assign_by_spills call.
6528 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6529 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6530 usage_insns if it is not NULL.
6531 (spill_hard_reg_in_range): New function.
6532 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6533 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6534 function prototypes.
6535 (lra_assign): Change prototype.
6536 * lra.c (lra): Add code to deal with fails by splitting hard reg
6539 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6541 * config/riscv/riscv.opt (mrelax): New option.
6542 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6543 "norelax" when riscv_mrelax is disabled.
6544 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6546 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6549 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6550 reassociation for int modes.
6552 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6554 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6555 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6557 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6558 * config/aarch64/aarch64-sve.md
6559 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6560 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6561 (*extend<mode><Vwide>2): Rename to...
6562 (aarch64_sve_extend<mode><Vwide>2): ...this.
6563 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6564 renaming the old pattern to...
6565 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6567 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6568 define_expand, renaming the old pattern to...
6569 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6570 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6571 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6572 account when deciding which SVE instruction the optab should use.
6573 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6575 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6577 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6578 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6579 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6580 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6581 (tlsdesc_small_<mode>): Turn a define_expand and use
6582 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6583 (tlsdesc_small_advsimd_<mode>): ...this.
6584 (tlsdesc_small_sve_<mode>): New pattern.
6586 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6588 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6589 (UNSPEC_UMUL_HIGHPART): New constants.
6590 (MUL_HIGHPART): New int iteraor.
6591 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6592 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6594 (*<su>mul<mode>3_highpart): New define_insn.
6596 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6599 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6602 2018-03-13 Martin Liska <mliska@suse.cz>
6605 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6607 (sem_item_optimizer::~sem_item_optimizer): Release it.
6608 (sem_item_optimizer::merge_classes): Register variable aliases.
6609 (sem_item_optimizer::fixup_pt_set): New function.
6610 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6611 * ipa-icf.h: Declare new variables and functions.
6613 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6616 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6617 integer_pow2p@2 and test integer_pow2p in condition.
6618 (A < 0 ? C : 0): Similarly for @1.
6621 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6622 characters starting at p contain '\0' character, don't look beyond
6626 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6627 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6630 * reg-stack.c (change_stack): Change update_end var from int to
6631 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6632 also call set_block_for_insn on the newly added insns and rescan.
6635 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6636 on the last operand.
6639 * tree.c (stabilize_reference_1): Return save_expr (e) for
6640 STATEMENT_LIST even if it doesn't have side-effects.
6642 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6644 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6646 2018-03-12 Renlin Li <renlin.li@arm.com>
6648 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6649 aarch64_output_scalar_simd_mov_immediate.
6651 2018-03-12 Martin Sebor <msebor@redhat.com>
6653 PR tree-optimization/83456
6654 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6655 for perfectly overlapping calls to memcpy.
6656 (gimple_fold_builtin_memory_chk): Same.
6657 (gimple_fold_builtin_strcpy): Handle no-warning.
6658 (gimple_fold_builtin_stxcpy_chk): Same.
6659 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6661 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6663 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6664 parameter. Use it for SFmode.
6665 (rs6000_function_arg_advance_1): Adjust.
6666 (rs6000_function_arg): Adjust.
6667 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6669 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6671 PR rtl-optimization/84169
6672 PR rtl-optimization/84780
6673 * combine.c (can_combine_p): Check for a 2-insn combination whether
6674 the destination register is used between the two insns, too.
6676 2018-03-12 Richard Biener <rguenther@suse.de>
6678 PR tree-optimization/84803
6679 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6680 for refs DR analysis didn't process.
6682 2018-03-12 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/84777
6685 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6686 force-vectorize loops ignore whether we are optimizing for size.
6688 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6690 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6691 (TARGET_MD_ASM_ADJUST): Define.
6693 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6694 Kito Cheng <kito.cheng@gmail.com>
6695 Chung-Ju Wu <jasonwucj@gmail.com>
6697 * config/nds32/nds32.c (nds32_compute_stack_frame,
6698 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6699 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6700 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6701 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6702 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6703 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6704 * config/nds32/nds32.md (prologue, epilogue): Use macro
6705 NDS32_V3PUSH_AVAILABLE_P to do checking.
6707 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6710 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6711 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6712 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6713 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6714 addition of most attributes on !orig_type_die or the attribute not
6715 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6717 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6718 Chung-Ju Wu <jasonwucj@gmail.com>
6720 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6722 * config/nds32/nds32.opt (mvh): New option.
6724 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6725 Chung-Ju Wu <jasonwucj@gmail.com>
6727 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6729 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6730 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6733 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6734 Chung-Ju Wu <jasonwucj@gmail.com>
6736 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6738 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6739 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6741 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6742 Kito Cheng <kito.cheng@gmail.com>
6743 Chung-Ju Wu <jasonwucj@gmail.com>
6745 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6746 UNSPEC_FFMISM and UNSPEC_FLMISM.
6747 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6748 for ffb, ffmism and flmism.
6749 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6750 (unspec_ffmism): Ditto.
6751 (unspec_flmism): Ditto.
6752 (nds32_expand_builtin_impl): Check if string extension is available.
6753 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6754 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6756 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6759 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6762 * lra-assigns.c (assign_by_spills): Return a flag of reload
6763 assignment failure. Do not process the reload assignment
6764 failures. Do not spill other reload pseudos if they has the same
6766 (lra_assign): Add a return arg. Set up from the result of
6767 assign_by_spills call.
6768 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6769 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6770 usage_insns if it is not NULL.
6771 (spill_hard_reg_in_range): New function.
6772 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6773 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6774 function prototypes.
6775 (lra_assign): Change prototype.
6776 * lra.c (lra): Add code to deal with fails by splitting hard reg
6779 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6782 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6784 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6787 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6788 (dw_val_node): Add val_symbolic_view.
6789 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6790 (symview_upper_bound): New.
6791 (new_line_info_table): Initialize symviews_since_reset.
6792 (dwarf2out_source_line): Count symviews_since_reset and set
6793 symview_upper_bound.
6794 (dw_val_equal_p): Handle symview.
6795 (add_AT_symview): New.
6796 (print_dw_val): Handle symview.
6797 (attr_checksum, attr_checksum_ordered): Likewise.
6798 (same_dw_val_p, size_of_die): Likewise.
6799 (value_format, output_die): Likewise.
6800 (add_high_low_attributes): Use add_AT_symview for entry_view.
6801 (dwarf2out_finish): Reset symview_upper_bound, clear
6804 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6807 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6808 Add strict argument and use it.
6809 (rs6000_split_multireg_move): Update for new strict argument.
6810 (mem_operand_gpr): Disallow all non-offsettable addresses.
6811 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6813 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6816 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6817 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6818 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6821 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6822 decl, use remap_type if we want to use the type.
6824 2018-03-09 Martin Sebor <msebor@redhat.com>
6826 PR tree-optimization/84526
6827 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6829 (builtin_access::generic_overlap): Be prepared to handle non-array
6832 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6834 PR rtl-optimization/84682
6835 * lra-constraints.c (process_address_1): Check is_address flag
6836 for address constraints.
6837 (process_alt_operands): Likewise.
6838 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6839 preprocess_constraints.
6840 * recog.h (preprocess_constraints): Add oploc parameter.
6842 * recog.c (preprocess_constraints): Test address_operand for
6843 CT_ADDRESS constraints.
6845 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6848 * lra-assigns.c (assign_by_spills): Return a flag of reload
6849 assignment failure. Do not process the reload assignment
6850 failures. Do not spill other reload pseudos if they has the same
6852 (lra_assign): Add a return arg. Set up from the result of
6853 assign_by_spills call.
6854 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6855 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6856 usage_insns if it is not NULL.
6857 (spill_hard_reg_in_range): New function.
6858 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6859 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6860 function prototypes.
6861 (lra_assign): Change prototype.
6862 * lra.c (lra): Add code to deal with fails by splitting hard reg
6865 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6868 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6869 Accept complain bool parameter. Only emit errors if it is true.
6870 (arm_parse_cpu_option_name): Likewise.
6871 (arm_target_thumb_only): Adjust callers of the above.
6872 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6873 prototype to take a default true bool parameter.
6874 (arm_parse_arch_option_name): Likewise.
6876 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6877 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6881 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6882 * configure: Regenerate.
6883 * configure.ac ("linker --version-script option"): New.
6884 ("linker soname option"): New.
6886 2018-03-09 Richard Biener <rguenther@suse.de>
6888 PR tree-optimization/84775
6889 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6890 immediate uses of predicate stmts and mark them modified.
6893 PR tree-optimization/84178
6894 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6896 (version_loop_for_if_conversion): Delay update_ssa call.
6897 (tree_if_conversion): Delay update_ssa until after predicate
6900 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6903 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6904 when the function accesses prior frames.
6906 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6909 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6910 gen_llsym, otherwise call maybe_gen_llsym.
6913 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6914 has ',' character inside of it.
6916 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6919 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6920 as clobbering CC_REGNUM.
6922 2018-03-08 Richard Biener <rguenther@suse.de>
6925 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6926 (follow_copies_to_constant): Do not follow SSA names registered
6929 2018-03-08 Richard Biener <rguenther@suse.de>
6931 PR tree-optimization/84178
6932 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6934 (version_loop_for_if_conversion): Delay update_ssa call.
6935 (tree_if_conversion): Delay update_ssa until after predicate
6938 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6940 PR tree-optimization/84178
6941 * tree-if-conv.c (release_bb_predicate): Remove the
6942 the assertion that the stmts have NULL use_ops.
6943 Discard the statements, asserting that they haven't
6944 yet been added to a BB.
6946 2018-03-08 Richard Biener <rguenther@suse.de>
6948 PR tree-optimization/84746
6949 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6950 (phi_translate): Pass in destination ANTIC_OUT set.
6951 (phi_translate_1): Likewise. For a simplified result lookup
6952 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6953 (phi_translate_set): Adjust.
6954 (do_pre_regular_insertion): Likewise.
6955 (do_pre_partial_partial_insertion): Likewise.
6957 2018-03-08 Martin Liska <mliska@suse.cz>
6959 PR gcov-profile/84735
6960 * doc/gcov.texi: Document usage of profile files.
6961 * gcov-io.h: Document changes in the format.
6963 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6967 * dwarf2out.c (struct dw_line_info_table): Update comments for
6969 (FORCE_RESET_NEXT_VIEW): New.
6970 (FORCE_RESETTING_VIEW_P): New.
6971 (RESETTING_VIEW_P): Check for -1 too.
6972 (ZERO_VIEW_P): Likewise.
6973 (new_line_info_table): Force-reset next view.
6974 (dwarf2out_begin_function): Likewise.
6975 (dwarf2out_source_line): Simplify zero_view_p initialization.
6976 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6977 view directly. Omit view when omitting .loc at line 0.
6979 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6981 PR tree-optimization/84740
6982 * tree-switch-conversion.c (process_switch): Call build_constructors
6983 only if info.phi_count is non-zero.
6985 PR tree-optimization/84739
6986 * tree-tailcall.c (find_tail_calls): Check call arguments against
6987 DECL_ARGUMENTS (current_function_decl) rather than
6988 DECL_ARGUMENTS (func) when checking for tail recursion.
6990 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6992 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6993 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6994 Volker Reichelt's entry and add entries for people that perform
6995 GCC fuzzy testing and report numerous bugs.
6997 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7000 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7001 readonly data in sdata, if that is disabled.
7002 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7003 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7004 -mreadonly-in-sdata option.
7006 2018-03-07 Martin Sebor <msebor@redhat.com>
7008 PR tree-optimization/84468
7009 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7010 basic block when looking for nul assignment.
7012 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7015 * except.h (output_function_exception_table): Adjust prototype.
7016 * except.c (output_function_exception_table): Remove FNNAME parameter
7017 and add SECTION parameter. Ouput one part of the table at a time.
7018 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7019 the first part of the exception table and emit unwind directives.
7020 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7021 (i386_pe_seh_cold_init): Likewise.
7022 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7023 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7024 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7025 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7026 * config/i386/winnt.c: Include except.h.
7027 (struct seh_frame_state): Add reg_offset, after_prologue and
7028 in_cold_section fields.
7029 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7030 (i386_pe_seh_cold_init): New function.
7031 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7032 to seh->in_cold_section.
7033 (seh_emit_push): Record the offset of the push.
7034 (seh_emit_save): Record the offet of the save.
7035 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7036 Test seh->after_prologue to disregard the epilogue.
7037 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7038 (i386_pe_end_cold_function): New function.
7040 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7043 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7044 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7047 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7049 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7050 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7053 * multiple_target.c: Include tree-inline.h and intl.h.
7054 (expand_target_clones): Diagnose and fail if node->definition and
7055 !tree_versionable_function_p (node->decl).
7057 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7059 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7061 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7062 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7063 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7065 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7068 * combine.c (try_combine): Use reg_or_subregno instead of handling
7069 just paradoxical SUBREGs and REGs.
7071 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7073 * config/arc/arc.c (arc_finalize_pic): Remove function.
7074 (arc_must_save_register): We use single base PIC register, remove
7075 checks to save/restore the PIC register.
7076 (arc_expand_prologue): Likewise.
7077 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7079 (arc_verify_short): Likewise.
7080 (arc_attr_type): Likewise.
7081 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7082 (walk_stores): Likewise.
7083 (arc_address_cost): Make it static.
7084 (arc_verify_short): Likewise.
7085 (branch_dest): Likewise.
7086 (arc_attr_type): Likewise.
7087 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7088 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7089 (arc_final_prescan_insn): Remove inserting the nops due to
7090 hardware hazards. It is done in reorg step.
7091 (insn_length_variant_t): Remove.
7092 (insn_length_parameters_t): Likewise.
7093 (arc_insn_length_parameters): Likewise.
7094 (arc_get_insn_variants): Likewise.
7095 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7097 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7100 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7103 PR tree-optimization/84687
7104 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7106 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7108 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7110 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7111 Rename to ppc_speculation_barrier.
7112 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7113 __builtin_ppc_speculation_barrier.
7115 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7118 * combine.c (combine_simplify_rtx): Don't try to simplify if
7119 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7122 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7124 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7125 to 32 bytes when compiling for POWER9.
7127 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7130 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7131 regparm >= 3 with no arg reg available also for calls with
7132 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7135 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7137 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7139 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7142 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7144 2018-03-05 Richard Biener <rguenther@suse.de>
7146 PR tree-optimization/84486
7147 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7148 When inserting a __builtin_assume_aligned call set the LHS
7149 SSA name alignment info accordingly.
7151 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7153 PR tree-optimization/84114
7154 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7155 Avoid reassociation of FLOAT_MODE addition.
7157 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7159 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7160 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7161 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7162 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7163 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7164 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7165 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7167 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7168 __builtin_ia32_wbinvd): New builtins.
7169 (SPECIAL_ARGS2): New.
7170 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7171 (SPECIAL_ARGS2): New.
7172 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7173 (ix86_valid_target_attribute_inner_p): Ditto.
7174 (ix86_init_mmx_sse_builtins): Add special_args2.
7175 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7176 TARGET_WBNOINVD_P): New.
7177 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7178 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7179 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7180 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7181 * config/i386/pconfigintrin.h: New file.
7182 * config/i386/wbnoinvdintrin.h: Ditto.
7183 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7184 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7186 2018-03-05 Richard Biener <rguenther@suse.de>
7188 PR tree-optimization/84670
7189 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7191 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7192 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7193 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7194 doesn't grow if all successors (recursively) were visited at least
7197 2018-03-05 Richard Biener <rguenther@suse.de>
7199 PR tree-optimization/84650
7200 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7201 if executed in the loop pipeline.
7203 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7205 * doc/configfiles.texi (Configuration Files): Move info about
7206 conditionalizing $target-protos.h to...
7207 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7208 differs from $target-protos.h.
7210 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7211 Chung-Ju Wu <jasonwucj@gmail.com>
7213 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7214 * config/nds32/nds32-multiple.md (setmemsi): Define.
7215 * config/nds32/nds32-memory-manipulation.c
7216 (nds32_gen_dup_4_byte_to_word_value): New.
7217 (emit_setmem_word_loop): New.
7218 (emit_setmem_byte_loop): New.
7219 (nds32_expand_setmem_loop): New.
7220 (nds32_expand_setmem_loop_v3m): New.
7221 (nds32_expand_setmem_unroll): New.
7222 (nds32_expand_setmem): New.
7224 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7225 Chung-Ju Wu <jasonwucj@gmail.com>
7227 * config/nds32/nds32-memory-manipulation.c
7228 (nds32_emit_load_store): New.
7229 (nds32_emit_post_inc_load_store): New.
7230 (nds32_emit_mem_move): New.
7231 (nds32_emit_mem_move_block): New.
7232 (nds32_expand_movmemsi_loop_unknown_size): New.
7233 (nds32_expand_movmemsi_loop_known_size): New.
7234 (nds32_expand_movmemsi_loop): New.
7235 (nds32_expand_movmemsi_unroll): New.
7236 (nds32_expand_movmemqi): Rename ...
7237 (nds32_expand_movmemsi): ... to this.
7238 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7239 (movmemsi): ... to this.
7240 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7241 (nds32_expand_movmemsi): ... to this.
7243 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7244 Monk Chiang <sh.chiang04@gmail.com>
7245 Chung-Ju Wu <jasonwucj@gmail.com>
7247 * config/nds32/nds32-protos.h
7248 (nds32_expand_load_multiple): New arguments.
7249 (nds32_expand_store_multiple): Ditto.
7250 (nds32_valid_multiple_load_store): Rename ...
7251 (nds32_valid_multiple_load_store_p): ... to this.
7252 * config/nds32/nds32-memory-manipulation.c
7253 (nds32_expand_load_multiple): Refine implementation.
7254 (nds32_expand_store_multiple): Ditto.
7255 * config/nds32/nds32-multiple.md
7256 (load_multiple): Update nds32_expand_load_multiple interface.
7257 (store_multiple): Update nds32_expand_store_multiple interface.
7258 * config/nds32/nds32-predicates.c
7259 (nds32_valid_multiple_load_store): Rename ...
7260 (nds32_valid_multiple_load_store_p): ... to this and refine
7262 * config/nds32/predicates.md
7263 (nds32_load_multiple_and_update_address_operation): New predicate.
7264 (nds32_store_multiple_and_update_address_operation): New predicate.
7266 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7267 Chung-Ju Wu <jasonwucj@gmail.com>
7269 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7270 (combo): New attribute.
7271 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7273 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7275 * config/nds32/nds32.opt: Change -mcmodel= default value.
7277 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7278 Monk Chiang <sh.chiang04@gmail.com>
7279 Chung-Ju Wu <jasonwucj@gmail.com>
7281 * config/nds32/constants.md (unspec_element): New enum.
7282 * config/nds32/constraints.md (Umw): New constraint.
7283 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7284 * config/nds32/nds32-intrinsic.md: Likewise.
7285 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7286 (nds32_valid_smw_lwm_base_p): New.
7287 (nds32_output_smw_single_word): New.
7288 (nds32_output_lmw_single_word): New.
7289 (nds32_expand_unaligned_load): New.
7290 (nds32_expand_unaligned_store): New.
7291 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7292 (nds32_output_smw_single_word): Declare.
7293 (nds32_output_lmw_single_word): Declare.
7294 (nds32_expand_unaligned_load): Declare.
7295 (nds32_expand_unaligned_store): Declare.
7296 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7297 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7298 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7299 NDS32_BUILTIN_UASTORE_DW.
7300 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7303 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7304 Kito Cheng <kito.cheng@gmail.com>
7305 Chung-Ju Wu <jasonwucj@gmail.com>
7307 * config/nds32/nds32-intrinsic.c
7308 (nds32_expand_builtin_null_ftype_reg): Delete.
7309 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7310 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7311 (nds32_read_argument): New.
7312 (nds32_legitimize_target): Ditto.
7313 (nds32_legitimize_argument): Ditto.
7314 (nds32_check_constant_argument): Ditto.
7315 (nds32_expand_unop_builtin): Ditto.
7316 (nds32_expand_unopimm_builtin): Ditto.
7317 (nds32_expand_binop_builtin): Ditto.
7318 (nds32_builtin_decl_impl): Ditto.
7319 (builtin_description): Ditto.
7320 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7321 (nds32_init_builtins_impl): Ditto.
7322 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7323 (nds32_builtin_decl): New.
7324 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7325 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7327 2018-03-02 Jeff Law <law@redhat.com>
7329 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7330 (redundant_insn, fill_simple_delay_slots): Likewise.
7331 (fill_slots_from_thread): Likewise.
7332 * resource.c (mark_referenced_resources): Likewise.
7333 (mark_set_resources, find_dead_or_set_registers): Likewise.
7335 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7337 * substring-locations.h (format_warning_va): Formatting fix for
7339 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7341 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7342 * substring-locations.c: Include intl.h.
7343 (format_warning_va): Turned into small wrapper around
7344 format_warning_n_va, renamed to ...
7345 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7346 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7348 (format_warning_at_substring_n): New function.
7349 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7350 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7351 format_warning_at_substring with just a shorter name instead of
7352 const function pointer.
7353 (fmtwarn_n): New function.
7354 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7355 appropriate, get rid of all the fmtstr temporaries, move conditionals
7356 with G_() wrapped string literals directly into fmtwarn arguments,
7357 cast dir.len to (int), formatting fixes.
7359 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7361 * doc/invoke.texi: Remove "Cilk Plus" references.
7363 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7364 Richard Biener <rguenther@suse.de>
7367 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7368 for error or warning attributes if CALL_FROM_THUNK_P is set.
7371 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7374 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7375 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7378 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7379 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7381 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7382 instead of -1U in last predictors element's probability member.
7384 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7387 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7388 arguments if they are comparable.
7390 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7392 PR tree-optimization/84634
7393 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7394 masks and masked_loop_p with a single loop_masks, making sure it's
7395 null for bb vectorization.
7397 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7399 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7400 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7401 loop->force_vectorize to check whether there is no alias.
7403 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7406 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7408 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7410 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7411 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7412 instead of a loop around prev_real_insn.
7413 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7417 * config/i386/i386.c (ix86_print_operand): Use conditional
7418 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7421 2018-03-02 Richard Biener <rguenther@suse.de>
7423 PR tree-optimization/84427
7424 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7425 (bitmap_set_subtract_values): Rewrite to handle multiple
7428 (prune_clobbered_mems): Likewise.
7429 (phi_translate): Take edge instead of pred/phiblock.
7430 (phi_translate_1): Likewise.
7431 (phi_translate_set): Likewise. Insert all translated
7432 exprs for a value into the set, keeping possibly multiple
7433 expressions per value.
7434 (compute_antic_aux): Adjust for phi_translate changes.
7435 When intersecting union the expressions and prune those
7436 not in the final value set, keeping possibly multiple
7437 expressions per value. Do not use value-insertion
7438 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7439 all expressions. Add verification that the value-sets
7440 only shrink during iteration.
7441 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7442 (do_pre_regular_insertion): Likewise.
7443 (do_pre_partial_partial_insertion): Likewise.
7445 2018-03-02 Richard Biener <rguenther@suse.de>
7448 * config/darwin.c (saved_debug_info_level): New static global.
7449 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7450 (darwin_asm_lto_end): Restore debug info generation settings.
7452 2018-03-01 Martin Liska <mliska@suse.cz>
7455 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7458 2018-03-01 Richard Biener <rguenther@suse.de>
7461 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7462 type annotation with LTO when debug was disabled at compile-time.
7464 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7466 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7468 (mips_final_postscan_insn): Likewise.
7470 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7472 PR rtl-optimization/84528
7473 * alias.c (init_alias_target): Add commentary.
7474 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7475 a unique base value if the frame pointer is not eliminated
7476 to the stack pointer.
7478 2018-03-01 Tom de Vries <tom@codesourcery.com>
7480 PR rtl-optimization/83327
7481 * lra-int.h (hard_regs_spilled_into): Declare.
7482 * lra.c (hard_regs_spilled_into): Define.
7483 (init_reg_info): Init hard_regs_spilled_into.
7484 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7485 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7486 (process_bb_lives): Handle hard_regs_spilled_into.
7487 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7488 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7490 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7492 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7493 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7494 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7495 * config/rs6000/aix72.h: New file.
7497 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7499 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7500 instead of warning_at with conditional singular and plural messages
7504 * stor-layout.c (update_alignment_for_field): For
7505 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7506 && !DECL_PACKED (field), do the alignment update, just use
7507 only desired_align instead of MAX (type_align, desired_align)
7509 (place_field): Don't do known_align < desired_align handling
7510 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7511 is non-NULL, instead do it after rli->prev_field handling and
7512 only if not within a bitfield word. For DECL_PACKED (field)
7513 use type_align of BITS_PER_UNIT.
7515 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7517 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7518 superfluous parentheses and trailing spaces.
7520 2018-02-28 Richard Biener <rguenther@suse.de>
7522 PR tree-optimization/84584
7523 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7524 SCoPs with fake exit edge.
7526 2018-02-28 Martin Liska <mliska@suse.cz>
7529 * timevar.c (timer::print): Fix format to properly print 100%
7532 2018-02-28 Richard Biener <rguenther@suse.de>
7535 * genmatch.c (capture_info::walk_match): Do not mark
7536 captured expressions without operands as expr_p given
7537 they act more like predicates and should be subject to
7538 "lost tail" side-effect preserving.
7540 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7542 PR rtl-optimization/81611
7543 * auto-inc-dec.c (attempt_change): Move dead note from
7544 mem_insn if it's the next use of regno
7545 (find_address): Take address use of reg holding
7546 non-incremented value. Add parm to limit search to the named
7548 (merge_in_block): Attempt to use a mem insn that is the next
7549 use of the original regno.
7551 2018-02-27 Martin Sebor <msebor@redhat.com>
7554 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7555 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7557 2018-02-27 Martin Sebor <msebor@redhat.com>
7559 PR translation/84207
7560 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7561 n argument to unsigned HOST_WIDE_INT.
7562 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7563 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7564 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7565 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7567 2018-02-27 Richard Biener <rguenther@suse.de>
7569 PR tree-optimization/84512
7570 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7571 Do not use the estimate returned from record_stmt_cost for
7572 the scalar iteration cost but sum properly using add_stmt_cost.
7574 2018-02-27 Richard Biener <rguenther@suse.de>
7576 PR tree-optimization/84466
7577 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7578 Adjust last change to less strictly validate use operands.
7580 2018-02-27 Martin Liska <mliska@suse.cz>
7582 PR gcov-profile/84548
7583 * gcov.c (process_file): Allow partial overlap and consider it
7584 also as group functions.
7585 (output_lines): Properly calculate range of lines for a group.
7587 2018-02-27 Martin Liska <mliska@suse.cz>
7589 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7590 'ggc' suffixes. Change first column width.
7591 (timer::print): Fix formatting of the column.
7593 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7595 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7596 preserve inline entry blocks for the sake of debug inline
7597 entry point markers alone.
7598 (remove_unused_locals): Suggest in comments a better place to
7599 force the preservation of inline entry blocks that are
7600 otherwise unused, but do not preserve them.
7602 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7604 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7606 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7609 * config/i386/constraints.md (Bs): Replace
7610 ix86_indirect_branch_register with
7611 TARGET_INDIRECT_BRANCH_REGISTER.
7613 * config/i386/i386.md (indirect_jump): Likewise.
7614 (tablejump): Likewise.
7615 (*sibcall_memory): Likewise.
7616 (*sibcall_value_memory): Likewise.
7617 Peepholes of indirect call and jump via memory: Likewise.
7618 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7619 (*sibcall_value_GOT_32): Likewise.
7620 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7621 (GOT_memory_operand): Likewise.
7622 (call_insn_operand): Likewise.
7623 (sibcall_insn_operand): Likewise.
7624 (GOT32_symbol_operand): Likewise.
7625 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7627 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7629 PR rtl-optimization/83496
7630 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7631 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7632 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7633 redundant insn, if any.
7634 (relax_delay_slots): Likewise.
7635 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7637 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7639 PR tree-optimization/83965
7640 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7641 that grouped statements are part of a reduction chain. Return
7642 true if the statement is not marked as a reduction itself but
7644 (vect_recog_dot_prod_pattern): Don't check whether the statement
7645 is part of a group here.
7646 (vect_recog_sad_pattern): Likewise.
7647 (vect_recog_widen_sum_pattern): Likewise.
7649 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7652 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7654 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7657 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7659 (ix86_output_indirect_function_return): New prototype.
7660 (ix86_split_simple_return_pop_internal): Likewise.
7661 * config/i386/i386.c (indirect_return_via_cx): New.
7662 (indirect_return_via_cx_bnd): Likewise.
7663 (indirect_thunk_name): Handle return va CX_REG.
7664 (output_indirect_thunk_function): Create alias for
7665 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7666 (ix86_output_indirect_jmp): Remove the bool argument.
7667 (ix86_output_indirect_function_return): New function.
7668 (ix86_split_simple_return_pop_internal): Likewise.
7669 * config/i386/i386.md (*indirect_jump): Don't pass false
7670 to ix86_output_indirect_jmp.
7671 (*tablejump_1): Likewise.
7672 (simple_return_pop_internal): Change it to define_insn_and_split.
7673 Call ix86_split_simple_return_pop_internal to split it for
7675 (simple_return_indirect_internal): Call
7676 ix86_output_indirect_function_return instead of
7677 ix86_output_indirect_jmp.
7679 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7682 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7683 memset and value initialization afterwards.
7685 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7687 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7689 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7692 * common/config/aarch64/aarch64-common.c
7693 (aarch_option_optimization_table[]): Switch
7694 off fomit-frame-pointer
7696 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7697 Chung-Ju Wu <jasonwucj@gmail.com>
7699 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7701 (store_multiple): Ditto.
7703 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7705 * config.gcc: Add --with-cpu support for nds32 target.
7706 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7707 * config/nds32/nds32.opt: Add -mcpu= option.
7709 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7711 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7712 isel=yes): Warn for these deprecated options.
7714 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7716 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7717 ISA_2_5_MASKS_EMBEDDED.
7719 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7721 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7722 p->max as pointers rather than using iterative_hash_expr.
7724 2018-02-23 Carl Love <cel@us.ibm.com>
7726 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7727 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7729 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7730 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7731 P8V_BUILTIN_VEC_VUNSIGNED2.
7733 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7736 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7737 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7739 * lra-constraints.c (curr_insn_transform): Set up
7740 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7742 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7743 * lra-eliminations.c (spill_pseudos): Ditto.
7744 (process_insn_for_elimination): Ditto.
7745 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7746 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7747 LRA_NON_CLOBBERED_ALT.
7749 2018-02-22 Martin Sebor <msebor@redhat.com>
7751 PR tree-optimization/84480
7752 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7753 to maybe_diag_stxncpy_trunc. Call it.
7754 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7755 from gimple_fold_builtin_strcpy. Print inlining stack.
7756 (handle_builtin_stxncpy): Print inlining stack.
7757 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7759 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7762 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7763 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7764 and -fcheck-pointer-bounds are used together.
7765 (indirect_thunk_prefix): New enum.
7766 (indirect_thunk_need_prefix): New function.
7767 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7768 "_nt" instead of "_bnd" for NOTRACK prefix.
7769 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7770 (output_indirect_thunk_function): Likewise.
7772 (ix86_code_end): Update output_indirect_thunk_function calls.
7773 (ix86_output_indirect_branch_via_reg): Replace
7774 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7775 (ix86_output_indirect_branch_via_push): Likewise.
7776 (ix86_output_function_return): Likewise.
7777 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7778 incompatible with -fcf-protection=branch and
7779 -fcheck-pointer-bounds.
7781 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7784 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7785 Change gcc_assert call to output_operand_lossage.
7787 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7789 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7791 2018-02-22 DJ Delorie <dj@redhat.com>
7792 Sebastian Perta <sebastian.perta@renesas.com>
7793 Oleg Endo <olegendo@gcc.gnu.org>
7795 * config/rx/rx.c (rx_rtx_costs): New function.
7796 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7798 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7800 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7802 2018-02-22 Martin Liska <mliska@suse.cz>
7805 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7806 Add "native" as a possible value.
7808 2018-02-22 Martin Liska <mliska@suse.cz>
7811 * config/i386/i386.c (ix86_option_override_internal):
7812 Add "native" as a possible value for -march and -mtune.
7814 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7817 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7818 to all type variants.
7820 PR tree-optimization/84503
7821 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7822 width as info->bitpos + info->bitsize - start.
7823 (merged_store_group::merge_overlapping): Simplify width computation.
7824 (check_no_overlap): New function.
7825 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7826 start + width and last_order of the group, fail if check_no_overlap
7828 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7829 to group if check_no_overlap fails.
7831 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7833 * config/rs6000/altivec.md: Delete contraint arguments to
7834 define_expand, define_split, and define_peephole2, and in
7835 define_insn_and_split if always unused.
7836 * config/rs6000/darwin.md: Ditto.
7837 * config/rs6000/dfp.md: Ditto.
7838 * config/rs6000/rs6000.md: Ditto.
7839 * config/rs6000/sync.md: Ditto.
7840 * config/rs6000/vector.md: Ditto.
7841 * config/rs6000/vsx.md: Ditto.
7843 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7845 * config/rs6000/altivec.md: Write output control strings as braced
7846 blocks instead of double-quoted strings.
7847 * config/rs6000/darwin.md: Ditto.
7848 * config/rs6000/rs6000.md: Ditto.
7849 * config/rs6000/vector.md: Ditto.
7850 * config/rs6000/vsx.md: Ditto.
7852 2018-02-21 Jason Merrill <jason@redhat.com>
7854 PR c++/84314 - ICE with templates and fastcall attribute.
7855 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7857 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7859 * ipa-cp.c (determine_versionability): Fix comment typos.
7861 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7864 * ipa-cp.c (determine_versionability): Do not version functions caling
7867 2018-02-21 Martin Liska <mliska@suse.cz>
7870 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7871 Add "native" as a possible value.
7872 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7873 the macro when native cpu detection is available.
7875 2018-02-21 Martin Liska <mliska@suse.cz>
7878 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7879 Add "native" as a possible value.
7880 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7881 when native cpu detection is available.
7883 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7884 Martin Sebor <msebor@redhat.com>
7886 PR tree-optimization/84478
7887 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7889 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7890 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7891 support which is conservatively correct, for 2 only stay conservative
7892 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7893 argument to the 2 argument get_range_strlen, adjust 6 arg
7894 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7896 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7897 (gimple_fold_builtin_strlen): Pass true as last argument to
7900 2018-02-20 Martin Sebor <msebor@redhat.com>
7903 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7904 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7905 (builtin_memref::builtin_memref): Factor out parts into
7906 set_base_and_offset and call it.
7908 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7911 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7912 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7913 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7914 search at the associated MODE_INT.
7916 2018-02-20 Jeff Law <law@redhat.com>
7919 PR tree-optimization/81592
7921 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7922 for range data rather than using global data.
7923 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7924 range data rather than using global data.
7925 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7926 pass it to children as needed.
7927 (struct directive::fmtresult): Similarly.
7928 (struct directive::set_width): Similarly.
7929 (struct directive::set_precision): Similarly.
7930 (format_integer, format_directive, parse_directive): Similarly.
7931 (format_none): Accept unnamed vr_values parameter.
7932 (format_percent, format_floating, format_character): Similarly.
7933 (format_string, format_plain): Similarly.
7934 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7935 the EVRP range analyzer for range data rather than using global data.
7936 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7937 gimple-ssa-evrp-analyze.h
7938 (class sprintf_dom_walker): Add after_dom_children member function.
7939 Add evrp_range_analyzer member.
7940 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7941 range analyzer as needed.
7942 (sprintf_dom_walker::after_dom_children): New member function.
7943 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7945 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7946 (evrp_range_analyzer::pop_to_marker): Likewise.
7948 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7950 PR tree-optimization/84419
7951 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7952 with the required type if its current type is compatible but
7955 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7958 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7959 after vectorization.
7961 2018-02-20 Martin Liska <mliska@suse.cz>
7964 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7965 possible values if we don't have a hint.
7967 2018-02-20 Martin Liska <mliska@suse.cz>
7971 * final.c (shorten_branches): Build align_tab array with one
7973 * opts.c (finish_options): Add alignment option limit check.
7974 (MAX_CODE_ALIGN): Likewise.
7975 (MAX_CODE_ALIGN_VALUE): Likewise.
7976 * doc/invoke.texi: Document maximum allowed option value for
7977 all -falign-* options.
7979 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7982 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7983 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7984 * var-tracking.c (emit_note_insn_var_location): Remove all references
7985 to NOTE_INSN_CALL_ARG_LOCATION.
7986 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7987 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7988 Use copy_rtx_if_shared.
7989 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7990 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7991 (dwarf2out_var_location): Remove handling of
7992 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7994 * final.c (final_scan_insn): Remove all references to
7995 NOTE_INSN_CALL_ARG_LOCATION.
7996 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7997 before dumping final insns.
7998 * except.c (emit_note_eh_region_end): Remove all references to
7999 NOTE_INSN_CALL_ARG_LOCATION.
8000 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8001 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8002 * config/arc/arc.c (hwloop_optimize): Likewise.
8003 * config/arm/arm.c (create_fix_barrier): Likewise.
8004 * config/s390/s390.c (s390_chunkify_start): Likewise.
8005 * config/sh/sh.c (find_barrier): Likewise.
8006 * config/i386/i386.c (rest_of_insert_endbranch,
8007 ix86_seh_fixup_eh_fallthru): Likewise.
8008 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8009 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8010 * config/frv/frv.c (frv_function_prologue): Likewise.
8011 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8013 (note_outside_basic_block_p): Remove all references to
8014 NOTE_INSN_CALL_ARG_LOCATION.
8015 * gengtype.c (adjust_field_rtx_def): Likewise.
8016 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8018 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8019 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8022 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8025 PR tree-optimization/84452
8026 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8027 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8030 2018-02-19 Martin Liska <mliska@suse.cz>
8033 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8035 2018-02-19 Martin Liska <mliska@suse.cz>
8036 Richard Sandiford <richard.sandiford@linaro.org>
8038 PR tree-optimization/82491
8039 * gimple-fold.c (get_base_constructor): Make earlier bail out
8042 2018-02-19 Carl Love <cel@us.ibm.com>
8044 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8045 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8047 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8048 P8V_BUILTIN_VEC_NEG.
8050 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8052 * config/rl78/rl78.md (movdf): New define expand.
8054 2018-02-19 Martin Liska <mliska@suse.cz>
8057 * doc/invoke.texi: Fix typo.
8058 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8060 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8062 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8063 handle rs6000_single_float and rs6000_double_float specially for
8066 2018-02-16 Jeff Law <law@redhat.com>
8068 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8071 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8074 * ipa-inline.c (inline_small_functions): Fix a typo.
8076 2018-02-16 Nathan Sidwell <nathan@acm.org>
8078 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8080 2018-02-16 Carl Love <cel@us.ibm.com>
8082 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8083 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8084 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8085 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8086 expansion to P8V_BUILTIN_VEC_FLOAT2.
8088 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8090 PR rtl-optimization/70023
8091 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8092 src_regno into account.
8094 2018-02-16 Carl Love <cel@us.ibm.com>
8096 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8097 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8098 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8099 * config/rs6000/rs6000.c: Remove case statements for
8100 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8101 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8102 and P9V_BUILTIN_VEC_VINSERT4B.
8103 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8104 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8105 * config/rs6000/vsx.md:
8106 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8109 2018-02-16 Carl Love <cel@us.ibm.com>
8111 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8113 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8115 * config/rs6000/rs6000-c.c: Add the definitions for
8116 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8117 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8118 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8119 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8120 definition for insert4b and define insn *insert3b_internal.
8121 * doc/extend.texi: Add documentation for vec_extract4b.
8123 2018-02-16 Nathan Sidwell <nathan@acm.org>
8125 * doc/extend.texi (Backwards Compatibility): Mention friend
8126 injection. Note for-scope is deprecated.
8127 * doc/invoke.texi (-ffriend-injection): Deprecate.
8129 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8131 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8132 that moved to I2, also allow destinations that are a paradoxical
8133 subreg (instead of a normal reg).
8135 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8138 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8141 2018-02-16 Richard Biener <rguenther@suse.de>
8143 PR tree-optimization/84037
8144 PR tree-optimization/84016
8146 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8147 Adjust vec_construct for the fact we need additional higher latency
8148 128bit inserts for AVX256 and AVX512 vector builds.
8149 (ix86_add_stmt_cost): Scale vector construction cost for
8152 2018-02-16 Richard Biener <rguenther@suse.de>
8154 PR tree-optimization/84417
8155 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8156 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8157 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8159 2018-02-16 Martin Liska <mliska@suse.cz>
8162 * internal-fn.def (ASAN_CHECK): Set proper flags.
8163 (ASAN_MARK): Likewise.
8165 2018-02-16 Julia Koval <julia.koval@intel.com>
8167 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8168 from PTA_CANNONLAKE.
8170 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8173 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8174 Use ++iter rather than iter++ for std::list iterators.
8175 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8176 defer deleting them until all nodes in the forest are processed. Do
8177 free even leaf nodes. Change to_process into auto_vec.
8180 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8181 * vec.h (vec_default_construct): Use memset instead of placement new
8182 if BROKEN_VALUE_INITIALIZATION is defined.
8183 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8184 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8187 PR rtl-optimization/83723
8188 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8189 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8190 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8192 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8194 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8196 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8198 PR rtl-optimization/81443
8199 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8200 from inner REGs to paradoxical SUBREGs.
8202 2018-02-16 Richard Biener <rguenther@suse.de>
8204 PR tree-optimization/84399
8205 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8206 For operands we can analyze at their definition make sure we can
8207 analyze them at each use as well.
8209 2018-02-16 Richard Biener <rguenther@suse.de>
8211 PR tree-optimization/84190
8212 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8213 volatile accesses if the decl isn't volatile.
8215 2018-02-15 Jason Merrill <jason@redhat.com>
8217 PR c++/84314 - ICE with templates and fastcall attribute.
8218 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8219 TYPE_CANONICAL on an existing type.
8221 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8223 PR tree-optimization/84383
8224 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8225 dstoff nor call operand_equal_p if dstbase is NULL.
8227 PR tree-optimization/84334
8228 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8229 also a CONSTANT_CLASS_P, punt.
8231 2018-02-14 Jim Wilson <jimw@sifive.com>
8233 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8234 first SMALL_OPERAND check. New local min_second_step. Move assert
8235 to where locals are set. Add TARGET_RVC support.
8236 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8238 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8240 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8242 2018-02-14 Martin Sebor <msebor@redhat.com>
8244 PR tree-optimization/83698
8245 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8246 arrays constrain the offset range to their bounds.
8247 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8248 (builtin_access::overlap): Avoid setting the size of overlap if it's
8250 (maybe_diag_overlap): Also consider arrays when deciding what values
8251 of offsets to include in diagnostics.
8253 2018-02-14 Martin Sebor <msebor@redhat.com>
8256 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8257 that correspond to the kind of a declaration.
8259 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8262 * config/pa/pa.md: Load address of PIC label using the linkage table
8263 if the label is nonlocal.
8265 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8267 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8268 warning message if user requests -maltivec=be.
8269 * doc/invoke.texi: Document deprecation of -maltivec=be.
8271 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8274 * config/rs6000/rs6000-c.c: Update definitions for
8275 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8276 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8278 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8281 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8282 add _get_ssp intrinsics. Remove argument from
8283 __builtin_ia32_rdssp[d|q].
8284 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8285 * config/i386/i386-builtin.def: Remove argument from
8286 __builtin_ia32_rdssp[d|q].
8287 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8288 ix86_expand_special_args_builtin for _rdssp[d|q].
8289 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8290 Clear register before usage.
8291 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8292 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8294 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8296 PR tree-optimization/84357
8297 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8298 operand 1 of an ARRAY_REF too.
8300 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8303 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8304 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8306 (set_of_reg): New struct.
8307 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8308 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8309 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8311 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8312 Split into bitclr, bitset, bitinvert patterns if appropriate.
8313 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8314 use rx_fuse_in_memory_bitop.
8315 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8316 to named insn, correct maximum insn length.
8318 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8321 * machmode.def: Define a complex mode for PARTIAL_INT.
8322 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8324 * doc/rtl.texi: Document CSPImode.
8325 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8327 (msp430_hard_regno_nregs_with_padding): Likewise.
8329 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8332 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8334 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8336 PR rtl-optimization/84169
8337 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8338 we generated a parallel as new i3 and we split that to new i2 and i3
8339 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8340 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8341 those to i2, not i1. Partially rewrite this scan code.
8343 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8346 * stor-layout.c (place_field): For variable length fields, adjust
8347 offset_align afterwards not just based on the field's alignment,
8348 but also on the size.
8351 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8352 of exps and logs in the use_exp2 case.
8354 2018-02-13 Jeff Law <law@redhat.com>
8356 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8359 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8362 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8366 * common.opt (gas-loc-support, gas-locview-support): New.
8367 (ginline-points, ginternal-reset-location-views): New.
8368 * doc/invoke.texi: Document them. Use @itemx where intended.
8369 (gvariable-location-views): Adjust.
8370 * target.def (reset_location_view): New.
8371 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8372 (TARGET_RESET_LOCATION_VIEW): New.
8373 * doc/tm.texi: Rebuilt.
8374 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8375 (dwarf2out_default_as_locview_support): New.
8376 (output_asm_line_debug_info): Use option variables.
8377 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8378 (output_loc_list): Likewise.
8379 (add_high_low_attributes): Check option variables.
8380 Don't output entry view attribute in strict mode.
8381 (gen_inlined_subroutine_die): Check option variables.
8382 (dwarf2out_inline_entry): Likewise.
8383 (init_sections_and_labels): Likewise.
8384 (dwarf2out_early_finish): Likewise.
8385 (maybe_reset_location_view): New, from...
8386 (dwarf2out_var_location): ... here. Call it.
8387 * debug.h (dwarf2out_default_as_loc_support): Declare.
8388 (dwarf2out_default_as_locview_support): Declare.
8389 * hooks.c (hook_int_rtx_insn_0): New.
8390 * hooks.h (hook_int_rtx_insn_0): Declare.
8391 * toplev.c (process_options): Take -gas-loc-support and
8392 -gas-locview-support from dwarf2out. Enable
8393 -gvariable-location-views by default only with locview
8394 assembler support. Enable -ginternal-reset-location-views by
8395 default only if the target defines the corresponding hook.
8396 Enable -ginline-points by default if location views are
8397 enabled; force it disabled if statement frontiers are
8399 * tree-inline.c (expand_call_inline): Check option variables.
8400 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8402 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8404 PR tree-optimization/84321
8405 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8406 handling. Also check whether the anti-range contains any values
8407 that satisfy the mask; switch to a VR_RANGE if not.
8409 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8412 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8414 2018-02-13 Martin Jambor <mjambor@suse.cz>
8417 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8418 of call statements, also set location of a load to a temporary.
8420 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8422 * config/rl78/rl78.c (add_vector_labels): New function.
8423 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8424 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8425 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8426 which checks that no arguments are passed.
8427 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8428 * doc/extend.texi: Documentation for the new attribute.
8430 2018-02-13 Andreas Schwab <schwab@suse.de>
8432 * config/riscv/linux.h (CPP_SPEC): Define.
8434 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8437 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8438 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8439 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8440 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8441 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8442 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8443 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8444 temporarily for AES and PCLMUL builtins.
8446 PR tree-optimization/84339
8447 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8448 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8452 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8453 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8454 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8456 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8458 * omp-simd-clone.h: New file.
8459 * omp-simd-clone.c: Include omp-simd-clone.h.
8460 (expand_simd_clones): No longer static.
8461 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8462 cgraph.h and omp-simd-clone.h.
8463 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8464 (vect_recog_widen_shift_pattern): Formatting fix.
8465 (vect_pattern_recog_1): Don't check optab for calls.
8468 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8469 operands[2] into a REG before using gen_lowpart on it.
8471 2018-02-12 Jeff Law <law@redhat.com>
8474 * config/sh/sh.c (find_barrier): Consider a sibling call
8477 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8478 successfully back substituting a reg.
8480 2018-02-12 Richard Biener <rguenther@suse.de>
8482 PR tree-optimization/84037
8483 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8484 parameter, move visited init to caller.
8485 (vect_slp_analyze_operations): Separate cost from validity
8486 check, initialize visited once for all instances.
8487 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8489 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8491 (vect_model_promotion_demotion_cost): Likewise.
8492 (vectorizable_bswap): Guard cost modeling with !slp_node
8493 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8495 (vectorizable_call): Likewise.
8496 (vectorizable_conversion): Likewise.
8497 (vectorizable_assignment): Likewise.
8498 (vectorizable_shift): Likewise.
8499 (vectorizable_operation): Likewise.
8500 (vectorizable_store): Likewise.
8501 (vectorizable_load): Likewise.
8502 (vectorizable_condition): Likewise.
8503 (vectorizable_comparison): Likewise.
8505 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8508 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8509 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8510 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8512 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8515 * params.def (inline-min-speedup): Increase from 8 to 15.
8516 (max-inline-insns-auto): Decrease from 40 to 30.
8517 * ipa-split.c (consider_split): Add some buffer for function to
8518 be considered inlining candidate.
8519 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8522 2018-02-12 Richard Biener <rguenther@suse.de>
8524 PR tree-optimization/84037
8525 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8526 matched stmts if we cannot swap the non-matched ones.
8528 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8530 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8531 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8532 _mm_maskz_scalef_round_ss): New intrinsics.
8533 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8534 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8535 __builtin_ia32_scalefss_round): Remove.
8536 (__builtin_ia32_scalefsd_mask_round,
8537 __builtin_ia32_scalefss_mask_round): New intrinsics.
8538 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8539 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8540 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8541 "<round_constraint>")): Changed to ...
8542 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8543 "<round_scalar_constraint>")): ... this.
8544 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8545 %0, %1, %2<round_op3>}"): Changed to ...
8546 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8547 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8548 %2<round_scalar_mask_op3>}"): ... this.
8549 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8551 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8553 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8554 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8555 (_mm_maskz_sqrt_round_ss): New intrinsics.
8556 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8557 (__builtin_ia32_sqrtsd_mask_round)
8558 (__builtin_ia32_sqrtss_mask_round): New builtins.
8559 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8560 (__builtin_ia32_sqrtss_round): Remove.
8561 (__builtin_ia32_sqrtsd_mask_round)
8562 (__builtin_ia32_sqrtss_mask_round): New builtins.
8563 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8564 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8565 ((match_operand:VF_128 1 "vector_operand"
8566 "xBm,<round_constraint>")): Changed to ...
8567 ((match_operand:VF_128 1 "vector_operand"
8568 "xBm,<round_scalar_constraint>")): ... this.
8569 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8570 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8571 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8572 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8573 %<iptr>1<round_scalar_mask_op3>}): ... this.
8574 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8575 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8577 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8580 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8581 Cast vec_cmpeq result to correct type.
8582 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8583 Cast vec_cmpgt result to correct type.
8585 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8587 * final.c (final_scan_insn_1): Renamed from...
8588 (final_scan_insn): ... this. New wrapper, to recover
8589 seen from the outermost call in recursive ones.
8590 * config/sparc/sparc.c (output_return): Drop seen from call.
8591 (output_sibcall): Likewise.
8592 * config/visium/visium.c (output_branch): Likewise.
8594 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8596 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8599 2018-02-10 Alan Modra <amodra@gmail.com>
8602 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8603 Specify LR as an input.
8605 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8608 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8609 remove_member_access_dummy_vars): New functions.
8610 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8611 lower_omp_1, execute_lower_omp): Use them.
8613 PR rtl-optimization/84308
8614 * shrink-wrap.c (spread_components): Release todo vector.
8616 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8618 PR rtl-optimization/57193
8619 * ira-color.c (struct allocno_color_data): Add member
8620 conflict_allocno_hard_prefs.
8621 (update_conflict_allocno_hard_prefs): New.
8622 (bucket_allocno_compare_func): Add a preference based on
8623 conflict_allocno_hard_prefs.
8624 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8625 (color_allocnos): Remove a dead code. Initiate
8626 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8628 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8631 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8632 constraint from =wa to wa. Avoid a subreg on the output operand,
8633 instead use a pseudo and subreg it in a move.
8634 (p9_xxbrd_<mode>): Changed to ...
8635 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8636 (p9_xxbrd_v2df): New expander.
8637 (p9_xxbrw_<mode>): Changed to ...
8638 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8639 (p9_xxbrw_v4sf): New expander.
8641 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8643 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8645 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8648 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8649 multiply in 32-bit mode.
8650 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8651 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8654 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8656 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8657 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8658 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8659 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8661 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8664 * dwarf2out.c (is_trivial_indirect_ref): New function.
8665 (dwarf2out_late_global_decl): Do not generate a location
8666 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8667 and that are not defined in the current unit.
8669 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8671 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8672 instead of a libcall for UNORDERED.
8674 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8677 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8678 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8680 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8683 * config/s390/s390.c (s390_set_current_function): Invoke
8684 s390_indirect_branch_settings also if fndecl didn't change.
8686 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8688 * config/rs6000/rs6000.md (blockage): Set length to zero.
8690 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8692 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8694 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8697 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8698 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8702 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8703 PARALLEL incoming that failed vt_get_decl_and_offset check.
8706 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8707 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8709 (get_variable_section): For decls in named .bss* sections pass true as
8710 second argument to bss_initializer_p.
8712 2018-02-09 Marek Polacek <polacek@redhat.com>
8713 Jakub Jelinek <jakub@redhat.com>
8716 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8717 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8718 Sync some changes from cxx_fold_indirect_ref.
8720 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8722 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8724 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8725 (BLOCK_INLINE_ENTRY_LABEL): New.
8726 (dwarf2out_var_location): Disregard inline entry markers.
8727 (inline_entry_data): New struct.
8728 (inline_entry_data_hasher): New hashtable type.
8729 (inline_entry_data_hasher::hash): New.
8730 (inline_entry_data_hasher::equal): New.
8731 (inline_entry_data_table): New variable.
8732 (add_high_low_attributes): Add DW_AT_entry_pc and
8733 DW_AT_GNU_entry_view attributes if a pending entry is found
8734 in inline_entry_data_table. Add old entry_pc attribute only
8735 if debug nonbinding markers are disabled.
8736 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8737 markers are enabled.
8738 (block_within_block_p, dwarf2out_inline_entry): New.
8739 (dwarf2out_finish): Check that no entries remained in
8740 inline_entry_data_table.
8741 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8742 (final_scan_insn, notice_source_line): Likewise.
8743 (rest_of_clean_state): Skip inline entry markers.
8744 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8746 * gimple.c (gimple_build_debug_inline_entry): New.
8747 * gimple.h (enum gimple_debug_subcode): Add
8748 GIMPLE_DEBUG_INLINE_ENTRY.
8749 (gimple_build_debug_inline_entry): Declare.
8750 (gimple_debug_inline_entry_p): New.
8751 (gimple_debug_nonbind_marker_p): Adjust.
8752 * insn-notes.def (INLINE_ENTRY): New.
8753 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8754 inline entry marker notes.
8755 (print_insn): Likewise.
8756 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8757 (INSN_DEBUG_MARKER_KIND): Likewise.
8758 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8759 * tree-inline.c (expand_call_inline): Build and insert
8760 debug_inline_entry stmt.
8761 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8762 inline entry blocks early, if nonbind markers are enabled.
8763 (dump_scope_block): Dump fragment info.
8764 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8765 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8766 (gimple_build_debug_inline_entry): New.
8767 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8768 Enable/disable inline entry points too.
8769 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8770 (DEBUG_INSN): Describe inline entry markers.
8772 * common.opt (gvariable-location-views): New.
8773 (gvariable-location-views=incompat5): New.
8774 * config.in: Rebuilt.
8775 * configure: Rebuilt.
8776 * configure.ac: Test assembler for view support.
8777 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8778 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8779 * dwarf2out.c (var_loc_view): New typedef.
8780 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8781 (dwarf2out_locviews_in_attribute): New.
8782 (dwarf2out_locviews_in_loclist): New.
8783 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8784 (enum dw_line_info_opcode): Add LI_adv_address.
8785 (struct dw_line_info_table): Add view.
8786 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8787 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8788 (zero_view_p): New variable.
8789 (ZERO_VIEW_P): New macro.
8790 (output_asm_line_debug_info): New.
8791 (struct var_loc_node): Add view.
8792 (add_AT_view_list, AT_loc_list): New.
8793 (add_var_loc_to_decl): Add view param. Test it against last.
8794 (new_loc_list): Add view params. Record them.
8795 (AT_loc_list_ptr): Handle loc and view lists.
8796 (view_list_to_loc_list_val_node): New.
8797 (print_dw_val): Handle dw_val_class_view_list.
8798 (size_of_die): Likewise.
8799 (value_format): Likewise.
8800 (loc_list_has_views): New.
8801 (gen_llsym): Set vl_symbol too.
8802 (maybe_gen_llsym, skip_loc_list_entry): New.
8803 (dwarf2out_maybe_output_loclist_view_pair): New.
8804 (output_loc_list): Output view list or entries too.
8805 (output_view_list_offset): New.
8806 (output_die): Handle dw_val_class_view_list.
8807 (output_dwarf_version): New.
8808 (output_compilation_unit_header): Use it.
8809 (output_skeleton_debug_sections): Likewise.
8810 (output_rnglists, output_line_info): Likewise.
8811 (output_pubnames, output_aranges): Update version comments.
8812 (output_one_line_info_table): Output view numbers in asm comments.
8813 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8814 Call maybe_gen_llsym.
8815 (loc_list_from_tree_1): Adjust.
8816 (add_AT_location_description): Create view list attribute if
8817 needed, check it's absent otherwise.
8818 (convert_cfa_to_fb_loc_list): Adjust.
8819 (maybe_emit_file): Call output_asm_line_debug_info for test.
8820 (dwarf2out_var_location): Reset views as needed. Precompute
8821 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8822 attribute. Set view.
8823 (new_line_info_table): Reset next view.
8824 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8825 (dwarf2out_source_line): Likewise. Output view resets and labels to
8826 the assembler, or select appropriate line info opcodes.
8827 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8828 (optimize_string_length): Catch it. Adjust.
8829 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8830 dw_val_class_view_list, and remove it if no longer needed.
8831 (hash_loc_list): Hash view numbers.
8832 (loc_list_hasher::equal): Compare them.
8833 (optimize_location_lists): Check whether a view list symbol is
8834 needed, and whether the locview attribute is present, and
8835 whether they match. Remove the locview attribute if no longer
8837 (index_location_lists): Call skip_loc_list_entry for test.
8838 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8839 Use output_dwarf_version.
8840 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8841 (struct dw_val_node): Add val_view_list.
8842 * final.c (SEEN_NEXT_VIEW): New.
8843 (set_next_view_needed): New.
8844 (clear_next_view_needed): New.
8845 (maybe_output_next_view): New.
8846 (final_start_function): Rename to...
8847 (final_start_function_1): ... this. Take pointer to FIRST,
8848 add SEEN parameter. Emit param bindings in the initial view.
8849 (final_start_function): Reintroduce SEEN-less interface.
8850 (final): Rename to...
8851 (final_1): ... this. Take SEEN parameter. Output final pending
8852 next view at the end.
8853 (final): Reintroduce seen-less interface.
8854 (final_scan_insn): Output pending next view before switching
8855 sections or ending a block. Mark the next view as needed when
8856 outputting variable locations. Notify debug backend of section
8857 changes, and of location view changes.
8858 (rest_of_handle_final): Adjust.
8859 * toplev.c (process_options): Autodetect value for debug variable
8860 location views option. Warn on incompat5 without -gdwarf-5.
8861 * doc/invoke.texi (gvariable-location-views): New.
8862 (gvariable-location-views=incompat5): New.
8863 (gno-variable-location-views): New.
8865 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8867 PR tree-optimization/84136
8868 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8869 that the result of find_edge is non-NULL.
8871 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8874 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8875 storing integer register in SImode. Fix cost of 256 and 512
8876 byte aligned SSE register store.
8878 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8880 * config/i386/i386.c (ix86_multiplication_cost): Fix
8881 multiplication cost for TARGET_AVX512DQ.
8883 2018-02-08 Marek Polacek <polacek@redhat.com>
8885 PR tree-optimization/84238
8886 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8889 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8891 PR tree-optimization/84265
8892 * tree-vect-stmts.c (vectorizable_store): Don't treat
8893 VMAT_CONTIGUOUS accesses as grouped.
8894 (vectorizable_load): Likewise.
8896 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8898 PR tree-optimization/81635
8899 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8900 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8901 (test_round_for_mask): New functions.
8902 (wide_int_cc_tests): Call test_round_for_mask.
8903 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8904 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8905 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8906 range returned by get_range_info.
8908 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8911 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8912 * symtab.c: Include builtins.h
8913 (symtab_node::output_to_lto_symbol_table_p): Move here
8914 from lto-streamer-out.c:output_symbol_p.
8915 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8916 (output_symbol_p): Move all logic to symtab.c
8917 (produce_symtab): Update.
8919 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8921 * config/s390/s390-opts.h (enum indirect_branch): Define.
8922 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8923 (s390_indirect_branch_via_thunk)
8924 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8925 (enum s390_indirect_branch_type): Define.
8926 * config/s390/s390.c (struct s390_frame_layout, struct
8927 machine_function): Remove.
8928 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8929 (indirect_branch_table_label_no, indirect_branch_table_name):
8931 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8932 (enum s390_indirect_branch_option): Define.
8933 (s390_return_addr_from_memory): New function.
8934 (s390_handle_string_attribute): New function.
8935 (s390_attribute_table): Add new attribute handler.
8936 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8937 (s390_indirect_branch_via_thunk): New function.
8938 (s390_indirect_branch_via_inline_thunk): New function.
8939 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8940 sibling call optimization for non z10 compiles.
8941 (s390_emit_call): Force indirect branch target to be a single
8942 register. Add r1 clobber for non-z10 compiles.
8943 (s390_emit_epilogue): Emit return jump via return_use expander.
8944 (s390_reorg): Handle JUMP_INSNs as execute targets.
8945 (s390_option_override_internal): Perform validity checks for the
8946 new command line options.
8947 (s390_indirect_branch_attrvalue): New function.
8948 (s390_indirect_branch_settings): New function.
8949 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8950 (s390_output_indirect_thunk_function): New function.
8951 (s390_code_end): Implement target hook.
8952 (s390_case_values_threshold): Implement target hook.
8953 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8955 * config/s390/s390.h (struct s390_frame_layout)
8956 (struct machine_function): Move here from s390.c.
8957 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8958 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8959 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8960 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8961 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8962 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8963 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8964 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8965 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8966 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8967 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8968 (mnemonic attribute): Add values which aren't recognized
8970 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8971 pattern for branch conversion. Fix mnemonic attribute.
8972 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8973 indirect branch via thunk if requested.
8974 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8975 ("*indirect_jump"): Disable for branch conversion using out of
8977 ("indirect_jump_via_thunk<mode>_z10")
8978 ("indirect_jump_via_thunk<mode>")
8979 ("indirect_jump_via_inlinethunk<mode>_z10")
8980 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8981 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8982 ("casesi_jump_via_inlinethunk<mode>_z10")
8983 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8984 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8985 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8986 ("*indirect2_jump"): Disable for branch conversion.
8987 ("casesi_jump"): Turn into expander and expand patterns for branch
8989 ("return_use"): New expander.
8990 ("*return"): Emit return via thunk and rename it to ...
8991 ("*return<mode>"): ... this one.
8992 * config/s390/s390.opt: Add new options and and enum for the
8995 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8997 * lra-constraints.c (match_reload): Unconditionally use
8998 gen_lowpart_SUBREG, rather than selecting between that
8999 and equivalent gen_rtx_SUBREG code.
9001 2018-02-08 Richard Biener <rguenther@suse.de>
9003 PR tree-optimization/84233
9004 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9005 changed flag instead of boguously re-using phi_inserted.
9007 2018-02-08 Martin Jambor <mjambor@suse.cz>
9009 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9010 static local variables.
9012 2018-02-08 Richard Biener <rguenther@suse.de>
9014 PR tree-optimization/84278
9015 * tree-vect-stmts.c (vectorizable_store): When looking for
9016 smaller vector types to perform grouped strided loads/stores
9017 make sure the mode is supported by the target.
9018 (vectorizable_load): Likewise.
9020 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9022 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9023 Increase LDP/STP opportunities by adding adjacent callee-saves.
9025 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9027 PR rtl-optimization/84068
9028 PR rtl-optimization/83459
9029 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9031 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9033 PR tree-optimization/84224
9034 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9035 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9038 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9041 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9042 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9043 count, remove one USE.
9045 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9047 * doc/install.texi (Configuration): Document the
9048 --with-long-double-format={ibm,ieee} PowerPC configuration
9052 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9053 Convert from define_expand to be define_insn_and_split. Rework
9054 float/double/_Float128 conversions to QI/HI/SImode to work with
9055 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9056 conversions to QI/HImode types did a store and then a load to
9057 truncate the value. For conversions to VSX registers, don't split
9058 the insn, instead emit the code directly. Use the code iterator
9059 any_fix to combine signed and unsigned conversions.
9060 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9061 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9062 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9063 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9064 (fix_<mode>di2_hw): Likewise.
9065 (fixuns_<mode>di2_hw): Likewise.
9066 (fix_<mode>si2_hw): Likewise.
9067 (fixuns_<mode>si2_hw): Likewise.
9068 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9069 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9070 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9071 fix<uns>_trunc<SFDF:mode>si2_p8.
9072 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9074 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9075 (fix<uns>_<mode>_mem): Likewise.
9076 (fctiw<u>z_<mode>_mem): Likewise.
9077 (fix<uns>_<mode>_mem): Likewise.
9078 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9079 the register allocator from doing a direct move to the GPRs to do
9080 a store, and instead use the ISA 3.0 store byte/half-word from
9081 vector register instruction. For IEEE 128-bit floating point,
9082 also optimize stores of 32-bit ints.
9083 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9085 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9087 * genextract.c (push_pathstr_operand): New function to support
9089 (walk_rtx): Call push_pathstr_operand.
9090 (print_path): Support [a-zA-Z].
9092 2018-02-07 Richard Biener <rguenther@suse.de>
9094 PR tree-optimization/84037
9095 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9096 (cse_and_gimplify_to_preheader): Declare.
9097 (vect_get_place_in_interleaving_chain): Likewise.
9098 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9100 (_loop_vec_info::~_loop_vec_info): Delete it.
9101 (cse_and_gimplify_to_preheader): New function.
9102 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9103 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9104 (vectorizable_load): Likewise. For grouped stores always base
9105 the IV on the first element.
9106 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9107 condition before gimplifying.
9109 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9111 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9112 *DIV_EXPR and *MOD_EXPR.
9114 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9117 * config/i386/i386.c (ix86_option_override_internal): Mask out
9118 the CF_SET bit when checking -fcf-protection.
9120 2018-02-07 Tom de Vries <tom@codesourcery.com>
9123 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9126 2018-02-07 Richard Biener <rguenther@suse.de>
9128 PR tree-optimization/84204
9129 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9132 PR tree-optimization/84205
9133 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9134 special-case isl_ast_op_zdiv_r.
9136 PR tree-optimization/84223
9137 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9138 Only add conditions from within the region.
9139 (gather_bbs::after_dom_children): Adjust.
9141 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9144 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9145 * config/avr/avr.md: Only post-reload split REG-REG moves if
9146 either register is GENERAL_REG_P.
9148 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9150 PR tree-optimization/84235
9151 * tree-ssa-scopedtables.c
9152 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9153 if the subtraction is performed in floating point type where NaNs are
9154 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9155 build 1. Formatting fix.
9157 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9160 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9161 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9162 and skip it regardless of bb boundaries. Use CALL_P macro,
9163 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9164 unnecessarily, formatting fix.
9166 2018-02-06 Michael Collison <michael.collison@arm.com>
9168 * config/arm/thumb2.md:
9169 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9170 (*thumb_mov_notscc): Ditto.
9172 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9175 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9176 unsigned_fix, not "s".
9178 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9180 * configure.ac (gcc_fn_eh_frame_ro): New function.
9181 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9182 correct .eh_frame permissions.
9183 * configure: Regenerate.
9185 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9187 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9190 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9192 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9193 Display warning message for -mno-speculate-indirect-jumps.
9195 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9197 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9199 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9201 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9203 PR tree-optimization/84225
9204 * tree-eh.c (find_trapping_overflow): Only call
9205 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9207 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9210 * config/i386/i386.c: Reimplement the check of possible options
9211 -mibt/-mshstk conbination. Change error messages.
9212 * doc/invoke.texi: Fix a typo: remove extra '='.
9214 2018-02-06 Marek Polacek <polacek@redhat.com>
9216 PR tree-optimization/84228
9217 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9219 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9222 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9223 emitted arch directives.
9224 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9225 __ARM_FEATURE_COPROC before changing architectures.
9227 2018-02-06 Richard Biener <rguenther@suse.de>
9229 * config/i386/i386.c (print_reg): Fix typo.
9230 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9232 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9234 * configure: Regenerate.
9236 2018-02-05 Martin Sebor <msebor@redhat.com>
9238 PR tree-optimization/83369
9239 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9242 2018-02-05 Martin Liska <mliska@suse.cz>
9244 * doc/invoke.texi: Cherry-pick upstream r323995.
9246 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9248 * ira.c (ira_init_register_move_cost): Adjust comment.
9250 2018-02-05 Martin Liska <mliska@suse.cz>
9252 PR gcov-profile/84137
9253 * doc/gcov.texi: Fix typo in documentation.
9255 2018-02-05 Martin Liska <mliska@suse.cz>
9257 PR gcov-profile/83879
9258 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9259 using dlopen functionality.
9261 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9263 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9264 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9265 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9266 _mm_maskz_range_round_ss): New intrinsics.
9267 (__builtin_ia32_rangesd128_round)
9268 (__builtin_ia32_rangess128_round): Remove.
9269 (__builtin_ia32_rangesd128_mask_round,
9270 __builtin_ia32_rangess128_mask_round): New builtins.
9271 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9272 __builtin_ia32_rangess128_round): Remove.
9273 (__builtin_ia32_rangesd128_mask_round,
9274 __builtin_ia32_rangess128_mask_round): New builtins.
9275 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9276 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9277 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9278 "<round_saeonly_constraint>")): Changed to ...
9279 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9280 "<round_saeonly_scalar_constraint>")): ... this.
9281 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9282 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9283 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9284 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9285 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9287 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9289 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9291 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9292 Remove all values except native, 8540 and 8548.
9294 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9296 * config/i386/i386.c (ix86_output_function_return): Pass
9297 INVALID_REGNUM, instead of -1, as invalid register number to
9298 indirect_thunk_name and output_indirect_thunk.
9300 2018-02-02 Julia Koval <julia.koval@intel.com>
9302 * config.gcc: Add -march=icelake.
9303 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9304 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9305 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9306 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9307 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9308 (processor_target_table): Add icelake.
9309 (ix86_option_override_internal): Handle new PTAs.
9310 (get_builtin_code_for_version): Handle icelake.
9311 (M_INTEL_COREI7_ICELAKE): New.
9312 (fold_builtin_cpu): Handle icelake.
9313 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9314 * doc/invoke.texi: Add -march=icelake.
9316 2018-02-02 Julia Koval <julia.koval@intel.com>
9318 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9319 to wide_int_bitmask.
9320 * wide-int-bitmask.h: New.
9322 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9325 * config/i386/i386.md: Replace Pmode with word_mode in
9326 builtin_setjmp_setup and builtin_longjmp to support x32.
9328 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9332 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9334 (rs6000_supported_cpu_names): New static variable.
9335 (linux_cpu_translation_table): Likewise.
9336 (elf_platform) <cpu>: Define new static variable and use it.
9337 Translate kernel AT_PLATFORM name to canonical name if needed.
9338 Error if platform name is unknown.
9340 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9343 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9345 2018-02-01 Jeff Law <law@redhat.com>
9348 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9349 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9350 the scratch if RELEASE_VIA_POP is false.
9351 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9352 If we have to save a temporary register, decrement SIZE appropriately.
9353 Pass new arguments to release_scratch_register_on_entry.
9354 (ix86_adjust_stack_and_probe): Likewise.
9355 (ix86_emit_probe_stack_range): Pass new arguments to
9356 release_scratch_register_on_entry.
9358 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9360 PR rtl-optimization/84157
9361 * combine.c (change_zero_ext): Use REG_P predicate in
9362 front of HARD_REGISTER_P predicate.
9364 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9366 * config/avr/avr.c (avr_option_override): Move disabling of
9367 -fdelete-null-pointer-checks to...
9368 * common/config/avr/avr-common.c (avr_option_optimization_table):
9371 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9373 PR tree-optimization/81635
9374 * tree-data-ref.c (split_constant_offset_1): For types that
9375 wrap on overflow, try to use range info to prove that wrapping
9378 2018-02-01 Renlin Li <renlin.li@arm.com>
9381 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9383 (aarch64_register_move_cost): Likewise.
9384 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9386 (REG_CLASS_NAMES): Likewise.
9387 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9388 TAILCALL_ADDR_REGS. Remove IP registers.
9389 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9391 2018-02-01 Richard Biener <rguenther@suse.de>
9393 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9394 for specifying RPO order and allow NULL for that.
9395 * domwalk.c (dom_walker::dom_walker): Likewise.
9396 (dom_walker::walk): Handle NULL RPO order.
9397 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9399 (rewrite_update_dom_walker): Likewise.
9400 (mark_def_dom_walker): Likewise.
9402 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9404 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9405 (aarch64_maybe_expand_sve_subreg_move): Declare.
9406 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9407 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9409 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9410 that are semantically a reverse operation.
9411 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9412 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9413 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9415 (aarch64_can_change_mode_class): For big-endian, forbid changes
9416 between two SVE modes if they have different element sizes.
9418 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9420 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9421 the TImode handling for big-endian targets.
9423 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9425 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9426 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9428 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9429 Remove BSWAP handing for big-endian targets and use the form of
9430 LD1RQ appropariate for the mode.
9432 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9434 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9435 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9438 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9441 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9442 check for operands that need to go through aarch64_sve_reload_be.
9444 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9446 PR tree-optimization/81661
9447 PR tree-optimization/84117
9448 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9449 * tree-eh.c: Include gimplify.h.
9450 (find_trapping_overflow, replace_trapping_overflow,
9451 rewrite_to_non_trapping_overflow): New functions.
9452 * tree-vect-loop.c: Include tree-eh.h.
9453 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9454 * tree-data-ref.c: Include tree-eh.h.
9455 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9457 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9459 PR rtl-optimization/84123
9460 * combine.c (change_zero_ext): Check if hard register satisfies
9461 can_change_dest_mode before calling gen_lowpart_SUBREG.
9463 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9466 * ira.c (ira_init_register_move_cost): Remove assert.
9468 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9470 PR rtl-optimization/84071
9471 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9472 * doc/tm.texi: Regenerate.
9474 2018-01-31 Richard Biener <rguenther@suse.de>
9476 PR tree-optimization/84132
9477 * tree-data-ref.c (analyze_miv_subscript): Properly
9478 check whether evolution_function_is_affine_multivariate_p
9479 before calling gcd_of_steps_may_divide_p.
9481 2018-01-31 Julia Koval <julia.koval@intel.com>
9484 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9485 * config/i386/i386.md (rdpid_rex64) New.
9486 (rdpid): Make 32bit only.
9488 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9491 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9492 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9494 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9497 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9499 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9501 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9503 PR rtl-optimization/84071
9504 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9505 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9507 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9509 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9510 (arc_attribute_table): Add 'aux' attribute.
9511 (arc_in_small_data_p): Consider aux like variables.
9512 (arc_is_aux_reg_p): New function.
9513 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9514 (arc_get_aux_arg): New function.
9515 (prepare_move_operands): Handle aux-register access.
9516 (arc_handle_aux_attribute): New function.
9517 * doc/extend.texi (ARC Variable attributes): Add subsection.
9519 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9521 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9522 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9523 (arc_attribute_table): Add 'uncached' attribute.
9524 (arc_print_operand): Print '.di' flag for uncached memory
9526 (arc_in_small_data_p): Do not consider for small data the uncached
9528 (arc_is_uncached_mem_p): New function.
9529 * config/arc/predicates.md (compact_store_memory_operand): Check
9530 for uncached memory accesses.
9531 (nonvol_nonimm_operand): Likewise.
9532 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9534 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9537 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9538 falign-loops=): Add Optimization flag.
9540 2018-01-30 Jeff Law <law@redhat.com>
9543 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9544 INT_REGISTERS_SAVED. Check it prior to calling
9545 get_scratch_register_on_entry.
9546 (ix86_adjust_stack_and_probe): Similarly.
9547 (ix86_emit_probe_stack_range): Similarly.
9548 (ix86_expand_prologue): Corresponding changes.
9550 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9553 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9554 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9556 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9559 * lra-constraints.c (curr_insn_transform): Process AND in the
9562 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9564 PR rtl-optimization/83986
9565 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9566 dependence against last_pending_memory_flush in addition to
9569 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9571 PR tree-optimization/81611
9572 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9575 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9578 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9581 2018-01-30 Richard Biener <rguenther@suse.de>
9582 Jakub Jelinek <jakub@redhat.com>
9584 PR tree-optimization/84111
9585 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9586 inner loops added during recursion, as they don't have up-to-date
9589 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9592 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9593 (can_inline_edge_by_limits_p): ... here.
9594 (can_early_inline_edge_p, check_callers,
9595 update_caller_keys, update_callee_keys, recursive_inlining,
9596 add_new_edges_to_heap, speculation_useful_p,
9597 inline_small_functions,
9598 inline_small_functions, flatten_function,
9599 inline_to_all_callers_1): Update.
9601 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9603 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9606 2018-01-30 Richard Biener <rguenther@suse.de>
9608 PR tree-optimization/83008
9609 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9610 invariant and constant vector uses in stmts when they need
9613 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9616 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9617 * configure: Regenerate.
9619 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9621 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9623 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9624 Use gen_rtx_REG rather than gen_lowpart.
9626 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9628 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9629 rather than 0 when creating partial subregs.
9631 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9633 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9636 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9639 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9640 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9641 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9642 flags. This restores the settings used before the 2017-07-24.
9643 Turning off pre increment/decrement/modify allows IVOPTS to
9644 optimize DF/SF loops where the index is an int.
9646 2018-01-29 Richard Biener <rguenther@suse.de>
9647 Kelvin Nilsen <kelvin@gcc.gnu.org>
9650 * tree-vect-stmts.c (vectorizable_call): Don't call
9651 targetm.vectorize_builtin_md_vectorized_function if callee is
9654 2018-01-22 Carl Love <cel@us.ibm.com>
9656 * doc/extend.tex: Fix typo in second arg in
9657 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9659 2018-01-29 Richard Biener <rguenther@suse.de>
9661 PR tree-optimization/84086
9662 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9663 (flush_ssaname_freelist): When SSA names were released reset
9664 the SCEV hash table.
9666 2018-01-29 Richard Biener <rguenther@suse.de>
9668 PR tree-optimization/84057
9669 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9670 removed paths when removing edges.
9672 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9674 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9675 -mfunction-return=@var{choice}.
9677 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9680 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9682 (layout::print_source_line): Likewise.
9683 (test_get_line_width_without_trailing_whitespace): Add test cases.
9685 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9688 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9691 2018-01-26 Jim Wilson <jimw@sifive.com>
9693 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9695 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9698 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9700 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9701 and CMP + SUB-immediate -> SUBS.
9703 2018-01-26 Martin Sebor <msebor@redhat.com>
9705 PR tree-optimization/83896
9706 * tree-ssa-strlen.c (get_string_len): Rename...
9707 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9708 Avoid assuming length is constant.
9709 (handle_char_store): Use HOST_WIDE_INT for string length.
9711 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9714 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9715 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9717 2018-01-26 Richard Biener <rguenther@suse.de>
9719 PR rtl-optimization/84003
9720 * dse.c (record_store): Only record redundant stores when
9721 the earlier store aliases at least all accesses the later one does.
9723 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9725 PR rtl-optimization/83985
9726 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9727 REG_CFA_RESTORE insns.
9728 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9729 REG_CFA_RESTORE insns here.
9732 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9733 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9735 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9737 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9738 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9739 (arc_init): Likewise.
9740 (arc_override_options): Likewise.
9741 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9743 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9745 * config/arc/arc.h (TARGET_DBNZ): Define.
9746 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9747 properly set the tune attribute.
9748 (dbnz): Use TARGET_DBNZ guard.
9749 * config/arc/arc.opt (mtune): Add core3 option.
9751 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9753 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9754 recognize new pic like addresses.
9755 (arc_delegitimize_address): Clean up.
9757 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9759 * config/arc/arc-arches.def: Option mrf16 valid for all
9761 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9762 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9763 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9764 * config/arc/arc-tables.opt: Regenerate.
9765 * config/arc/arc.c (arc_conditional_register_usage): Handle
9766 reduced register file case.
9767 (arc_file_start): Set must have build attributes.
9768 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9770 * config/arc/arc.opt (mrf16): Add new option.
9771 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9772 * config/arc/genmultilib.awk: Handle new mrf16 option.
9773 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9774 * config/arc/t-multilib: Regenerate.
9775 * doc/invoke.texi (ARC Options): Document mrf16 option.
9777 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9779 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9780 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9781 (arc_attribute_table): Add 'secure_call' attribute.
9782 (arc_print_operand): Print secure call operand.
9783 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9785 (arc_is_secure_call_p): New function. * config/arc/arc.md
9786 (call_i): Add support for sjli instruction.
9787 (call_value_i): Likewise.
9788 * config/arc/constraints.md (Csc): New constraint.
9790 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9791 John Eric Martin <John.Martin@emmicro-us.com>
9793 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9794 * config/arc/arc.c (_arc_jli_section): New struct.
9795 (arc_jli_section): New type.
9796 (rc_jli_sections): New static variable.
9797 (arc_handle_jli_attribute): New function.
9798 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9799 (arc_file_end): New function.
9800 (TARGET_ASM_FILE_END): Define.
9801 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9802 (arc_add_jli_section): New function.
9803 (jli_call_scan): Likewise.
9804 (arc_reorg): Call jli_call_scan.
9805 (arc_output_addsi): Remove 'S' from printing asm operand.
9806 (arc_is_jli_call_p): New function.
9807 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9809 (movhi_insn): Likewise.
9810 (movsi_insn): Likewise.
9811 (movsi_set_cc_insn): Likewise.
9812 (loadqi_update): Likewise.
9813 (load_zeroextendqisi_update): Likewise.
9814 (load_signextendqisi_update): Likewise.
9815 (loadhi_update): Likewise.
9816 (load_zeroextendhisi_update): Likewise.
9817 (load_signextendhisi_update): Likewise.
9818 (loadsi_update): Likewise.
9819 (loadsf_update): Likewise.
9820 (movsicc_insn): Likewise.
9821 (bset_insn): Likewise.
9822 (bxor_insn): Likewise.
9823 (bclr_insn): Likewise.
9824 (bmsk_insn): Likewise.
9825 (bicsi3_insn): Likewise.
9826 (cmpsi_cc_c_insn): Likewise.
9827 (movsi_ne): Likewise.
9828 (movsi_cond_exec): Likewise.
9829 (clrsbsi2): Likewise.
9840 (call_i): Remove 'S' asm letter, add jli instruction.
9841 (call_value_i): Likewise.
9842 * config/arc/arc.op (mjli-always): New option.
9843 * config/arc/constraints.md (Cji): New constraint.
9844 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9846 (subsf3_fpx): Likewise.
9847 (mulsf3_fpx): Likewise.
9848 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9850 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9852 * doc/invoke.texi (ARC): Document mjli-always option.
9854 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9856 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9857 avoid addition with 0 and use incw and decw where possible.
9859 2018-01-26 Richard Biener <rguenther@suse.de>
9861 PR tree-optimization/81082
9862 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9863 association if it requires casting to unsigned.
9864 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9865 from fold_plusminus_mult_expr to catch important cases late when
9866 range info is available.
9868 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9870 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9871 * configure.ac (hidden_linkonce): New test.
9872 * configure: Regenerate.
9873 * config.in: Regenerate.
9875 2018-01-26 Julia Koval <julia.koval@intel.com>
9877 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9878 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9879 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9880 _mm_mask_bitshuffle_epi64_mask): Fix type.
9881 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9882 USI_FTYPE_V4DI_V4DI_USI): Remove.
9883 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9884 __builtin_ia32_vpshufbitqmb256_mask,
9885 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9886 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9887 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9889 2018-01-26 Alan Modra <amodra@gmail.com>
9892 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9893 UNSPEC_VBPERMQ. Sort other unspecs.
9895 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9897 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9899 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9902 * predict.c (drop_profile): Do not push/pop cfun; update also
9904 (handle_missing_profiles): Fix logic looking for zero profiles.
9906 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9909 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9910 on functions with #pragma omp declare simd or functions with simd
9912 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9913 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9914 Remove trailing \n from warning_at calls.
9916 2018-01-25 Tom de Vries <tom@codesourcery.com>
9919 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9920 for neutered workers.
9922 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9925 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9926 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9928 2018-01-24 Jeff Law <law@redhat.com>
9931 * i386.c (get_probe_interval): Move to earlier point.
9932 (ix86_compute_frame_layout): If -fstack-clash-protection and
9933 the frame is larger than the probe interval, then use pushes
9934 to save registers rather than reg->mem moves.
9935 (ix86_expand_prologue): Remove conditional for int_registers_saved
9938 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9941 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9942 min/max for never referenced object.
9944 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9947 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9949 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9950 attributes from DECL_ATTRIBUTES (decl) without affecting
9951 DECL_ATTRIBUTES (current_function_decl).
9952 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9953 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9955 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9957 PR tree-optimization/83979
9958 * fold-const.c (fold_comparison): Use constant_boolean_node
9959 instead of boolean_{true,false}_node.
9961 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9963 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9966 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9968 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9969 Simplify the clause that sets the length attribute.
9970 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9971 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9972 clause that sets the length attribute.
9973 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9975 2018-01-24 Tom de Vries <tom@codesourcery.com>
9978 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9979 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9980 Add strict parameter.
9981 (prevent_branch_around_nothing): Insert dummy insn between branch to
9982 label and label with no ptx insn inbetween.
9983 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9985 2018-01-24 Tom de Vries <tom@codesourcery.com>
9988 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9989 for neutered threads in warp.
9990 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9992 2018-01-24 Richard Biener <rguenther@suse.de>
9994 PR tree-optimization/83176
9995 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9998 2018-01-24 Richard Biener <rguenther@suse.de>
10000 PR tree-optimization/82819
10001 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10002 code generating pluses that are no-ops in the target precision.
10004 2018-01-24 Richard Biener <rguenther@suse.de>
10006 PR middle-end/84000
10007 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10009 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10011 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10012 to merge probabilities.
10013 * predict.c (probably_never_executed): Also mark as cold functions
10014 with global 0 profile and guessed local profile.
10015 * profile-count.c (profile_probability::combine_with_count): New
10017 * profile-count.h (profile_probability::operator*,
10018 profile_probability::operator*=, profile_probability::operator/,
10019 profile_probability::operator/=): Reduce precision to adjusted
10020 and set value to guessed on contradictory divisions.
10021 (profile_probability::combine_with_freq): Remove.
10022 (profile_probability::combine_wiht_count): Declare.
10023 (profile_count::force_nonzero):: Set to adjusted.
10024 (profile_count::probability_in):: Set quality to adjusted.
10025 * tree-ssa-tail-merge.c (replace_block_by): Use
10026 combine_with_count.
10028 2018-01-23 Andrew Waterman <andrew@sifive.com>
10029 Jim Wilson <jimw@sifive.com>
10031 * config/riscv/riscv.c (riscv_stack_boundary): New.
10032 (riscv_option_override): Set riscv_stack_boundary. Handle
10033 riscv_preferred_stack_boundary_arg.
10034 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10035 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10036 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10037 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10038 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10039 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10041 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10044 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10045 of struct ix86_frame.
10046 (ix86_expand_epilogue): Likewise. Add a local variable for
10047 the reg_save_offset field in struct ix86_frame.
10049 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10051 PR tree-optimization/82604
10052 * tree-loop-distribution.c (enum partition_kind): New enum item
10053 PKIND_PARTIAL_MEMSET.
10054 (partition_builtin_p): Support above new enum item.
10055 (generate_code_for_partition): Ditto.
10056 (compute_access_range): Differentiate cases that equality can be
10057 proven at all loops, the innermost loops or no loops.
10058 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10059 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10060 (finalize_partitions, distribute_loop): Don't fuse partition of
10061 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10062 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10063 parloop is enabled.
10065 2018-01-23 Martin Liska <mliska@suse.cz>
10067 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10068 order to ignore the predictor.
10069 (PRED_POLYMORPHIC_CALL): Likewise.
10070 (PRED_RECURSIVE_CALL): Likewise.
10072 2018-01-23 Martin Liska <mliska@suse.cz>
10074 * tree-profile.c (tree_profiling): Print function header to
10075 aware reader which function we are working on.
10076 * value-prof.c (gimple_find_values_to_profile): Do not print
10077 not interesting value histograms.
10079 2018-01-23 Martin Liska <mliska@suse.cz>
10081 * profile-count.h (enum profile_quality): Add
10082 profile_uninitialized as the first value. Do not number values
10083 as they are zero based.
10084 (profile_count::verify): Update sanity check.
10085 (profile_probability::verify): Likewise.
10087 2018-01-23 Nathan Sidwell <nathan@acm.org>
10089 * doc/invoke.texi (ffor-scope): Deprecate.
10091 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10093 PR tree-optimization/83510
10094 * domwalk.c (set_all_edges_as_executable): New function.
10095 (dom_walker::dom_walker): Convert bool param
10096 "skip_unreachable_blocks" to enum reachability. Move setup of
10097 edge flags to set_all_edges_as_executable and only do it when
10098 reachability is REACHABLE_BLOCKS.
10099 * domwalk.h (enum dom_walker::reachability): New enum.
10100 (dom_walker::dom_walker): Convert bool param
10101 "skip_unreachable_blocks" to enum reachability.
10102 (set_all_edges_as_executable): New decl.
10103 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10104 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10106 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10107 but converting true to REACHABLE_BLOCKS.
10108 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10110 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10111 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10112 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10114 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10115 if check_all_array_refs will be called.
10117 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10119 * tree.c (selftest::test_location_wrappers): Add more test
10122 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10124 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10125 (selftest::test_bit_in_range): Likewise.
10127 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10130 * doc/sourcebuild.texi (vect_float): Say that the selector
10131 only describes the situation when -funsafe-math-optimizations is on.
10132 (vect_float_strict): Document.
10134 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10136 PR tree-optimization/83965
10137 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10138 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10139 instead of checking only for a reduction.
10140 (vect_recog_widen_sum_pattern): Likewise.
10142 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10144 * predict.c (probably_never_executed): Only use precise profile info.
10145 (compute_function_frequency): Skip after inlining hack since we now
10146 have quality checking.
10148 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10150 * profile-count.h (profile_probability::very_unlikely,
10151 profile_probability::unlikely, profile_probability::even): Set
10152 precision to guessed.
10154 2018-01-23 Richard Biener <rguenther@suse.de>
10156 PR tree-optimization/83963
10157 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10158 Properly terminate dominator walk when crossing the exit edge not
10159 when visiting its source block.
10161 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10164 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10165 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10167 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10169 PR tree-optimization/83957
10170 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10171 semicolon after for body surrounded by braces.
10173 PR tree-optimization/83081
10174 * profile-count.h (profile_probability::split): New method.
10175 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10176 Use profile_probability::split.
10177 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10178 when splitting a single conditional jump into 2.
10180 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10182 PR tree-optimization/69452
10183 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10186 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10188 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10189 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10190 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10192 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10194 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10195 * config/rl78/rl78.md (movdi): New define_expand.
10196 * config/rl78/rl78.c (rl78_split_movdi): New function.
10198 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10201 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10203 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10204 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10205 128-bit to produce an UNSPEC move to get the double word with the
10206 signbit and then a shift directly to do signbit.
10207 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10208 implementation with a new version that just does either a direct
10209 move or a regular move. Move memory interface to separate insns.
10210 Move insns so they are next to the expander.
10211 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10212 with signbit move. Split big and little endian case.
10213 (signbit<mode>2_dm_mem_le): Likewise.
10214 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10215 (signbit<mode>2_dm2): Likewise.
10217 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10219 * config/rl78/rl78.md (anddi3): New define_expand.
10221 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10223 * config/rl78/rl78.md (umindi3): New define_expand.
10225 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10227 * config/rl78/rl78.md (smindi3): New define_expand.
10229 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10231 * config/rl78/rl78.md (smaxdi3): New define_expand.
10233 2018-01-22 Carl Love <cel@us.ibm.com>
10235 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10236 LVX_V1TI): Add macro expansion.
10237 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10238 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10239 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10240 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10241 Change check to determine if the instruction is a byte reversing
10242 entry. Fix typo in comment.
10243 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10244 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10245 Add def_builtin calls for new builtins.
10246 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10247 Add define_insn expansion.
10249 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10251 * config/rl78/rl78.md (umaxdi3): New define_expand.
10253 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10255 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10256 for non-QImode registers.
10258 2018-01-22 Richard Biener <rguenther@suse.de>
10260 PR tree-optimization/83963
10261 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10262 including the loop exit block.
10263 (scop_detection::merge_sese): Likewise.
10264 (scop_detection::add_scop): Do it here instead.
10266 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10268 * doc/sourcebuild.texi (arm_softfloat): Document.
10270 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10273 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10274 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10275 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10277 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10278 David Edelsohn <dje.gcc@gmail.com>
10281 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10282 Change "crset eq" to "crset 2".
10283 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10284 (*call_indirect_aix<mode>_nospec): Likewise.
10285 (*call_value_indirect_aix<mode>_nospec): Likewise.
10286 (*call_indirect_elfv2<mode>_nospec): Likewise.
10287 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10288 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10289 change assembly output from . to $.
10290 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10291 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10292 (*tablejump<mode>_internal1_nospec): Likewise.
10294 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10297 * config/sh/sh_optimize_sett_clrt.cc:
10298 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10300 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10302 PR tree-optimization/83940
10303 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10304 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10305 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10306 use it to pass back the definition type.
10307 (vect_check_store_rhs): Likewise rhs_dt_out.
10308 (vect_build_gather_load_calls): Add a mask_dt argument and use
10309 it instead of a call to vect_is_simple_use.
10310 (vectorizable_store): Update calls to vect_check_load_store_mask
10311 and vect_check_store_rhs. Use the dt returned by the latter instead
10312 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10313 instead of calls to vect_is_simple_use. Pass the scalar rather
10314 than the vector operand to vect_is_simple_use when handling
10315 second and subsequent copies of an rhs value.
10316 (vectorizable_load): Update calls to vect_check_load_store_mask
10317 and vect_build_gather_load_calls. Use the cached mask_dt and
10318 gs_info.offset_dt instead of calls to vect_is_simple_use.
10320 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10322 PR middle-end/83945
10323 * tree-emutls.c: Include gimplify.h.
10324 (lower_emutls_2): New function.
10325 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10326 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10327 it before further processing.
10330 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10331 UINTVAL (trueop1) instead of INTVAL (op1).
10333 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10337 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10338 INCOMING_FRAME_SP_OFFSET if not defined.
10339 (scan_trace): Add ENTRY argument. If true and
10340 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10341 emit a note to adjust the CFA offset.
10342 (create_cfi_notes): Adjust scan_trace callers.
10343 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10344 INCOMING_FRAME_SP_OFFSET in the CIE.
10345 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10346 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10348 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10349 * doc/tm.texi: Regenerated.
10351 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10353 PR rtl-optimization/83147
10354 * lra-constraints.c (remove_inheritance_pseudos): Use
10355 lra_substitute_pseudo_within_insn.
10357 2018-01-19 Tom de Vries <tom@codesourcery.com>
10358 Cesar Philippidis <cesar@codesourcery.com>
10361 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10363 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10366 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10367 spaces for function labels.
10369 2018-01-19 Martin Liska <mliska@suse.cz>
10371 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10372 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10373 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10374 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10375 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10376 (PRED_CONST_RETURN): Change from 69 to 65.
10377 (PRED_NULL_RETURN): Change from 91 to 71.
10378 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10379 (PRED_LOOP_GUARD): Change from 66 to 73.
10381 2018-01-19 Martin Liska <mliska@suse.cz>
10383 * predict.c (predict_insn_def): Add new assert.
10384 (struct branch_predictor): Change type to signed integer.
10385 (test_prediction_value_range): Amend test to cover
10386 PROB_UNINITIALIZED.
10387 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10388 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10389 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10390 (PRED_LOOP_IV_COMPARE): Likewise.
10391 * predict.h (PROB_UNINITIALIZED): Define new constant.
10393 2018-01-19 Martin Liska <mliska@suse.cz>
10395 * predict.c (dump_prediction): Add new format for
10396 analyze_brprob.py script which is enabled with -details
10398 * profile-count.h (precise_p): New function.
10400 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10402 PR tree-optimization/83922
10403 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10404 there are no statements that need masking.
10405 (vect_active_double_reduction_p): New function.
10406 (vect_analyze_loop_operations): Use it when handling phis that
10407 are not in the loop header.
10409 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10411 PR tree-optimization/83914
10412 * tree-vect-loop.c (vectorizable_induction): Don't convert
10413 init_expr or apply the peeling adjustment for inductions
10414 that are nested within the vectorized loop.
10416 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10418 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10421 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10425 * function.h (gimplify_parameters): Add gimple_seq * argument.
10426 * function.c: Include gimple.h and options.h.
10427 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10428 for the added local temporaries if needed.
10429 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10430 if there are any parameter cleanups, wrap whole body into a
10431 try/finally with the cleanups.
10433 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10436 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10437 Use GET_MODE_CLASS for scalar floating point.
10439 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10443 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10444 Fix call of call_cgraph_insertion_hooks.
10446 2018-01-18 Martin Sebor <msebor@redhat.com>
10448 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10450 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10453 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10456 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10459 * common.opt: (-ffile-prefix-map): New option.
10460 * opts.c (common_handle_option): Defer it.
10461 * opts-global.c (handle_common_deferred_options): Handle it.
10462 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10463 * file-prefix-map.h: New file.
10464 (remap_debug_filename, add_debug_prefix_map): ...here.
10465 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10466 * final.c (debug_prefix_map, add_debug_prefix_map
10467 remap_debug_filename): Move to...
10468 * file-prefix-map.c: New file.
10469 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10470 generalize, get rid of alloca(), use strrchr() instead of strchr().
10471 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10472 Implement in terms of add_prefix_map().
10473 (remap_macro_filename, remap_debug_filename): Implement in term of
10475 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10476 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10477 * dbxout.c: Include file-prefix-map.h.
10478 * varasm.c: Likewise.
10479 * vmsdbgout.c: Likewise.
10480 * xcoffout.c: Likewise.
10481 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10482 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10483 * doc/invoke.texi (-ffile-prefix-map): Document.
10484 (-fdebug-prefix-map): Update description.
10486 2018-01-18 Martin Liska <mliska@suse.cz>
10488 * config/i386/i386.c (indirect_thunk_name): Document that also
10490 (output_indirect_thunk): Document why both instructions
10491 (pause and lfence) are generated.
10493 2018-01-18 Richard Biener <rguenther@suse.de>
10495 PR tree-optimization/83887
10496 * graphite-scop-detection.c
10497 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10498 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10499 (scop_detection::merge_sese): Re-implement with a flood-fill
10500 algorithm that properly finds a SESE region if it exists.
10502 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10505 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10506 pointer_diff optimizations use view_convert instead of convert.
10508 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10510 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10511 Generate different code for -mno-speculate-indirect-jumps.
10512 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10513 (*call_indirect_aix<mode>): Disable for
10514 -mno-speculate-indirect-jumps.
10515 (*call_indirect_aix<mode>_nospec): New define_insn.
10516 (*call_value_indirect_aix<mode>): Disable for
10517 -mno-speculate-indirect-jumps.
10518 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10519 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10520 -mno-speculate-indirect-jumps.
10521 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10523 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10525 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10526 long double type, set the flags for noting the default long double
10527 type, even if we don't pass or return a long double type.
10529 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10532 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10535 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10537 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10538 support for merge[hl].
10539 (fold_mergehl_helper): New helper function.
10540 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10541 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10542 (altivec_vmrglw_direct): Add xxmrglw insn.
10544 2018-01-17 Andrew Waterman <andrew@sifive.com>
10546 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10547 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10549 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10552 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10553 call the lto_location_cache before reading the
10554 DECL_SOURCE_LOCATION of the types.
10556 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10557 Richard Sandiford <richard.sandiford@linaro.org>
10559 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10560 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10561 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10562 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10563 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10565 * config/aarch64/constraints.md (aarch64_movti_operand):
10567 * config/aarch64/predicates.md (Uti): Add new constraint.
10569 2018-01-17 Carl Love <cel@us.ibm.com>
10571 * config/rs6000/vsx.md (define_expand xl_len_r,
10572 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10573 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10575 (define_expand, define_insn): Move the shift left from the
10576 define_insn to the define_expand for lxvl and stxvl instructions.
10577 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10578 and XL_LEN_R definitions to PURE.
10580 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10582 * config/i386/i386.c (indirect_thunk_name): Declare regno
10583 as unsigned int. Compare regno with INVALID_REGNUM.
10584 (output_indirect_thunk): Ditto.
10585 (output_indirect_thunk_function): Ditto.
10586 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10587 in the call to output_indirect_thunk_function.
10589 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10591 PR middle-end/83884
10592 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10593 rather than the size of inner_type to determine the stack slot size
10594 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10596 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10599 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10602 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10604 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10605 endian Linux systems to optionally enable multilibs for selecting
10606 the long double type if the user configured an explicit type.
10607 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10608 have no long double multilibs if not defined.
10609 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10610 warn if the user used -mabi={ieee,ibm}longdouble and we built
10611 multilibs for long double.
10612 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10613 appropriate multilib option.
10614 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10616 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10617 for building long double multilibs.
10618 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10620 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10622 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10625 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10627 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10630 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10633 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10636 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10638 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10639 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10641 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10643 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10644 different rtl trees depending on TARGET_64BIT.
10645 (rs6000_gen_lvx): Likewise.
10647 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10649 * config/visium/visium.md (nop): Tweak comment.
10650 (hazard_nop): Likewise.
10652 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10654 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10655 -mspeculate-indirect-jumps.
10656 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10657 for -mno-speculate-indirect-jumps.
10658 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10659 (*call_value_indirect_elfv2<mode>): Disable for
10660 -mno-speculate-indirect-jumps.
10661 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10662 (indirect_jump): Emit different RTL for
10663 -mno-speculate-indirect-jumps.
10664 (*indirect_jump<mode>): Disable for
10665 -mno-speculate-indirect-jumps.
10666 (*indirect_jump<mode>_nospec): New define_insn.
10667 (tablejump): Emit different RTL for
10668 -mno-speculate-indirect-jumps.
10669 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10670 (tablejumpsi_nospec): New define_expand.
10671 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10672 (tablejumpdi_nospec): New define_expand.
10673 (*tablejump<mode>_internal1): Disable for
10674 -mno-speculate-indirect-jumps.
10675 (*tablejump<mode>_internal1_nospec): New define_insn.
10676 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10679 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10681 * caller-save.c (insert_save): Drop unnecessary parameter. All
10684 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10685 Richard Biener <rguenth@suse.de>
10688 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10689 return early, inline manually is_gimple_sizepos. Make sure if we
10690 call gimplify_expr we don't end up with a gimple constant.
10691 * tree.c (variably_modified_type_p): Don't return true for
10692 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10693 * gimplify.h (is_gimple_sizepos): Remove.
10695 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10697 PR tree-optimization/83857
10698 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10699 vectorizable_live_operation for pure SLP statements.
10700 (vectorizable_live_operation): Handle PHIs.
10702 2018-01-16 Richard Biener <rguenther@suse.de>
10704 PR tree-optimization/83867
10705 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10706 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10708 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10711 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10712 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10713 If off is not INTEGER_CST, issue a may not be aligned warning
10714 rather than isn't aligned. Use isn%'t rather than isn't.
10715 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10717 <case MULT_EXPR>: Improve the case when bottom and one of the
10718 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10719 operand, in that case check if the other operand is multiple of
10720 bottom divided by the INTEGER_CST operand.
10722 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10725 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10726 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10727 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10728 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10729 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10730 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10731 (pa_function_arg_size): New function.
10733 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10735 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10736 in a separate statement.
10738 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10740 PR tree-optimization/83847
10741 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10742 group gathers and scatters.
10744 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10746 PR rtl-optimization/86620
10747 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10749 PR rtl-optimization/83213
10750 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10751 to last if both are JUMP_INSNs.
10753 PR tree-optimization/83843
10754 * gimple-ssa-store-merging.c
10755 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10756 store_immediate_info for bswap/nop orig_stores.
10758 2018-01-15 Andrew Waterman <andrew@sifive.com>
10760 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10762 <UDIV>: Increase cost if !TARGET_DIV.
10764 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10766 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10767 (define_attr "cr_logical_3op"): New.
10768 (cceq_ior_compare): Adjust.
10769 (cceq_ior_compare_complement): Adjust.
10770 (*cceq_rev_compare): Adjust.
10771 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10772 (is_cracked_insn): Adjust.
10773 (insn_must_be_first_in_group): Adjust.
10774 * config/rs6000/40x.md: Adjust.
10775 * config/rs6000/440.md: Adjust.
10776 * config/rs6000/476.md: Adjust.
10777 * config/rs6000/601.md: Adjust.
10778 * config/rs6000/603.md: Adjust.
10779 * config/rs6000/6xx.md: Adjust.
10780 * config/rs6000/7450.md: Adjust.
10781 * config/rs6000/7xx.md: Adjust.
10782 * config/rs6000/8540.md: Adjust.
10783 * config/rs6000/cell.md: Adjust.
10784 * config/rs6000/e300c2c3.md: Adjust.
10785 * config/rs6000/e500mc.md: Adjust.
10786 * config/rs6000/e500mc64.md: Adjust.
10787 * config/rs6000/e5500.md: Adjust.
10788 * config/rs6000/e6500.md: Adjust.
10789 * config/rs6000/mpc.md: Adjust.
10790 * config/rs6000/power4.md: Adjust.
10791 * config/rs6000/power5.md: Adjust.
10792 * config/rs6000/power6.md: Adjust.
10793 * config/rs6000/power7.md: Adjust.
10794 * config/rs6000/power8.md: Adjust.
10795 * config/rs6000/power9.md: Adjust.
10796 * config/rs6000/rs64.md: Adjust.
10797 * config/rs6000/titan.md: Adjust.
10799 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10801 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10802 ix86_indirect_branch_register logic.
10804 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10806 * config/i386/constraints.md (Bs): Update
10807 ix86_indirect_branch_register check. Don't check
10808 ix86_indirect_branch_register with GOT_memory_operand.
10810 * config/i386/predicates.md (GOT_memory_operand): Don't check
10811 ix86_indirect_branch_register here.
10812 (GOT32_symbol_operand): Likewise.
10814 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10816 * config/i386/predicates.md (constant_call_address_operand):
10817 Rewrite ix86_indirect_branch_register logic.
10818 (sibcall_insn_operand): Likewise.
10820 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10822 * config/i386/constraints.md (Bs): Replace
10823 ix86_indirect_branch_thunk_register with
10824 ix86_indirect_branch_register.
10826 * config/i386/i386.md (indirect_jump): Likewise.
10827 (tablejump): Likewise.
10828 (*sibcall_memory): Likewise.
10829 (*sibcall_value_memory): Likewise.
10830 Peepholes of indirect call and jump via memory: Likewise.
10831 * config/i386/i386.opt: Likewise.
10832 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10833 (GOT_memory_operand): Likewise.
10834 (call_insn_operand): Likewise.
10835 (sibcall_insn_operand): Likewise.
10836 (GOT32_symbol_operand): Likewise.
10838 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10840 PR middle-end/83837
10841 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10842 type rather than type addr's type points to.
10843 (expand_omp_atomic_mutex): Likewise.
10844 (expand_omp_atomic): Likewise.
10846 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10849 * config/i386/i386.c (output_indirect_thunk_function): Use
10850 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10851 for __x86_return_thunk.
10853 2018-01-15 Richard Biener <rguenther@suse.de>
10855 PR middle-end/83850
10856 * expmed.c (extract_bit_field_1): Fix typo.
10858 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10861 * config/arm/iterators.md (VF): New mode iterator.
10862 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10863 Remove integer-related logic from pattern.
10864 (neon_vabd<mode>_3): Likewise.
10866 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10868 PR middle-end/82694
10869 * common.opt (fstrict-overflow): No longer an alias.
10870 (fwrapv-pointer): New option.
10871 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10872 also for pointer types based on flag_wrapv_pointer.
10873 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10874 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10875 opts->x_flag_wrapv got set.
10876 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10877 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10878 POINTER_TYPE_OVERFLOW_UNDEFINED.
10879 * match.pd: Likewise in address comparison pattern.
10880 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10882 2018-01-15 Richard Biener <rguenther@suse.de>
10885 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10886 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10887 Reset type names to their identifier if their TYPE_DECL doesn't
10888 have linkage (and thus is used for ODR and devirt).
10889 (save_debug_info_for_decl): Remove.
10890 (save_debug_info_for_type): Likewise.
10891 (add_tree_to_fld_list): Adjust.
10892 * tree-pretty-print.c (dump_generic_node): Make dumping of
10893 type names more robust.
10895 2018-01-15 Richard Biener <rguenther@suse.de>
10897 * BASE-VER: Bump to 8.0.1.
10899 2018-01-14 Martin Sebor <msebor@redhat.com>
10902 * builtins.c (check_access): Avoid warning when the no-warning bit
10905 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10907 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10908 * ira-color (allocno_hard_regs_compare): Likewise.
10910 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10913 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10914 Use .pushsection/.popsection.
10916 2018-01-14 Martin Sebor <msebor@redhat.com>
10919 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10921 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10923 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10924 entry from extra_headers.
10925 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10926 extra_headers, make the list bitwise identical to the i?86-*-* one.
10928 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10930 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10931 -mcmodel=large with -mindirect-branch=thunk,
10932 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10933 -mfunction-return=thunk-extern.
10934 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10935 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10936 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10938 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10940 * config/i386/i386.c (print_reg): Print the name of the full
10941 integer register without '%'.
10942 (ix86_print_operand): Handle 'V'.
10943 * doc/extend.texi: Document 'V' modifier.
10945 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10947 * config/i386/constraints.md (Bs): Disallow memory operand for
10948 -mindirect-branch-register.
10950 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10951 (GOT_memory_operand): Likewise.
10952 (call_insn_operand): Likewise.
10953 (sibcall_insn_operand): Likewise.
10954 (GOT32_symbol_operand): Likewise.
10955 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10956 for -mindirect-branch-register.
10957 (tablejump): Likewise.
10958 (*sibcall_memory): Likewise.
10959 (*sibcall_value_memory): Likewise.
10960 Disallow peepholes of indirect call and jump via memory for
10961 -mindirect-branch-register.
10962 (*call_pop): Replace m with Bw.
10963 (*call_value_pop): Likewise.
10964 (*sibcall_pop_memory): Replace m with Bs.
10965 * config/i386/i386.opt (mindirect-branch-register): New option.
10966 * doc/invoke.texi: Document -mindirect-branch-register option.
10968 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10970 * config/i386/i386-protos.h (ix86_output_function_return): New.
10971 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10972 set function_return_type.
10973 (indirect_thunk_name): Add ret_p to indicate thunk for function
10975 (output_indirect_thunk_function): Pass false to
10976 indirect_thunk_name.
10977 (ix86_output_indirect_branch_via_reg): Likewise.
10978 (ix86_output_indirect_branch_via_push): Likewise.
10979 (output_indirect_thunk_function): Create alias for function
10980 return thunk if regno < 0.
10981 (ix86_output_function_return): New function.
10982 (ix86_handle_fndecl_attribute): Handle function_return.
10983 (ix86_attribute_table): Add function_return.
10984 * config/i386/i386.h (machine_function): Add
10985 function_return_type.
10986 * config/i386/i386.md (simple_return_internal): Use
10987 ix86_output_function_return.
10988 (simple_return_internal_long): Likewise.
10989 * config/i386/i386.opt (mfunction-return=): New option.
10990 (indirect_branch): Mention -mfunction-return=.
10991 * doc/extend.texi: Document function_return function attribute.
10992 * doc/invoke.texi: Document -mfunction-return= option.
10994 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10996 * config/i386/i386-opts.h (indirect_branch): New.
10997 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10998 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10999 with local indirect jump when converting indirect call and jump.
11000 (ix86_set_indirect_branch_type): New.
11001 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11002 (indirectlabelno): New.
11003 (indirect_thunk_needed): Likewise.
11004 (indirect_thunk_bnd_needed): Likewise.
11005 (indirect_thunks_used): Likewise.
11006 (indirect_thunks_bnd_used): Likewise.
11007 (INDIRECT_LABEL): Likewise.
11008 (indirect_thunk_name): Likewise.
11009 (output_indirect_thunk): Likewise.
11010 (output_indirect_thunk_function): Likewise.
11011 (ix86_output_indirect_branch_via_reg): Likewise.
11012 (ix86_output_indirect_branch_via_push): Likewise.
11013 (ix86_output_indirect_branch): Likewise.
11014 (ix86_output_indirect_jmp): Likewise.
11015 (ix86_code_end): Call output_indirect_thunk_function if needed.
11016 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11018 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11019 (ix86_attribute_table): Add indirect_branch.
11020 * config/i386/i386.h (machine_function): Add indirect_branch_type
11021 and has_local_indirect_jump.
11022 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11024 (tablejump): Likewise.
11025 (*indirect_jump): Use ix86_output_indirect_jmp.
11026 (*tablejump_1): Likewise.
11027 (simple_return_indirect_internal): Likewise.
11028 * config/i386/i386.opt (mindirect-branch=): New option.
11029 (indirect_branch): New.
11032 (thunk-inline): Likewise.
11033 (thunk-extern): Likewise.
11034 * doc/extend.texi: Document indirect_branch function attribute.
11035 * doc/invoke.texi: Document -mindirect-branch= option.
11037 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11040 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11042 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11044 * ipa-inline.c (want_inline_small_function_p): Return false if
11045 inlining has already failed with CIF_FINAL_ERROR.
11046 (update_caller_keys): Call want_inline_small_function_p before
11048 (update_callee_keys): Likewise.
11050 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11052 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11054 (rs6000_quadword_masked_address_p): Likewise.
11055 (quad_aligned_load_p): Likewise.
11056 (quad_aligned_store_p): Likewise.
11057 (const_load_sequence_p): Add comment to describe the outer-most loop.
11058 (mimic_memory_attributes_and_flags): New function.
11059 (rs6000_gen_stvx): Likewise.
11060 (replace_swapped_aligned_store): Likewise.
11061 (rs6000_gen_lvx): Likewise.
11062 (replace_swapped_aligned_load): Likewise.
11063 (replace_swapped_load_constant): Capitalize argument name in
11064 comment describing this function.
11065 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11066 and stores that access quad-word aligned addresses and replace
11067 with stvx or lvx instructions when appropriate.
11068 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11069 New function prototype.
11070 (rs6000_quadword_masked_address_p): Likewise.
11071 (rs6000_gen_lvx): Likewise.
11072 (rs6000_gen_stvx): Likewise.
11073 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11074 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11075 when memory address is aligned.
11076 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11077 this split to select lvx instruction when memory address is aligned.
11078 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11079 instruction when memory address is aligned.
11080 (*vsx_le_perm_load_v16qi): Likewise.
11081 (four unnamed splitters): Modify to select the stvx instruction
11082 when memory is aligned.
11084 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11086 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11087 which appears in the queue multiple times.
11089 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11090 Alan Hayward <alan.hayward@arm.com>
11091 David Sherwood <david.sherwood@arm.com>
11093 * tree-vectorizer.h (vec_lower_bound): New structure.
11094 (_loop_vec_info): Add check_nonzero and lower_bounds.
11095 (LOOP_VINFO_CHECK_NONZERO): New macro.
11096 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11097 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11098 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11099 fields. Make seg_len the distance travelled, not including the
11101 (dr_direction_indicator): Declare.
11102 (dr_zero_step_indicator): Likewise.
11103 (dr_known_forward_stride_p): Likewise.
11104 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11106 (runtime_alias_check_p): Allow runtime alias checks with
11108 (operator ==): Compare access_size and align.
11109 (prune_runtime_alias_test_list): Rework for new distinction between
11110 the access_size and seg_len.
11111 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11113 (get_segment_min_max): New function.
11114 (create_intersect_range_checks): Use it.
11115 (dr_step_indicator): New function.
11116 (dr_direction_indicator): Likewise.
11117 (dr_zero_step_indicator): Likewise.
11118 (dr_known_forward_stride_p): Likewise.
11119 * tree-loop-distribution.c (data_ref_segment_size): Return
11120 DR_STEP * (niters - 1).
11121 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11123 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11124 (vect_preserves_scalar_order_p): New function, split out from...
11125 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11126 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11127 (vect_vfa_access_size): New function.
11128 (vect_vfa_align): Likewise.
11129 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11130 (dump_lower_bound): New function.
11131 (vect_check_lower_bound): Likewise.
11132 (vect_small_gap_p): Likewise.
11133 (vectorizable_with_step_bound_p): Likewise.
11134 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11135 depencies if the vectorization factor is 1. Convert the checks
11136 for nonzero steps into checks on the bounds of DR_STEP. Try using
11137 a bunds check for variable steps if the minimum required step is
11138 relatively small. Update calls to the dr_with_seg_len
11139 constructor and to vect_compile_time_alias.
11140 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11142 (vect_loop_versioning): Call it.
11143 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11145 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11148 Alan Hayward <alan.hayward@arm.com>
11149 David Sherwood <david.sherwood@arm.com>
11151 * doc/sourcebuild.texi (vect_scatter_store): Document.
11152 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11154 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11156 * genopinit.c (main): Add supports_vec_scatter_store and
11157 supports_vec_scatter_store_cached to target_optabs.
11158 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11159 IFN_MASK_SCATTER_STORE.
11160 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11162 * internal-fn.h (internal_store_fn_p): Declare.
11163 (internal_fn_stored_value_index): Likewise.
11164 * internal-fn.c (scatter_store_direct): New macro.
11165 (expand_scatter_store_optab_fn): New function.
11166 (direct_scatter_store_optab_supported_p): New macro.
11167 (internal_store_fn_p): New function.
11168 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11169 IFN_MASK_SCATTER_STORE.
11170 (internal_fn_mask_index): Likewise.
11171 (internal_fn_stored_value_index): New function.
11172 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11173 for scatter stores.
11174 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11175 * optabs-query.c (supports_vec_scatter_store_p): New function.
11176 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11177 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11178 true for scatter stores.
11179 (vect_gather_scatter_fn_p): Handle scatter stores too.
11180 (vect_check_gather_scatter): Consider using scatter stores if
11181 supports_vec_scatter_store_p.
11182 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11183 scatter stores too.
11184 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11185 internal_fn_stored_value_index.
11186 (check_load_store_masking): Handle scatter stores too.
11187 (vect_get_store_rhs): Make public.
11188 (vectorizable_call): Use internal_store_fn_p.
11189 (vectorizable_store): Handle scatter store internal functions.
11190 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11191 when deciding whether the end of the group has been reached.
11192 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11193 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11194 (mask_scatter_store<mode>): New insns.
11196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11197 Alan Hayward <alan.hayward@arm.com>
11198 David Sherwood <david.sherwood@arm.com>
11200 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11201 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11202 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11204 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11205 Use vect_truncate_gather_scatter_offset if we can't treat the
11206 operation as a normal gather load or scatter store.
11207 (get_group_load_store_type): Take the gather_scatter_info
11208 as argument. Try using a gather load or scatter store for
11209 single-element groups.
11210 (get_load_store_type): Update calls to get_group_load_store_type
11211 and vect_use_strided_gather_scatters_p.
11213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11214 Alan Hayward <alan.hayward@arm.com>
11215 David Sherwood <david.sherwood@arm.com>
11217 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11218 optional tree argument.
11219 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11221 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11222 but continue to use the current value as a fallback.
11223 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11224 to compare the updates.
11225 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11226 (get_load_store_type): Use it when handling a strided access.
11227 (vect_get_strided_load_store_ops): New function.
11228 (vect_get_data_ptr_increment): Likewise.
11229 (vectorizable_load): Handle strided gather loads. Always pass
11230 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11233 Alan Hayward <alan.hayward@arm.com>
11234 David Sherwood <david.sherwood@arm.com>
11236 * doc/md.texi (gather_load@var{m}): Document.
11237 (mask_gather_load@var{m}): Likewise.
11238 * genopinit.c (main): Add supports_vec_gather_load and
11239 supports_vec_gather_load_cached to target_optabs.
11240 * optabs-tree.c (init_tree_optimization_optabs): Use
11241 ggc_cleared_alloc to allocate target_optabs.
11242 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11243 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11245 * internal-fn.h (internal_load_fn_p): Declare.
11246 (internal_gather_scatter_fn_p): Likewise.
11247 (internal_fn_mask_index): Likewise.
11248 (internal_gather_scatter_fn_supported_p): Likewise.
11249 * internal-fn.c (gather_load_direct): New macro.
11250 (expand_gather_load_optab_fn): New function.
11251 (direct_gather_load_optab_supported_p): New macro.
11252 (direct_internal_fn_optab): New function.
11253 (internal_load_fn_p): Likewise.
11254 (internal_gather_scatter_fn_p): Likewise.
11255 (internal_fn_mask_index): Likewise.
11256 (internal_gather_scatter_fn_supported_p): Likewise.
11257 * optabs-query.c (supports_at_least_one_mode_p): New function.
11258 (supports_vec_gather_load_p): Likewise.
11259 * optabs-query.h (supports_vec_gather_load_p): Declare.
11260 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11261 and memory_type field.
11262 (NUM_PATTERNS): Bump to 15.
11263 * tree-vect-data-refs.c: Include internal-fn.h.
11264 (vect_gather_scatter_fn_p): New function.
11265 (vect_describe_gather_scatter_call): Likewise.
11266 (vect_check_gather_scatter): Try using internal functions for
11267 gather loads. Recognize existing calls to a gather load function.
11268 (vect_analyze_data_refs): Consider using gather loads if
11269 supports_vec_gather_load_p.
11270 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11271 (vect_get_gather_scatter_offset_type): Likewise.
11272 (vect_convert_mask_for_vectype): Likewise.
11273 (vect_add_conversion_to_patterm): Likewise.
11274 (vect_try_gather_scatter_pattern): Likewise.
11275 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11276 (vect_vect_recog_func_ptrs): Add it.
11277 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11278 internal_fn_mask_index and internal_gather_scatter_fn_p.
11279 (check_load_store_masking): Take the gather_scatter_info as an
11280 argument and handle gather loads.
11281 (vect_get_gather_scatter_ops): New function.
11282 (vectorizable_call): Check internal_load_fn_p.
11283 (vectorizable_load): Likewise. Handle gather load internal
11285 (vectorizable_store): Update call to check_load_store_masking.
11286 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11287 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11288 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11289 (aarch64_gather_scale_operand_d): New predicates.
11290 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11291 (mask_gather_load<mode>): New insns.
11293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11294 Alan Hayward <alan.hayward@arm.com>
11295 David Sherwood <david.sherwood@arm.com>
11297 * optabs.def (fold_left_plus_optab): New optab.
11298 * doc/md.texi (fold_left_plus_@var{m}): Document.
11299 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11300 * internal-fn.c (fold_left_direct): Define.
11301 (expand_fold_left_optab_fn): Likewise.
11302 (direct_fold_left_optab_supported_p): Likewise.
11303 * fold-const-call.c (fold_const_fold_left): New function.
11304 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11305 * tree-parloops.c (valid_reduction_p): New function.
11306 (gather_scalar_reductions): Use it.
11307 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11308 (vect_finish_replace_stmt): Declare.
11309 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11310 (needs_fold_left_reduction_p): New function, split out from...
11311 (vect_is_simple_reduction): ...here. Accept reductions that
11312 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11313 (vect_force_simple_reduction): Also store the reduction type in
11314 the assignment's STMT_VINFO_REDUC_TYPE.
11315 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11316 (merge_with_identity): New function.
11317 (vect_expand_fold_left): Likewise.
11318 (vectorize_fold_left_reduction): Likewise.
11319 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11320 scalar phi in place for it. Check for target support and reject
11321 cases that would reassociate the operation. Defer the transform
11322 phase to vectorize_fold_left_reduction.
11323 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11324 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11325 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11329 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11330 call to ifc_temp_var.
11332 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11333 Alan Hayward <alan.hayward@arm.com>
11334 David Sherwood <david.sherwood@arm.com>
11336 * target.def (legitimize_address_displacement): Take the original
11337 offset as a poly_int.
11338 * targhooks.h (default_legitimize_address_displacement): Update
11340 * targhooks.c (default_legitimize_address_displacement): Likewise.
11341 * doc/tm.texi: Regenerate.
11342 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11343 as an argument, moving assert of ad->disp == ad->disp_term to...
11344 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11345 Try calling targetm.legitimize_address_displacement before expanding
11346 the address rather than afterwards, and adjust for the new interface.
11347 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11348 Match the new hook interface. Handle SVE addresses.
11349 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11350 new hook interface.
11352 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11354 * Makefile.in (OBJS): Add early-remat.o.
11355 * target.def (select_early_remat_modes): New hook.
11356 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11357 * doc/tm.texi: Regenerate.
11358 * targhooks.h (default_select_early_remat_modes): Declare.
11359 * targhooks.c (default_select_early_remat_modes): New function.
11360 * timevar.def (TV_EARLY_REMAT): New timevar.
11361 * passes.def (pass_early_remat): New pass.
11362 * tree-pass.h (make_pass_early_remat): Declare.
11363 * early-remat.c: New file.
11364 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11366 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11368 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11369 Alan Hayward <alan.hayward@arm.com>
11370 David Sherwood <david.sherwood@arm.com>
11372 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11373 vfm1 with a bound_epilog parameter.
11374 (vect_do_peeling): Update calls accordingly, and move the prologue
11375 call earlier in the function. Treat the base bound_epilog as 0 for
11376 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11377 this base when peeling for gaps.
11378 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11379 with fully-masked loops.
11380 (vect_estimate_min_profitable_iters): Handle the single peeled
11381 iteration in that case.
11383 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11384 Alan Hayward <alan.hayward@arm.com>
11385 David Sherwood <david.sherwood@arm.com>
11387 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11388 single-element interleaving even if the size is not a power of 2.
11389 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11390 accesses for single-element interleaving if the group size is
11393 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11394 Alan Hayward <alan.hayward@arm.com>
11395 David Sherwood <david.sherwood@arm.com>
11397 * doc/md.texi (fold_extract_last_@var{m}): Document.
11398 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11399 * optabs.def (fold_extract_last_optab): New optab.
11400 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11401 * internal-fn.c (fold_extract_direct): New macro.
11402 (expand_fold_extract_optab_fn): Likewise.
11403 (direct_fold_extract_optab_supported_p): Likewise.
11404 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11405 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11406 EXTRACT_LAST_REDUCTION.
11407 (get_initial_def_for_reduction): Do not create an initial vector
11408 for EXTRACT_LAST_REDUCTION reductions.
11409 (vectorizable_reduction): Leave the scalar phi in place for
11410 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11411 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11412 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11413 transform phase to vectorizable_condition.
11414 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11416 (vect_finish_stmt_generation): ...here.
11417 (vect_finish_replace_stmt): New function.
11418 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11419 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11421 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11424 Alan Hayward <alan.hayward@arm.com>
11425 David Sherwood <david.sherwood@arm.com>
11427 * doc/md.texi (extract_last_@var{m}): Document.
11428 * optabs.def (extract_last_optab): New optab.
11429 * internal-fn.def (EXTRACT_LAST): New internal function.
11430 * internal-fn.c (cond_unary_direct): New macro.
11431 (expand_cond_unary_optab_fn): Likewise.
11432 (direct_cond_unary_optab_supported_p): Likewise.
11433 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11434 loops using EXTRACT_LAST.
11435 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11436 (extract_last_<mode>): ...this optab.
11437 (vec_extract<mode><Vel>): Update accordingly.
11439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11440 Alan Hayward <alan.hayward@arm.com>
11441 David Sherwood <david.sherwood@arm.com>
11443 * target.def (empty_mask_is_expensive): New hook.
11444 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11445 * doc/tm.texi: Regenerate.
11446 * targhooks.h (default_empty_mask_is_expensive): Declare.
11447 * targhooks.c (default_empty_mask_is_expensive): New function.
11448 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11449 if the target says that empty masks are expensive.
11450 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11452 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11454 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11455 Alan Hayward <alan.hayward@arm.com>
11456 David Sherwood <david.sherwood@arm.com>
11458 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11459 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11460 (vect_use_loop_mask_for_alignment_p): New function.
11461 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11462 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11463 niters_skip argument. Make sure that the first niters_skip elements
11464 of the first iteration are inactive.
11465 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11466 Update call to vect_set_loop_masks_directly.
11467 (get_misalign_in_elems): New function, split out from...
11468 (vect_gen_prolog_loop_niters): ...here.
11469 (vect_update_init_of_dr): Take a code argument that specifies whether
11470 the adjustment should be added or subtracted.
11471 (vect_update_init_of_drs): Likewise.
11472 (vect_prepare_for_masked_peels): New function.
11473 (vect_do_peeling): Skip prologue peeling if we're using a mask
11474 instead. Update call to vect_update_inits_of_drs.
11475 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11477 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11478 alignment. Do not include the number of peeled iterations in
11479 the minimum threshold in that case.
11480 (vectorizable_induction): Adjust the start value down by
11481 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11482 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11483 Take the number of skipped iterations into account when calculating
11485 * tree-vect-stmts.c (vect_gen_while_not): New function.
11487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11488 Alan Hayward <alan.hayward@arm.com>
11489 David Sherwood <david.sherwood@arm.com>
11491 * doc/sourcebuild.texi (vect_fully_masked): Document.
11492 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11493 default value to 0.
11494 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11496 (vect_analyze_loop_2): ...here. Don't check the vectorization
11497 factor against the number of loop iterations if the loop is
11500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11501 Alan Hayward <alan.hayward@arm.com>
11502 David Sherwood <david.sherwood@arm.com>
11504 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11505 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11506 (dump_groups): Update accordingly.
11507 (iv_use::mem_type): New member variable.
11508 (address_p): New function.
11509 (record_use): Add a mem_type argument and initialize the new
11511 (record_group_use): Add a mem_type argument. Use address_p.
11512 Remove obsolete null checks of base_object. Update call to record_use.
11513 (find_interesting_uses_op): Update call to record_group_use.
11514 (find_interesting_uses_cond): Likewise.
11515 (find_interesting_uses_address): Likewise.
11516 (get_mem_type_for_internal_fn): New function.
11517 (find_address_like_use): Likewise.
11518 (find_interesting_uses_stmt): Try find_address_like_use before
11519 calling find_interesting_uses_op.
11520 (addr_offset_valid_p): Use the iv mem_type field as the type
11521 of the addressed memory.
11522 (add_autoinc_candidates): Likewise.
11523 (get_address_cost): Likewise.
11524 (split_small_address_groups_p): Use address_p.
11525 (split_address_groups): Likewise.
11526 (add_iv_candidate_for_use): Likewise.
11527 (autoinc_possible_for_pair): Likewise.
11528 (rewrite_groups): Likewise.
11529 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11530 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11531 (get_alias_ptr_type_for_ptr_address): New function.
11532 (rewrite_use_address): Rewrite address uses in calls that were
11533 identified by find_address_like_use.
11535 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11536 Alan Hayward <alan.hayward@arm.com>
11537 David Sherwood <david.sherwood@arm.com>
11539 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11541 * gimple-expr.h (is_gimple_addressable: Likewise.
11542 * gimple-expr.c (is_gimple_address): Likewise.
11543 * internal-fn.c (expand_call_mem_ref): New function.
11544 (expand_mask_load_optab_fn): Use it.
11545 (expand_mask_store_optab_fn): Likewise.
11547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11548 Alan Hayward <alan.hayward@arm.com>
11549 David Sherwood <david.sherwood@arm.com>
11551 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11552 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11553 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11554 (cond_umax@var{mode}): Document.
11555 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11556 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11557 (cond_umin_optab, cond_umax_optab): New optabs.
11558 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11559 (COND_IOR, COND_XOR): New internal functions.
11560 * internal-fn.h (get_conditional_internal_fn): Declare.
11561 * internal-fn.c (cond_binary_direct): New macro.
11562 (expand_cond_binary_optab_fn): Likewise.
11563 (direct_cond_binary_optab_supported_p): Likewise.
11564 (get_conditional_internal_fn): New function.
11565 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11566 Cope with reduction statements that are vectorized as calls rather
11568 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11569 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11570 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11571 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11572 (UNSPEC_COND_EOR): New unspecs.
11573 (optab): Add mappings for them.
11574 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11575 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
11582 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11583 * internal-fn.def (WHILE_ULT): New internal function.
11584 * internal-fn.h (direct_internal_fn_supported_p): New override
11585 that takes two types as argument.
11586 * internal-fn.c (while_direct): New macro.
11587 (expand_while_optab_fn): New function.
11588 (convert_optab_supported_p): Likewise.
11589 (direct_while_optab_supported_p): New macro.
11590 * wide-int.h (wi::udiv_ceil): New function.
11591 * tree-vectorizer.h (rgroup_masks): New structure.
11592 (vec_loop_masks): New typedef.
11593 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11594 and fully_masked_p.
11595 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11596 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11597 (vect_max_vf): New function.
11598 (slpeel_make_loop_iterate_ntimes): Delete.
11599 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11600 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11601 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11602 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11603 internal-fn.h, stor-layout.h and optabs-query.h.
11604 (vect_set_loop_mask): New function.
11605 (add_preheader_seq): Likewise.
11606 (add_header_seq): Likewise.
11607 (interleave_supported_p): Likewise.
11608 (vect_maybe_permute_loop_masks): Likewise.
11609 (vect_set_loop_masks_directly): Likewise.
11610 (vect_set_loop_condition_masked): Likewise.
11611 (vect_set_loop_condition_unmasked): New function, split out from
11612 slpeel_make_loop_iterate_ntimes.
11613 (slpeel_make_loop_iterate_ntimes): Rename to..
11614 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11615 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11616 (vect_do_peeling): Update call accordingly.
11617 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11619 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11620 mask_compare_type, can_fully_mask_p and fully_masked_p.
11621 (release_vec_loop_masks): New function.
11622 (_loop_vec_info): Use it to free the loop masks.
11623 (can_produce_all_loop_masks_p): New function.
11624 (vect_get_max_nscalars_per_iter): Likewise.
11625 (vect_verify_full_masking): Likewise.
11626 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11627 retries, and free the mask rgroups before retrying. Check loop-wide
11628 reasons for disallowing fully-masked loops. Make the final decision
11629 about whether use a fully-masked loop or not.
11630 (vect_estimate_min_profitable_iters): Do not assume that peeling
11631 for the number of iterations will be needed for fully-masked loops.
11632 (vectorizable_reduction): Disable fully-masked loops.
11633 (vectorizable_live_operation): Likewise.
11634 (vect_halve_mask_nunits): New function.
11635 (vect_double_mask_nunits): Likewise.
11636 (vect_record_loop_mask): Likewise.
11637 (vect_get_loop_mask): Likewise.
11638 (vect_transform_loop): Handle the case in which the final loop
11639 iteration might handle a partial vector. Call vect_set_loop_condition
11640 instead of slpeel_make_loop_iterate_ntimes.
11641 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11642 (check_load_store_masking): New function.
11643 (prepare_load_store_mask): Likewise.
11644 (vectorizable_store): Handle fully-masked loops.
11645 (vectorizable_load): Likewise.
11646 (supportable_widening_operation): Use vect_halve_mask_nunits for
11648 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11649 (vect_gen_while): New function.
11650 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11651 (aarch64_uqdec<mode>): New insn.
11653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11654 Alan Hayward <alan.hayward@arm.com>
11655 David Sherwood <david.sherwood@arm.com>
11657 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11658 (reduc_xor_scal_optab): New optabs.
11659 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11660 (reduc_xor_scal_@var{m}): Document.
11661 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11662 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11663 internal functions.
11664 * fold-const-call.c (fold_const_call): Handle them.
11665 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11666 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11667 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11668 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11669 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11670 (UNSPEC_XORV): New unspecs.
11671 (optab): Add entries for them.
11672 (BITWISEV): New int iterator.
11673 (bit_reduc_op): New int attributes.
11675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11676 Alan Hayward <alan.hayward@arm.com>
11677 David Sherwood <david.sherwood@arm.com>
11679 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11680 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11681 * optabs.def (vec_shl_insert_optab): New optab.
11682 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11683 (duplicate_and_interleave): Likewise.
11684 * tree-vect-loop.c: Include internal-fn.h.
11685 (neutral_op_for_slp_reduction): New function, split out from
11686 get_initial_defs_for_reduction.
11687 (get_initial_def_for_reduction): Handle option 2 for variable-length
11688 vectors by loading the neutral value into a vector and then shifting
11689 the initial value into element 0.
11690 (get_initial_defs_for_reduction): Replace the code argument with
11691 the neutral value calculated by neutral_op_for_slp_reduction.
11692 Use gimple_build_vector for constant-length vectors.
11693 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11694 but the first group_size elements have a neutral value.
11695 Use duplicate_and_interleave otherwise.
11696 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11697 Update call to get_initial_defs_for_reduction. Handle SLP
11698 reductions for variable-length vectors by creating one vector
11699 result for each scalar result, with the elements associated
11700 with other scalar results stubbed out with the neutral value.
11701 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11702 Require IFN_VEC_SHL_INSERT for double reductions on
11703 variable-length vectors, or SLP reductions that have
11704 a neutral value. Require can_duplicate_and_interleave_p
11705 support for variable-length unchained SLP reductions if there
11706 is no neutral value, such as for MIN/MAX reductions. Also require
11707 the number of vector elements to be a multiple of the number of
11708 SLP statements when doing variable-length unchained SLP reductions.
11709 Update call to vect_create_epilog_for_reduction.
11710 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11711 and remove initial values.
11712 (duplicate_and_interleave): Make public.
11713 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11714 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11717 Alan Hayward <alan.hayward@arm.com>
11718 David Sherwood <david.sherwood@arm.com>
11720 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11721 (can_duplicate_and_interleave_p): New function.
11722 (vect_get_and_check_slp_defs): Take the vector of statements
11723 rather than just the current one. Remove excess parentheses.
11724 Restriction rejectinon of vect_constant_def and vect_external_def
11725 for variable-length vectors to boolean types, or types for which
11726 can_duplicate_and_interleave_p is false.
11727 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11728 (duplicate_and_interleave): New function.
11729 (vect_get_constant_vectors): Use gimple_build_vector for
11730 constant-length vectors and suitable variable-length constant
11731 vectors. Use duplicate_and_interleave for other variable-length
11732 vectors. Don't defer the update when inserting new statements.
11734 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11735 Alan Hayward <alan.hayward@arm.com>
11736 David Sherwood <david.sherwood@arm.com>
11738 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11739 min_profitable_iters doesn't go negative.
11741 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11742 Alan Hayward <alan.hayward@arm.com>
11743 David Sherwood <david.sherwood@arm.com>
11745 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11746 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11747 * optabs.def (vec_mask_load_lanes_optab): New optab.
11748 (vec_mask_store_lanes_optab): Likewise.
11749 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11750 (MASK_STORE_LANES): Likewise.
11751 * internal-fn.c (mask_load_lanes_direct): New macro.
11752 (mask_store_lanes_direct): Likewise.
11753 (expand_mask_load_optab_fn): Handle masked operations.
11754 (expand_mask_load_lanes_optab_fn): New macro.
11755 (expand_mask_store_optab_fn): Handle masked operations.
11756 (expand_mask_store_lanes_optab_fn): New macro.
11757 (direct_mask_load_lanes_optab_supported_p): Likewise.
11758 (direct_mask_store_lanes_optab_supported_p): Likewise.
11759 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11761 (vect_load_lanes_supported): Likewise.
11762 * tree-vect-data-refs.c (strip_conversion): New function.
11763 (can_group_stmts_p): Likewise.
11764 (vect_analyze_data_ref_accesses): Use it instead of checking
11765 for a pair of assignments.
11766 (vect_store_lanes_supported): Take a masked_p parameter.
11767 (vect_load_lanes_supported): Likewise.
11768 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11769 vect_store_lanes_supported and vect_load_lanes_supported.
11770 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11771 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11772 parameter. Don't allow gaps for masked accesses.
11773 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11774 and vect_load_lanes_supported.
11775 (get_load_store_type): Take a masked_p parameter and update
11776 call to get_group_load_store_type.
11777 (vectorizable_store): Update call to get_load_store_type.
11778 Handle IFN_MASK_STORE_LANES.
11779 (vectorizable_load): Update call to get_load_store_type.
11780 Handle IFN_MASK_LOAD_LANES.
11782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11783 Alan Hayward <alan.hayward@arm.com>
11784 David Sherwood <david.sherwood@arm.com>
11786 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11788 * config/aarch64/aarch64-protos.h
11789 (aarch64_sve_struct_memory_operand_p): Declare.
11790 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11791 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11792 (VPRED, vpred): Handle SVE structure modes.
11793 * config/aarch64/constraints.md (Utx): New constraint.
11794 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11795 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11796 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11797 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11798 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11799 structure modes. Split into pieces after RA.
11800 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11801 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11803 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11804 SVE structure modes.
11805 (aarch64_classify_address): Likewise.
11806 (sizetochar): Move earlier in file.
11807 (aarch64_print_operand): Handle SVE register lists.
11808 (aarch64_array_mode): New function.
11809 (aarch64_sve_struct_memory_operand_p): Likewise.
11810 (TARGET_ARRAY_MODE): Redefine.
11812 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11813 Alan Hayward <alan.hayward@arm.com>
11814 David Sherwood <david.sherwood@arm.com>
11816 * target.def (array_mode): New target hook.
11817 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11818 * doc/tm.texi: Regenerate.
11819 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11820 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11821 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11822 targetm.array_mode.
11823 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11826 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11827 Alan Hayward <alan.hayward@arm.com>
11828 David Sherwood <david.sherwood@arm.com>
11830 * fold-const.c (fold_binary_loc): Check the argument types
11831 rather than the result type when testing for a vector operation.
11833 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11835 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11836 * doc/tm.texi: Regenerate.
11838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11839 Alan Hayward <alan.hayward@arm.com>
11840 David Sherwood <david.sherwood@arm.com>
11842 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11843 (sve): Document new AArch64 extension.
11844 * doc/md.texi (w): Extend the description of the AArch64
11845 constraint to include SVE vectors.
11846 (Upl, Upa): Document new AArch64 predicate constraints.
11847 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11849 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11850 (msve-vector-bits=): New option.
11851 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11852 SVE when these are disabled.
11853 (sve): New extension.
11854 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11855 modes. Adjust their number of units based on aarch64_sve_vg.
11856 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11857 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11858 aarch64_addr_query_type.
11859 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11860 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11861 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11862 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11863 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11864 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11865 (aarch64_simd_imm_zero_p): Delete.
11866 (aarch64_check_zero_based_sve_index_immediate): Declare.
11867 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11868 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11869 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11870 (aarch64_sve_float_mul_immediate_p): Likewise.
11871 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11872 rather than an rtx.
11873 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11874 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11875 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11876 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11877 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11878 (aarch64_regmode_natural_size): Likewise.
11879 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11880 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11882 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11883 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11884 for VG and the SVE predicate registers.
11885 (V_ALIASES): Add a "z"-prefixed alias.
11886 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11887 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11888 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11889 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11890 (REG_CLASS_NAMES): Add entries for them.
11891 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11892 and the predicate registers.
11893 (aarch64_sve_vg): Declare.
11894 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11895 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11896 (REGMODE_NATURAL_SIZE): Define.
11897 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11899 * config/aarch64/aarch64.c: Include cfgrtl.h.
11900 (simd_immediate_info): Add a constructor for series vectors,
11901 and an associated step field.
11902 (aarch64_sve_vg): New variable.
11903 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11904 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11905 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11906 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11907 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11908 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11909 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11910 (aarch64_get_mask_mode): New functions.
11911 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11912 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11913 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11914 predicate modes and predicate registers. Explicitly restrict
11915 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11916 to store a vector mode if it is recognized by
11917 aarch64_classify_vector_mode.
11918 (aarch64_regmode_natural_size): New function.
11919 (aarch64_hard_regno_caller_save_mode): Return the original mode
11921 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11922 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11923 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11924 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11926 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11927 does not overlap dest if the function is frame-related. Handle
11929 (aarch64_split_add_offset): New function.
11930 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11931 them aarch64_add_offset.
11932 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11933 and update call to aarch64_sub_sp.
11934 (aarch64_add_cfa_expression): New function.
11935 (aarch64_expand_prologue): Pass extra temporary registers to the
11936 functions above. Handle the case in which we need to emit new
11937 DW_CFA_expressions for registers that were originally saved
11938 relative to the stack pointer, but now have to be expressed
11939 relative to the frame pointer.
11940 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11942 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11943 IP0 and IP1 values for SVE frames.
11944 (aarch64_expand_vec_series): New function.
11945 (aarch64_expand_sve_widened_duplicate): Likewise.
11946 (aarch64_expand_sve_const_vector): Likewise.
11947 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11948 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11949 into the register, rather than emitting a SET directly.
11950 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11951 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11952 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11953 (offset_9bit_signed_scaled_p): New functions.
11954 (aarch64_replicate_bitmask_imm): New function.
11955 (aarch64_bitmask_imm): Use it.
11956 (aarch64_cannot_force_const_mem): Reject expressions involving
11957 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11958 (aarch64_classify_index): Handle SVE indices, by requiring
11959 a plain register index with a scale that matches the element size.
11960 (aarch64_classify_address): Handle SVE addresses. Assert that
11961 the mode of the address is VOIDmode or an integer mode.
11962 Update call to aarch64_classify_symbol.
11963 (aarch64_classify_symbolic_expression): Update call to
11964 aarch64_classify_symbol.
11965 (aarch64_const_vec_all_in_range_p): New function.
11966 (aarch64_print_vector_float_operand): Likewise.
11967 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11968 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11969 and the FP immediates 1.0 and 0.5.
11970 (aarch64_print_address_internal): Handle SVE addresses.
11971 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11972 (aarch64_regno_regclass): Handle predicate registers.
11973 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11975 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11976 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11977 (aarch64_convert_sve_vector_bits): New function.
11978 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11979 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11980 rather than an rtx.
11981 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11982 Handle SVE vector and predicate modes. Accept VL-based constants
11983 that need only one temporary register, and VL offsets that require
11984 no temporary registers.
11985 (aarch64_conditional_register_usage): Mark the predicate registers
11986 as fixed if SVE isn't available.
11987 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11988 Return true for SVE vector and predicate modes.
11989 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11990 rather than an unsigned int. Handle SVE modes.
11991 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11993 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11995 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11996 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11997 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11998 (aarch64_sve_float_mul_immediate_p): New functions.
11999 (aarch64_sve_valid_immediate): New function.
12000 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12001 Explicitly reject structure modes. Check for INDEX constants.
12002 Handle PTRUE and PFALSE constants.
12003 (aarch64_check_zero_based_sve_index_immediate): New function.
12004 (aarch64_simd_imm_zero_p): Delete.
12005 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12006 vector modes. Accept constants in the range of CNT[BHWD].
12007 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12008 ask for an Advanced SIMD mode.
12009 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12010 (aarch64_simd_vector_alignment): Handle SVE predicates.
12011 (aarch64_vectorize_preferred_vector_alignment): New function.
12012 (aarch64_simd_vector_alignment_reachable): Use it instead of
12014 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12015 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12017 (MAX_VECT_LEN): Delete.
12018 (expand_vec_perm_d): Add a vec_flags field.
12019 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12020 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12021 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12023 (aarch64_evpc_rev): Rename to...
12024 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12025 (aarch64_evpc_rev_global): New function.
12026 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12027 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12029 (aarch64_evpc_sve_tbl): New function.
12030 (aarch64_expand_vec_perm_const_1): Update after rename of
12031 aarch64_evpc_rev. Handle SVE permutes too, trying
12032 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12033 than aarch64_evpc_tbl.
12034 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12035 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12036 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12037 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12038 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12039 (aarch64_expand_sve_vcond): New functions.
12040 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12041 of aarch64_vector_mode_p.
12042 (aarch64_dwarf_poly_indeterminate_value): New function.
12043 (aarch64_compute_pressure_classes): Likewise.
12044 (aarch64_can_change_mode_class): Likewise.
12045 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12046 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12047 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12048 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12049 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12050 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12051 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12052 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12054 (Dn, Dl, Dr): Accept const as well as const_vector.
12055 (Dz): Likewise. Compare against CONST0_RTX.
12056 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12057 of "vector" where appropriate.
12058 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12059 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12060 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12061 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12062 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12063 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12064 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12065 (v_int_equiv): Extend to SVE modes.
12066 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12068 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12069 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12070 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12071 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12072 (SVE_COND_FP_CMP): New int iterators.
12073 (perm_hilo): Handle the new unpack unspecs.
12074 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12076 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12077 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12078 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12079 (aarch64_equality_operator, aarch64_constant_vector_operand)
12080 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12081 (aarch64_sve_nonimmediate_operand): Likewise.
12082 (aarch64_sve_general_operand): Likewise.
12083 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12084 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12085 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12086 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12087 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12088 (aarch64_sve_float_arith_immediate): Likewise.
12089 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12090 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12091 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12092 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12093 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12094 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12095 (aarch64_sve_float_arith_operand): Likewise.
12096 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12097 (aarch64_sve_float_mul_operand): Likewise.
12098 (aarch64_sve_vec_perm_operand): Likewise.
12099 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12100 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12101 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12102 as well as const_vector.
12103 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12104 in file. Use CONST0_RTX and CONSTM1_RTX.
12105 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12106 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12107 Use aarch64_simd_imm_zero.
12108 * config/aarch64/aarch64-sve.md: New file.
12109 * config/aarch64/aarch64.md: Include it.
12110 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12111 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12112 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12113 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12114 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12115 (sve): New attribute.
12116 (enabled): Disable instructions with the sve attribute unless
12118 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12119 aarch64_expand_mov_immediate.
12120 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12121 CNT[BHSD] immediates.
12122 (movti): Split CONST_POLY_INT moves into two halves.
12123 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12124 Split additions that need a temporary here if the destination
12125 is the stack pointer.
12126 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12127 (*add<mode>3_poly_1): New instruction.
12128 (set_clobber_cc): New expander.
12130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12132 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12133 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12134 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12135 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12136 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12137 Change innermode from fixed_mode_size to machine_mode.
12138 (simplify_subreg): Update call accordingly. Handle a constant-sized
12139 subreg of a variable-length CONST_VECTOR.
12141 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12142 Alan Hayward <alan.hayward@arm.com>
12143 David Sherwood <david.sherwood@arm.com>
12145 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12146 (add_offset_to_base): New function, split out from...
12147 (create_mem_ref): ...here. When handling a scale other than 1,
12148 check first whether the address is valid without the offset.
12149 Add it into the base if so, leaving the index and scale as-is.
12151 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12154 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12155 fold_for_warn before checking if arg2 is INTEGER_CST.
12157 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12159 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12160 (store_multiple_operation): Delete.
12161 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12162 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12163 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12164 guarded by TARGET_STRING.
12165 (rs6000_output_load_multiple): Delete.
12166 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12167 OPTION_MASK_STRING / TARGET_STRING handling.
12168 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12169 (const rs6000_opt_masks) <"string">: Change mask to 0.
12170 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12171 (MASK_STRING): Delete.
12172 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12174 (load_multiple): Delete.
12181 (store_multiple): Delete.
12188 (movmemsi_8reg): Delete.
12189 (corresponding unnamed define_insn): Delete.
12190 (movmemsi_6reg): Delete.
12191 (corresponding unnamed define_insn): Delete.
12192 (movmemsi_4reg): Delete.
12193 (corresponding unnamed define_insn): Delete.
12194 (movmemsi_2reg): Delete.
12195 (corresponding unnamed define_insn): Delete.
12196 (movmemsi_1reg): Delete.
12197 (corresponding unnamed define_insn): Delete.
12198 * config/rs6000/rs6000.opt (mno-string): New.
12199 (mstring): Replace by deprecation warning stub.
12200 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12202 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12204 * regrename.c (regrename_do_replace): If replacing the same
12205 reg multiple times, try to reuse last created gen_raw_REG.
12208 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12209 main to workaround a bug in GDB.
12211 2018-01-12 Tom de Vries <tom@codesourcery.com>
12214 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12216 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12218 PR rtl-optimization/80481
12219 * ira-color.c (get_cap_member): New function.
12220 (allocnos_conflict_by_live_ranges_p): Use it.
12221 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12222 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12224 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12227 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12228 (*saddl_se_1): Ditto.
12229 (*ssubsi_1): Ditto.
12230 (*ssubl_se_1): Ditto.
12232 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12234 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12235 rather than wi::to_widest for DR_INITs.
12236 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12237 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12238 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12240 (vect_analyze_group_access_1): Note that here.
12242 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12244 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12245 polynomial type sizes.
12247 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12249 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12250 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12251 (gimple_add_tmp_var): Likewise.
12253 2018-01-12 Martin Liska <mliska@suse.cz>
12255 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12256 (gimple_alloc_sizes): Likewise.
12257 (dump_gimple_statistics): Use PRIu64 in printf format.
12258 * gimple.h: Change uint64_t to int.
12260 2018-01-12 Martin Liska <mliska@suse.cz>
12262 * tree-core.h: Use uint64_t instead of int.
12263 * tree.c (tree_node_counts): Likewise.
12264 (tree_node_sizes): Likewise.
12265 (dump_tree_statistics): Use PRIu64 in printf format.
12267 2018-01-12 Martin Liska <mliska@suse.cz>
12269 * Makefile.in: As qsort_chk is implemented in vec.c, add
12270 vec.o to linkage of gencfn-macros.
12271 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12272 passing the info to record_node_allocation_statistics.
12273 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12275 * ggc-common.c (struct ggc_usage): Add operator== and use
12276 it in operator< and compare function.
12277 * mem-stats.h (struct mem_usage): Likewise.
12278 * vec.c (struct vec_usage): Remove operator< and compare
12279 function. Can be simply inherited.
12281 2018-01-12 Martin Jambor <mjambor@suse.cz>
12284 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12285 * tree-ssa-math-opts.c: Include domwalk.h.
12286 (convert_mult_to_fma_1): New function.
12287 (fma_transformation_info): New type.
12288 (fma_deferring_state): Likewise.
12289 (cancel_fma_deferring): New function.
12290 (result_of_phi): Likewise.
12291 (last_fma_candidate_feeds_initial_phi): Likewise.
12292 (convert_mult_to_fma): Added deferring logic, split actual
12293 transformation to convert_mult_to_fma_1.
12294 (math_opts_dom_walker): New type.
12295 (math_opts_dom_walker::after_dom_children): New method, body moved
12296 here from pass_optimize_widening_mul::execute, added deferring logic
12298 (pass_optimize_widening_mul::execute): Moved most of code to
12299 math_opts_dom_walker::after_dom_children.
12300 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12301 * config/i386/i386.c (ix86_option_override_internal): Added
12302 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12304 2018-01-12 Richard Biener <rguenther@suse.de>
12307 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12308 inline instance vars.
12310 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12313 * config/rx/rx.c (rx_is_restricted_memory_address):
12314 Handle SUBREG case.
12316 2018-01-12 Richard Biener <rguenther@suse.de>
12318 PR tree-optimization/80846
12319 * target.def (split_reduction): New target hook.
12320 * targhooks.c (default_split_reduction): New function.
12321 * targhooks.h (default_split_reduction): Declare.
12322 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12323 target requests first reduce vectors by combining low and high
12325 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12326 (get_vectype_for_scalar_type_and_size): Export.
12327 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12328 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12329 * doc/tm.texi: Regenerate.
12330 * config/i386/i386.c (ix86_split_reduction): Implement
12331 TARGET_VECTORIZE_SPLIT_REDUCTION.
12333 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12336 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12337 in PIC mode except for TARGET_VXWORKS_RTP.
12338 * config/sparc/sparc.c: Include cfgrtl.h.
12339 (TARGET_INIT_PIC_REG): Define.
12340 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12341 (sparc_pic_register_p): New predicate.
12342 (sparc_legitimate_address_p): Use it.
12343 (sparc_legitimize_pic_address): Likewise.
12344 (sparc_delegitimize_address): Likewise.
12345 (sparc_mode_dependent_address_p): Likewise.
12346 (gen_load_pcrel_sym): Remove 4th parameter.
12347 (load_got_register): Adjust call to above. Remove obsolete stuff.
12348 (sparc_expand_prologue): Do not call load_got_register here.
12349 (sparc_flat_expand_prologue): Likewise.
12350 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12351 (sparc_use_pseudo_pic_reg): New function.
12352 (sparc_init_pic_reg): Likewise.
12353 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12354 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12356 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12358 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12359 Add item for branch_cost.
12361 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12363 PR rtl-optimization/83565
12364 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12365 not extend the result to a larger mode for rotate operations.
12366 (num_sign_bit_copies1): Likewise.
12368 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12371 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12373 Use values-Xc.o for -pedantic.
12374 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12376 2018-01-12 Martin Liska <mliska@suse.cz>
12379 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12381 (possible_polymorphic_call_targets): Use it.
12382 (ipa_devirt): Likewise.
12384 2018-01-12 Martin Liska <mliska@suse.cz>
12386 * profile-count.h (enum profile_quality): Use 0 as invalid
12387 enum value of profile_quality.
12389 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12391 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12392 -mext-string options.
12394 2018-01-12 Richard Biener <rguenther@suse.de>
12396 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12397 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12398 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12400 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12402 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12404 * configure.ac (--with-long-double-format): Add support for the
12405 configuration option to change the default long double format on
12407 * config.gcc (powerpc*-linux*-*): Likewise.
12408 * configure: Regenerate.
12409 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12410 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12411 used without modification.
12413 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12415 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12416 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12417 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12418 MISC_BUILTIN_SPEC_BARRIER.
12419 (rs6000_init_builtins): Likewise.
12420 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12422 (speculation_barrier): New define_insn.
12423 * doc/extend.texi: Document __builtin_speculation_barrier.
12425 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12428 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12429 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12430 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12432 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12433 integral modes instead of "ss" and "sd".
12434 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12435 vectors with 32-bit and 64-bit elements.
12436 (vecdupssescalarmodesuffix): New mode attribute.
12437 (vec_dup<mode>): Use it.
12439 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12442 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12443 frame if argument is passed on stack.
12445 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12448 * ree.c (combine_reaching_defs): Optimize also
12449 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12450 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12452 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12454 PR middle-end/83189
12455 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12457 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12459 PR middle-end/83718
12460 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12461 after they are computed.
12463 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12465 PR tree-optimization/83695
12466 * gimple-loop-linterchange.cc
12467 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12468 reset cached scev information after interchange.
12469 (pass_linterchange::execute): Remove call to scev_reset_htab.
12471 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12473 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12474 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12475 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12476 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12477 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12478 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12479 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12480 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12481 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12482 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12483 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12484 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12485 (V_lane_reg): Likewise.
12486 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12488 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12489 (vfmal_lane_low<mode>_intrinsic,
12490 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12491 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12492 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12493 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12494 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12495 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12497 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12499 * config/arm/arm-cpus.in (fp16fml): New feature.
12500 (ALL_SIMD): Add fp16fml.
12501 (armv8.2-a): Add fp16fml as an option.
12502 (armv8.3-a): Likewise.
12503 (armv8.4-a): Add fp16fml as part of fp16.
12504 * config/arm/arm.h (TARGET_FP16FML): Define.
12505 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12507 * config/arm/arm-modes.def (V2HF): Define.
12508 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12509 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12510 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12511 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12512 vfmsl_low, vfmsl_high): New set of builtins.
12513 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12514 (vfml_op): New code attribute.
12515 (VFMLHALVES): New int iterator.
12516 (VFML, VFMLSEL): New mode attributes.
12517 (V_reg): Define mapping for V2HF.
12518 (V_hi, V_lo): New mode attributes.
12519 (VF_constraint): Likewise.
12520 (vfml_half, vfml_half_selector): New int attributes.
12521 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12523 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12524 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12526 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12527 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12528 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12529 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12531 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12532 Document new effective target and option set.
12534 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12536 * config/arm/arm-cpus.in (armv8_4): New feature.
12537 (ARMv8_4a): New fgroup.
12538 (armv8.4-a): New arch.
12539 * config/arm/arm-tables.opt: Regenerate.
12540 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12541 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12542 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12543 Add matching rules for -march=armv8.4-a and extensions.
12544 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12546 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12549 * config/rx/rx.md (BW): New mode attribute.
12550 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12552 2018-01-11 Richard Biener <rguenther@suse.de>
12554 PR tree-optimization/83435
12555 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12556 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12557 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12559 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12560 Alan Hayward <alan.hayward@arm.com>
12561 David Sherwood <david.sherwood@arm.com>
12563 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12565 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12566 (aarch64_print_address_internal): Use it to check for a zero offset.
12568 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12569 Alan Hayward <alan.hayward@arm.com>
12570 David Sherwood <david.sherwood@arm.com>
12572 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12573 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12574 Return a poly_int64 rather than a HOST_WIDE_INT.
12575 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12576 rather than a HOST_WIDE_INT.
12577 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12578 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12579 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12580 final_offset from HOST_WIDE_INT to poly_int64.
12581 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12582 to_constant when getting the number of units in an Advanced SIMD
12584 (aarch64_builtin_vectorized_function): Check for a constant number
12586 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12588 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12589 attribute instead of GET_MODE_NUNITS.
12590 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12591 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12592 GET_MODE_SIZE for fixed-size registers.
12593 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12594 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12595 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12596 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12597 (aarch64_print_operand, aarch64_print_address_internal)
12598 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12599 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12600 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12601 Handle polynomial GET_MODE_SIZE.
12602 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12603 wider than SImode without modification.
12604 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12605 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12606 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12607 passing and returning SVE modes.
12608 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12609 rather than GEN_INT.
12610 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12611 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12612 (aarch64_allocate_and_probe_stack_space): Likewise.
12613 (aarch64_layout_frame): Cope with polynomial offsets.
12614 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12615 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12616 polynomial offsets.
12617 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12618 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12619 poly_int64 rather than a HOST_WIDE_INT.
12620 (aarch64_get_separate_components, aarch64_process_components)
12621 (aarch64_expand_prologue, aarch64_expand_epilogue)
12622 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12623 (aarch64_anchor_offset): New function, split out from...
12624 (aarch64_legitimize_address): ...here.
12625 (aarch64_builtin_vectorization_cost): Handle polynomial
12626 TYPE_VECTOR_SUBPARTS.
12627 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12629 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12630 number of elements from the PARALLEL rather than the mode.
12631 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12632 rather than GET_MODE_BITSIZE.
12633 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12634 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12635 (aarch64_expand_vec_perm_const_1): Handle polynomial
12636 d->perm.length () and d->perm elements.
12637 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12638 Apply to_constant to d->perm elements.
12639 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12640 polynomial CONST_VECTOR_NUNITS.
12641 (aarch64_move_pointer): Take amount as a poly_int64 rather
12643 (aarch64_progress_pointer): Avoid temporary variable.
12644 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12645 the mode attribute instead of GET_MODE.
12647 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12648 Alan Hayward <alan.hayward@arm.com>
12649 David Sherwood <david.sherwood@arm.com>
12651 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12652 x exists before using it.
12653 (aarch64_add_constant_internal): Rename to...
12654 (aarch64_add_offset_1): ...this. Replace regnum with separate
12655 src and dest rtxes. Handle the case in which they're different,
12656 including when the offset is zero. Replace scratchreg with an rtx.
12657 Use 2 additions if there is no spare register into which we can
12658 move a 16-bit constant.
12659 (aarch64_add_constant): Delete.
12660 (aarch64_add_offset): Replace reg with separate src and dest
12661 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12662 Use aarch64_add_offset_1.
12663 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12664 an rtx rather than an int. Take the delta as a poly_int64
12665 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12666 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12667 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12668 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12669 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12670 and aarch64_add_sp.
12671 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12672 aarch64_add_constant.
12674 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12676 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12677 Use scalar_float_mode.
12679 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12681 * config/aarch64/aarch64-simd.md
12682 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12683 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12684 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12685 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12686 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12687 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12688 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12689 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12690 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12691 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12693 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12696 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12697 targ_options->x_arm_arch_string is non NULL.
12699 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12701 * config/aarch64/aarch64.h
12702 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12704 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12707 * expmed.c (emit_store_flag_force): Swap if const op0
12708 and change VOIDmode to mode of op0.
12710 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12712 PR rtl-optimization/83761
12713 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12714 than bytes to mode_for_size.
12716 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12718 PR middle-end/83189
12719 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12720 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12723 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12725 PR middle-end/83575
12726 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12727 when in layout mode.
12728 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12729 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12732 2018-01-10 Michael Collison <michael.collison@arm.com>
12734 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12735 * config/aarch64/aarch64-option-extension.def: Add
12736 AARCH64_OPT_EXTENSION of 'fp16fml'.
12737 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12738 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12739 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12740 * config/aarch64/constraints.md (Ui7): New constraint.
12741 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12742 (VFMLA_SEL_W): Ditto.
12745 (VFMLA16_LOW): New int iterator.
12746 (VFMLA16_HIGH): Ditto.
12747 (UNSPEC_FMLAL): New unspec.
12748 (UNSPEC_FMLSL): Ditto.
12749 (UNSPEC_FMLAL2): Ditto.
12750 (UNSPEC_FMLSL2): Ditto.
12751 (f16mac): New code attribute.
12752 * config/aarch64/aarch64-simd-builtins.def
12753 (aarch64_fmlal_lowv2sf): Ditto.
12754 (aarch64_fmlsl_lowv2sf): Ditto.
12755 (aarch64_fmlalq_lowv4sf): Ditto.
12756 (aarch64_fmlslq_lowv4sf): Ditto.
12757 (aarch64_fmlal_highv2sf): Ditto.
12758 (aarch64_fmlsl_highv2sf): Ditto.
12759 (aarch64_fmlalq_highv4sf): Ditto.
12760 (aarch64_fmlslq_highv4sf): Ditto.
12761 (aarch64_fmlal_lane_lowv2sf): Ditto.
12762 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12763 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12764 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12765 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12766 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12767 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12768 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12769 (aarch64_fmlal_lane_highv2sf): Ditto.
12770 (aarch64_fmlsl_lane_highv2sf): Ditto.
12771 (aarch64_fmlal_laneq_highv2sf): Ditto.
12772 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12773 (aarch64_fmlalq_lane_highv4sf): Ditto.
12774 (aarch64_fmlsl_lane_highv4sf): Ditto.
12775 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12776 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12777 * config/aarch64/aarch64-simd.md:
12778 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12779 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12780 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12781 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12782 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12783 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12784 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12785 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12786 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12787 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12788 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12789 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12790 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12791 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12792 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12793 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12794 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12795 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12796 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12797 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12798 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12799 (vfmlsl_low_u32): Ditto.
12800 (vfmlalq_low_u32): Ditto.
12801 (vfmlslq_low_u32): Ditto.
12802 (vfmlal_high_u32): Ditto.
12803 (vfmlsl_high_u32): Ditto.
12804 (vfmlalq_high_u32): Ditto.
12805 (vfmlslq_high_u32): Ditto.
12806 (vfmlal_lane_low_u32): Ditto.
12807 (vfmlsl_lane_low_u32): Ditto.
12808 (vfmlal_laneq_low_u32): Ditto.
12809 (vfmlsl_laneq_low_u32): Ditto.
12810 (vfmlalq_lane_low_u32): Ditto.
12811 (vfmlslq_lane_low_u32): Ditto.
12812 (vfmlalq_laneq_low_u32): Ditto.
12813 (vfmlslq_laneq_low_u32): Ditto.
12814 (vfmlal_lane_high_u32): Ditto.
12815 (vfmlsl_lane_high_u32): Ditto.
12816 (vfmlal_laneq_high_u32): Ditto.
12817 (vfmlsl_laneq_high_u32): Ditto.
12818 (vfmlalq_lane_high_u32): Ditto.
12819 (vfmlslq_lane_high_u32): Ditto.
12820 (vfmlalq_laneq_high_u32): Ditto.
12821 (vfmlslq_laneq_high_u32): Ditto.
12822 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12823 (AARCH64_FL_FOR_ARCH8_4): New.
12824 (AARCH64_ISA_F16FML): New ISA flag.
12825 (TARGET_F16FML): New feature flag for fp16fml.
12826 (doc/invoke.texi): Document new fp16fml option.
12828 2018-01-10 Michael Collison <michael.collison@arm.com>
12830 * config/aarch64/aarch64-builtins.c:
12831 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12832 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12833 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12834 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12835 (AARCH64_ISA_SHA3): New ISA flag.
12836 (TARGET_SHA3): New feature flag for sha3.
12837 * config/aarch64/iterators.md (sha512_op): New int attribute.
12838 (CRYPTO_SHA512): New int iterator.
12839 (UNSPEC_SHA512H): New unspec.
12840 (UNSPEC_SHA512H2): Ditto.
12841 (UNSPEC_SHA512SU0): Ditto.
12842 (UNSPEC_SHA512SU1): Ditto.
12843 * config/aarch64/aarch64-simd-builtins.def
12844 (aarch64_crypto_sha512hqv2di): New builtin.
12845 (aarch64_crypto_sha512h2qv2di): Ditto.
12846 (aarch64_crypto_sha512su0qv2di): Ditto.
12847 (aarch64_crypto_sha512su1qv2di): Ditto.
12848 (aarch64_eor3qv8hi): Ditto.
12849 (aarch64_rax1qv2di): Ditto.
12850 (aarch64_xarqv2di): Ditto.
12851 (aarch64_bcaxqv8hi): Ditto.
12852 * config/aarch64/aarch64-simd.md:
12853 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12854 (aarch64_crypto_sha512su0qv2di): Ditto.
12855 (aarch64_crypto_sha512su1qv2di): Ditto.
12856 (aarch64_eor3qv8hi): Ditto.
12857 (aarch64_rax1qv2di): Ditto.
12858 (aarch64_xarqv2di): Ditto.
12859 (aarch64_bcaxqv8hi): Ditto.
12860 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12861 (vsha512h2q_u64): Ditto.
12862 (vsha512su0q_u64): Ditto.
12863 (vsha512su1q_u64): Ditto.
12864 (veor3q_u16): Ditto.
12865 (vrax1q_u64): Ditto.
12866 (vxarq_u64): Ditto.
12867 (vbcaxq_u16): Ditto.
12868 * config/arm/types.md (crypto_sha512): New type attribute.
12869 (crypto_sha3): Ditto.
12870 (doc/invoke.texi): Document new sha3 option.
12872 2018-01-10 Michael Collison <michael.collison@arm.com>
12874 * config/aarch64/aarch64-builtins.c:
12875 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12876 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12877 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12878 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12879 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12880 (AARCH64_ISA_SM4): New ISA flag.
12881 (TARGET_SM4): New feature flag for sm4.
12882 * config/aarch64/aarch64-simd-builtins.def
12883 (aarch64_sm3ss1qv4si): Ditto.
12884 (aarch64_sm3tt1aq4si): Ditto.
12885 (aarch64_sm3tt1bq4si): Ditto.
12886 (aarch64_sm3tt2aq4si): Ditto.
12887 (aarch64_sm3tt2bq4si): Ditto.
12888 (aarch64_sm3partw1qv4si): Ditto.
12889 (aarch64_sm3partw2qv4si): Ditto.
12890 (aarch64_sm4eqv4si): Ditto.
12891 (aarch64_sm4ekeyqv4si): Ditto.
12892 * config/aarch64/aarch64-simd.md:
12893 (aarch64_sm3ss1qv4si): Ditto.
12894 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12895 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12896 (aarch64_sm4eqv4si): Ditto.
12897 (aarch64_sm4ekeyqv4si): Ditto.
12898 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12899 (sm3part_op): Ditto.
12900 (CRYPTO_SM3TT): Ditto.
12901 (CRYPTO_SM3PART): Ditto.
12902 (UNSPEC_SM3SS1): New unspec.
12903 (UNSPEC_SM3TT1A): Ditto.
12904 (UNSPEC_SM3TT1B): Ditto.
12905 (UNSPEC_SM3TT2A): Ditto.
12906 (UNSPEC_SM3TT2B): Ditto.
12907 (UNSPEC_SM3PARTW1): Ditto.
12908 (UNSPEC_SM3PARTW2): Ditto.
12909 (UNSPEC_SM4E): Ditto.
12910 (UNSPEC_SM4EKEY): Ditto.
12911 * config/aarch64/constraints.md (Ui2): New constraint.
12912 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12913 * config/arm/types.md (crypto_sm3): New type attribute.
12914 (crypto_sm4): Ditto.
12915 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12916 (vsm3tt1aq_u32): Ditto.
12917 (vsm3tt1bq_u32): Ditto.
12918 (vsm3tt2aq_u32): Ditto.
12919 (vsm3tt2bq_u32): Ditto.
12920 (vsm3partw1q_u32): Ditto.
12921 (vsm3partw2q_u32): Ditto.
12922 (vsm4eq_u32): Ditto.
12923 (vsm4ekeyq_u32): Ditto.
12924 (doc/invoke.texi): Document new sm4 option.
12926 2018-01-10 Michael Collison <michael.collison@arm.com>
12928 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12929 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12930 (AARCH64_FL_FOR_ARCH8_4): New.
12931 (AARCH64_FL_V8_4): New flag.
12932 (doc/invoke.texi): Document new armv8.4-a option.
12934 2018-01-10 Michael Collison <michael.collison@arm.com>
12936 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12937 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12938 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12939 * config/aarch64/aarch64-option-extension.def: Add
12940 AARCH64_OPT_EXTENSION of 'sha2'.
12941 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12942 (crypto): Disable sha2 and aes if crypto disabled.
12943 (crypto): Enable aes and sha2 if enabled.
12944 (simd): Disable sha2 and aes if simd disabled.
12945 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12947 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12948 (TARGET_SHA2): New feature flag for sha2.
12949 (TARGET_AES): New feature flag for aes.
12950 * config/aarch64/aarch64-simd.md:
12951 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12952 conditional on TARGET_AES.
12953 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12954 (aarch64_crypto_sha1hsi): Make pattern conditional
12956 (aarch64_crypto_sha1hv4si): Ditto.
12957 (aarch64_be_crypto_sha1hv4si): Ditto.
12958 (aarch64_crypto_sha1su1v4si): Ditto.
12959 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12960 (aarch64_crypto_sha1su0v4si): Ditto.
12961 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12962 (aarch64_crypto_sha256su0v4si): Ditto.
12963 (aarch64_crypto_sha256su1v4si): Ditto.
12964 (doc/invoke.texi): Document new aes and sha2 options.
12966 2018-01-10 Martin Sebor <msebor@redhat.com>
12968 PR tree-optimization/83781
12969 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12972 2018-01-11 Martin Sebor <msebor@gmail.com>
12973 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12975 PR tree-optimization/83501
12976 PR tree-optimization/81703
12978 * tree-ssa-strlen.c (get_string_cst): Rename...
12979 (get_string_len): ...to this. Handle global constants.
12980 (handle_char_store): Adjust.
12982 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12983 Jim Wilson <jimw@sifive.com>
12985 * config/riscv/riscv-protos.h (riscv_output_return): New.
12986 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12987 (riscv_attribute_table, riscv_output_return),
12988 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12989 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12990 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12992 (riscv_expand_prologue): Add early return for naked function.
12993 (riscv_expand_epilogue): Likewise.
12994 (riscv_function_ok_for_sibcall): Return false for naked function.
12995 (riscv_set_current_function): New.
12996 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12997 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12998 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12999 * doc/extend.texi (RISC-V Function Attributes): New.
13001 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13003 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13004 check for 128-bit long double before checking TCmode.
13005 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13006 128-bit long doubles before checking TFmode or TCmode.
13007 (FLOAT128_IBM_P): Likewise.
13009 2018-01-10 Martin Sebor <msebor@redhat.com>
13011 PR tree-optimization/83671
13012 * builtins.c (c_strlen): Unconditionally return zero for the empty
13014 Use -Warray-bounds for warnings.
13015 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13016 for non-constant array indices with COMPONENT_REF, arrays of
13017 arrays, and pointers to arrays.
13018 (gimple_fold_builtin_strlen): Determine and set length range for
13019 non-constant character arrays.
13021 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13023 PR middle-end/81897
13024 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13027 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13029 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13031 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13034 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13035 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13036 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13037 indexed_or_indirect_operand predicate.
13038 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13039 (*vsx_le_perm_load_v8hi): Likewise.
13040 (*vsx_le_perm_load_v16qi): Likewise.
13041 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13042 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13043 (*vsx_le_perm_store_v8hi): Likewise.
13044 (*vsx_le_perm_store_v16qi): Likewise.
13045 (eight unnamed splitters): Likewise.
13047 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13049 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13050 * config/rs6000/emmintrin.h: Likewise.
13051 * config/rs6000/mmintrin.h: Likewise.
13052 * config/rs6000/xmmintrin.h: Likewise.
13054 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13057 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13059 * tree.c (tree_nop_conversion): Return true for location wrapper
13061 (maybe_wrap_with_location): New function.
13062 (selftest::check_strip_nops): New function.
13063 (selftest::test_location_wrappers): New function.
13064 (selftest::tree_c_tests): Call it.
13065 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13066 (maybe_wrap_with_location): New decl.
13067 (EXPR_LOCATION_WRAPPER_P): New macro.
13068 (location_wrapper_p): New inline function.
13069 (tree_strip_any_location_wrapper): New inline function.
13071 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13074 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13075 stack_realign_offset for the largest alignment of stack slot
13077 (ix86_find_max_used_stack_alignment): New function.
13078 (ix86_finalize_stack_frame_flags): Use it. Set
13079 max_used_stack_alignment if we don't realign stack.
13080 * config/i386/i386.h (machine_function): Add
13081 max_used_stack_alignment.
13083 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13085 * config/arm/arm.opt (-mbranch-cost): New option.
13086 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13089 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13092 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13093 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13095 2018-01-10 Richard Biener <rguenther@suse.de>
13098 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13099 early out so it also covers the case where we have a non-NULL
13102 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13104 PR tree-optimization/83753
13105 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13106 for non-strided grouped accesses if the number of elements is 1.
13108 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13111 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13112 * i386.h (TARGET_USE_GATHER): Define.
13113 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13115 2018-01-10 Martin Liska <mliska@suse.cz>
13118 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13119 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13121 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13122 CLEANUP_NO_PARTITIONING is not set.
13124 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13126 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13127 for vectors, as a partial revert of r254296.
13128 * rtl.h (const_vec_p): Delete.
13129 (const_vec_duplicate_p): Don't test for vector CONSTs.
13130 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13131 * expmed.c (make_tree): Likewise.
13134 * common.md (E, F): Use CONSTANT_P instead of checking for
13136 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13137 checking for CONST_VECTOR.
13139 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13141 PR middle-end/83575
13142 * predict.c (force_edge_cold): Handle in more sane way edges
13143 with no prediction.
13145 2018-01-09 Carl Love <cel@us.ibm.com>
13147 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13149 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13150 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13151 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13152 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13153 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13154 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13155 * config/rs6000/rs6000-protos.h: Add extern defition for
13156 rs6000_generate_float2_double_code.
13157 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13159 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13160 (float2_v2df): Add define_expand.
13162 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13165 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13166 op_mode in the force_to_mode call.
13168 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13170 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13171 instead of checking each element individually.
13172 (aarch64_evpc_uzp): Likewise.
13173 (aarch64_evpc_zip): Likewise.
13174 (aarch64_evpc_ext): Likewise.
13175 (aarch64_evpc_rev): Likewise.
13176 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13177 instead of checking each element individually. Return true without
13179 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13180 whether all selected elements come from the same input, instead of
13181 checking each element individually. Remove calls to gen_rtx_REG,
13182 start_sequence and end_sequence and instead assert that no rtl is
13185 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13187 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13188 order of HIGH and CONST checks.
13190 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13192 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13193 if the destination isn't an SSA_NAME.
13195 2018-01-09 Richard Biener <rguenther@suse.de>
13197 PR tree-optimization/83668
13198 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13200 (canonicalize_loop_form): ... here, renamed from ...
13201 (canonicalize_loop_closed_ssa_form): ... this and amended to
13202 swap successor edges for loop exit blocks to make us use
13203 the RPO order we need for initial schedule generation.
13205 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13207 PR tree-optimization/64811
13208 * match.pd: When optimizing comparisons with Inf, avoid
13209 introducing or losing exceptions from comparisons with NaN.
13211 2018-01-09 Martin Liska <mliska@suse.cz>
13214 * asan.c (shadow_mem_size): Add gcc_assert.
13216 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13218 Don't save registers in main().
13221 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13222 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13223 * config/avr/avr.c (avr_set_current_function): Don't error if
13224 naked, OS_task or OS_main are specified at the same time.
13225 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13227 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13229 * common/config/avr/avr-common.c (avr_option_optimization_table):
13230 Switch on -mmain-is-OS_task for optimizing compilations.
13232 2018-01-09 Richard Biener <rguenther@suse.de>
13234 PR tree-optimization/83572
13235 * graphite.c: Include cfganal.h.
13236 (graphite_transform_loops): Connect infinite loops to exit
13237 and remove fake edges at the end.
13239 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13241 * ipa-inline.c (edge_badness): Revert accidental checkin.
13243 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13246 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13247 symbols; not inline clones.
13249 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13252 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13253 hard registers. Formatting fixes.
13255 PR preprocessor/83722
13256 * gcc.c (try_generate_repro): Pass
13257 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13258 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13261 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13262 Kito Cheng <kito.cheng@gmail.com>
13264 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13265 (riscv_leaf_function_p): Delete.
13266 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13268 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13270 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13272 (do_ifelse): New function.
13273 (do_isel): New function.
13274 (do_sub3): New function.
13275 (do_add3): New function.
13276 (do_load_mask_compare): New function.
13277 (do_overlap_load_compare): New function.
13278 (expand_compare_loop): New function.
13279 (expand_block_compare): Call expand_compare_loop() when appropriate.
13280 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13281 option description.
13282 (-mblock-compare-inline-loop-limit): New option.
13284 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13287 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13288 Reverse order of second and third operands in first alternative.
13289 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13290 of first and second elements in UNSPEC_VPERMR vector.
13291 (altivec_expand_vec_perm_le): Likewise.
13293 2018-01-08 Jeff Law <law@redhat.com>
13295 PR rtl-optimizatin/81308
13296 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13297 (process_switch): If group_case_labels makes a change, then set
13299 (pass_convert_switch::execute): If a switch is converted, then
13300 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13302 PR rtl-optimization/81308
13303 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13306 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13308 PR target/83663 - Revert r255946
13309 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13310 generation for cases where splatting a value is not useful.
13311 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13312 across a vec_duplicate and a paradoxical subreg forming a vector
13313 mode to a vec_concat.
13315 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13317 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13318 -march=armv8.3-a variants.
13319 * config/arm/t-multilib: Likewise.
13320 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13322 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13324 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13326 (cceq_ior_compare_complement): Give it a name so I can use it, and
13327 change boolean_or_operator predicate to boolean_operator so it can
13328 be used to generate a crand.
13329 (eqne): New code iterator.
13330 (bd/bd_neg): New code_attrs.
13331 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13332 a single define_insn.
13333 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13334 decrement (bdnzt/bdnzf/bdzt/bdzf).
13335 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13336 with the new names of the branch decrement patterns, and added the
13337 names of the branch decrement conditional patterns.
13339 2018-01-08 Richard Biener <rguenther@suse.de>
13341 PR tree-optimization/83563
13342 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13345 2018-01-08 Richard Biener <rguenther@suse.de>
13347 PR middle-end/83713
13348 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13350 2018-01-08 Richard Biener <rguenther@suse.de>
13352 PR tree-optimization/83685
13353 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13354 references to abnormals.
13356 2018-01-08 Richard Biener <rguenther@suse.de>
13359 * dwarf2out.c (output_indirect_strings): Handle empty
13360 skeleton_debug_str_hash.
13361 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13363 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13365 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13366 (emit_store_direct): Likewise.
13367 (arc_trampoline_adjust_address): Likewise.
13368 (arc_asm_trampoline_template): New function.
13369 (arc_initialize_trampoline): Use asm_trampoline_template.
13370 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13371 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13372 * config/arc/arc.md (flush_icache): Delete pattern.
13374 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13376 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13377 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13380 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13383 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13384 by not USED_FOR_TARGET.
13385 (make_pass_resolve_sw_modes): Likewise.
13387 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13389 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13392 2018-01-08 Richard Biener <rguenther@suse.de>
13394 PR middle-end/83580
13395 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13397 2018-01-08 Richard Biener <rguenther@suse.de>
13399 PR middle-end/83517
13400 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13402 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13404 PR middle-end/81897
13405 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13406 basic blocks with a small number of successors.
13407 (convert_control_dep_chain_into_preds): Improve handling of
13409 (dump_predicates): Split apart into...
13410 (dump_pred_chain): ...here...
13411 (dump_pred_info): ...and here.
13412 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13413 (can_chain_union_be_invalidated_p): Improve check for invalidation
13415 (uninit_uses_cannot_happen): Avoid unnecessary if
13416 convert_control_dep_chain_into_preds yielded nothing.
13418 2018-01-06 Martin Sebor <msebor@redhat.com>
13420 PR tree-optimization/83640
13421 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13422 subtracting negative offset from size.
13423 (builtin_access::overlap): Adjust offset bounds of the access to fall
13424 within the size of the object if possible.
13426 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13428 PR rtl-optimization/83699
13429 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13430 extract_bit_field_as_subreg to cases in which the extracted
13431 value is also a vector.
13433 * lra-constraints.c (process_alt_operands): Test for the equivalence
13434 substitutions when detecting a possible reload cycle.
13436 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13439 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13440 by default if flag_selective_schedling{,2}. Formatting fixes.
13442 PR rtl-optimization/83682
13443 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13444 if it has non-VECTOR_MODE element mode.
13445 (vec_duplicate_p): Likewise.
13447 PR middle-end/83694
13448 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13449 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13451 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13454 * config/i386/i386-builtin.def
13455 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13456 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13457 Require also OPTION_MASK_ISA_AVX512F in addition to
13458 OPTION_MASK_ISA_GFNI.
13459 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13460 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13461 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13462 to OPTION_MASK_ISA_GFNI.
13463 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13464 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13465 OPTION_MASK_ISA_AVX512BW.
13466 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13467 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13468 addition to OPTION_MASK_ISA_GFNI.
13469 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13470 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13471 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13472 to OPTION_MASK_ISA_GFNI.
13473 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13474 a requirement for all ISAs rather than any of them with a few
13476 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13478 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13479 bitmasks to be enabled with 3 exceptions, instead of requiring any
13480 enabled ISA with lots of exceptions.
13481 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13482 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13483 Change avx512bw in isa attribute to avx512f.
13484 * config/i386/sgxintrin.h: Add license boilerplate.
13485 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13486 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13487 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13488 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13490 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13491 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13492 temporarily sse2 rather than sse if not enabled already.
13495 * config/i386/sse.md (VI248_VLBW): Rename to ...
13496 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13497 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13498 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13499 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13500 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13501 mode iterator instead of VI248_VLBW.
13503 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13505 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13506 (record_modified): Skip clobbers; add debug output.
13507 (param_change_prob): Use sreal frequencies.
13509 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13511 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13512 punt for user-aligned variables.
13514 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13516 * tree-chrec.c (chrec_contains_symbols): Return true for
13519 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13522 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13523 of (x|y) == x for BICS pattern.
13525 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13527 PR tree-optimization/83605
13528 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13529 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13532 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13534 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13535 * config/epiphany/rtems.h: New file.
13537 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13538 Uros Bizjak <ubizjak@gmail.com>
13541 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13542 QIreg_operand instead of register_operand predicate.
13543 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13544 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13545 comments instead of -fmitigate[-_]rop.
13547 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13550 * cgraphunit.c (symbol_table::compile): Switch to text_section
13551 before calling assembly_start debug hook.
13552 * run-rtl-passes.c (run_rtl_passes): Likewise.
13555 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13557 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13558 range_int_cst_p rather than !symbolic_range_p before calling
13559 extract_range_from_multiplicative_op_1.
13561 2018-01-04 Jeff Law <law@redhat.com>
13563 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13564 redundant test in assertion.
13566 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13568 * doc/rtl.texi: Document machine_mode wrapper classes.
13570 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13572 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13573 using tree_to_uhwi.
13575 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13577 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13578 the VEC_PERM_EXPR fold to fail.
13580 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13583 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13584 to switched_sections.
13586 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13589 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13590 test for d.testing.
13592 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13595 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13596 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13598 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13601 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13602 is BLKmode and bitpos not zero or mode change is needed.
13604 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13607 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13610 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13613 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13614 instead of MULT rtx. Update all corresponding splitters.
13615 (*saddl_se): Ditto.
13616 (*ssub<modesuffix>): Ditto.
13617 (*ssubl_se): Ditto.
13618 (*cmp_sadd_di): Update split patterns.
13619 (*cmp_sadd_si): Ditto.
13620 (*cmp_sadd_sidi): Ditto.
13621 (*cmp_ssub_di): Ditto.
13622 (*cmp_ssub_si): Ditto.
13623 (*cmp_ssub_sidi): Ditto.
13624 * config/alpha/predicates.md (const23_operand): New predicate.
13625 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13626 Look for ASHIFT, not MULT inner operand.
13627 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13629 2018-01-04 Martin Liska <mliska@suse.cz>
13631 PR gcov-profile/83669
13632 * gcov.c (output_intermediate_file): Add version to intermediate
13634 * doc/gcov.texi: Document new field 'version' in intermediate
13635 file format. Fix location of '-k' option of gcov command.
13637 2018-01-04 Martin Liska <mliska@suse.cz>
13640 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13642 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13644 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13646 2018-01-03 Martin Sebor <msebor@redhat.com>
13648 PR tree-optimization/83655
13649 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13650 checking calls with invalid arguments.
13652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13654 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13655 (vectorizable_mask_load_store): Delete.
13656 (vectorizable_call): Return false for masked loads and stores.
13657 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13658 instead of gimple_assign_rhs1.
13659 (vectorizable_load): Handle IFN_MASK_LOAD.
13660 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13664 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13666 (vectorizable_mask_load_store): ...here.
13667 (vectorizable_load): ...and here.
13669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13671 * tree-vect-stmts.c (vect_build_all_ones_mask)
13672 (vect_build_zero_merge_argument): New functions, split out from...
13673 (vectorizable_load): ...here.
13675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13679 (vectorizable_mask_load_store): ...here.
13680 (vectorizable_store): ...and here.
13682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13684 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13686 (vectorizable_mask_load_store): ...here.
13688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13690 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13691 (vect_model_store_cost): Take a vec_load_store_type instead of a
13693 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13694 (vect_model_store_cost): Take a vec_load_store_type instead of a
13696 (vectorizable_mask_load_store): Update accordingly.
13697 (vectorizable_store): Likewise.
13698 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13703 IFN_MASK_LOAD calls here rather than...
13704 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13707 Alan Hayward <alan.hayward@arm.com>
13708 David Sherwood <david.sherwood@arm.com>
13710 * expmed.c (extract_bit_field_1): For vector extracts,
13711 fall back to extract_bit_field_as_subreg if vec_extract
13714 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13715 Alan Hayward <alan.hayward@arm.com>
13716 David Sherwood <david.sherwood@arm.com>
13718 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13719 they are variable or constant sized.
13720 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13721 slots for constant-sized data.
13723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13724 Alan Hayward <alan.hayward@arm.com>
13725 David Sherwood <david.sherwood@arm.com>
13727 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13728 handling COND_EXPRs with boolean comparisons, try to find a better
13729 basis for the mask type than the boolean itself.
13731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13733 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13734 is calculated and how it can be overridden.
13735 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13736 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13738 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13742 Alan Hayward <alan.hayward@arm.com>
13743 David Sherwood <david.sherwood@arm.com>
13745 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13746 Remove the mode argument.
13747 (aarch64_simd_valid_immediate): Remove the mode and inverse
13749 * config/aarch64/iterators.md (bitsize): New iterator.
13750 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13751 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13752 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13753 aarch64_simd_valid_immediate.
13754 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13755 (aarch64_reg_or_bic_imm): Likewise.
13756 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13757 with an insn_type enum and msl with a modifier_type enum.
13758 Replace element_width with a scalar_mode. Change the shift
13759 to unsigned int. Add constructors for scalar_float_mode and
13760 scalar_int_mode elements.
13761 (aarch64_vect_float_const_representable_p): Delete.
13762 (aarch64_can_const_movi_rtx_p)
13763 (aarch64_simd_scalar_immediate_valid_for_move)
13764 (aarch64_simd_make_constant): Update call to
13765 aarch64_simd_valid_immediate.
13766 (aarch64_advsimd_valid_immediate_hs): New function.
13767 (aarch64_advsimd_valid_immediate): Likewise.
13768 (aarch64_simd_valid_immediate): Remove mode and inverse
13769 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13770 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13771 and aarch64_float_const_representable_p on the result.
13772 (aarch64_output_simd_mov_immediate): Remove mode argument.
13773 Update call to aarch64_simd_valid_immediate and use of
13774 simd_immediate_info.
13775 (aarch64_output_scalar_simd_mov_immediate): Update call
13778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13779 Alan Hayward <alan.hayward@arm.com>
13780 David Sherwood <david.sherwood@arm.com>
13782 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13783 (mode_nunits): Likewise CONST_MODE_NUNITS.
13784 * machmode.def (ADJUST_NUNITS): Document.
13785 * genmodes.c (mode_data::need_nunits_adj): New field.
13786 (blank_mode): Update accordingly.
13787 (adj_nunits): New variable.
13788 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13790 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13791 listed in adj_nunits.
13792 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13793 listed in adj_nunits. Don't emit case statements for such modes.
13794 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13795 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13796 nothing if adj_nunits is nonnull.
13797 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13798 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13799 (emit_mode_fbit): Update use of print_maybe_const_decl.
13800 (emit_move_size): Likewise. Treat the array as non-const
13802 (emit_mode_adjustments): Handle adj_nunits.
13804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13806 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13807 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13808 (VECTOR_MODES): Use it.
13809 (make_vector_modes): Take the prefix as an argument.
13811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13812 Alan Hayward <alan.hayward@arm.com>
13813 David Sherwood <david.sherwood@arm.com>
13815 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13816 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13817 for MODE_VECTOR_BOOL.
13818 * machmode.def (VECTOR_BOOL_MODE): Document.
13819 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13820 (make_vector_bool_mode): New function.
13821 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13823 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13824 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13826 * stor-layout.c (int_mode_for_mode): Likewise.
13827 * tree.c (build_vector_type_for_mode): Likewise.
13828 * varasm.c (output_constant_pool_2): Likewise.
13829 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13830 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13831 for MODE_VECTOR_BOOL.
13832 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13833 of mode class checks.
13834 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13835 instead of a list of mode class checks.
13836 (expand_vector_scalar_condition): Likewise.
13837 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13839 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13840 Alan Hayward <alan.hayward@arm.com>
13841 David Sherwood <david.sherwood@arm.com>
13843 * machmode.h (mode_size): Change from unsigned short to
13845 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13846 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13847 or if measurement_type is not polynomial.
13848 (fixed_size_mode::includes_p): Check for constant-sized modes.
13849 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13850 return a poly_uint16 rather than an unsigned short.
13851 (emit_mode_size): Change the type of mode_size from unsigned short
13852 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13853 (emit_mode_adjustments): Cope with polynomial vector sizes.
13854 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13856 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13858 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13859 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13860 * caller-save.c (setup_save_areas): Likewise.
13861 (replace_reg_with_saved_mem): Likewise.
13862 * calls.c (emit_library_call_value_1): Likewise.
13863 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13864 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13865 (gen_lowpart_for_combine): Likewise.
13866 * convert.c (convert_to_integer_1): Likewise.
13867 * cse.c (equiv_constant, cse_insn): Likewise.
13868 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13869 (cselib_subst_to_values): Likewise.
13870 * dce.c (word_dce_process_block): Likewise.
13871 * df-problems.c (df_word_lr_mark_ref): Likewise.
13872 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13873 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13874 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13875 (rtl_for_decl_location): Likewise.
13876 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13877 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13878 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13879 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13880 (expand_expr_real_1): Likewise.
13881 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13882 (pad_below): Likewise.
13883 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13884 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13885 * ira.c (get_subreg_tracking_sizes): Likewise.
13886 * ira-build.c (ira_create_allocno_objects): Likewise.
13887 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13888 (ira_sort_regnos_for_alter_reg): Likewise.
13889 * ira-costs.c (record_operand_costs): Likewise.
13890 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13891 (resolve_simple_move): Likewise.
13892 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13893 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13894 (lra_constraints): Likewise.
13895 (CONST_POOL_OK_P): Reject variable-sized modes.
13896 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13897 (add_pseudo_to_slot, lra_spill): Likewise.
13898 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13899 * optabs-query.c (get_best_extraction_insn): Likewise.
13900 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13901 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13902 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13903 * recog.c (offsettable_address_addr_space_p): Likewise.
13904 * regcprop.c (maybe_mode_change): Likewise.
13905 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13906 * regrename.c (build_def_use): Likewise.
13907 * regstat.c (dump_reg_info): Likewise.
13908 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13909 (find_reloads, find_reloads_subreg_address): Likewise.
13910 * reload1.c (eliminate_regs_1): Likewise.
13911 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13912 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13913 (simplify_binary_operation_1, simplify_subreg): Likewise.
13914 * targhooks.c (default_function_arg_padding): Likewise.
13915 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13916 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13917 (verify_gimple_assign_ternary): Likewise.
13918 * tree-inline.c (estimate_move_cost): Likewise.
13919 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13920 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13921 (get_address_cost_ainc): Likewise.
13922 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13923 (vect_supportable_dr_alignment): Likewise.
13924 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13925 (vectorizable_reduction): Likewise.
13926 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13927 (vectorizable_operation, vectorizable_load): Likewise.
13928 * tree.c (build_same_sized_truth_vector_type): Likewise.
13929 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13930 * var-tracking.c (emit_note_insn_var_location): Likewise.
13931 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13932 (ADDR_VEC_ALIGN): Likewise.
13934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13935 Alan Hayward <alan.hayward@arm.com>
13936 David Sherwood <david.sherwood@arm.com>
13938 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13940 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13941 or if measurement_type is polynomial.
13942 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13943 * combine.c (make_extraction): Likewise.
13944 * dse.c (find_shift_sequence): Likewise.
13945 * dwarf2out.c (mem_loc_descriptor): Likewise.
13946 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13947 (extract_bit_field, extract_low_bits): Likewise.
13948 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13949 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13950 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13951 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13952 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13953 * reload.c (find_reloads): Likewise.
13954 * reload1.c (alter_reg): Likewise.
13955 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13956 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13957 * tree-if-conv.c (predicate_mem_writes): Likewise.
13958 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13959 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13960 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13961 * valtrack.c (dead_debug_insert_temp): Likewise.
13962 * varasm.c (mergeable_constant_section): Likewise.
13963 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13966 Alan Hayward <alan.hayward@arm.com>
13967 David Sherwood <david.sherwood@arm.com>
13969 * expr.c (expand_assignment): Cope with polynomial mode sizes
13970 when assigning to a CONCAT.
13972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13973 Alan Hayward <alan.hayward@arm.com>
13974 David Sherwood <david.sherwood@arm.com>
13976 * machmode.h (mode_precision): Change from unsigned short to
13978 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13980 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13981 or if measurement_type is not polynomial.
13982 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13983 in which the mode is already known to be a scalar_int_mode.
13984 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13985 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13987 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13988 for GET_MODE_PRECISION.
13989 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13990 for GET_MODE_PRECISION.
13991 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13993 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13994 (expand_field_assignment, make_extraction): Likewise.
13995 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13996 (get_last_value): Likewise.
13997 * convert.c (convert_to_integer_1): Likewise.
13998 * cse.c (cse_insn): Likewise.
13999 * expr.c (expand_expr_real_1): Likewise.
14000 * lra-constraints.c (simplify_operand_subreg): Likewise.
14001 * optabs-query.c (can_atomic_load_p): Likewise.
14002 * optabs.c (expand_atomic_load): Likewise.
14003 (expand_atomic_store): Likewise.
14004 * ree.c (combine_reaching_defs): Likewise.
14005 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14006 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14007 * tree.h (type_has_mode_precision_p): Likewise.
14008 * ubsan.c (instrument_si_overflow): Likewise.
14010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14011 Alan Hayward <alan.hayward@arm.com>
14012 David Sherwood <david.sherwood@arm.com>
14014 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14015 polynomial numbers of units.
14016 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14017 (valid_vector_subparts_p): New function.
14018 (build_vector_type): Remove temporary shim and take the number
14019 of units as a poly_uint64 rather than an int.
14020 (build_opaque_vector_type): Take the number of units as a
14021 poly_uint64 rather than an int.
14022 * tree.c (build_vector_from_ctor): Handle polynomial
14023 TYPE_VECTOR_SUBPARTS.
14024 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14025 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14026 (build_vector_from_val): If the number of units is variable,
14027 use build_vec_duplicate_cst for constant operands and
14028 VEC_DUPLICATE_EXPR otherwise.
14029 (make_vector_type): Remove temporary is_constant ().
14030 (build_vector_type, build_opaque_vector_type): Take the number of
14031 units as a poly_uint64 rather than an int.
14032 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14034 * cfgexpand.c (expand_debug_expr): Likewise.
14035 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14036 (store_constructor, expand_expr_real_1): Likewise.
14037 (const_scalar_mask_from_tree): Likewise.
14038 * fold-const-call.c (fold_const_reduction): Likewise.
14039 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14040 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14041 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14042 (fold_relational_const): Likewise.
14043 (native_interpret_vector): Likewise. Change the size from an
14044 int to an unsigned int.
14045 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14046 TYPE_VECTOR_SUBPARTS.
14047 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14048 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14049 duplicating a non-constant operand into a variable-length vector.
14050 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14051 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14052 * ipa-icf.c (sem_variable::equals): Likewise.
14053 * match.pd: Likewise.
14054 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14055 * print-tree.c (print_node): Likewise.
14056 * stor-layout.c (layout_type): Likewise.
14057 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14058 * tree-cfg.c (verify_gimple_comparison): Likewise.
14059 (verify_gimple_assign_binary): Likewise.
14060 (verify_gimple_assign_ternary): Likewise.
14061 (verify_gimple_assign_single): Likewise.
14062 * tree-pretty-print.c (dump_generic_node): Likewise.
14063 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14064 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14065 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14066 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14067 (vect_shift_permute_load_chain): Likewise.
14068 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14069 (expand_vector_condition, optimize_vector_constructor): Likewise.
14070 (lower_vec_perm, get_compute_type): Likewise.
14071 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14072 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14073 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14074 (vect_recog_mask_conversion_pattern): Likewise.
14075 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14076 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14077 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14078 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14079 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14080 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14081 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14082 (supportable_widening_operation): Likewise.
14083 (supportable_narrowing_operation): Likewise.
14084 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14086 * varasm.c (output_constant): Likewise.
14088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14089 Alan Hayward <alan.hayward@arm.com>
14090 David Sherwood <david.sherwood@arm.com>
14092 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14093 so that both the length == 3 and length != 3 cases set up their
14094 own permute vectors. Add comments explaining why we know the
14095 number of elements is constant.
14096 (vect_permute_load_chain): Likewise.
14098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14099 Alan Hayward <alan.hayward@arm.com>
14100 David Sherwood <david.sherwood@arm.com>
14102 * machmode.h (mode_nunits): Change from unsigned char to
14104 (ONLY_FIXED_SIZE_MODES): New macro.
14105 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14106 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14107 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14109 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14110 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14111 or if measurement_type is not polynomial.
14112 * genmodes.c (ZERO_COEFFS): New macro.
14113 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14115 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14116 Use ZERO_COEFFS when emitting initializers.
14117 * data-streamer.h (bp_pack_poly_value): New function.
14118 (bp_unpack_poly_value): Likewise.
14119 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14120 for GET_MODE_NUNITS.
14121 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14122 for GET_MODE_NUNITS.
14123 * tree.c (make_vector_type): Remove temporary shim and make
14124 the real function take the number of units as a poly_uint64
14125 rather than an int.
14126 (build_vector_type_for_mode): Handle polynomial nunits.
14127 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14128 * emit-rtl.c (const_vec_series_p_1): Likewise.
14129 (gen_rtx_CONST_VECTOR): Likewise.
14130 * fold-const.c (test_vec_duplicate_folding): Likewise.
14131 * genrecog.c (validate_pattern): Likewise.
14132 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14133 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14134 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14135 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14136 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14137 * rtlanal.c (subreg_get_info): Likewise.
14138 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14139 (vect_grouped_load_supported): Likewise.
14140 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14141 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14142 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14143 (simplify_const_unary_operation, simplify_binary_operation_1)
14144 (simplify_const_binary_operation, simplify_ternary_operation)
14145 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14146 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14147 instead of CONST_VECTOR_NUNITS.
14148 * varasm.c (output_constant_pool_2): Likewise.
14149 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14150 explicit-encoded elements in the XVEC for variable-length vectors.
14152 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14154 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14157 Alan Hayward <alan.hayward@arm.com>
14158 David Sherwood <david.sherwood@arm.com>
14160 * coretypes.h (fixed_size_mode): Declare.
14161 (fixed_size_mode_pod): New typedef.
14162 * builtins.h (target_builtins::x_apply_args_mode)
14163 (target_builtins::x_apply_result_mode): Change type to
14164 fixed_size_mode_pod.
14165 * builtins.c (apply_args_size, apply_result_size, result_vector)
14166 (expand_builtin_apply_args_1, expand_builtin_apply)
14167 (expand_builtin_return): Update accordingly.
14169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14171 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14172 * cselib.c (cselib_hash_rtx): Likewise.
14173 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14174 CONST_VECTOR encoding.
14176 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14177 Jeff Law <law@redhat.com>
14180 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14181 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14182 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14183 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14186 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14187 explicitly probe *sp in a noreturn function if there were any callee
14188 register saves or frame pointer is needed.
14190 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14193 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14194 BLKmode for ternary, binary or unary expressions.
14197 * var-tracking.c (delete_vta_debug_insn): New inline function.
14198 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14199 insns from get_insns () to NULL instead of each bb separately.
14200 Use delete_vta_debug_insn. No longer static.
14201 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14202 delete_vta_debug_insns callers.
14203 * rtl.h (delete_vta_debug_insns): Declare.
14204 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14205 instead of variable_tracking_main.
14207 2018-01-03 Martin Sebor <msebor@redhat.com>
14209 PR tree-optimization/83603
14210 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14211 arguments past the endof the argument list in functions declared
14212 without a prototype.
14213 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14214 Avoid checking when arguments are null.
14216 2018-01-03 Martin Sebor <msebor@redhat.com>
14219 * doc/extend.texi (attribute const): Fix a typo.
14220 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14221 issuing -Wsuggest-attribute for void functions.
14223 2018-01-03 Martin Sebor <msebor@redhat.com>
14225 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14226 offset_int::from instead of wide_int::to_shwi.
14227 (maybe_diag_overlap): Remove assertion.
14228 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14229 * gimple-ssa-sprintf.c (format_directive): Same.
14230 (parse_directive): Same.
14231 (sprintf_dom_walker::compute_format_length): Same.
14232 (try_substitute_return_value): Same.
14234 2018-01-03 Jeff Law <law@redhat.com>
14236 PR middle-end/83654
14237 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14238 non-constant residual for zero at runtime and avoid probing in
14239 that case. Reorganize code for trailing problem to mirror handling
14242 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14244 PR tree-optimization/83501
14245 * tree-ssa-strlen.c (get_string_cst): New.
14246 (handle_char_store): Call get_string_cst.
14248 2018-01-03 Martin Liska <mliska@suse.cz>
14250 PR tree-optimization/83593
14251 * tree-ssa-strlen.c: Include tree-cfg.h.
14252 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14253 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14254 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14256 (strlen_dom_walker::before_dom_children): Call
14257 gimple_purge_dead_eh_edges. Dump tranformation with details
14259 (strlen_dom_walker::before_dom_children): Update call by adding
14260 new argument cleanup_eh.
14261 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14263 2018-01-03 Martin Liska <mliska@suse.cz>
14266 * cif-code.def (VARIADIC_THUNK): New enum value.
14267 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14270 2018-01-03 Jan Beulich <jbeulich@suse.com>
14272 * sse.md (mov<mode>_internal): Tighten condition for when to use
14273 vmovdqu<ssescalarsize> for TI and OI modes.
14275 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14277 Update copyright years.
14279 2018-01-03 Martin Liska <mliska@suse.cz>
14282 * ipa-visibility.c (function_and_variable_visibility): Skip
14283 functions with noipa attribure.
14285 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14287 * gcc.c (process_command): Update copyright notice dates.
14288 * gcov-dump.c (print_version): Ditto.
14289 * gcov.c (print_version): Ditto.
14290 * gcov-tool.c (print_version): Ditto.
14291 * gengtype.c (create_file): Ditto.
14292 * doc/cpp.texi: Bump @copying's copyright year.
14293 * doc/cppinternals.texi: Ditto.
14294 * doc/gcc.texi: Ditto.
14295 * doc/gccint.texi: Ditto.
14296 * doc/gcov.texi: Ditto.
14297 * doc/install.texi: Ditto.
14298 * doc/invoke.texi: Ditto.
14300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14302 * vector-builder.h (vector_builder::m_full_nelts): Change from
14303 unsigned int to poly_uint64.
14304 (vector_builder::full_nelts): Update prototype accordingly.
14305 (vector_builder::new_vector): Likewise.
14306 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14307 (vector_builder::operator ==): Likewise.
14308 (vector_builder::finalize): Likewise.
14309 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14310 Take the number of elements as a poly_uint64 rather than an
14312 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14313 from unsigned int to poly_uint64.
14314 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14315 (vec_perm_indices::new_vector): Likewise.
14316 (vec_perm_indices::length): Likewise.
14317 (vec_perm_indices::nelts_per_input): Likewise.
14318 (vec_perm_indices::input_nelts): Likewise.
14319 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14320 number of elements per input as a poly_uint64 rather than an
14321 unsigned int. Use the original encoding for variable-length
14322 vectors, rather than clamping each individual element.
14323 For the second and subsequent elements in each pattern,
14324 clamp the step and base before clamping their sum.
14325 (vec_perm_indices::series_p): Handle polynomial element counts.
14326 (vec_perm_indices::all_in_range_p): Likewise.
14327 (vec_perm_indices_to_tree): Likewise.
14328 (vec_perm_indices_to_rtx): Likewise.
14329 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14330 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14331 (tree_vector_builder::new_binary_operation): Handle polynomial
14332 element counts. Return false if we need to know the number
14333 of elements at compile time.
14334 * fold-const.c (fold_vec_perm): Punt if the number of elements
14335 isn't known at compile time.
14337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14339 * vec-perm-indices.h (vec_perm_builder): Change element type
14340 from HOST_WIDE_INT to poly_int64.
14341 (vec_perm_indices::element_type): Update accordingly.
14342 (vec_perm_indices::clamp): Handle polynomial element_types.
14343 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14344 (vec_perm_indices::all_in_range_p): Likewise.
14345 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14347 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14348 polynomial vec_perm_indices element types.
14349 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14350 * fold-const.c (fold_vec_perm): Likewise.
14351 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14352 * tree-vect-generic.c (lower_vec_perm): Likewise.
14353 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14354 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14355 element type to HOST_WIDE_INT.
14357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14358 Alan Hayward <alan.hayward@arm.com>
14359 David Sherwood <david.sherwood@arm.com>
14361 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14362 rather than an int. Use plus_constant.
14363 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14364 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14367 Alan Hayward <alan.hayward@arm.com>
14368 David Sherwood <david.sherwood@arm.com>
14370 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14371 a HOST_WIDE_INT to a poly_int64.
14373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14374 Alan Hayward <alan.hayward@arm.com>
14375 David Sherwood <david.sherwood@arm.com>
14377 * calls.c (load_register_parameters): Cope with polynomial
14378 mode sizes. Require a constant size for BLKmode parameters
14379 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14380 forces a parameter to be padded at the lsb end in order to
14381 fill a complete number of words, require the parameter size
14382 to be ordered wrt UNITS_PER_WORD.
14384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14385 Alan Hayward <alan.hayward@arm.com>
14386 David Sherwood <david.sherwood@arm.com>
14388 * reload1.c (spill_stack_slot_width): Change element type
14389 from unsigned int to poly_uint64_pod.
14390 (alter_reg): Treat mode sizes as polynomial.
14392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14393 Alan Hayward <alan.hayward@arm.com>
14394 David Sherwood <david.sherwood@arm.com>
14396 * reload.c (complex_word_subreg_p): New function.
14397 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
14404 operands whose sizes aren't ordered.
14405 (match_reload): Refer to this check here.
14407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14408 Alan Hayward <alan.hayward@arm.com>
14409 David Sherwood <david.sherwood@arm.com>
14411 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14412 that the mode size is in the set {1, 2, 4, 8, 16}.
14414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14415 Alan Hayward <alan.hayward@arm.com>
14416 David Sherwood <david.sherwood@arm.com>
14418 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14419 Use plus_constant instead of gen_rtx_PLUS.
14421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14422 Alan Hayward <alan.hayward@arm.com>
14423 David Sherwood <david.sherwood@arm.com>
14425 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14426 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14427 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14428 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14429 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14430 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14431 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14432 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14433 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14434 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14436 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14437 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14438 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14439 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14440 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14441 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14442 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14443 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14444 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14445 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14447 * expr.c (emit_move_resolve_push): Treat the input and result
14448 of PUSH_ROUNDING as a poly_int64.
14449 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14450 (emit_push_insn): Likewise.
14451 * lra-eliminations.c (mark_not_eliminable): Likewise.
14452 * recog.c (push_operand): Likewise.
14453 * reload1.c (elimination_effects): Likewise.
14454 * rtlanal.c (nonzero_bits1): Likewise.
14455 * calls.c (store_one_arg): Likewise. Require the padding to be
14456 known at compile time.
14458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14459 Alan Hayward <alan.hayward@arm.com>
14460 David Sherwood <david.sherwood@arm.com>
14462 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14463 Use plus_constant instead of gen_rtx_PLUS.
14465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14466 Alan Hayward <alan.hayward@arm.com>
14467 David Sherwood <david.sherwood@arm.com>
14469 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14470 rather than an int.
14472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14473 Alan Hayward <alan.hayward@arm.com>
14474 David Sherwood <david.sherwood@arm.com>
14476 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14477 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14478 via stack temporaries. Treat the mode size as polynomial too.
14480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14481 Alan Hayward <alan.hayward@arm.com>
14482 David Sherwood <david.sherwood@arm.com>
14484 * expr.c (expand_expr_real_2): When handling conversions involving
14485 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14486 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14487 as a poly_uint64 too.
14489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14490 Alan Hayward <alan.hayward@arm.com>
14491 David Sherwood <david.sherwood@arm.com>
14493 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14496 Alan Hayward <alan.hayward@arm.com>
14497 David Sherwood <david.sherwood@arm.com>
14499 * combine.c (can_change_dest_mode): Handle polynomial
14500 REGMODE_NATURAL_SIZE.
14501 * expmed.c (store_bit_field_1): Likewise.
14502 * expr.c (store_constructor): Likewise.
14503 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14504 and polynomial REGMODE_NATURAL_SIZE.
14505 (gen_lowpart_common): Likewise.
14506 * reginfo.c (record_subregs_of_mode): Likewise.
14507 * rtlanal.c (read_modify_subreg_p): Likewise.
14509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14510 Alan Hayward <alan.hayward@arm.com>
14511 David Sherwood <david.sherwood@arm.com>
14513 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14514 numbers of elements.
14516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14517 Alan Hayward <alan.hayward@arm.com>
14518 David Sherwood <david.sherwood@arm.com>
14520 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14527 in a POINTER_PLUS_EXPR.
14529 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14530 Alan Hayward <alan.hayward@arm.com>
14531 David Sherwood <david.sherwood@arm.com>
14533 * omp-simd-clone.c (simd_clone_subparts): New function.
14534 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14535 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14542 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14543 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14544 (expand_vector_condition, vector_element): Likewise.
14545 (subparts_gt): New function.
14546 (get_compute_type): Use subparts_gt.
14547 (count_type_subparts): Delete.
14548 (expand_vector_operations_1): Use subparts_gt instead of
14549 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14556 (vect_compile_time_alias): ...this new function. Do the calculation
14557 on poly_ints rather than trees.
14558 (vect_prune_runtime_alias_test_list): Update call accordingly.
14560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14566 (vect_schedule_slp_instance): Likewise.
14568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14569 Alan Hayward <alan.hayward@arm.com>
14570 David Sherwood <david.sherwood@arm.com>
14572 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14573 constant and extern definitions for variable-length vectors.
14574 (vect_get_constant_vectors): Note that the number of units
14575 is known to be constant.
14577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14578 Alan Hayward <alan.hayward@arm.com>
14579 David Sherwood <david.sherwood@arm.com>
14581 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14582 of units as polynomial. Choose between WIDE and NARROW based
14585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14586 Alan Hayward <alan.hayward@arm.com>
14587 David Sherwood <david.sherwood@arm.com>
14589 * tree-vect-stmts.c (simd_clone_subparts): New function.
14590 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14593 Alan Hayward <alan.hayward@arm.com>
14594 David Sherwood <david.sherwood@arm.com>
14596 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14597 vectors as polynomial. Use build_index_vector for
14598 IFN_GOMP_SIMD_LANE.
14600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14601 Alan Hayward <alan.hayward@arm.com>
14602 David Sherwood <david.sherwood@arm.com>
14604 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14605 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14606 for variable-length vectors.
14607 (vectorizable_mask_load_store): Treat the number of units as
14608 polynomial, asserting that it is constant if the condition has
14609 already been enforced.
14610 (vectorizable_store, vectorizable_load): Likewise.
14612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14613 Alan Hayward <alan.hayward@arm.com>
14614 David Sherwood <david.sherwood@arm.com>
14616 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14617 of units as polynomial. Punt if we can't tell at compile time
14618 which vector contains the final result.
14620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14621 Alan Hayward <alan.hayward@arm.com>
14622 David Sherwood <david.sherwood@arm.com>
14624 * tree-vect-loop.c (vectorizable_induction): Treat the number
14625 of units as polynomial. Punt on SLP inductions. Use an integer
14626 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14627 cast of such a series for variable-length floating-point
14630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14631 Alan Hayward <alan.hayward@arm.com>
14632 David Sherwood <david.sherwood@arm.com>
14634 * tree.h (build_index_vector): Declare.
14635 * tree.c (build_index_vector): New function.
14636 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14637 of units as polynomial, forcibly converting it to a constant if
14638 vectorizable_reduction has already enforced the condition.
14639 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14640 to create a {1,2,3,...} vector.
14641 (vectorizable_reduction): Treat the number of units as polynomial.
14642 Choose vectype_in based on the largest scalar element size rather
14643 than the smallest number of units. Enforce the restrictions
14646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14647 Alan Hayward <alan.hayward@arm.com>
14648 David Sherwood <david.sherwood@arm.com>
14650 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14651 number of units as polynomial.
14653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14654 Alan Hayward <alan.hayward@arm.com>
14655 David Sherwood <david.sherwood@arm.com>
14657 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14658 * target.def (autovectorize_vector_sizes): Return the vector sizes
14659 by pointer, using vector_sizes rather than a bitmask.
14660 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14661 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14662 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14664 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14665 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14666 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14667 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14668 * omp-general.c (omp_max_vf): Likewise.
14669 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14670 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14671 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14672 * tree-vect-slp.c (vect_slp_bb): Likewise.
14673 * doc/tm.texi: Regenerate.
14674 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14676 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14677 the vector size as a poly_uint64 rather than an unsigned int.
14678 (current_vector_size): Change from an unsigned int to a poly_uint64.
14679 (get_vectype_for_scalar_type): Update accordingly.
14680 * tree.h (build_truth_vector_type): Take the size and number of
14681 units as a poly_uint64 rather than an unsigned int.
14682 (build_vector_type): Add a temporary overload that takes
14683 the number of units as a poly_uint64 rather than an unsigned int.
14684 * tree.c (make_vector_type): Likewise.
14685 (build_truth_vector_type): Take the number of units as a poly_uint64
14686 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
14693 as poly_uint64s rather than unsigned ints.
14694 * targhooks.h (default_get_mask_mode): Update accordingly.
14695 * targhooks.c (default_get_mask_mode): Likewise.
14696 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14697 * doc/tm.texi: Regenerate.
14699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14700 Alan Hayward <alan.hayward@arm.com>
14701 David Sherwood <david.sherwood@arm.com>
14703 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14704 * omp-general.c (omp_max_vf): Likewise.
14705 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14706 (expand_omp_simd): Handle polynomial safelen.
14707 * omp-low.c (omplow_simd_context): Add a default constructor.
14708 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14709 (lower_rec_simd_input_clauses): Update accordingly.
14710 (lower_rec_input_clauses): Likewise.
14712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14713 Alan Hayward <alan.hayward@arm.com>
14714 David Sherwood <david.sherwood@arm.com>
14716 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14717 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14718 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14719 (vect_analyze_slp_cost): Likewise.
14720 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14721 (vect_model_load_cost): Likewise.
14723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14724 Alan Hayward <alan.hayward@arm.com>
14725 David Sherwood <david.sherwood@arm.com>
14727 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14728 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14729 from an unsigned int * to a poly_uint64_pod *.
14730 (calculate_unrolling_factor): New function.
14731 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14734 Alan Hayward <alan.hayward@arm.com>
14735 David Sherwood <david.sherwood@arm.com>
14737 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14738 from an unsigned int to a poly_uint64.
14739 (_loop_vec_info::slp_unrolling_factor): Likewise.
14740 (_loop_vec_info::vectorization_factor): Change from an int
14742 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14743 (vect_get_num_vectors): New function.
14744 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14745 (vect_get_num_copies): Use vect_get_num_vectors.
14746 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14747 to an unsigned int *.
14748 (vect_analyze_data_refs): Change min_vf from an int * to a
14750 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14751 than an unsigned HOST_WIDE_INT.
14752 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14753 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14754 to an unsigned int *.
14755 (vect_analyze_data_ref_dependences): Likewise.
14756 (vect_compute_data_ref_alignment): Handle polynomial vf.
14757 (vect_enhance_data_refs_alignment): Likewise.
14758 (vect_prune_runtime_alias_test_list): Likewise.
14759 (vect_shift_permute_load_chain): Likewise.
14760 (vect_supportable_dr_alignment): Likewise.
14761 (dependence_distance_ge_vf): Take the vectorization factor as a
14762 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14763 (vect_analyze_data_refs): Change min_vf from an int * to a
14765 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14766 vfm1 as a poly_uint64 rather than an int. Make the same change
14767 for the returned bound_scalar.
14768 (vect_gen_vector_loop_niters): Handle polynomial vf.
14769 (vect_do_peeling): Likewise. Update call to
14770 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14771 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14773 * tree-vect-loop.c (vect_determine_vectorization_factor)
14774 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14775 (vect_get_known_peeling_cost): Likewise.
14776 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14777 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14778 (vect_transform_loop): Likewise. Use the lowest possible VF when
14779 updating the upper bounds of the loop.
14780 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14781 rather than an int.
14782 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14783 polynomial unroll factors.
14784 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14785 (vect_make_slp_decision): Likewise.
14786 (vect_supported_load_permutation_p): Likewise, and polynomial
14788 (vect_analyze_slp_cost): Handle polynomial vf.
14789 (vect_slp_analyze_node_operations): Likewise.
14790 (vect_slp_analyze_bb_1): Likewise.
14791 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14792 than an unsigned HOST_WIDE_INT.
14793 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14794 (vectorizable_load): Handle polynomial vf.
14795 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14797 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14800 Alan Hayward <alan.hayward@arm.com>
14801 David Sherwood <david.sherwood@arm.com>
14803 * match.pd: Handle bit operations involving three constants
14804 and try to fold one pair.
14806 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14808 * tree-vect-loop-manip.c: Include gimple-fold.h.
14809 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14810 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14811 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14812 Add a path that uses a step of VF instead of 1, but disable it
14814 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14815 and niters_no_overflow parameters. Update calls to
14816 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14817 Create a new SSA name if the latter choses to use a ste other
14818 than zero, and return it via niters_vector_mult_vf_var.
14819 * tree-vect-loop.c (vect_transform_loop): Update calls to
14820 vect_do_peeling, vect_gen_vector_loop_niters and
14821 slpeel_make_loop_iterate_ntimes.
14822 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14823 (vect_gen_vector_loop_niters): Update declarations after above changes.
14825 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14827 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14828 128-bit round to integer instructions.
14829 (ceil<mode>2): Likewise.
14830 (btrunc<mode>2): Likewise.
14831 (round<mode>2): Likewise.
14833 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14835 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14836 unaligned VSX load/store on P8/P9.
14837 (expand_block_clear): Allow the use of unaligned VSX
14838 load/store on P8/P9.
14840 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14842 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14844 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14845 swap associated with both a load and a store.
14847 2018-01-02 Andrew Waterman <andrew@sifive.com>
14849 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14850 * config/riscv/riscv.md (clear_cache): Use it.
14852 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14854 * web.c: Remove out-of-date comment.
14856 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14858 * expr.c (fixup_args_size_notes): Check that any existing
14859 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14860 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14861 (emit_single_push_insn): ...here.
14863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14865 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14866 (const_vector_encoded_nelts): New function.
14867 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14868 (const_vector_int_elt, const_vector_elt): Declare.
14869 * emit-rtl.c (const_vector_int_elt_1): New function.
14870 (const_vector_elt): Likewise.
14871 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14872 of CONST_VECTOR_ELT.
14874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14876 * expr.c: Include rtx-vector-builder.h.
14877 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14878 directly on the tree encoding.
14879 (const_vector_from_tree): Likewise.
14880 * optabs.c: Include rtx-vector-builder.h.
14881 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14882 sequence of "u" values.
14883 * vec-perm-indices.c: Include rtx-vector-builder.h.
14884 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14885 directly on the vec_perm_indices encoding.
14887 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14889 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14890 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14891 * rtx-vector-builder.h: New file.
14892 * rtx-vector-builder.c: Likewise.
14893 * rtl.h (rtx_def::u2): Add a const_vector field.
14894 (CONST_VECTOR_NPATTERNS): New macro.
14895 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14896 (CONST_VECTOR_DUPLICATE_P): Likewise.
14897 (CONST_VECTOR_STEPPED_P): Likewise.
14898 (CONST_VECTOR_ENCODED_ELT): Likewise.
14899 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14900 (unwrap_const_vec_duplicate): Likewise.
14901 (const_vec_series_p): Check for a non-duplicated vector encoding.
14902 Say that the function only returns true for integer vectors.
14903 * emit-rtl.c: Include rtx-vector-builder.h.
14904 (gen_const_vec_duplicate_1): Delete.
14905 (gen_const_vector): Call gen_const_vec_duplicate instead of
14906 gen_const_vec_duplicate_1.
14907 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14908 (gen_const_vec_duplicate): Use rtx_vector_builder.
14909 (gen_const_vec_series): Likewise.
14910 (gen_rtx_CONST_VECTOR): Likewise.
14911 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14912 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14913 Build a new vector rather than modifying a CONST_VECTOR in-place.
14914 (handle_special_swappables): Update call accordingly.
14915 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14916 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14917 Build a new vector rather than modifying a CONST_VECTOR in-place.
14918 (handle_special_swappables): Update call accordingly.
14920 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14922 * simplify-rtx.c (simplify_const_binary_operation): Use
14923 CONST_VECTOR_ELT instead of XVECEXP.
14925 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14927 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14928 the selector elements to be different from the data elements
14929 if the selector is a VECTOR_CST.
14930 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14931 ssizetype for the selector.
14933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14935 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14936 before testing each element individually.
14937 * tree-vect-generic.c (lower_vec_perm): Likewise.
14939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14941 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14942 * selftest-run-tests.c (selftest::run_tests): Call it.
14943 * vector-builder.h (vector_builder::operator ==): New function.
14944 (vector_builder::operator !=): Likewise.
14945 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14946 (vec_perm_indices::all_from_input_p): New function.
14947 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14948 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14949 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14950 instead of reading the VECTOR_CST directly. Detect whether both
14951 vector inputs are the same before constructing the vec_perm_indices,
14952 and update the number of inputs argument accordingly. Use the
14953 utility functions added above. Only construct sel2 if we need to.
14955 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14957 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14958 the broadcast of the low byte.
14959 (expand_mult_highpart): Use an explicit encoding for the permutes.
14960 * optabs-query.c (can_mult_highpart_p): Likewise.
14961 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14962 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14963 (vectorizable_bswap): Likewise.
14964 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14965 explicit encoding for the power-of-2 permutes.
14966 (vect_permute_store_chain): Likewise.
14967 (vect_grouped_load_supported): Likewise.
14968 (vect_permute_load_chain): Likewise.
14970 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14972 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14973 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14974 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14975 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14976 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14977 (vect_gen_perm_mask_any): Likewise.
14979 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14981 * int-vector-builder.h: New file.
14982 * vec-perm-indices.h: Include int-vector-builder.h.
14983 (vec_perm_indices): Redefine as an int_vector_builder.
14984 (auto_vec_perm_indices): Delete.
14985 (vec_perm_builder): Redefine as a stand-alone class.
14986 (vec_perm_indices::vec_perm_indices): New function.
14987 (vec_perm_indices::clamp): Likewise.
14988 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14989 (vec_perm_indices::new_vector): New function.
14990 (vec_perm_indices::new_expanded_vector): Update for new
14991 vec_perm_indices class.
14992 (vec_perm_indices::rotate_inputs): New function.
14993 (vec_perm_indices::all_in_range_p): Operate directly on the
14994 encoded form, without computing elided elements.
14995 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14996 encoding. Update for new vec_perm_indices class.
14997 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14998 the given vec_perm_builder.
14999 (expand_vec_perm_var): Update vec_perm_builder constructor.
15000 (expand_mult_highpart): Use vec_perm_builder instead of
15001 auto_vec_perm_indices.
15002 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15003 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15004 or double series encoding as appropriate.
15005 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15006 vec_perm_indices instead of auto_vec_perm_indices.
15007 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15008 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15009 (vect_permute_store_chain): Likewise.
15010 (vect_grouped_load_supported): Likewise.
15011 (vect_permute_load_chain): Likewise.
15012 (vect_shift_permute_load_chain): Likewise.
15013 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15014 (vect_transform_slp_perm_load): Likewise.
15015 (vect_schedule_slp_instance): Likewise.
15016 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15017 (vectorizable_mask_load_store): Likewise.
15018 (vectorizable_bswap): Likewise.
15019 (vectorizable_store): Likewise.
15020 (vectorizable_load): Likewise.
15021 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15022 vec_perm_indices instead of auto_vec_perm_indices. Use
15023 tree_to_vec_perm_builder to read the vector from a tree.
15024 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15025 vec_perm_builder instead of a vec_perm_indices.
15026 (have_whole_vector_shift): Use vec_perm_builder and
15027 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15028 truncation to calc_vec_perm_mask_for_shift.
15029 (vect_create_epilog_for_reduction): Likewise.
15030 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15031 from auto_vec_perm_indices to vec_perm_indices.
15032 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15033 instead of changing individual elements.
15034 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15035 the vector in d.perm.
15036 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15037 from auto_vec_perm_indices to vec_perm_indices.
15038 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15039 instead of changing individual elements.
15040 (arm_vectorize_vec_perm_const): Use new_vector to install
15041 the vector in d.perm.
15042 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15043 Update vec_perm_builder constructor.
15044 (rs6000_expand_interleave): Likewise.
15045 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15046 (rs6000_expand_interleave): Likewise.
15048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15050 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15051 to qimode could truncate the indices.
15052 * optabs.c (expand_vec_perm_var): Likewise.
15054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15056 * Makefile.in (OBJS): Add vec-perm-indices.o.
15057 * vec-perm-indices.h: New file.
15058 * vec-perm-indices.c: Likewise.
15059 * target.h (vec_perm_indices): Replace with a forward class
15061 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15062 * optabs.h: Include vec-perm-indices.h.
15063 (expand_vec_perm): Delete.
15064 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15065 (expand_vec_perm_const): Declare.
15066 * target.def (vec_perm_const_ok): Replace with...
15067 (vec_perm_const): ...this new hook.
15068 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15069 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15070 * doc/tm.texi: Regenerate.
15071 * optabs.def (vec_perm_const): Delete.
15072 * doc/md.texi (vec_perm_const): Likewise.
15073 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15074 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15075 expand_vec_perm for constant permutation vectors. Assert that
15076 the mode of variable permutation vectors is the integer equivalent
15077 of the mode that is being permuted.
15078 * optabs-query.h (selector_fits_mode_p): Declare.
15079 * optabs-query.c: Include vec-perm-indices.h.
15080 (selector_fits_mode_p): New function.
15081 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15082 is defined, instead of checking whether the vec_perm_const_optab
15083 exists. Use targetm.vectorize.vec_perm_const instead of
15084 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15085 fit in the vector mode before using a variable permute.
15086 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15087 vec_perm_indices instead of an rtx.
15088 (expand_vec_perm): Replace with...
15089 (expand_vec_perm_const): ...this new function. Take the selector
15090 as a vec_perm_indices rather than an rtx. Also take the mode of
15091 the selector. Update call to shift_amt_for_vec_perm_mask.
15092 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15093 Use vec_perm_indices::new_expanded_vector to expand the original
15094 selector into bytes. Check whether the indices fit in the vector
15095 mode before using a variable permute.
15096 (expand_vec_perm_var): Make global.
15097 (expand_mult_highpart): Use expand_vec_perm_const.
15098 * fold-const.c: Includes vec-perm-indices.h.
15099 * tree-ssa-forwprop.c: Likewise.
15100 * tree-vect-data-refs.c: Likewise.
15101 * tree-vect-generic.c: Likewise.
15102 * tree-vect-loop.c: Likewise.
15103 * tree-vect-slp.c: Likewise.
15104 * tree-vect-stmts.c: Likewise.
15105 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15107 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15108 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15109 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15110 (aarch64_vectorize_vec_perm_const): ...this new function.
15111 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15112 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15113 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15114 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15115 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15116 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15117 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15119 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15120 check for NEON modes.
15121 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15122 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15123 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15124 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15126 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15127 the old VEC_PERM_CONST conditions.
15128 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15129 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15130 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15131 (ia64_vectorize_vec_perm_const_ok): Merge into...
15132 (ia64_vectorize_vec_perm_const): ...this new function.
15133 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15134 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15135 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15136 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15137 * config/mips/mips.c (mips_expand_vec_perm_const)
15138 (mips_vectorize_vec_perm_const_ok): Merge into...
15139 (mips_vectorize_vec_perm_const): ...this new function.
15140 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15141 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15142 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15143 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15144 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15145 (rs6000_expand_vec_perm_const): Delete.
15146 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15148 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15149 (altivec_expand_vec_perm_const_le): Take each operand individually.
15150 Operate on constant selectors rather than rtxes.
15151 (altivec_expand_vec_perm_const): Likewise. Update call to
15152 altivec_expand_vec_perm_const_le.
15153 (rs6000_expand_vec_perm_const): Delete.
15154 (rs6000_vectorize_vec_perm_const_ok): Delete.
15155 (rs6000_vectorize_vec_perm_const): New function.
15156 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15157 an element count and rtx array.
15158 (rs6000_expand_extract_even): Update call accordingly.
15159 (rs6000_expand_interleave): Likewise.
15160 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15161 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15162 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15163 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15164 (rs6000_expand_vec_perm_const): Delete.
15165 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15166 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15167 (altivec_expand_vec_perm_const_le): Take each operand individually.
15168 Operate on constant selectors rather than rtxes.
15169 (altivec_expand_vec_perm_const): Likewise. Update call to
15170 altivec_expand_vec_perm_const_le.
15171 (rs6000_expand_vec_perm_const): Delete.
15172 (rs6000_vectorize_vec_perm_const_ok): Delete.
15173 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15174 reference to the SPE evmerge intructions.
15175 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15176 an element count and rtx array.
15177 (rs6000_expand_extract_even): Update call accordingly.
15178 (rs6000_expand_interleave): Likewise.
15179 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15180 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15182 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15184 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15186 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15187 vector mode and that that mode matches the mode of the data
15189 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15190 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15191 directly using expand_vec_perm_1 when forcing selectors into
15193 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15195 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15197 * optabs-query.h (can_vec_perm_p): Delete.
15198 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15199 * optabs-query.c (can_vec_perm_p): Split into...
15200 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15201 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15202 particular selector is valid.
15203 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15204 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15205 (vect_grouped_load_supported): Likewise.
15206 (vect_shift_permute_load_chain): Likewise.
15207 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15208 (vect_transform_slp_perm_load): Likewise.
15209 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15210 (vectorizable_bswap): Likewise.
15211 (vect_gen_perm_mask_checked): Likewise.
15212 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15213 implementations of variable permutation vectors into account
15214 when deciding which selector to use.
15215 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15216 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15217 with a false third argument.
15218 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15219 to test whether the constant selector is valid and can_vec_perm_var_p
15220 to test whether a variable selector is valid.
15222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15224 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15225 * optabs-query.c (can_vec_perm_p): Likewise.
15226 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15227 instead of vec_perm_indices.
15228 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15229 (vect_gen_perm_mask_checked): Likewise,
15230 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15231 (vect_gen_perm_mask_checked): Likewise,
15233 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15235 * optabs-query.h (qimode_for_vec_perm): Declare.
15236 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15237 (qimode_for_vec_perm): ...this new function.
15238 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15240 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15242 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15243 does not have a conditional at the top.
15245 2018-01-02 Richard Biener <rguenther@suse.de>
15247 * ipa-inline.c (big_speedup_p): Fix expression.
15249 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15252 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15255 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15259 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15260 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15261 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15262 cond_taken_branch_cost 3->4.
15264 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15266 PR tree-optimization/83581
15267 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15268 TODO_cleanup_cfg if any changes have been made.
15270 PR middle-end/83608
15271 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15272 convert_modes if target mode has the right side, but different mode
15275 PR middle-end/83609
15276 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15277 last argument when extracting from CONCAT. If either from_real or
15278 from_imag is NULL, use expansion through memory. If result is not
15279 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15280 the parts directly to inner mode, if even that fails, use expansion
15283 PR middle-end/83623
15284 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15285 check for bswap in mode rather than HImode and use that in expand_unop
15288 Copyright (C) 2018 Free Software Foundation, Inc.
15290 Copying and distribution of this file, with or without modification,
15291 are permitted in any medium without royalty provided the copyright
15292 notice and this notice are preserved.