1 2018-06-05 Martin Liska <mliska@suse.cz>
4 * doc/invoke.texi: Document how -fprofile-dir format
7 2018-06-05 Richard Biener <rguenther@suse.de>
9 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10 removal pretend DOM info isn't available so we do not update
11 it and only remove edges, not dominated blocks. Actually free
12 DOM info in case we removed something. Remove unreachable blocks.
13 (mfb_keep_latches): Work with either DOM info or marked backedges.
14 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
15 first. Mark backedges if DOM info isn't available.
16 (Re-)compute DOM info after cleanup_control_flow_pre.
18 2018-06-05 Richard Biener <rguenther@suse.de>
20 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
21 (locus_discrim_hasher::hash): Adjust.
22 (locus_discrim_hasher::equal): Likewise.
23 (next_discriminator_for_locus): Work on line directly.
24 (same_line_p): Pass in expanded locus1 as well.
25 (assign_discriminators): Avoid redundant location expansions.
27 2018-06-05 Richard Biener <rguenther@suse.de>
29 PR tree-optimization/86046
30 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
31 if required after clearing TREE_ADDRESSABLE.
33 2018-06-05 Richard Biener <rguenther@suse.de>
35 PR tree-optimization/86047
36 * tree-ssa-loop.c (for_each_index): Glob handling of all
37 decls and constants and really handle all of them.
39 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
42 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
43 qualifier_void_pointer and qualifier_const_void_pointer.
44 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
45 (arm_init_builtins): Handle the above.
46 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
47 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
48 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
51 2018-06-05 Martin Liska <mliska@suse.cz>
53 * auto-profile.c (read_autofdo_file): Do not use
54 gcov_ctr_summary struct.
55 (afdo_callsite_hot_enough_for_early_inline): Likewise.
56 * coverage.c (struct counts_entry): Likewise.
57 (read_counts_file): Read just single summary entry.
58 (get_coverage_counts): Use gcov_summary struct.
59 * coverage.h (get_coverage_counts): Likewise.
60 * gcov-dump.c (dump_working_sets): Likewise.
61 (tag_summary): Dump just single summary.
62 * gcov-io.c (gcov_write_summary): Write just histogram
64 (gcov_read_summary): Read just single summary.
65 (compute_working_sets): Use gcov_summary struct.
66 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
67 of GCOV_COUNTERS_SUMMABLE.
68 (GCOV_COUNTERS_SUMMABLE): Remove.
69 (GCOV_FIRST_VALUE_COUNTER): Replace with
70 GCOV_COUNTER_V_INTERVAL.
71 (struct gcov_ctr_summary): Remove.
72 (struct gcov_summary): Directly use fields of former
74 (compute_working_sets): Use gcov_summary struct.
75 * gcov.c (read_count_file): Do not use ctrs fields.
76 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
78 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
80 * profile.c: Likewise.
81 * profile.h: Likewise.
83 2018-06-05 Martin Liska <mliska@suse.cz>
86 * gcov.c (output_lines): Print working directory only
87 in intermediate format.
89 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
91 * config/s390/s390-builtin-types.def: Add void function type.
92 * config/s390/s390-builtins.def: Use the function type for the
95 2018-06-04 Jim Wilson <jimw@sifive.com>
97 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
99 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
100 and maybe_eh_return. Change regno to unsigned int. Use new args to
101 handle EH_RETURN_DATA_REGNO registers properly.
102 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
103 (riscv_expand_epilogue): Update comment. Change argument name and
104 type. Update code to use new name and type. Pass new args to
105 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
107 * config/riscv/riscv.md (NORMAL_RETURN): New.
108 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
109 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
110 (eh_return): Call gen_eh_return_internal and emit barrier.
111 (eh_return_internal): Call riscv_expand_epilogue.
113 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
115 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
116 bit_insertion field and declare can_be_merged_into method.
117 (merged_store_group::can_be_merged_into): New method.
118 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
119 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
120 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
122 2018-06-04 Richard Biener <rguenther@suse.de>
124 PR tree-optimization/85955
125 * builtins.c (fold_builtin_sincos): Convert pointers to
126 destination to appropriate type before dereferencing.
128 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
130 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
132 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
134 * expr.c (expand_expr_real_1): Force the operand into memory if
135 its TYPE_MODE is BLKmode and if there is no integer mode for
136 the number of bits being extracted.
138 2018-06-04 Jakub Jelinek <jakub@redhat.com>
142 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
143 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
145 2018-06-04 Richard Biener <rguenther@suse.de>
147 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
148 (cleanup_tree_cfg_noloop): ... single caller. Do
149 start_recording_case_labels later.
151 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
153 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
154 to _IMMINTRIN_H_INCLUDED.
155 * config/i386/pconfigintrin.h: Ditto.
156 * config/i386/waitpkgintrin.h: Ditto.
157 * config/i386/immintrin.h: Add includes for sgxintrin.h,
158 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
159 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
160 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
161 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
162 waitpkgintrin.h and cldemoteintrin.h.
164 2018-06-04 Richard Biener <rguenther@suse.de>
166 PR tree-optimization/86038
167 * tracer.c (find_best_successor): Check probability for
168 being initialized, bail out if not.
170 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
173 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
174 of bits to ignore when comparing architectures.
176 2018-06-04 Jakub Jelinek <jakub@redhat.com>
178 PR tree-optimization/69615
179 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
180 maximum or minimum of the type, try to merge it also as if
181 range1 is + [-, x - 1] or + [x + 1, -].
184 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
186 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
188 PR tree-optimization/86034
189 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
190 the unsigned bitfield type in a bit insertion sequence if it does not
191 have a larger precision than the bitfield size.
192 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
194 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
196 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
198 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
200 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
201 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
202 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
203 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
205 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
207 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
208 Disable -fdelete-null-pointer-checks for ELF toolchain.
210 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
211 Kito Cheng <kito.cheng@gmail.com>
213 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
214 (nds32le-*-*, nds32be-*-*): Integrate checking process.
215 (nds32*-*-*): Add glibc and uclibc conditions.
216 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
217 (TARGET_EXCEPT_UNWIND_INFO): Define.
218 * config/nds32/elf.h: New file.
219 * config/nds32/linux.h: New file.
220 * config/nds32/nds32-elf.opt: New file.
221 * config/nds32/nds32-linux.opt: New file.
222 * config/nds32/nds32-fp-as-gp.c
223 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
224 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
226 (nds32_asm_file_end): Ditto.
227 (nds32_print_operand): Ditto.
228 (nds32_insert_attributes): Ditto.
229 (nds32_init_libfuncs): New function.
230 (TARGET_HAVE_TLS): Define.
231 (TARGET_INIT_LIBFUNCS): Define.
232 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
234 (TARGET_ELF): Apply different mcmodel setting.
235 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
236 been migrated into elf.h and linux.h files.
237 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
238 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
239 (mcmodel): The content has been migrated into nds32-elf.opt and
240 nds32-linux.opt files.
241 * config/nds32/t-elf: New file.
242 * config/nds32/t-linux: New file.
244 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
245 Shiva Chen <shiva0217@gmail.com>
247 * config/nds32/constants.md (unspec_volatile_element): Add
248 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
249 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
251 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
252 (make_pass_nds32_fp_as_gp): Declare.
253 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
255 (nds32_asm_function_end_prologue): Remove unused asm output.
256 (nds32_asm_function_begin_epilogue): Remove unused asm output.
257 (nds32_asm_file_start): Output necessary fp_as_gp information.
258 (nds32_option_override): Adjust register usage.
259 (nds32_expand_prologue): Consider fp_as_gp situation.
260 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
261 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
264 (simple_return): Ditto.
265 (omit_fp_begin): Output special directive for fp_as_gp.
266 (omit_fp_end): Output special directive for fp_as_gp.
267 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
268 mforbid-fp-as-gp): New options.
270 2018-06-01 Mark Wielaard <mark@klomp.org>
272 * dwarf2out.c (dwarf2out_finish): Remove generation of
275 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
277 * gimple-ssa-store-merging.c: Include gimple-fold.h.
278 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
279 (struct merged_store_group): Add bit_insertion field.
280 (dump_char_array): Use standard hexadecimal format.
281 (merged_store_group::merged_store_group): Set bit_insertion to false.
282 (merged_store_group::apply_stores): Use optimal buffer size. Deal
283 with BIT_INSERT_EXPR stores. Move up code updating the mask and
284 also print the mask in the dump file.
285 (pass_store_merging::gate): Minor tweak.
286 (imm_store_chain_info::coalesce_immediate): Fix wrong association
287 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
288 stores with INTEGER_CST stores.
289 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
290 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
291 and use it throughout. Generate bit insertion sequences if need be.
292 (pass_store_merging::process_store): Remove redundant condition.
293 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
295 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
297 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
298 the 128-bit floating point types. Fix function comment.
300 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
302 * config/aarch64/aarch64-simd.md
303 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
305 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
308 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
310 PR tree-optimization/85989
311 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
313 (backprop::intersect_uses): Check it when deciding whether this
314 is a backedge reference.
315 (backprop::process_block): Add each phi to m_visited_phis
316 after visiting it, then clear it at the end.
318 2018-06-01 Richard Biener <rguenther@suse.de>
320 * tree-vectorizer.h (vect_dr_stmt): New function.
321 (vect_get_load_cost): Adjust.
322 (vect_get_store_cost): Likewise.
323 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
324 Use vect_dr_stmt instead of DR_SMTT.
325 (vect_record_base_alignments): Likewise.
326 (vect_calculate_target_alignment): Likewise.
327 (vect_compute_data_ref_alignment): Likewise and make static.
328 (vect_update_misalignment_for_peel): Likewise.
329 (vect_verify_datarefs_alignment): Likewise.
330 (vector_alignment_reachable_p): Likewise.
331 (vect_get_data_access_cost): Likewise. Pass down
332 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
333 (vect_get_peeling_costs_all_drs): Likewise.
334 (vect_peeling_hash_get_lowest_cost): Likewise.
335 (vect_enhance_data_refs_alignment): Likewise.
336 (vect_find_same_alignment_drs): Likewise.
337 (vect_analyze_data_refs_alignment): Likewise.
338 (vect_analyze_group_access_1): Likewise.
339 (vect_analyze_group_access): Likewise.
340 (vect_analyze_data_ref_access): Likewise.
341 (vect_analyze_data_ref_accesses): Likewise.
342 (vect_vfa_segment_size): Likewise.
343 (vect_small_gap_p): Likewise.
344 (vectorizable_with_step_bound_p): Likewise.
345 (vect_prune_runtime_alias_test_list): Likewise.
346 (vect_analyze_data_refs): Likewise.
347 (vect_supportable_dr_alignment): Likewise.
348 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
349 (vect_gen_prolog_loop_niters): Likewise.
350 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
351 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
353 (vect_recog_mask_conversion_pattern): Likewise.
354 (vect_try_gather_scatter_pattern): Likewise.
355 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
356 to vect_get_store_cost.
357 (vect_get_store_cost): Get stmt_info instead of DR.
358 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
359 (vect_get_load_cost): Get stmt_info instead of DR.
361 2018-06-01 Richard Biener <rguenther@suse.de>
364 * gimple-fold.c (var_decl_component_p): Also allow offsetted
365 vars wrapped in MEM_REFs.
367 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
369 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
370 Fix subreg tests so that we only return a choice between
371 GENERAL_REGS and FP_REGS if the original classes included both.
373 2018-06-01 Richard Biener <rguenther@suse.de>
376 * tree-ssa-structalias.c (get_function_part_constraint):
377 Handle NULL fi->decl.
378 (find_func_aliases_for_call): Properly handle indirect
380 (find_func_clobbers): Likewise.
381 (ipa_pta_execute): Likewise.
382 (create_variable_info_for): For functions that are ifunc_resolver
383 resolve to a varinfo that contains the result of the resolver
385 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
388 2018-05-31 Michael Collison <michael.collison@arm.com>
390 * config/aarch64/aarch64.md:
391 (*fix_to_zero_extenddfdi2): New pattern.
392 * gcc.target/aarch64/fix_extend1.c: New testcase.
394 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
398 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
399 and BUILT_IN_STRNCMP_EQ.
400 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
402 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
403 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
404 (gimple_fold_builtin): Likewise.
405 * tree-ssa-strlen.c (compute_string_length): New function.
406 (determine_min_obsize): New function.
407 (handle_builtin_string_cmp): New function to handle calls to
408 string compare functions.
409 (strlen_optimize_stmt): Add handling to builtin string compare
411 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
412 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
413 * tree.c (build_common_builtin_nodes): Add new defines of
414 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
416 2018-05-31 Jakub Jelinek <jakub@redhat.com>
419 * bb-reorder.c (pass_partition_blocks::gate): Return false for
420 functions with naked attribute.
422 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
424 * config/i386/sse.md (avx_vec_concat<mode>):
425 Substitute concat_tg_mode mode attribute with xtg_mode.
426 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
427 (concat_tg_mode): Remove mode attribute.
429 2018-05-31 Martin Sebor <msebor@redhat.com>
432 * calls.c (alloc_max_size): Correct a logic error/typo.
433 Treat excessive arguments as infinite. Warn for invalid arguments.
434 * doc/invoke.texi (-Walloc-size-larger-than): Update.
436 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
439 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
440 and movx for Haswell.
442 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
443 Cesar Philippidis <cesar@codesourcery.com>
446 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
447 when emitting error on private/firstprivate reductions.
448 * omp-low.c (lower_omp_target): Avoid reference-type processing
449 on pointers for firstprivate clause.
451 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
453 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
456 * config/aarch64/aarch64-simd.md
457 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
458 (aarch64_ld1_x3_<mode>): Likewise
459 (aarch64_st1x2<VALLDIF:mode>): Likewise
460 (aarch64_st1_x2_<mode>): Likewise
461 (aarch64_st1x3<VALLDIF:mode>): Likewise
462 (aarch64_st1_x3_<mode>): Likewise
463 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
464 (vld1_s8_x3): Likewise.
465 (vld1_u16_x3): Likewise.
466 (vld1_s16_x3): Likewise.
467 (vld1_u32_x3): Likewise.
468 (vld1_s32_x3): Likewise.
469 (vld1_u64_x3): Likewise.
470 (vld1_s64_x3): Likewise.
471 (vld1_f16_x3): Likewise.
472 (vld1_f32_x3): Likewise.
473 (vld1_f64_x3): Likewise.
474 (vld1_p8_x3): Likewise.
475 (vld1_p16_x3): Likewise.
476 (vld1_p64_x3): Likewise.
477 (vld1q_u8_x3): Likewise.
478 (vld1q_s8_x3): Likewise.
479 (vld1q_u16_x3): Likewise.
480 (vld1q_s16_x3): Likewise.
481 (vld1q_u32_x3): Likewise.
482 (vld1q_s32_x3): Likewise.
483 (vld1q_u64_x3): Likewise.
484 (vld1q_s64_x3): Likewise.
485 (vld1q_f16_x3): Likewise.
486 (vld1q_f32_x3): Likewise.
487 (vld1q_f64_x3): Likewise.
488 (vld1q_p8_x3): Likewise.
489 (vld1q_p16_x3): Likewise.
490 (vld1q_p64_x3): Likewise.
491 (vst1_s64_x2): Likewise.
492 (vst1_u64_x2): Likewise.
493 (vst1_f64_x2): Likewise.
494 (vst1_s8_x2): Likewise.
495 (vst1_p8_x2): Likewise.
496 (vst1_s16_x2): Likewise.
497 (vst1_p16_x2): Likewise.
498 (vst1_s32_x2): Likewise.
499 (vst1_u8_x2): Likewise.
500 (vst1_u16_x2): Likewise.
501 (vst1_u32_x2): Likewise.
502 (vst1_f16_x2): Likewise.
503 (vst1_f32_x2): Likewise.
504 (vst1_p64_x2): Likewise.
505 (vst1q_s8_x2): Likewise.
506 (vst1q_p8_x2): Likewise.
507 (vst1q_s16_x2): Likewise.
508 (vst1q_p16_x2): Likewise.
509 (vst1q_s32_x2): Likewise.
510 (vst1q_s64_x2): Likewise.
511 (vst1q_u8_x2): Likewise.
512 (vst1q_u16_x2): Likewise.
513 (vst1q_u32_x2): Likewise.
514 (vst1q_u64_x2): Likewise.
515 (vst1q_f16_x2): Likewise.
516 (vst1q_f32_x2): Likewise.
517 (vst1q_f64_x2): Likewise.
518 (vst1q_p64_x2): Likewise.
519 (vst1_s64_x3): Likewise.
520 (vst1_u64_x3): Likewise.
521 (vst1_f64_x3): Likewise.
522 (vst1_s8_x3): Likewise.
523 (vst1_p8_x3): Likewise.
524 (vst1_s16_x3): Likewise.
525 (vst1_p16_x3): Likewise.
526 (vst1_s32_x3): Likewise.
527 (vst1_u8_x3): Likewise.
528 (vst1_u16_x3): Likewise.
529 (vst1_u32_x3): Likewise.
530 (vst1_f16_x3): Likewise.
531 (vst1_f32_x3): Likewise.
532 (vst1_p64_x3): Likewise.
533 (vst1q_s8_x3): Likewise.
534 (vst1q_p8_x3): Likewise.
535 (vst1q_s16_x3): Likewise.
536 (vst1q_p16_x3): Likewise.
537 (vst1q_s32_x3): Likewise.
538 (vst1q_s64_x3): Likewise.
539 (vst1q_u8_x3): Likewise.
540 (vst1q_u16_x3): Likewise.
541 (vst1q_u32_x3): Likewise.
542 (vst1q_u64_x3): Likewise.
543 (vst1q_f16_x3): Likewise.
544 (vst1q_f32_x3): Likewise.
545 (vst1q_f64_x3): Likewise.
546 (vst1q_p64_x3): Likewise.
548 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
550 * config/msp430/msp430.c (msp430_output_labelref): Prepend
551 user_label_prefix to name.
553 * tree-core.h: Update comment about the format of NAME string
554 passed to handler in attribute_spec.
556 * config/msp430/msp430.md: Remove erroneous subreg expression from
557 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
558 zero_extend{q,h}isi2.
560 2018-05-30 Borislav Petkov <bp@suse.de>
562 * doc/extend.texi: Document some architecture specific
563 constraints and sort entries.
565 2018-05-30 Martin Sebor <msebor@redhat.com>
568 * builtins.c (expand_builtin_stpcpy_1): New function.
569 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
570 only if the former succeeds.
572 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
574 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
577 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
579 * doc/invoke.texi (-flinker-output): Document
581 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
583 * passes.c (ipa_write_summaries): Only modify statements if body
585 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
586 incrementally linking.
587 (ipa_passes): Likewise.
588 * lto-cgraph.c (lto_output_node): When incrementally linking do not
589 pass down resolution info.
590 * common.opt (flag_incremental_link): Update info.
591 * gcc.c (plugin specs): Turn flinker-output=* to
592 -plugin-opt=-linker-output-known
593 * toplev.c (compile_file): Also cut compilation when doing incremental
595 * flag-types. (enum lto_partition_model): Add
596 LTO_LINKER_OUTPUT_NOLTOREL.
597 (invoke.texi): Add -flinker-output docs.
598 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
599 link same way as WPA; do not stream in dead initializers.
601 * dwarf2out.c (dwarf2out_die_ref_for_decl,
602 darf2out_register_external_decl): Support incremental link.
604 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
606 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
608 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
610 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
611 it down to simple_object_copy_lto_debug_sections.
612 (run_gcc): Determine incremental LTO link time and configure
613 lto1 into non-wpa mode, disable renaming of debug sections.
615 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
617 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
618 descriptions of various incorrectly documented functions.
620 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
623 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
624 address check not strict.
626 2018-05-30 Richard Biener <rguenther@suse.de>
628 PR tree-optimization/85964
629 * tracer.c (better_p): Drop initialized count check, we only
630 call the function with initialized counts now.
631 (find_best_successor): Do find a best edge if one
632 has uninitialized count.
633 (find_best_predecessor): Likewise. Do BB frequency check only
634 if count is initialized.
636 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
638 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
639 (aarch64_ldrstr_offset_compare): New.
640 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
641 load/store orderings.
642 (aarch64_gen_adjusted_ldpstp): Likewise.
644 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
646 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
647 Check for subset of GENERAL_REGS and FP_REGS.
648 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
651 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
653 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
654 and wi::to_poly_offset. Add the current offset and then check
655 whether the sum fits, rather than using an unchecked addition of
656 a checked term. Check for a shwi rather than a uhwi.
657 * expr.c (get_bit_range): Use tree_to_poly_uint64.
658 (store_constructor): Use poly_int_tree_p.
659 (expand_expr_real_1): Likewise.
660 * function.c (assign_temp): Likewise.
661 * fold-const.c (const_binop): Use poly_int_tree_p and
663 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
665 * ipa-icf-gimple.c (func_checker::compare_operand): Use
666 to_poly_offset for MEM offsets.
667 * ipa-icf.c (sem_variable::equals): Likewise.
668 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
669 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
670 wi::to_poly_offset for BIT_FIELD_REF offsets.
671 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
673 * var-tracking.c (emit_note_insn_var_location): Use
676 2018-05-29 Jim Wilson <jimw@sifive.com>
678 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
680 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
683 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
684 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
686 (sse4_1_round<mode>2): Use nonimmediate_operand
687 for operand 1 predicate.
689 2018-05-29 Martin Sebor <msebor@redhat.com>
690 Richard Biener <rguenther@suse.de>
693 * calls.c (get_size_range): Call determine_value_range instead
695 * tree-vrp.h (determine_value_range): Declared new function.
696 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
698 2018-05-29 Richard Biener <rguenther@suse.de>
700 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
701 sure to use non-pattern stmts for get_earlier_stmt arguments.
702 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
703 called on pattern stmts.
704 (get_later_stmt): Likewise.
706 2018-05-29 Martin Liska <mliska@suse.cz>
708 PR gcov-profile/85759
709 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
712 2018-05-29 Jakub Jelinek <jakub@redhat.com>
714 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
716 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
720 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
721 VEC_PACK_FLOAT_EXPR): New tree codes.
722 * tree-pretty-print.c (op_code_prio): Handle
723 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
724 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
725 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
726 * tree-inline.c (estimate_operator_cost): Likewise.
727 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
728 * fold-const.c (const_binop): Likewise.
729 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
730 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
731 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
732 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
733 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
734 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
735 * expr.c (expand_expr_real_2): Likewise.
736 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
737 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
738 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
740 * optabs.c (expand_widen_pattern_expr): For
741 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
742 sign from result type rather than operand's type.
743 (expand_binop_directly): For vec_packu_float_optab and
744 vec_packs_float_optab allow result type to be different from operand's
746 * optabs-tree.c (optab_for_tree_code): Handle
747 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
748 VEC_PACK_FLOAT_EXPR. Formatting fixes.
749 * tree-vect-generic.c (expand_vector_operations_1): Handle
750 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
752 * tree-vect-stmts.c (supportable_widening_operation): Handle
754 (supportable_narrowing_operation): Handle FLOAT_EXPR.
755 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
756 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
757 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
758 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
760 (vec_pack<floatprefix>_float_<mode>): New expander.
761 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
763 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
764 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
765 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
766 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
767 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
769 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
770 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
771 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
772 VEC_PACK_FLOAT_EXPR): Document.
774 2018-05-29 Richard Biener <rguenther@suse.de>
776 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
778 (stmt_vec_info_vec): Make pointer.
779 (init_stmt_vec_info_vec): Remove.
780 (free_stmt_vec_info_vec): Likewise.
781 (set_stmt_vec_info_vec): New function.
782 (free_stmt_vec_infos): Likewise.
783 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
784 (set_vinfo_for_stmt): Likewise.
785 (get_earlier_stmt): Likewise.
786 (get_later_stmt): Likewise.
787 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
788 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
789 (vec_info::~vec_info): Free stmt_vec_infos.
790 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
791 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
792 (pass_slp_vectorize::execute): Likewise.
793 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
794 (free_stmt_vec_info_vec): Likewise.
795 (set_stmt_vec_info_vec): New function.
796 (free_stmt_vec_infos): Likewise.
797 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
798 the global stmt_vec_info_vec.
799 * tree-parloops.c (gather_scalar_reductions): Use
800 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
803 2018-05-29 Richard Biener <rguenther@suse.de>
805 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
807 2018-05-29 Martin Liska <mliska@suse.cz>
808 David Malcolm <dmalcolm@redhat.com>
810 * vec.c (test_reverse): New.
811 (vec_c_tests): Add new test.
812 * vec.h (vl_ptr>::reverse): New function.
814 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
816 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
818 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
821 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
823 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
825 2018-05-28 Richard Biener <rguenther@suse.de>
827 PR tree-optimization/85933
828 * tree-vect-data-refs.c (vect_record_base_alignments): Only
829 look at stmts marked as vectorizable.
831 2018-05-28 Richard Biener <rguenther@suse.de>
833 PR tree-optimization/85934
834 * tree-vect-generic.c (expand_vector_operations_1): Hoist
835 vector boolean check before scalar optimization.
837 2018-05-28 Jakub Jelinek <jakub@redhat.com>
839 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
842 2018-05-28 Mark Wielaard <mark@klomp.org>
844 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
845 if it is an expression containing a minus sign.
847 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
849 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
851 2018-05-27 Paul Koning <ni1d@arrl.net>
853 * config/pdp11/pdp11.md (truncsihi2): Remove.
855 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
856 Chung-Ju Wu <jasonwucj@gmail.com>
858 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
860 (unaligned_store_dw): Ditto.
861 * config/nds32/nds32-memory-manipulation.c
862 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
863 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
864 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
865 (emit_setmem_word_loop): Rename to ...
866 (emit_setmem_doubleword_loop): ... this.
867 (nds32_gen_dup_4_byte_to_word_value): New function.
868 (nds32_gen_dup_8_byte_to_double_word_value): New function.
869 (nds32_expand_setmem_loop): Refine implementation.
870 (nds32_expand_setmem_loop_v3m): Ditto.
871 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
874 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
876 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
878 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
880 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
881 (nds32_init_machine_status): Initialize machine->attr_naked_p and
882 machine->attr_no_prologue_p.
883 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
884 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
885 (nds32_expand_epilogue): Consider attr_naked_p.
886 (nds32_expand_epilogue_v3pop): Likewise.
887 (nds32_can_use_return_insn): Likewise.
888 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
889 attr_no_prologue_p fields.
890 * config/nds32/nds32.opt (mret-in-naked-func): New option.
892 2018-05-27 Jakub Jelinek <jakub@redhat.com>
895 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
898 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
900 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
902 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
904 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
906 (*<floatsuffix>floatv2div2sf2): Rename to ...
907 (*float<floatunssuffix>v2div2sf2): ... this.
908 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
909 (float<floatunssuffix>v2div2sf2_mask): ... this.
910 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
911 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
912 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
914 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
916 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
918 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
920 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
922 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
924 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
925 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
926 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
927 gen_ufix_truncv8dfv8si2.
928 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
929 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
930 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
931 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
932 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
933 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
934 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
935 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
937 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
941 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
943 2018-05-25 Jim Wilson <jimw@sifive.com>
945 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
946 * config/riscv/riscv.c (struct machine_function): Add
947 interrupt_handler_p and attribute_checked_p fields.
948 (riscv_attribute_table): Add interrupt.
949 (riscv_interrupt_type_p): New.
950 (riscv_save_reg_p): Save extra regs for interrupt handler.
951 (riscv_use_save_libcall): Return false for interrupt handler.
952 (riscv_first_stack_step): Add forward declaration.
953 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
954 for interrupt handler with large frame. Use it for saved reg list.
955 (riscv_expand_prologue): Move flag_stack_usage_info support to
956 eliminate duplication.
957 (riscv_expand_epilogue): Generate mret for interrupt handler.
958 (riscv_epilogue_uses): New.
959 (riscv_can_use_return_insn): Return false for interrupt handler.
960 (riscv_function_ok_for_sibcall): Likewise.
961 (riscv_set_current_function): Add interrupt handler support.
962 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
963 * config/riscv/riscv.md (UNSPECV_MRET): New.
965 (riscv_frflags, riscv_fsflags): Use tab after opcode.
967 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
969 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
971 PR tree-optimization/85712
972 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
973 this candidate has already been replaced in-situ by a copy.
975 2018-05-25 Jason Merrill <jason@redhat.com>
977 PR c++/80485 - inline function non-zero address.
978 * symtab.c (nonzero_address): Check DECL_COMDAT.
980 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
983 * config/alpha/alpha.md (ashlsi3): New insn pattern.
984 (*ashlsi_se): Rename from *ashldi_se. Define as sign
985 extension of SImode operation. Use const123_operand predicate.
987 (*saddl_se_1): Ditto.
989 (*ssubl_se_1): Ditto.
990 * config/alpha/predicates.md (const123_operand): New predicate.
991 * config/alpha/constraints.md (P): Use IN_RANGE.
993 2018-05-25 Richard Biener <rguenther@suse.de>
995 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
997 (ref_maybe_used_by_stmt_p): Likewise.
998 (stmt_may_clobber_ref_p): Likewise.
999 (stmt_may_clobber_ref_p_1): Likewise.
1000 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
1002 (ref_maybe_used_by_stmt_p): Likewise.
1003 (stmt_may_clobber_ref_p): Likewise.
1004 (stmt_may_clobber_ref_p_1): Likewise.
1005 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
1006 the alias oracle to disambiguate DRs with stmts DR analysis
1008 (vect_analyze_data_refs): Do not give up on not analyzable
1009 DRs for BB vectorization. Remove code truncating the dataref
1012 2018-05-25 Jakub Jelinek <jakub@redhat.com>
1015 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1016 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
1017 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
1019 2018-05-25 Richard Biener <rguenther@suse.de>
1021 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
1022 function, combining stmt data ref gathering and fatal analysis
1024 (vect_analyze_data_refs): Remove now redudnant code and simplify.
1025 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
1026 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
1027 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
1028 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
1030 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1032 PR tree-optimization/85720
1033 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
1034 SCC if all partitions are builtins.
1035 (version_loop_by_alias_check): New parameter. Generate cancelable
1036 runtime alias check if all partitions are builtins.
1037 (distribute_loop): Update call to above function.
1039 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1041 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1042 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1043 (parm_default_def_partition_arg): Ditto.
1044 (set_parm_default_def_partition): Ditto.
1045 (get_parm_default_def_partitions): Ditto and make it static.
1046 (get_undefined_value_partitions): Ditto and make it static.
1047 (remove_ssa_form): Refactor call to init_var_map here.
1048 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1049 computation for loop region.
1050 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1051 (register_default_def): Delete.
1052 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1053 (parm_default_def_partition_arg): Ditto.
1054 (set_parm_default_def_partition): Ditto.
1055 (get_parm_default_def_partitions): Ditto and make it static.
1056 (get_undefined_value_partitions): Ditto and make it static.
1057 (coalesce_with_default, coalesce_with_default): Update comment.
1058 (create_coalesce_list_for_region): New func factored out from
1059 create_outofssa_var_map.
1060 (populate_coalesce_list_for_outofssa): New func factored out from
1061 create_outofssa_var_map and coalesce_ssa_name.
1062 (create_outofssa_var_map): Delete.
1063 (coalesce_ssa_name): Refactor to support live range computation.
1064 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1065 (get_parm_default_def_partitions): Delete.
1066 (get_undefined_value_partitions): Ditto.
1067 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1068 computation for loop region.
1069 (new_tree_live_info, loe_visit_block): Ditto.
1070 (live_worklist, set_var_live_on_entry): Ditto.
1071 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1072 * tree-ssa-live.h (struct _var_map): New fields.
1073 (init_var_map): Change decl.
1074 (region_contains_p): New.
1076 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1078 * tree-ssa-live.h (live_merge_and_clear): Delete.
1080 2018-05-25 Richard Biener <rguenther@suse.de>
1083 * tree-dump.c (dequeue_and_dump): Remove access to removed
1084 operand 2 of a SWITCH_EXPR.
1086 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1088 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1089 multiplication and division.
1090 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1091 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1092 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1093 (cond_udiv_optab, cond_umod_optab): New optabs.
1094 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1095 (IFN_COND_RDIV): New internal functions.
1096 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1097 TRUNC_MOD_EXPR and RDIV_EXPR.
1098 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1099 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1101 (SVE_INT_BINARY): Include mult.
1102 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1103 (optab, sve_int_op): Handle mult.
1104 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1106 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1107 for SVE_INT_BINARY_SD.
1109 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1111 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1112 (optab, sve_int_op): Handle div and udiv.
1113 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1114 for SVE_INT_BINARY_SD.
1115 (*<optab><mode>3): New insn for the same.
1117 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1119 * tree-vect-patterns.c: Include predict.h.
1120 (vect_recog_divmod_pattern): Restrict check for division support
1121 to when optimizing for size.
1123 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1125 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1126 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1127 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1128 (gimple_match_op::set_op): Likewise.
1129 (gimple_resimplify4): Declare.
1130 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1131 (expr::gen_transform): Likewise.
1132 (decision_tree::gen): Generate a simplification routine for 4 operands.
1133 * gimple-match-head.c (gimple_simplify): Add an overload for
1134 4 operands. In the top-level function, handle up to 4 call
1135 arguments and call gimple_resimplify4.
1136 (gimple_resimplify4): New function.
1137 (build_call_internal): Pass a fourth operand.
1138 (maybe_push_to_seq): Likewise.
1139 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1140 Fold VEC_COND_EXPRs of an operation and a default value into
1141 an IFN_COND_* function if possible.
1142 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1144 (SVE_COND_FP_BINARY): Include them.
1145 (optab, sve_fp_op): Handle them.
1146 (SVE_INT_BINARY_REV): New code iterator.
1147 (SVE_COND_FP_BINARY_REV): New int iterator.
1148 (commutative): New int attribute.
1149 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1151 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1153 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1154 (*cond_<optab><mode>): New patterns for reversed operands.
1156 2018-05-25 Richard Biener <rguenther@suse.de>
1158 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1159 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1160 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1161 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1162 * tree-vect-data-refs.c (everywhere): Adjust users.
1163 * tree-vect-loop.c (everywhere): Likewise.
1164 * tree-vect-slp.c (everywhere): Likewise.
1165 * tree-vect-stmts.c (everywhere): Likewise.
1166 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1168 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1170 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1172 (gcc_cv_as_section_exclude): ... this.
1173 Try Solaris as #exclude syntax.
1174 * configure: Regenerate.
1175 * config.in: Regenerate.
1176 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1178 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1179 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1181 * varasm.c (default_elf_asm_named_section): Don't check if
1182 HAVE_GAS_SECTION_EXCLUDE is defined.
1184 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1186 * doc/md.texi: Update the documentation of the cond_* optabs
1187 to mention the new final operand. Fix GET_MODE_NUNITS call.
1188 Describe the scalar case too.
1189 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1190 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1192 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1193 (get_conditional_internal_fn): Update comment.
1194 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1195 accumulator value as a final argument to conditional functions.
1196 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1197 a define_expand and add an "else" operand. Assert for now that
1198 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1199 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1200 (*cond_<optab><mode>): New patterns.
1201 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1202 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1203 (UNSPEC_COND_EOR): Delete.
1204 (optab): Remove associated mappings.
1205 (SVE_INT_BINARY): New code iterator.
1206 (sve_int_op): Remove int attribute and add "minus" to the code
1208 (SVE_COND_INT_OP): Delete.
1209 (SVE_COND_FP_OP): Rename to...
1210 (SVE_COND_FP_BINARY): ...this.
1212 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1214 * optabs.c (can_reuse_operands_p): New function.
1215 (maybe_legitimize_operands): Try to reuse the results for
1218 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1220 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1221 Add {q} suffix to insn mnemonic.
1223 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1225 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1226 (msp430_warn_func_return): New.
1228 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1230 * fold-const.c (tree_nonzero_bits): New function.
1231 * fold-const.h (tree_nonzero_bits): Likewise.
1232 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1233 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1235 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1239 * varasm.c (assemble_alias): Check ifunc_resolver only on
1242 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1245 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1246 when memory input operand is handled.
1248 2018-05-24 Luis Machado <luis.machado@linaro.org>
1250 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1252 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1254 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1256 * match.pd: Delay FMA folds until after vectorization.
1258 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1261 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1262 address check not strict.
1264 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1266 * gimple-match.h (gimple_match_op): New class.
1267 (mprts_hook): Replace parameters with a gimple_match_op *.
1268 (maybe_build_generic_op): Likewise.
1269 (gimple_simplified_result_is_gimple_val): Replace parameters with
1270 a const gimple_match_op *.
1271 (gimple_simplify): Replace code_helper * and tree * parameters with
1272 a gimple_match_op * parameter.
1273 (gimple_resimplify1): Replace code_helper *, tree and tree *
1274 parameters with a gimple_match_op * parameter.
1275 (gimple_resimplify2): Likewise.
1276 (gimple_resimplify3): Likewise.
1277 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1278 parameters with a gimple_match_op * parameter.
1279 * gimple-match-head.c (gimple_simplify): Change prototypes of
1280 auto-generated functions to take a gimple_match_op * instead of
1281 separate code_helper * and tree * parameters. Make the same
1282 change in the top-level overload and update calls to the
1283 gimple_resimplify routines. Update calls to the auto-generated
1284 functions and to maybe_push_res_to_seq in the publicly-facing
1285 operation-specific gimple_simplify overloads.
1286 (gimple_match_op::MAX_NUM_OPS): Define.
1287 (gimple_resimplify1): Replace rcode and ops with a single res_op
1288 parameter. Update call to gimple_simplify.
1289 (gimple_resimplify2): Likewise.
1290 (gimple_resimplify3): Likewise.
1291 (mprts_hook): Replace parameters with a gimple_match_op *.
1292 (maybe_build_generic_op): Likewise.
1293 (build_call_internal): Replace type, nargs and ops with
1294 a gimple_match_op *.
1295 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1296 with a single gimple_match_op *. Update calls to mprts_hook,
1297 build_call_internal and gimple_simplified_result_is_gimple_val.
1298 Factor out code that is common to the tree_code and combined_fn cases.
1299 * genmatch.c (expr::gen_transform): Replace tem_code and
1300 tem_ops with a gimple_match_op called tem_op. Update calls
1301 to the gimple_resimplify functions and maybe_push_res_to_seq.
1302 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1303 res_ops. Update call to the gimple_resimplify functions.
1304 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1305 (decision_tree::gen): Make the functions take a gimple_match_op *
1306 called res_op instead of separate res_code and res_ops parameters.
1307 Update call accordingly.
1308 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1309 and ops with a single res_op parameter. Update calls to
1310 maybe_build_generic_op and maybe_push_res_to_seq.
1311 (fold_stmt_1): Update calls to gimple_simplify and
1312 replace_stmt_with_simplification.
1313 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1314 and gimple_simplified_result_is_gimple_val.
1315 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1317 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1318 with a gimple_match_op *.
1319 (vn_nary_build_or_lookup): Likewise. Update call to
1320 vn_nary_build_or_lookup_1.
1321 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1322 gimple_match_op *. Update calls to the gimple_resimplify routines
1323 and to gimple_simplified_result_is_gimple_val.
1324 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1325 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1326 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1327 (visit_nary_op): Likewise.
1328 (visit_reference_op_load): Likewise.
1330 2018-05-23 Luis Machado <luis.machado@linaro.org>
1332 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1333 modifier for printing the step amount.
1335 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1338 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1341 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1343 * doc/sourcebuild.texi (Endianness): New subsubsection.
1345 2018-05-23 Luis Machado <luis.machado@linaro.org>
1347 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1348 <prefetch_dynamic_strides>: New const bool field.
1349 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1350 prefetch_dynamic_strides.
1351 (exynosm1_prefetch_tune): Likewise.
1352 (thunderxt88_prefetch_tune): Likewise.
1353 (thunderx_prefetch_tune): Likewise.
1354 (thunderx2t99_prefetch_tune): Likewise.
1355 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1357 (aarch64_override_options_internal): Update to set
1358 PARAM_PREFETCH_DYNAMIC_STRIDES.
1359 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1360 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1361 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1362 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1363 prefetch-dynamic-strides setting.
1365 2018-05-23 Luis Machado <luis.machado@linaro.org>
1367 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1368 <minimum_stride>: New const int field.
1369 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1370 minimum_stride field defaulting to -1.
1371 (exynosm1_prefetch_tune): Likewise.
1372 (thunderxt88_prefetch_tune): Likewise.
1373 (thunderx_prefetch_tune): Likewise.
1374 (thunderx2t99_prefetch_tune): Likewise.
1375 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1376 <default_opt_level>: Set to 3.
1377 (aarch64_override_options_internal): Update to set
1378 PARAM_PREFETCH_MINIMUM_STRIDE.
1379 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1380 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1381 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1382 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1383 stride is constant and is below the minimum stride threshold.
1385 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1387 * config/arm/arm-cpus.in (mode26): Delete.
1388 (armv4): Delete mode26 reference.
1389 * config/arm/arm.c (arm_configure_build_target): Delete use of
1392 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1394 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1396 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1397 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1399 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1400 (floatunsdidf2): Ditto.
1402 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1404 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1405 (fixuns_trunc<mode>si2_avx512f): Ditto.
1406 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1407 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1408 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1410 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1412 PR rtl-optimization/79985
1413 * df-scan.c (df_insn_refs_collect): Remove special case for
1414 global registers and asm statements.
1416 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1418 * extend.texi (Global Register Variables): Rewrite the bullet list.
1419 Note that the register is available for allocation. Note that access
1420 via inline asm must use constraints. Add note about async-signal
1421 handlers. Remove paragraph about automagic register selection.
1423 2018-05-23 Richard Biener <rguenther@suse.de>
1425 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1426 of fixed offset from memset VN.
1428 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1430 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1432 (alloc_cand_and_find_basis): Initialize first_interp field.
1433 (slsr_process_mul): Modify first_interp field.
1434 (slsr_process_add): Likewise.
1435 (slsr_process_cast): Modify first_interp field for each new
1437 (slsr_process_copy): Likewise.
1438 (dump_candidate): Dump first_interp field.
1439 (replace_mult_candidate): Process all interpretations, not just
1441 (replace_rhs_if_not_dup): Likewise.
1442 (replace_one_candidate): Likewise.
1444 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1446 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1448 (aarch64_needs_frame_chain): New function.
1449 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1451 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1454 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1455 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1456 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1457 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1458 as true for strict-align.
1459 (aarch64_can_inline_p): Perform checks even when callee has no
1460 attributes to check for strict alignment.
1461 * doc/extend.texi (AArch64 Function Attributes): Document
1463 * doc/invoke.texi: (AArch64 Options): Likewise.
1465 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1467 PR tree-optimization/85853
1468 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1469 the handling of the root of the node to...
1470 (vect_slp_analyze_node_operations_1): ...this new function,
1471 and run the whole thing with the child nodes' def types
1472 set according to their SLP node's def type.
1474 2018-05-23 Richard Biener <rguenther@suse.de>
1477 * tree-data-ref.c (create_runtime_alias_checks): Defer
1478 and ignore overflow warnings.
1480 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1482 PR tree-optimization/85822
1483 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1486 2018-05-23 Richard Biener <rguenther@suse.de>
1488 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1489 memset constants via native_interpret_expr.
1491 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1494 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1496 (cgraph_node::create_alias): Likewise.
1497 (cgraph_node::get_availability): Check ifunc_resolver instead
1498 of looking up ifunc attribute.
1499 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1500 * varasm.c (do_assemble_alias): Likewise.
1501 (assemble_alias): Likewise.
1502 (default_binds_local_p_3): Likewise.
1503 * cgraph.h (cgraph_node): Add ifunc_resolver.
1504 (cgraph_node::only_called_directly_or_aliased_p): Return false
1506 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1508 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1509 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1510 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1511 instead of looking up ifunc attribute.
1513 2018-05-22 Luis Machado <luis.machado@linaro.org>
1515 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1517 2018-05-22 Martin Sebor <msebor@redhat.com>
1520 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1521 only when expasion succeeds.
1522 (expand_builtin_strcmp): Same.
1523 (expand_builtin_strncmp): Same.
1525 2018-05-22 Martin Sebor <msebor@redhat.com>
1527 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1529 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1530 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1532 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1533 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1535 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1536 Add check for legitimate_address.
1537 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1538 (aarch64_swap_ldrstr_operands): New.
1539 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1542 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1543 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1545 * config/aarch64/aarch64.md: New patterns to generate stp
1547 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1548 single words and double words.
1549 (load_pair_sw, load_pair_dw): Likewise.
1550 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1552 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1554 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1555 for different mode ldpstp and add peephole for merged zero stores.
1557 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1559 (aarch64_gen_store_pair): Rename calls to match new patterns.
1560 (aarch64_gen_load_pair): Rename calls to match new patterns.
1561 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1562 (load_pair<DREG:mode><DREG2:mode>): ... This.
1563 (store_pair<mode>): Rename to...
1564 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1565 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1567 (V_INT_EQUIV): Handle SImode.
1568 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1571 2018-05-22 Martin Sebor <msebor@redhat.com>
1574 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1575 or ajust the presumed bound on an operation to avoid unnecessary
1578 2018-05-22 Martin Sebor <msebor@redhat.com>
1580 PR tree-optimization/85826
1581 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1582 assuming that a DECL necesarily has a constant size.
1584 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1587 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1589 2018-05-22 Richard Biener <rguenther@suse.de>
1591 PR tree-optimization/85834
1592 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1593 non-constant and non-zero memset arguments.
1595 2018-05-22 Martin Liska <mliska@suse.cz>
1598 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1600 2018-05-22 Richard Biener <rguenther@suse.de>
1602 PR tree-optimization/85863
1603 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1604 comparisons when vectype is specified.
1605 (vectorizable_condition): Do not specify vectype for
1606 vect_is_simple_cond when SLP vectorizing.
1608 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1611 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1612 define __ibm128 as long double.
1613 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1614 as a distinct type when IEEE 128-bit support is enabled.
1615 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1616 128-bit types to use the correct functions.
1617 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1618 convert between 128-bit floating point types that have different
1619 modes but the same representation, instead of using gen_lowpart to
1621 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1623 (IFKF_reg): New attributes to give the register constraints for
1625 (extend<mode>tf2_internal): New insns to mark an explicit
1626 conversion between 128-bit floating point types that have a
1627 different mode but share the same representation.
1629 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1631 PR tree-optimization/85814
1632 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1633 a null return from get_strinfo when unsharing the next
1634 strinfo in the chain.
1636 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1639 * varasm.c (weak_finish): Clean up weak_decls.
1641 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1643 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1644 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1645 UNSPEC_UADALP values.
1646 * config/aarch64/iterators.md (ABAL): New int iterator.
1649 (sur): Add mappings for the above.
1650 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1652 (aarch64_<sur>abal<mode>_4): Likewise.
1653 (aarch64_<sur>adalp<mode>_3): Likewise.
1654 (<sur>sadv16qi): New define_expand.
1656 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1658 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1659 (*movdf_internal): Ditto.
1660 (*rcpsf2_sse): Ditto.
1661 (*rsqrtsf2_sse): Ditto.
1662 (*sqrt<mode>2_sse): Ditto.
1664 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1666 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1668 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1669 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1670 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1671 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1673 * config/aarch64/arm_neon.h: Likewise.
1674 * config/aarch64/iterators.md (VQ_I): New.
1676 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1678 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1680 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1683 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1685 * config/nds32/constraints.md (S): New constraint.
1686 * config/nds32/nds32.md (call_internal): Use constraint S.
1687 (call_value_internal): Likewise.
1688 (sibcall_internal): Likewise.
1689 (sibcall_value_internal): Likewise.
1691 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1692 Chung-Ju Wu <jasonwucj@gmail.com>
1694 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1697 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1698 Chung-Ju Wu <jasonwucj@gmail.com>
1700 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1701 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1702 (nds32_rtx_costs_impl): Simplify.
1703 (nds32_address_cost_impl): Simplify.
1704 (nds32_init_rtx_costs): New function.
1705 (nds32_rtx_costs_speed_prefer): Likewise.
1706 (nds32_rtx_costs_size_prefer): Likewise.
1707 (nds32_address_cost_speed_prefer): Likewise.
1708 (nds32_address_cost_speed_fwprop): Likewise.
1709 (nds32_address_cost_size_prefer): Likewise.
1710 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1711 * config/nds32/nds32.c (nds32_option_override): Use
1712 nds32_init_rtx_costs function.
1714 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1716 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1717 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1718 (TARGET_PIPELINE_N8): Likewise.
1719 (TARGET_PIPELINE_N10): Likewise.
1720 (TARGET_PIPELINE_N13): Likewise.
1721 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1723 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1725 * config/nds32/nds32-fpu.md: Update copyright year.
1727 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1731 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1733 * config/nds32/nds32.c
1734 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1735 * config/nds32/nds32.opt (minline-asm-r15): New option.
1737 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1739 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1741 * config/nds32/nds32.md (abssi2): New pattern.
1743 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1745 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1746 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1747 Merge insn pattern from sse_cvtsi2ss<round_name> and
1748 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1749 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1750 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1751 using SWI48 mode iterator.
1752 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1753 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1754 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1755 pattern from sse_cvttss2si<round_saeonly_name>
1756 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1757 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1758 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1759 using SWI48 mode iterator.
1760 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1761 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1762 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1763 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1764 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1765 using SWI48 mode iterator.
1766 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1767 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1768 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1769 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1770 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1771 SWI48 mode iterator.
1772 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1773 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1774 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1775 pattern from sse_cvttsd2si<round_saeonly_name>
1776 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1778 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1780 * config/nds32/nds32-md-auxiliary.c
1781 (nds32_valid_smw_lwm_base_p): Refine.
1782 (nds32_output_smw_single_word): Refine.
1783 (nds32_output_smw_double_word): New.
1784 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1786 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1788 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1789 (nds32_output_stack_pop): Refine.
1790 (nds32_expand_unaligned_load): Refine.
1791 (nds32_expand_unaligned_store): Refine.
1793 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1794 Chung-Ju Wu <jasonwucj@gmail.com>
1796 * config/nds32/constants.md: Add TP_REGNUM constant.
1797 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1798 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1800 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1801 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1802 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1803 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1804 and PIC code generation.
1805 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1807 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1809 * config/nds32/nds32.md: Support TLS and PIC.
1810 * config/nds32/nds32.c: Support TLS and PIC.
1811 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1812 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1815 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1817 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1818 mode with E_ prefix.
1820 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1821 Chung-Ju Wu <jasonwucj@gmail.com>
1823 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1824 * config/nds32/nds32-md-auxiliary.c
1825 (symbolic_reference_mentioned_p): New.
1826 (nds32_legitimize_ict_address): New.
1827 (nds32_expand_ict_move): New.
1828 (nds32_indirect_call_referenced_p): New.
1829 (nds32_symbol_binds_local_p): Delete.
1830 (nds32_long_call_p): Modify.
1831 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1832 * config/nds32/nds32-protos.h
1833 (symbolic_reference_mentioned_p): Declare.
1834 (nds32_legitimize_ict_address): Declare.
1835 (nds32_expand_ict_move): Declare.
1836 (nds32_indirect_call_referenced_p): Declare.
1837 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1838 (nds32_relax_group): Use nds32_ict_const_p as condition.
1839 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1840 (nds32_asm_file_start): Output ict_model directive in asm code.
1841 (nds32_legitimate_address_p): Consider indirect call.
1842 (nds32_print_operand): Consider indirect call.
1843 (nds32_print_operand_address): Consider indirect call.
1844 (nds32_insert_attributes): Handle "indirect_call" attribute.
1845 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1846 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1847 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1848 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1849 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1850 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1851 (TARGET_ICT_MODEL_SMALL): Define.
1852 (TARGET_ICT_MODEL_LARGE): Define.
1853 * config/nds32/nds32.md (movsi): Consider ict model.
1854 (call, call_value): Consider ict model.
1855 (sibcall, sibcall_value): Consider ict model.
1856 * config/nds32/nds32.opt (mict-model): New option.
1857 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1860 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1861 Monk Chiang <sh.chiang04@gmail.com>
1862 Jim Wilson <jimw@sifive.com>
1864 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1865 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1866 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1867 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1868 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1869 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1870 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1871 compute save_libcall_adjustment properly.
1872 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1873 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1874 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1875 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1876 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1877 (ABI_SPEC): Handle mabi=ilp32e.
1878 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1879 (RVE): Add RVE mask.
1880 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1881 <-march>: Add rv32e as an example.
1883 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1886 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1887 (intra_create_variable_infos): Handle C++ constructors.
1889 2018-05-18 Martin Liska <mliska@suse.cz>
1891 * passes.def: Remove a redundant pass.
1893 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1896 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1898 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1900 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1902 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1904 (armv2, armv2a, armv3, armv3m): Delete architectures.
1905 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1906 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1907 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1909 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1910 (*mulsidi3adddi): Likewise.
1911 (mulsidi3): Likewise.
1912 (*mulsidi3_nov6): Likewise.
1913 (umulsidi3): Likewise.
1914 (umulsidi3_nov6): Likewise.
1915 (umaddsidi4): Likewise.
1916 (*umulsidi3adddi): Likewise.
1917 (smulsi3_highpart): Likewise.
1918 (*smulsi3_highpart_nov6): Likewise.
1919 (umulsi3_highpart): Likewise.
1920 (*umulsi3_highpart_nov6): Likewise.
1921 * config/arm/arm.h (arm_arch3m): Delete.
1922 * config/arm/arm.c (arm_arch3m): Delete.
1923 (arm_option_override_internal): Update armv3-related comment.
1924 (arm_configure_build_target): Delete use of isa_bit_mode32.
1925 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1926 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1927 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1929 (mulusa3): Likewise.
1930 * config/arm/arm-protos.h (arm_arch3m): Delete.
1931 * config/arm/arm-tables.opt: Regenerate.
1932 * config/arm/arm-tune.md: Likewise.
1933 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1934 deleted architectures.
1936 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1938 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1939 (armv5t, armv5te): New features.
1940 (ARMv5, ARMv5e): Delete fgroups.
1941 (ARMv5t, ARMv5te): Adjust for above changes.
1943 (armv5, armv5e): Delete arches.
1944 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1946 (*call_reg_arm): Likewise.
1947 (*call_value_reg_armv5): Likewise.
1948 (*call_value_reg_arm): Likewise.
1949 (*call_symbol): Likewise.
1950 (*call_value_symbol): Likewise.
1951 (*sibcall_insn): Likewise.
1952 (*sibcall_value_insn): Likewise.
1954 (prefetch): Likewise.
1955 (define_split and define_peephole2 dependent on arm_arch5):
1957 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1959 (TARGET_ARM_QBIT): Likewise.
1960 (TARGET_DSP_MULTIPLY): Likewise.
1961 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1962 (arm_arch5, arm_arch5e): Delete.
1963 (arm_arch5t, arm_arch5te): Declare.
1964 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1965 (arm_arch5t): Declare.
1966 (arm_option_reconfigure_globals): Update for the above.
1967 (arm_options_perform_arch_sanity_checks): Update comment, replace
1968 use of arm_arch5 with arm_arch5t.
1969 (use_return_insn): Likewise.
1970 (arm_emit_call_insn): Likewise.
1971 (output_return_instruction): Likewise.
1972 (arm_final_prescan_insn): Likewise.
1973 (arm_coproc_builtin_available): Likewise.
1974 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1975 arm_arch5e with arm_arch5t and arm_arch5te.
1976 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1977 (arm_arch5t, arm_arch5te): Declare.
1978 * config/arm/arm-tables.opt: Regenerate.
1979 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1980 * config/arm/t-multilib: Likewise.
1981 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1982 instead of arm_arch5.
1983 (*call_reg_thumb1): Likewise.
1984 (*call_value_reg_thumb1_v5): Likewise.
1985 (*call_value_reg_thumb1): Likewise.
1986 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1988 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1990 2018-05-18 Martin Liska <mliska@suse.cz>
1992 PR gcov-profile/84846
1993 * doc/gcov.texi: Document -t option of gcov tool.
1995 2018-05-18 Martin Liska <mliska@suse.cz>
1997 PR gcov-profile/84846
1998 * gcov.c (print_usage): Add new -t option.
1999 (process_args): Handle the option.
2000 (generate_results): Use stdout as output when requested by
2003 2018-05-18 Martin Liska <mliska@suse.cz>
2005 PR gcov-profile/84846
2006 * coverage.c (coverage_init): Write PWD to .gcno file.
2007 * doc/gcov.texi: Document how working directory is printed.
2008 * gcov-dump.c (dump_gcov_file): Print PWD.
2009 * gcov.c (output_intermediate_file): Likewise.
2010 (read_graph_file): Read PWD string.
2011 (output_lines): Print PWD.
2013 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2016 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
2017 for retval and return false if all args to phi are zero.
2019 2018-05-18 Richard Biener <rguenther@suse.de>
2021 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
2023 (evrp_dom_walker::before_dom_children): Call it.
2025 2018-05-18 Richard Biener <rguenther@suse.de>
2027 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
2028 results when processing array refs with variable index.
2030 2018-05-18 Toon Moene <toon@moene.org>
2032 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
2033 directly after that of -floop-interchange. Indicate that both
2034 options are enabled by default when specifying -O3.
2036 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2038 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2039 iterator. Delete separate integer-mode vec_set<mode> expander.
2040 (aarch64_simd_vec_setv2di): Delete.
2041 (vec_setv2di): Delete.
2042 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2043 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2044 the "w, r" alternative.
2046 2018-05-18 Martin Liska <mliska@suse.cz>
2048 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2049 * tree-pass.h (make_pass_lower_switch_O0): New function.
2050 * tree-switch-conversion.c (node_has_low_bound): Remove.
2051 (node_has_high_bound): Likewise.
2052 (node_is_bounded): Likewise.
2053 (class pass_lower_switch): Make it a template type and create
2055 (pass_lower_switch::execute): Add template argument.
2056 (make_pass_lower_switch): New function.
2057 (make_pass_lower_switch_O0): New function.
2058 (do_jump_if_equal): Remove.
2059 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2060 all the hard work to tree optimization passes.
2062 2018-05-18 Martin Liska <mliska@suse.cz>
2064 * dbgcnt.c (limit_low): Renamed from limit.
2065 (limit_high): New variable.
2066 (dbg_cnt_is_enabled): Check for upper limit.
2067 (dbg_cnt): Adjust dumping.
2068 (dbg_cnt_set_limit_by_index): Add new argument for high
2070 (dbg_cnt_set_limit_by_name): Likewise.
2071 (dbg_cnt_process_single_pair): Parse new format.
2072 (dbg_cnt_process_opt): Use strtok.
2073 (dbg_cnt_list_all_counters): Remove 'value' and add
2075 * doc/invoke.texi: Document changes.
2077 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2079 * doc/sourcebuild.texi (scalar_all_fma): Document.
2080 * tree.def (FMA_EXPR): Delete.
2081 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2082 * internal-fn.c (ternary_direct): New macro.
2083 (expand_ternary_optab_fn): Likewise.
2084 (direct_ternary_optab_supported_p): Likewise.
2085 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2086 * builtins.c (fold_builtin_fma): Delete.
2087 (fold_builtin_3): Don't call it.
2088 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2089 * expr.c (expand_expr_real_2): Likewise.
2090 * fold-const.c (operand_equal_p): Likewise.
2091 (fold_ternary_loc): Likewise.
2092 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2093 * gimple.c (DEFTREECODE): Likewise.
2094 * gimplify.c (gimplify_expr): Likewise.
2095 * optabs-tree.c (optab_for_tree_code): Likewise.
2096 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2097 * tree-eh.c (operation_could_trap_p): Likewise.
2098 (stmt_could_throw_1_p): Likewise.
2099 * tree-inline.c (estimate_operator_cost): Likewise.
2100 * tree-pretty-print.c (dump_generic_node): Likewise.
2101 (op_code_prio): Likewise.
2102 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2103 * tree-ssa-operands.c (get_expr_operands): Likewise.
2104 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2105 * fold-const-call.h (fold_fma): Delete.
2106 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2107 CFN_FNMA and CFN_FNMS.
2109 * genmatch.c (combined_fn): New enum.
2110 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2111 (commutative_op): New function.
2112 (commutate): Use it. Handle more than 2 operands.
2113 (dt_operand::gen_gimple_expr): Use commutative_op.
2114 (parser::parse_expr): Allow :c to be used with non-binary
2115 operators if the commutative operand is known.
2116 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2117 CFN_FMS, CFN_FNMA and CFN_FNMS.
2118 (backprop::process_assign_use): Remove FMA_EXPR handling.
2119 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2120 (gen_hsa_fma): New function.
2121 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2122 IFN_FNMA and IFN_FNMS.
2123 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2124 * gimple-fold.h (follow_all_ssa_edges): Declare.
2125 * gimple-fold.c (follow_all_ssa_edges): New function.
2126 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2127 gimple_build interface and use follow_all_ssa_edges to fold the result.
2128 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2129 instead of checking for optabs directly.
2130 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2131 rather than FMA_EXPRs.
2132 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2133 call to IFN_FMA instead of an FMA_EXPR.
2135 2018-05-17 Jim Wilson <jimw@sifive.com>
2137 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2138 sign extended promoted subreg, and the range does not have the sign bit
2139 set, then do a sign extend.
2141 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2142 test, check for sign extended subreg and/or constant operands, and
2143 do a sign extend in that case.
2145 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2147 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2148 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2150 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2151 Change logics_shift_reg to logics_shift_imm.
2152 (thunderx2t99_fp_loadpair_basic): Delete.
2153 (thunderx2t99_fp_storepair_basic): Delete.
2154 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2155 (thunderx2t99_asimd_polynomial): Delete.
2156 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2157 and neon_fp_mul_d_scalar_q.
2158 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2159 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2160 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2161 (thunderx2t99_asimd_lut): Add missing tbl types.
2162 (thunderx2t99_asimd_ext): Delete.
2163 (thunderx2t99_asimd_load1_1_mult): Delete.
2164 (thunderx2t99_asimd_load1_2_mult): Delete.
2165 (thunderx2t99_asimd_load1_ldp): New.
2166 (thunderx2t99_asimd_load1): New.
2167 (thunderx2t99_asimd_load2): Add missing *load2* types.
2168 (thunderx2t99_asimd_load3): New.
2169 (thunderx2t99_asimd_load4): New.
2170 (thunderx2t99_asimd_store1_1_mult): Delete.
2171 (thunderx2t99_asimd_store1_2_mult): Delete.
2172 (thunderx2t99_asimd_store2_mult): Delete.
2173 (thunderx2t99_asimd_store2_onelane): Delete.
2174 (thunderx2t99_asimd_store_stp): New.
2175 (thunderx2t99_asimd_store1): New.
2176 (thunderx2t99_asimd_store2): New.
2177 (thunderx2t99_asimd_store3): New.
2178 (thunderx2t99_asimd_store4): New.
2180 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2182 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2183 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2185 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2186 Segher Boessenkool <segher@kernel.crashing.org>
2189 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2192 2018-05-17 Richard Biener <rguenther@suse.de>
2194 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2195 for pruning loop and prune defs feeding only already visited PHIs.
2197 2018-05-17 Richard Biener <rguenther@suse.de>
2199 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2201 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2202 Richard Biener <rguenther@suse.de>
2204 PR tree-optimization/85793
2205 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2206 for VMAT_ELEMENTWISE.
2208 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2210 * internal-fn.h (lookup_internal_fn): Declare
2211 * internal-fn.c (lookup_internal_fn): New function.
2212 * gimple.c (gimple_build_call_from_tree): Handle calls to
2214 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2215 internal function names.
2216 * tree-pretty-print.c (dump_generic_node): Likewise.
2217 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2219 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2221 * gimple-fold.h (gimple_build): Make the function forms take
2222 combined_fn rather than built_in_function.
2223 (gimple_simplify): Likewise.
2224 * gimple-match-head.c (gimple_simplify): Likewise.
2225 * gimple-fold.c (gimple_build): Likewise.
2226 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2227 rather than gimple_build_call_internal.
2228 (get_initial_defs_for_reduction): Likewise.
2229 (vect_create_epilog_for_reduction): Likewise.
2230 (vectorizable_live_operation): Likewise.
2232 2018-05-17 Martin Liska <mliska@suse.cz>
2234 * gimple-ssa-sprintf.c (format_directive): Do not use
2235 space in between 'G_' and '('.
2237 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2240 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2241 even if the mask is not all ones.
2244 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2246 (ix86_gimple_fold_builtin): Likewise.
2249 * config/i386/i386.c: Include tree-vector-builder.h.
2250 (ix86_vector_shift_count): New function.
2251 (ix86_fold_builtin): Fold shift builtins by scalar count.
2252 (ix86_gimple_fold_builtin): Likewise.
2254 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2255 _mm512_setzero): New intrinsics.
2257 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2258 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2260 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2261 code generation for cases where splatting a value is not useful.
2262 * simplify-rtx.c (simplify_ternary_operation): Simplify
2263 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2264 mode to a vec_concat.
2266 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2268 * config.gcc: Support "goldmont-plus".
2269 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2271 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2272 PROCESSOR_GOLDMONT_PLUS.
2273 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2274 (processor_target_table): Add "goldmont-plus".
2275 (PTA_GOLDMONT_PLUS): Define.
2276 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2277 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2278 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2279 (fold_builtin_cpu): Add "goldmont-plus".
2280 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2281 (ix86_option_override_internal): Add "goldmont-plus".
2282 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2283 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2284 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2285 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2287 2018-05-17 Richard Biener <rguenther@suse.de>
2289 PR tree-optimization/85757
2290 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2291 remove defs that only feed that PHI from further processing.
2293 2018-05-16 Jim Wilson <jimw@sifive.com>
2295 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2297 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2299 2018-05-16 Mark Wielaard <mark@klomp.org>
2301 * dwarf2out.c (count_index_strings): New function.
2302 (output_indirect_strings): Call count_index_strings and generate
2303 header for dwarf_version >= 5.
2305 2018-05-16 Mark Wielaard <mark@klomp.org>
2307 * dwarf2out.c (dwarf_FORM): New function.
2308 (set_indirect_string): Use dwarf_FORM.
2309 (reset_indirect_string): Likewise.
2310 (size_of_die): Likewise.
2311 (value_format): Likewise.
2312 (output_die): Likewise.
2313 (add_skeleton_AT_string): Likewise.
2314 (output_macinfo_op): Likewise.
2315 (index_string): Likewise.
2316 (output_index_string_offset): Likewise.
2317 (output_index_string): Likewise.
2318 (count_index_strings): Likewise.
2320 2018-05-16 Carl Love <cel@us.ibm.com>
2322 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2323 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2325 2018-05-16 Martin Jambor <mjambor@suse.cz>
2327 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2328 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2330 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2332 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2333 (fnma<mode>4): Likewise.
2334 (fms<mode>4): Likewise.
2335 (fnms<mode>4): Likewise.
2336 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2337 (aarch64_fnma<mode>4): Likewise.
2338 (aarch64_fms<mode>4): Likewise.
2339 (aarch64_fnms<mode>4): Likewise.
2340 (aarch64_fnmadd<mode>4): Likewise.
2342 2018-05-16 Jason Merrill <jason@redhat.com>
2344 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2346 2018-05-16 Richard Biener <rguenther@suse.de>
2348 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2349 (dump_stmt_cost): Declare.
2350 (add_stmt_cost): Dump cost we add.
2351 (add_stmt_costs): New function.
2352 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2354 (vect_analyze_stmt): Adjust prototype.
2355 (vectorizable_condition): Likewise.
2356 (vectorizable_live_operation): Likewise.
2357 (vectorizable_reduction): Likewise.
2358 (vectorizable_induction): Likewise.
2359 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2360 cost vector to pass to vectorizable_ and record afterwards.
2361 (vect_model_reduction_cost): Take cost vector argument and adjust.
2362 (vect_model_induction_cost): Likewise.
2363 (vectorizable_reduction): Likewise.
2364 (vectorizable_induction): Likewise.
2365 (vectorizable_live_operation): Likewise.
2366 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2367 SLP_TREE_NUMBER_OF_VEC_STMTS.
2368 (vect_analyze_slp_cost_1): Remove.
2369 (vect_analyze_slp_cost): Likewise.
2370 (vect_slp_analyze_node_operations): Take visited args and
2371 a target cost vector. Avoid processing already visited stmt sets.
2372 (vect_slp_analyze_operations): Use a local cost vector to gather
2373 costs and register those of non-discarded instances.
2374 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2375 (vect_schedule_slp_instance): Remove copying of
2376 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2378 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2379 adding cost. Record cost entry location.
2380 (vect_prologue_cost_for_slp_op): Function to compute cost of
2381 a constant or invariant generated for SLP vect in the prologue,
2382 split out from vect_analyze_slp_cost_1.
2383 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2384 (vect_model_promotion_demotion_cost): Likewise.
2385 (vect_model_store_cost): Likewise, make static.
2386 (vect_model_load_cost): Likewise.
2387 (vectorizable_bswap): Add cost vector arg and adjust.
2388 (vectorizable_call): Likewise.
2389 (vectorizable_simd_clone_call): Likewise.
2390 (vectorizable_conversion): Likewise.
2391 (vectorizable_assignment): Likewise.
2392 (vectorizable_shift): Likewise.
2393 (vectorizable_operation): Likewise.
2394 (vectorizable_store): Likewise.
2395 (vectorizable_load): Likewise.
2396 (vectorizable_condition): Likewise.
2397 (vectorizable_comparison): Likewise.
2398 (can_vectorize_live_stmts): Likewise.
2399 (vect_analyze_stmt): Likewise.
2400 (vect_transform_stmt): Adjust calls to vectorizable_*.
2401 * tree-vectorizer.c: Include gimple-pretty-print.h.
2402 (dump_stmt_cost): New function.
2404 2018-05-16 Richard Biener <rguenther@suse.de>
2406 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2407 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2408 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2409 (check_name): New callback.
2410 (dse_classify_store): Track cycles via a visited bitmap of PHI
2411 defs and simplify handling of in-loop and across loop dead stores
2412 and properly fail for loop-variant refs. Handle byte-tracking with
2413 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2416 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2418 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2419 (vect_get_mask_type_for_stmt): Likewise.
2420 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2422 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2423 to determine the statement's vector type and the vector type that
2424 should be used for calculating nunits. Deal with cases in which
2425 the type has to be deferred.
2426 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2427 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2428 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2429 (vect_determine_vf_for_stmt): New functions, split out from...
2430 (vect_determine_vectorization_factor): ...here.
2431 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2432 (vect_get_mask_type_for_stmt): New functions, split out from
2433 vect_determine_vectorization_factor.
2435 2018-05-16 Richard Biener <rguenther@suse.de>
2437 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2438 verify the [VEC_]COND_EXPR embedded comparison.
2440 2018-05-15 Martin Sebor <msebor@redhat.com>
2442 PR tree-optimization/85753
2443 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2444 RECORD_TYPE in addition to ARRAY_TYPE.
2446 2018-05-15 Martin Sebor <msebor@redhat.com>
2449 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2451 2018-05-15 Richard Biener <rguenther@suse.de>
2453 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2454 add by_clobber_p one. Change algorithm to collect all defs
2455 representing uses we need to walk and try reducing them to
2456 a single one before failing.
2457 (dse_dom_walker::dse_optimize_stmt): Adjust.
2459 2018-05-13 Mark Wielaard <mark@klomp.org>
2461 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2462 (size_of_loc_descr): Likewise.
2463 (output_loc_operands): Likewise.
2464 (output_loc_operands_raw): Likewise.
2465 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2466 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2467 (hash_loc_operands): Likewise.
2468 (compare_loc_operands): Likewise.
2470 2018-05-14 Mark Wielaard <mark@klomp.org>
2472 * dwarf2out.c (count_index_addrs): New function.
2473 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2475 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2477 PR tree-optimization/83648
2478 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2479 return value as malloc candidate.
2481 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2484 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2485 as true in call to suggest_attribute.
2487 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2489 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2490 -mreadonly-in-sdata.
2492 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2494 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2496 (aarch64_crypto_aesd_fused): Likewise.
2498 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2500 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2501 (movsi_aarch64): Likewise.
2502 (load_pairsi): Likewise.
2503 (load_pairdi): Likewise.
2504 (store_pairsi): Likewise.
2505 (store_pairdi): Likewise.
2506 (load_pairsf): Likewise.
2507 (load_pairdf): Likewise.
2508 (store_pairsf): Likewise.
2509 (store_pairdf): Likewise.
2510 (zero_extend): Likewise.
2511 (trunc): Swap alternatives.
2512 (fcvt_target): Add '?' to prefer w over r.
2514 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2517 * config/i386/i386.md: Disallow non-commutative arithmetics in
2518 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2519 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2520 in the peephole2 before it.
2522 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2524 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2525 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2526 (ix86_handle_option): Handle -mcldemote.
2527 * config.gcc: New header.
2528 * config/i386/cldemoteintrin.h: New file.
2529 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2530 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2532 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2533 OPTION_MASK_ISA_CLDEMOTE.
2534 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2535 (ix86_valid_target_attribute_inner_p): Ditto.
2536 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2537 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2538 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2539 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2540 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2542 * config/i386/i386.opt: Add -mcldemote.
2543 * config/i386/x86intrin.h: New header.
2544 * doc/invoke.texi: Add -mcldemote.
2546 2018-05-14 Richard Biener <rguenther@suse.de>
2548 * doc/match-and-simplify.texi: Adjust :s documentation.
2550 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2552 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2553 intended memcpy size.
2554 (REORDER_45): Likewise.
2556 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2558 * sort.cc: New file.
2559 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2560 * vec.c (qsort_chk): Use gcc_qsort.
2561 * Makefile.in (OBJS-libcommon): Add sort.o.
2562 (build/sort.o): New target. Use it...
2563 (BUILD_RTL): ... here, and...
2564 (build/gencfn-macros): ... here, and...
2565 (build/genmatch): ... here.
2567 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2568 Chung-Ju Wu <jasonwucj@gmail.com>
2570 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2571 * config/nds32/nds32-graywolf.md: New file.
2572 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2573 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2575 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2576 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2577 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2578 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2579 * config/nds32/pipelines.md: Include n15 settings.
2581 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2582 Chung-Ju Wu <jasonwucj@gmail.com>
2584 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2585 * config/nds32/nds32-n13.md: New file.
2586 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2587 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2589 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2590 * config/nds32/nds32.md (pipeline_model): Add n13.
2591 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2592 * config/nds32/pipelines.md: Include n13 settings.
2594 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2595 Chung-Ju Wu <jasonwucj@gmail.com>
2597 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2598 * config/nds32/nds32-n10.md: New file.
2599 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2600 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2602 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2603 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2604 * config/nds32/nds32.md (pipeline_model): Add n10.
2605 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2606 * config/nds32/pipelines.md: Include n10 settings.
2608 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2609 Kito Cheng <kito.cheng@gmail.com>
2610 Chung-Ju Wu <jasonwucj@gmail.com>
2612 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2613 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2614 Add enum values for DSP extension instructions.
2615 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2617 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2618 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2620 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2621 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2622 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2623 * config/nds32/nds32-intrinsic.md: Likewise.
2624 * config/nds32/nds32_intrinsic.h: Likewise.
2625 * config/nds32/nds32-md-auxiliary.c: Likewise.
2626 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2627 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2628 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2629 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2630 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2631 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2632 TYPE_DMAC in switch statement.
2633 * config/nds32/nds32.c: New checking and implementation for DSP
2634 extension instructions.
2635 * config/nds32/nds32.h: Likewise.
2636 * config/nds32/nds32.md: Likewise.
2637 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2638 * config/nds32/predicates.md: Implement new predicates for DSP
2641 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2643 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2644 Reformat alternatives and attributes so it is easier to identify
2645 which constraints/attributes go with which instruction.
2646 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2647 (mov<mode>_softfloat32, FMOVE64): Likewise.
2648 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2649 (mov<mode>_softfloat64, FMOVE64): Likewise.
2651 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2653 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2655 (Basic PowerPC Built-in Functions): The new name of the
2656 subsection previously known as "PowerPC Built-in Functions".
2657 (Basic PowerPC Built-in Functions Available on all Configurations):
2659 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2660 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2661 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2662 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2664 2018-05-11 Martin Jambor <mjambor@suse.cz>
2667 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2670 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2673 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2675 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2677 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2678 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2679 (ix86_handle_option): Handle -mwaitpkg.
2680 * config.gcc: New header.
2681 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2682 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2683 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2685 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2686 OPTION_MASK_ISA_WAITPKG.
2687 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2688 (ix86_option_override_internal): Add PTA_WAITPKG.
2689 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2690 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2691 IX86_BUILTIN_TPAUSE.
2692 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2693 __builtin_ia32_umwait and __builtin_ia32_tpause.
2694 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2695 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2696 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2697 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2698 UNSPECV_TPAUSE): New.
2699 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2700 * config/i386/i386.opt: Add -mwaitpkg.
2701 * config/i386/waitpkgintrin.h: New file.
2702 * config/i386/x86intrin.h: New header.
2703 * doc/invoke.texi: Add -mwaitpkg.
2705 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2708 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2710 (cortex-m0): Use armv6s-m isa.
2711 (cortex-m0plus): Likewise.
2712 (cortex-m1): Likewise.
2713 (cortex-m0.small-multiply): Likewise.
2714 (cortex-m0plus.small-multiply): Likewise.
2715 (cortex-m1.small-multiply): Likewise.
2717 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2718 Jakub Jelinek <jakub@redhat.com>
2720 PR tree-optimization/85692
2721 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2722 source permute as well.
2724 2018-05-11 Martin Liska <mliska@suse.cz>
2727 * doc/extend.texi: Document LLVM style format for no_sanitize
2730 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2732 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2733 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2734 (mode_supports_vsx_dform_quad): Likewise.
2735 (mode_supports_vmx_dform): Move these functions to be next to the
2736 other mode_supports functions.
2737 (mode_supports_dq_form): Likewise.
2738 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2739 mode_supports_dq_form.
2740 (reg_offset_addressing_ok_p): Likewise.
2741 (offsettable_ok_by_alignment): Likewise.
2742 (rs6000_legitimate_offset_address_p): Likewise.
2743 (legitimate_lo_sum_address_p): Likewise.
2744 (rs6000_legitimize_address): Likewise.
2745 (rs6000_legitimize_reload_address): Likewise.
2746 (rs6000_secondary_reload_inner): Likewise.
2747 (rs6000_preferred_reload_class): Likewise.
2748 (rs6000_output_move_128bit): Likewise.
2750 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2752 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2753 Generate SImode target register for null target.
2754 <case IX86_BUILTIN_XGETBV>: Ditto.
2755 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2756 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2758 2018-05-10 Carl Love <cel@us.ibm.com>
2760 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2761 dcbtt and dcbtstt if operands[2] is 0.
2763 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2766 * config/i386/sse.md (usadv64qi): New expander.
2768 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2770 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2771 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2772 -maltivec=be support.
2773 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2774 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2775 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2776 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2777 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2778 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2779 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2780 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2781 altivec_vsumsws): Adjust.
2782 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2783 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2784 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2786 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2787 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2788 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2789 (altivec_lve<VI_char>x): Delete expand.
2790 (*altivec_lve<VI_char>x_internal): Rename to...
2791 (altivec_lve<VI_char>x): ... this.
2792 (altivec_lvxl_<mode>): Delete expand.
2793 (*altivec_lvxl_<mode>_internal): Rename to ...
2794 (altivec_lvxl_<mode>): ... this.
2795 (altivec_stvxl_<mode>): Delete expand.
2796 (*altivec_stvxl_<mode>_internal): Rename to ...
2797 (altivec_stvxl_<mode>): ... this.
2798 (altivec_stve<VI_char>x): Delete expand.
2799 (*altivec_stve<VI_char>x_internal): Rename to ...
2800 (altivec_stve<VI_char>x): ... this.
2801 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2802 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2803 reduc_plus_scal_<mode>): Adjust.
2804 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2806 (rs6000_cpu_cpp_builtins): Adjust.
2807 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2808 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2809 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2810 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2811 -maltivec=be support.
2812 (rs6000_split_vec_extract_var): Adjust.
2813 (rs6000_split_v4si_init): Adjust.
2814 (swap_selector_for_mode): Delete.
2815 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2816 altivec_expand_stvex_be): Delete.
2817 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2818 -maltivec=be support.
2819 (rs6000_gimple_fold_builtin): Ditto.
2820 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2822 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2823 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2824 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2825 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2826 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2827 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2828 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2829 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2830 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2831 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2832 anonymous split): Adjust.
2833 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2834 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2836 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2838 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2839 when --with-gxx-include-dir is also specified.
2840 * configure: Regenerate.
2842 2018-05-09 Jim Wilson <jimw@sifive.com>
2845 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2846 * config/riscv/t-withmultilib: New.
2847 * config/riscv/withmultilib.h: New.
2848 * doc/install.texi: Document RISC-V --with-multilib-list support.
2850 2018-05-09 Richard Biener <rguenther@suse.de>
2852 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2854 (vect_bb_vectorization_profitable_p): Adjust. Compute
2855 actual scalar cost using the cost vector and the add_stmt_cost
2858 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2860 PR rtl-optimization/85645
2861 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2862 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2864 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2866 PR rtl-optimization/85645
2867 * shrink-wrap.c (spread_components): Return a boolean saying if
2868 anything was changed.
2869 (try_shrink_wrapping_separate): Iterate spread_components until
2870 nothing changes anymore.
2872 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2874 PR rtl-optimization/85645
2875 * regrename.c (build_def_use): Also kill the chains that include the
2876 destination of a REG_CFA_REGISTER note.
2878 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2880 PR rtl-optimization/85645
2881 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2882 insn that has a REG_CFA_REGISTER note.
2884 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2886 * cfgexpand.c (expand_clobber): New function.
2887 (expand_gimple_stmt_1): Use it.
2888 * tree-vect-stmts.c (vect_clobber_variable): New function,
2890 (vectorizable_simd_clone_call): ...here.
2891 (vectorizable_store): Emit a clobber either side of an
2892 IFN_STORE_LANES sequence.
2893 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2895 2018-05-09 Tom de Vries <tom@codesourcery.com>
2898 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2899 (define_insn "trap_if_false"): Add exit after trap.
2901 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2903 PR rtl-optimization/85638
2904 * bb-reorder.c: Include common/common-target.h.
2905 (create_forwarder_block): New function extracted from...
2906 (fix_up_crossing_landing_pad): ...here. Rename into...
2907 (dw2_fix_up_crossing_landing_pad): ...this.
2908 (sjlj_fix_up_crossing_landing_pad): New function.
2909 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2910 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2911 from both partitions and exit the loop after one iteration.
2913 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2916 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2918 (Basic PowerPC Built-in Functions): The new name of the
2919 subsection previously known as "PowerPC Built-in Functions".
2920 (Basic PowerPC Built-in Functions Available on all Configurations):
2922 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2924 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2926 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2928 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2931 2018-05-08 Jim Wilson <jimw@sifive.com>
2933 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2934 (LD_EMUL_SUFFIX): New.
2935 (LINK_SPEC): Use it.
2937 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2939 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2941 (Basic PowerPC Built-in Functions): The new name of the
2942 subsection previously known as "PowerPC Built-in Functions".
2943 (Basic PowerPC Built-in Functions Available on all Configurations):
2945 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2947 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2949 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2951 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2954 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2957 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2958 after cmpelim optimization.
2960 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2962 * config.gcc: Support "goldmont".
2963 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2964 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2966 * config/i386/i386.c (m_GOLDMONT): Define.
2967 (processor_target_table): Add "goldmont".
2968 (PTA_GOLDMONT): Define.
2969 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2970 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2971 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2972 (fold_builtin_cpu): Add "goldmont".
2973 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2974 (ix86_option_override_internal): Add "goldmont".
2975 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2976 (processor_type): Add PROCESSOR_GOLDMONT.
2977 * config/i386/i386.md: Add CPU "glm".
2978 * config/i386/glm.md: New file.
2979 * config/i386/x86-tune.def: Add m_GOLDMONT.
2980 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2982 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2985 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2987 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2988 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2989 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2992 * config/i386/i386.c (ix86_fold_builtin): Handle
2993 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2996 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2997 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2999 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
3002 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
3003 (check_arch): Likewise.
3004 (check_fpu): Return the result rather than printing it.
3005 (end arch): Fix operator precedence.
3006 (end cpu): Likewise.
3007 (END): Print the result from check_fpu.
3009 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3010 Alan Hayward <alan.hayward@arm.com>
3011 David Sherwood <david.sherwood@arm.com>
3013 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
3014 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
3015 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
3016 (*fcmuo<mode>_and): New patterns.
3018 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
3020 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
3021 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
3022 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
3023 (cmp_op, sve_imm_con): New code attributes.
3024 (SVE_COND_INT_CMP, imm_con): Delete.
3025 (cmp_op): Remove above unspecs from int attribute.
3026 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
3028 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
3029 comparison-specific unspecs.
3030 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
3031 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
3032 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
3033 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
3034 (*vec_fcm<cmp_op><mode>): Rename to...
3035 (*fcm<cmp_op><mode>): ...this and adjust likewise.
3036 (*vec_fcmuo<mode>): Rename to...
3037 (*fcmuo<mode>): ...this and adjust likewise.
3038 (*pred_fcm<cmp_op><mode>): New pattern.
3039 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3040 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3042 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3044 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3045 (aarch64_emit_sve_predicated_cond): New function.
3046 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3047 (aarch64_emit_unspec_cond_or): Replace with...
3048 (aarch64_emit_sve_or_conds): ...this new function. Use
3049 aarch64_emit_sve_ptrue_op for the individual comparisons and
3050 aarch64_emit_binop to OR them together.
3051 (aarch64_emit_inverted_unspec_cond): Replace with...
3052 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3053 aarch64_emit_sve_ptrue_op for the comparison and
3054 aarch64_emit_unop to invert the result.
3055 (aarch64_expand_sve_vec_cmp_float): Update after the above
3056 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3058 2018-05-07 Nathan Sidwell <nathan@acm.org>
3060 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3061 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3062 (Backwards Compatibility): Likewise.
3064 2018-05-07 Luis Machado <luis.machado@linaro.org>
3068 2018-05-07 Luis Machado <luis.machado@linaro.org>
3070 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3071 <prefetch_dynamic_strides>: New const bool field.
3072 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3073 prefetch_dynamic_strides.
3074 (exynosm1_prefetch_tune): Likewise.
3075 (thunderxt88_prefetch_tune): Likewise.
3076 (thunderx_prefetch_tune): Likewise.
3077 (thunderx2t99_prefetch_tune): Likewise.
3078 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3079 (aarch64_override_options_internal): Update to set
3080 PARAM_PREFETCH_DYNAMIC_STRIDES.
3081 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3082 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3083 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3084 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3085 prefetch-dynamic-strides setting.
3087 2018-05-07 Luis Machado <luis.machado@linaro.org>
3089 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3090 <minimum_stride>: New const int field.
3091 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3092 minimum_stride field.
3093 (exynosm1_prefetch_tune): Likewise.
3094 (thunderxt88_prefetch_tune): Likewise.
3095 (thunderx_prefetch_tune): Likewise.
3096 (thunderx2t99_prefetch_tune): Likewise.
3097 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3098 (aarch64_override_options_internal): Update to set
3099 PARAM_PREFETCH_MINIMUM_STRIDE.
3100 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3101 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3102 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3103 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3104 stride is constant and is below the minimum stride threshold.
3106 2018-05-07 Luis Machado <luis.machado@linaro.org>
3108 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3111 2018-05-07 Luis Machado <luis.machado@linaro.org>
3113 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3114 <prefetch_dynamic_strides>: New const bool field.
3115 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3116 prefetch_dynamic_strides.
3117 (exynosm1_prefetch_tune): Likewise.
3118 (thunderxt88_prefetch_tune): Likewise.
3119 (thunderx_prefetch_tune): Likewise.
3120 (thunderx2t99_prefetch_tune): Likewise.
3121 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3122 (aarch64_override_options_internal): Update to set
3123 PARAM_PREFETCH_DYNAMIC_STRIDES.
3124 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3125 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3126 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3127 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3128 prefetch-dynamic-strides setting.
3130 2018-05-07 Luis Machado <luis.machado@linaro.org>
3132 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3133 <minimum_stride>: New const int field.
3134 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3135 minimum_stride field.
3136 (exynosm1_prefetch_tune): Likewise.
3137 (thunderxt88_prefetch_tune): Likewise.
3138 (thunderx_prefetch_tune): Likewise.
3139 (thunderx2t99_prefetch_tune): Likewise.
3140 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3141 (aarch64_override_options_internal): Update to set
3142 PARAM_PREFETCH_MINIMUM_STRIDE.
3143 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3144 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3145 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3146 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3147 stride is constant and is below the minimum stride threshold.
3149 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3152 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3153 the type is addressable. Don't force op into register if it has
3156 2018-05-05 Roland McGrath <mcgrathr@google.com>
3159 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3160 any section for which we don't know a specific type it should have,
3161 regardless of name. Previously this was done only for the exact
3162 names ".init_array", ".fini_array", and ".preinit_array".
3163 (default_elf_asm_named_section): Add comment about
3164 relationship with default_section_type_flags and SECTION_NOTYPE.
3165 (get_section): Don't consider it a type conflict if one side has
3166 SECTION_NOTYPE and the other doesn't, as long as neither has the
3167 SECTION_BSS et al used in the default_section_type_flags logic.
3169 2018-05-05 Tom de Vries <tom@codesourcery.com>
3172 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3173 (workaround_barsyncs): New function.
3174 (nvptx_reorg): Use workaround_barsyncs.
3175 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3176 (define_expand "nvptx_membar_cta"): New define_expand.
3177 (define_insn "*nvptx_membar_cta"): New insn.
3179 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3181 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3182 To improve optimization opportunities.
3183 * builtin-types.def: The new needed builtin types for the above.
3185 2018-05-04 Richard Biener <rguenther@suse.de>
3187 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3188 * gimple-ssa-store-merging.c
3189 (imm_store_chain_info::output_merged_store): Remove redundant create,
3190 release split_store vector contents on failure.
3191 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3192 scalar stmt vector on cache hit.
3194 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3196 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3198 * config.gcc (powerpc-xilinx-eabi*): Remove.
3199 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3201 (fusion_addis_mem_combo_load): Ditto.
3202 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3204 (rs6000_cpu_cpp_builtins): Ditto.
3205 * config/rs6000/rs6000-linux.c
3206 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3207 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3208 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3210 (rs6000_setup_reg_addr_masks): Ditto.
3211 (rs6000_init_hard_regno_mode_ok): Ditto.
3212 (rs6000_option_override_internal): Ditto.
3213 (legitimate_lo_sum_address_p): Ditto.
3214 (rs6000_legitimize_address): Ditto.
3215 (rs6000_legitimize_reload_address): Ditto.
3216 (rs6000_legitimate_address_p): Ditto.
3217 (abi_v4_pass_in_fpr): Ditto.
3218 (setup_incoming_varargs): Ditto.
3219 (rs6000_gimplify_va_arg): Ditto.
3220 (rs6000_split_multireg_move): Ditto.
3221 (rs6000_savres_strategy): Ditto.
3222 (rs6000_emit_prologue_components): Ditto.
3223 (rs6000_emit_epilogue_components): Ditto.
3224 (rs6000_emit_prologue): Ditto.
3225 (rs6000_emit_epilogue): Ditto.
3226 (rs6000_elf_file_end): Ditto.
3227 (rs6000_function_value): Ditto.
3228 (rs6000_libcall_value): Ditto.
3229 * config/rs6000/rs6000.h: Ditto.
3230 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3231 (TARGET_MINMAX): ... this. New.
3232 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3233 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3234 (*movsi_internal1_single): Delete.
3235 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3236 mfpu=, mxilinx-fpu): Delete.
3237 * config/rs6000/singlefp.h: Delete.
3238 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3239 * config/rs6000/t-rs6000: Ditto.
3240 * config/rs6000/t-xilinx: Delete.
3241 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3242 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3243 (VStype_simple): Delete.
3244 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3245 * config/rs6000/xfpu.h: Delete.
3246 * config/rs6000/xfpu.md: Delete.
3247 * config/rs6000/xilinx.h: Delete.
3248 * config/rs6000/xilinx.opt: Delete.
3249 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3250 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3252 2018-05-04 Tom de Vries <tom@codesourcery.com>
3255 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3258 2018-05-04 Richard Biener <rguenther@suse.de>
3261 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3262 (expand_complex_div_wide): Likewise.
3263 (expand_complex_operations_1): Likewise.
3264 (expand_complex_libcall): Preserve EH info of the original stmt.
3265 (tree_lower_complex): Handle removed blocks.
3266 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3267 on complex multiplication and division libcall builtins.
3269 2018-05-04 Richard Biener <rguenther@suse.de>
3272 * fold-const.c (negate_expr_p): Restrict negation of operand
3273 zero of a division to when we know that can happen without
3275 (fold_negate_expr_1): Likewise.
3277 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3280 * real.h (real_nextafter): Declare.
3281 * real.c (real_nextafter): New function.
3282 * fold-const-call.c (fold_const_nextafter): New function.
3283 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3284 CASE_CFN_NEXTTOWARD.
3285 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3286 even when arg1_mode is different from arg0_mode.
3288 2018-05-03 Nathan Sidwell <nathan@acm.org>
3290 * doc/extend.texi (Deprecated Features): Remove
3292 (Backwards Compatibility): Likewise.
3293 * doc/invoke.texi (C++ Language Options): Likewise.
3294 (C++ Dialect Options): Likewise.
3296 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3299 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3300 _mm512_mask_mullox_epi64): New intrinsics.
3302 2018-05-03 Tom de Vries <tom@codesourcery.com>
3305 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3306 dump files): Add offload-tree.
3308 2018-05-03 Richard Biener <rguenther@suse.de>
3310 PR tree-optimization/85615
3311 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3312 to loops not nested in BBs loop father to avoid creating multi-entry
3315 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3317 PR tree-optimization/70291
3318 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3319 arguments. Change return type to tree. Emit libcall as a new
3320 statement rather than replacing existing one when inplace_p is true.
3321 (expand_complex_multiplication_components): New function.
3322 (expand_complex_multiplication): Expand floating-point complex
3323 multiplication using the above.
3324 (expand_complex_division): Rename inner_type parameter to type.
3325 Update expand_complex_libcall call-site.
3326 (expand_complex_operations_1): Update expand_complex_multiplication
3327 and expand_complex_division call-sites.
3329 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3332 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3333 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3334 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3335 the highest significant bit of the shift count mask is clear. In
3336 check whether and[sq]i3 is needed verify that all significant bits
3337 of the shift count other than the highest are set.
3339 2018-05-02 Tom de Vries <tom@codesourcery.com>
3342 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3343 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3344 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3345 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3346 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3347 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3348 __builtin_goacc_parlevel_size.
3350 2018-05-02 Richard Biener <rguenther@suse.de>
3352 PR tree-optimization/85597
3353 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3354 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
3363 2018-05-02 Tom de Vries <tom@codesourcery.com>
3366 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3367 dump files): Add wpa-ipa.
3369 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3371 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3372 powerpc*-*-linux*paired* target.
3373 * config/rs6000/750cl.h: Delete.
3374 * config/rs6000/paired.h: Delete.
3375 * config/rs6000/paired.md: Delete.
3376 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3378 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3379 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3380 comment. Remove paired float support.
3381 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3382 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3384 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3385 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3387 * config/rs6000/rs6000.c: Remove paired float support.
3388 (paired_expand_vector_init, paired_expand_vector_move,
3389 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3390 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3391 paired_expand_builtin, paired_expand_predicate_builtin,
3392 paired_init_builtins): Delete.
3393 * config/rs6000/rs6000.h: Remove paired float support.
3394 * config/rs6000/rs6000.md: Remove paired float support.
3395 (move_from_CR_ov_bit): Delete.
3396 * config/rs6000/rs6000.opt (mpaired): Delete.
3397 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3398 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3400 2018-05-02 Richard Biener <rguenther@suse.de>
3403 * gimplify.c (gimplify_save_expr): When in SSA form allow
3404 SAVE_EXPRs to compute to SSA vars.
3406 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3409 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3410 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3411 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3412 clobber operands[2], instead use a new pseudo. Formatting fixes.
3414 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3416 PR tree-optimization/85586
3417 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3418 exit early for statements in the same group if the accesses are
3421 2018-05-02 Tom de Vries <tom@codesourcery.com>
3424 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3427 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3429 PR tree-optimization/85143
3430 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3432 2018-05-01 Tom de Vries <tom@codesourcery.com>
3435 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3436 not found" error message.
3438 2018-05-01 Tom de Vries <tom@codesourcery.com>
3441 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3442 * vec.c (test_ordered_remove_if): New function.
3443 (vec_c_tests): Call test_ordered_remove_if.
3444 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3445 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3446 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3447 VEC_ORDERED_REMOVE_IF.
3449 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3451 PR tree-optimization/82665
3452 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3453 pointer subtraction where arguments come from a memchr call.
3455 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3457 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3458 --push-state --as-needed and --pop-state instead of --as-needed and
3459 --no-as-needed if ld supports it.
3460 * configure: Regenerated.
3463 * doc/install.texi2html: Replace _002d with - and _002a with * in
3464 generated html files using sed.
3466 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3469 * gcc-rich-location.c (blank_line_before_p): New function.
3470 (use_new_line): New function.
3471 (gcc_rich_location::add_fixit_insert_formatted): New function.
3472 * gcc-rich-location.h
3473 (gcc_rich_location::add_fixit_insert_formatted): New function.
3475 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3477 * selftest.c (assert_streq): Rename "expected" and "actual" to
3478 "val1" and "val2". Extend NULL-handling to cover both inputs
3479 symmetrically, while still requiring both to be non-NULL for a pass.
3480 * selftest.h (assert_streq): Rename "expected" and "actual" to
3482 (ASSERT_EQ): Likewise.
3483 (ASSERT_EQ_AT): Likewise.
3484 (ASSERT_KNOWN_EQ): Likewise.
3485 (ASSERT_KNOWN_EQ_AT): Likewise.
3486 (ASSERT_NE): Likewise.
3487 (ASSERT_MAYBE_NE): Likewise.
3488 (ASSERT_MAYBE_NE_AT): Likewise.
3489 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3490 the assertion to pass.
3491 (ASSERT_STREQ_AT): Likewise.
3493 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3495 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3496 interaction with -pie.
3498 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3500 * selftest.h: Fix alphabetization of per-source-file selftest
3503 2018-04-30 Jason Merrill <jason@redhat.com>
3505 PR c++/61982 - dead stores to destroyed objects.
3506 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3509 2018-04-30 Jason Merrill <jason@redhat.com>
3511 * tree.c (build_clobber): New.
3512 * tree.h: Declare it.
3513 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3515 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3517 * diagnostic-show-locus.c (layout::layout): Update for
3518 location_get_source_line returning a char_span.
3519 (struct char_span): Move to input.h.
3520 (struct correction): Update for fields in char_span becoming
3522 (struct source_line): Update for location_get_source_line
3523 returning a char_span.
3524 (layout::print_line): Likewise.
3525 * edit-context.c (edited_file::print_content): Likewise.
3526 (edited_file::print_diff_hunk): Likewise.
3527 (edited_file::print_run_of_changed_lines): Likewise.
3528 (edited_file::get_num_lines): Likewise.
3529 (edited_line::edited_line): Likewise.
3530 * final.c (asm_show_source): Likewise.
3531 * input.c (location_get_source_line): Convert return type
3532 from const char * to char_span, losing the final "line_len"
3534 (dump_location_info): Update for the above.
3535 (get_substring_ranges_for_loc): Likewise. Use a char_span
3536 when handling the literal within the line.
3537 (test_reading_source_line): Update for location_get_source_line
3538 returning a char_span.
3539 * input.h (class char_span): Move here from
3540 diagnostic-show-locus.c, converting from a struct to a class.
3541 Make data members private.
3542 (char_span::operator bool): New.
3543 (char_span::length): New.
3544 (char_span::get_buffer): New.
3545 (char_span::operator[]): New.
3546 (char_span::subspan): Make const.
3547 (char_span::xstrdup): New.
3548 (location_get_source_line): Convert return type from const char *
3549 to char_span, losing the final "line_size" param.
3551 2018-04-30 Jan Hubicka <jh@suse.cz>
3553 * lto-wrapper.c (ltrans_priorities): New static var.
3554 (cmp_priority): New.
3555 (run_gcc): Read priorities and if doing parallel build order
3556 the Makefile by them.
3558 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3560 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3562 2018-04-30 Richard Biener <rguenther@suse.de>
3564 * tree-cfg.c (verify_address): Remove base argument, add
3565 flag whether to check TREE_ADDRESSABLE and do that.
3566 (verify_expr): Remove.
3567 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3568 (verify_gimple_assign_single): Likewise.
3569 (verify_gimple_switch): Likewise.
3570 (verify_expr_location_1): Dereference tp once. Add (disabled)
3571 piece from verify_expr.
3572 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3574 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3576 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3578 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3580 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3581 (small_data_pattern): Likewise.
3582 (arc_rewrite_small_data): Likewise.
3583 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3584 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3585 (get_symbol_alignment): New function.
3586 (legitimate_small_data_address_p): Likewise.
3587 (legitimate_scaled_address): Update, call
3588 legitimate_small_data_address_p.
3589 (output_sdata): New static variable.
3590 (arc_print_operand): Update how we handle small data operands.
3591 (arc_print_operand_address): Likewise.
3592 (arc_legitimate_address_p): Update, use
3593 legitimate_small_data_address_p.
3594 (arc_rewrite_small_data_p): Remove.
3595 (arc_rewrite_small_data_1): Likewise.
3596 (arc_rewrite_small_data): Likewise.
3597 (small_data_pattern): Likewise.
3598 (compact_sda_memory_operand): Update to use
3599 legitimate_small_data_address_p and get_symbol_alignment.
3600 (prepare_move_operands): Don't rewite sdata pattern.
3601 (prepare_extend_operands): Remove.
3602 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3604 (zero_extendqisi2): Likewise.
3605 (zero_extendhisi2): Likewise.
3606 (extendqihi2): Likewise.
3607 (extendqisi2): Likewise.
3608 (extendhisi2): Likewise.
3612 * config/arc/constraints.md (Usd): Change it to memory constraint.
3614 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3616 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3617 as source of std instructions.
3618 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3619 allow 6-bit constants as source for store instructions.
3620 (movdi_insn): Update instruction pattern to allow 6-bit constants
3621 as source for store instructions.
3623 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3625 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3627 2018-04-30 Nathan Sidwell <nathan@acm.org>
3628 Sandra Loosemore <sandra@codesourcery.com>
3630 * dumpfile.c (dump_open): Allow '-' for stdout.
3631 * doc/invoke.texi (Developer Options): Document dump filename
3632 determination early. Document stdin/stdout selection.
3634 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3636 Microblaze Target: PIC data text relative
3638 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3639 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3641 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3642 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3643 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3644 New addressing mode for data-text relative position indepenedent code.
3645 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3646 'ADDRESS_SYMBOLIC_TXT_REL'.
3647 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3648 (microblaze_legitimate_pic_operand): Exclude function calls from
3649 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3650 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3652 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3653 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3654 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3655 for 'address + offset'.
3656 (microblaze_expand_prologue): Add new function prologue call for
3658 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3659 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3660 table in case of TARGET_PIC_DATA_TEXT_REL.
3661 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3662 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3663 Add new macros 'UNSPEC_TEXT',
3664 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3665 + exclude function calls from 'UNSPEC_PLT' in case of data text
3667 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3668 new target hook for generating address diff vector tables in case of
3670 * doc/tm.texi : Regenerate.
3671 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3672 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3673 of addr diff vector generation.
3674 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3675 target hook definition.
3676 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3677 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3678 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3679 option for data text relative.
3681 2018-04-30 Richard Biener <rguenther@suse.de>
3683 * tree-chrec.h (evolution_function_is_constant_p): Remove
3685 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3687 2018-04-30 Richard Biener <rguenther@suse.de>
3690 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3692 2018-04-30 Richard Biener <rguenther@suse.de>
3694 PR tree-optimization/28364
3695 PR tree-optimization/85275
3696 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3697 copying first exit test.
3699 2018-04-28 Mark Wielaard <mark@klomp.org>
3701 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3703 (dwarf_AT): Handle DW_AT_addr_base.
3704 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3706 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3709 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3710 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3711 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3712 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3714 2018-04-28 Richard Biener <rguenther@suse.de>
3716 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3717 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3718 to reflect use. Only add interesting stmts.
3720 2018-04-27 Martin Jambor <mjambor@suse.cz>
3723 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3724 the jump function allows for passing through aggregate values.
3726 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3728 * input.h (in_system_header_at): Convert from macro to inline
3730 (from_macro_expansion_at): Likewise.
3731 (from_macro_definition_at): Likewise.
3733 2018-04-27 Jeff Law <law@redhat.com>
3735 * config.gcc: Mark tile* targets as deprecated/obsolete.
3737 2018-04-27 Richard Biener <rguenther@suse.de>
3739 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3742 2018-04-27 Richard Biener <rguenther@suse.de>
3744 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3746 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3748 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3749 with Yd constraint. Set "preferred_for_speed" attribute from
3750 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3752 (*movdi_internal): Ditto.
3753 (movti_interunit splitters): Remove
3754 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3755 (movdi_interunit splitters): Ditto.
3756 * config/i386/constraints.md (Ye): Remove.
3757 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3759 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3762 * config/aarch64/constraints.md (Usg): Limit to 31.
3765 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3767 PR tree-optimization/85529
3768 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3769 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3770 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3771 zero extension or masking of the MSB bit.
3772 (optimize_range_tests): Add FIRST_BB argument, pass it through
3773 to optimize_range_tests_var_bound.
3774 (maybe_optimize_range_tests, reassociate_bb): Adjust
3775 optimize_range_tests callers.
3777 2018-04-26 Richard Biener <rguenther@suse.de>
3778 Jakub Jelinek <jakub@redhat.com>
3780 * cgraph.h (symbol_table): Just declare debug method here.
3781 * symtab.c (symbol_table::debug): Define.
3783 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3785 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3787 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3789 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3790 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3791 (*movdi_internal): Substitute Yi and Yj constraint with x
3792 and Ym and Yn constraint with y constraint. Update "isa"
3793 attribute and set "preferred_for_speed" attribute from
3794 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3795 (*movsi_internal): Ditto.
3796 (*movdf_internal): Ditto.
3797 (*movsf_internal): Ditto.
3798 (*zero_extendsidi2): Ditto.
3799 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3800 (sse2_loadld): Ditto.
3801 (*vec_extract<ssevecmodelower>_0): Ditto.
3802 (*vec_extractv4si_0_zext_sse4): Ditto.
3803 (vec_concatv2di): Ditto.
3804 (*vec_dup<mode>): Ditto.
3805 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3806 * config/i386/constraints.md (Yi): Remove.
3811 2018-04-26 Nathan Sidwell <nathan@acm.org>
3813 * dumpfile.c (dump_open): New.
3814 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3815 (dump_finish): Detect stdio/stderr by value not name.
3817 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3819 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3821 2018-04-26 Tom de Vries <tom@codesourcery.com>
3824 * config/nvptx/nvptx.c (verify_neutering_jumps)
3825 (verify_neutering_labels): New function
3826 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3828 2018-04-26 Tom de Vries <tom@codesourcery.com>
3831 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3832 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3835 2018-04-26 Richard Biener <rguenther@suse.de>
3836 Tom de Vries <tom@codesourcery.com>
3839 * lto-streamer-out.c (output_function): Fixup loops if required to match
3840 discovery done in the reader.
3842 2018-04-26 Richard Biener <rguenther@suse.de>
3844 PR tree-optimization/85116
3845 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3846 have a loop exit from the single latch predecessor. Remove
3847 case of header with just condition.
3848 (ch_base::copy_headers): Exclude infinite loops from any
3850 (pass_ch::execute): Record exits.
3852 2018-04-26 Richard Biener <rguenther@suse.de>
3854 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3855 prologue cost vector and pass it to vect_get_load_cost.
3856 (vect_get_peeling_costs_all_drs): Likewise.
3857 (vect_peeling_hash_get_lowest_cost): Likewise.
3858 (vect_enhance_data_refs_alignment): Likewise.
3860 2018-04-26 Richard Biener <rguenther@suse.de>
3863 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3864 checking of integer<->pointer conversions.
3865 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3866 sign-/zero-extending pointer types.
3867 (expand_omp_for_static_chunk): Likewise.
3869 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3870 Jean Lee <xiaoyur347@gmail.com>
3872 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3873 (TARGET_ASAN_SHADOW_OFFSET): Define.
3874 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3875 true for -fsanitize=address.
3877 2018-04-25 Mark Wielaard <mark@klomp.org>
3879 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3882 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3884 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3885 than "alu", remove explicit "memory" and "imm_disp" attributes.
3886 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3889 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3890 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3891 gen_lowpart_no_emit.
3893 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3896 * config/i386/i386.c (ix86_expand_builtin): Change memory
3897 operand to XI, extend p0 to Pmode.
3898 * config/i386/i386.md: Change unspec volatile and operand
3899 1 mode to XI, change operand 0 mode to P.
3901 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3903 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3904 GET_MODE_MASK before any checking.
3905 (nds32_can_use_bset_p): Likewise.
3906 (nds32_can_use_btgl_p): Likewise.
3908 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32-doubleword.md: New define_split pattern for
3911 illegal register number.
3913 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3915 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3917 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3919 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3921 2018-04-25 Richard Biener <rguenther@suse.de>
3923 * lto-streamer.h (LTO_major_version): Bump to 8.
3925 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3927 * BASE-VER: Set to 9.0.0.
3929 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3931 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3932 in __abskf2 and __powikf2.
3934 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3937 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3938 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3939 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3940 Use the above on operand 2. Reindent.
3941 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3943 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3946 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3948 * config/i386/i386.opt (mcet): Removed.
3949 * doc/install.texi: Remove -mcet documentation.
3950 * doc/invoke.texi: Likewise.
3952 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3955 * doc/install.texi: Remove -mcet from bootstrap-cet.
3957 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3960 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3961 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3965 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3966 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3967 containing a CONST_VECTOR.
3969 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3971 * doc/install.texi: Update newlib dependency for nvptx.
3973 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3976 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3977 instead of INTVAL when shifting x left.
3979 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3981 PR tree-optimization/85478
3982 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3983 vect_grouped_store_supported for single element vectors.
3985 2018-04-24 Richard Biener <rguenther@suse.de>
3988 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3989 load cost increase to the case of non-constant step.
3991 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3994 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3995 destination if any_malformed_asm.
3997 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
4000 * expr.c (store_field): In the bitfield case, if the value comes from
4001 a function call and is returned in registers by means of a PARALLEL,
4002 do not change the mode of the temporary unless BLKmode and VOIDmode.
4004 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
4006 PR rtl-optimization/85423
4007 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
4008 dependencies to debug insns when the previous insn is non-debug.
4010 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4012 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
4013 enums into a single definition.
4014 (fls): Fix predicates and printing.
4017 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
4019 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
4020 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
4021 and short u6 immediate.
4022 (check_if_valid_sleep_operand): Remove.
4023 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
4025 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4027 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
4028 flag_always_save_lp condition.
4029 * config/nds32/nds32.opt (malways-save-lp): New option.
4031 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4033 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
4034 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
4035 * config/nds32/nds32.h
4036 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
4037 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4039 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4041 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4042 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4044 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4045 Chung-Ju Wu <jasonwucj@gmail.com>
4047 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4049 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4050 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4052 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4054 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4056 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4058 * config/nds32/nds32-protos.h (nds32_data_alignment,
4059 nds32_local_alignment): Declare.
4060 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4061 nds32_local_alignment): New functions.
4062 (TARGET_CONSTANT_ALIGNMENT): Define.
4063 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4065 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4067 * config/nds32/nds32.c
4068 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4069 (TARGET_MODES_TIEABLE_P): Likewise.
4071 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4073 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4076 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4077 Chung-Ju Wu <jasonwucj@gmail.com>
4079 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4080 for unaligned access.
4081 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4083 * config/nds32/nds32-intrinsic.md: Likewise.
4084 * config/nds32/nds32_intrinsic.h: Likewise.
4085 * config/nds32/nds32.h (nds32_builtins): Likewise.
4086 * config/nds32/nds32.opt (munaligned-access): New option.
4087 * config/nds32/nds32.c (nds32_asm_file_start): Display
4088 flag_unaligned_access status.
4090 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4092 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4093 -mno-relax is present.
4094 * config/riscv/linux.h (LINK_SPEC): Ditto.
4096 2018-04-20 Martin Sebor <msebor@redhat.com>
4099 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4101 (gimple_fold_builtin_stxcpy_chk): Same.
4102 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4104 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4107 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4108 __powikf2 when long double is IEEE 128-bit.
4110 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4112 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4113 step to make sure stack always aligned.
4115 2018-04-20 Carl Love <cel@us.ibm.com>
4118 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4119 size check for arg0.
4121 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4122 Tom de Vries <tom@codesourcery.com>
4125 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4126 Emit insns for calls too.
4127 (nvptx_find_par): Always look for worker-level predecessor insn.
4128 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4130 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4131 (nvptx_process_pars): Propagate frames for calls.
4133 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4136 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4138 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4139 (ix86_handle_option): Don't handle OPT_mibt.
4140 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4142 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4144 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4145 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4146 (ix86_target_macros): Define __CET__ with flag_cf_protection
4147 for -fcf-protection.
4148 * config/i386/i386.c (isa2_opts): Remove -mibt.
4149 * config/i386/i386.h (TARGET_IBT): Removed.
4150 (TARGET_IBT_P): Likewise.
4151 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4152 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4153 * config/i386/i386.opt (mcet): Update help message.
4156 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4157 -mcet as an alias for -mshstk.
4159 2018-04-20 Richard Biener <rguenther@suse.de>
4162 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4163 complexity by forcing a single use of the multiply operand.
4165 2018-04-20 Martin Jambor <mjambor@suse.cz>
4168 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4169 recursion dependency to only apply to non-clones.
4171 2018-04-20 Martin Jambor <mjambor@suse.cz>
4174 * ipa-cp.c (create_specialized_node): Check that clones of
4175 self-recursive edges exist during IPA-CP.
4177 2018-04-19 Toon Moene <toon@moene.org>
4179 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4182 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4184 PR tree-optimization/85467
4185 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4186 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4187 VECTOR_CST element to type.
4189 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4192 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4193 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4194 (builtin_longjmp): Likewise.
4195 (save_stack_nonlocal): New pattern.
4196 (restore_stack_nonlocal): Likewise.
4198 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4201 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4202 Replace ASM_OUTPUT_LABEL with fprintf.
4204 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4207 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4208 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4209 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4210 define __IBT__ and __SHSTK__ for -fcf-protection.
4211 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4213 (ix86_trampoline_init): Likewise.
4214 (x86_output_mi_thunk): Likewise.
4215 (ix86_notrack_prefixed_insn_p): Likewise.
4216 (ix86_option_override_internal): Don't disallow -fcf-protection.
4217 * config/i386/i386.md (rdssp<mode>): Also enable for
4219 (incssp<mode>): Likewise.
4220 (nop_endbr): Likewise.
4221 * config/i386/i386.opt (mcet): Change help message to built-in
4225 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4226 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4227 enable CET built-in functions.
4229 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4231 * common/config/i386/i386-common.c
4232 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4233 OPTION_MASK_ISA_MOVDIRI_UNSET,
4234 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4235 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4236 * config.gcc (movdirintrin.h): New header.
4237 * config/i386/cpuid.h (bit_MOVDIRI,
4238 bit_MOVDIR64B): New bits.
4239 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4241 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4242 (VOID, PVOID, PCVOID)): New function types.
4243 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4244 __builtin_ia32_directstoreu_u64,
4245 __builtin_ia32_movdir64b): New builtins.
4246 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4247 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4249 (ix86_valid_target_attribute_inner_p): Ditto.
4250 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4251 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4252 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4253 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4254 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4255 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4256 (movdiri<mode>, movdir64b_<mode>): New.
4257 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4258 * config/i386/immintrin.h: Include movdirintrin.h.
4259 * config/i386/movdirintrin.h: New file.
4260 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4262 2018-04-19 Richard Biener <rguenther@suse.de>
4265 * cfg.c (clear_bb_flags): When loop state says we have
4266 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4268 2018-04-19 Richard Biener <rguenther@suse.de>
4270 PR tree-optimization/84737
4271 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4272 copying restrict info.
4273 (vect_setup_realignment): Use it.
4274 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4275 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4276 the first DR to all generated stores.
4277 (vectorizable_load): Likewise for loads.
4279 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4281 PR tree-optimization/85446
4282 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4283 the integral and pointer types to have the same precision.
4285 * doc/install.texi: Document --disable-cet being the default and
4288 2018-04-18 Martin Liska <mliska@suse.cz>
4290 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4293 2018-04-18 Martin Liska <mliska@suse.cz>
4296 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4299 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4300 arguments if they are comparable.
4302 2018-04-18 Martin Liska <mliska@suse.cz>
4305 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4308 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4311 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4314 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4315 ENDBR after calling __morestack.
4317 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4320 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4321 by using gcc_base_ver to generate a gcc_driver_version, and use
4322 it when generating GCC_DRIVER_NAME.
4323 * configure: Regenerate.
4325 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4328 * config.gcc: Obsolete powerpc*-*-*spe*.
4330 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4333 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4334 (stabstr_D): Change type of unum from unsigned int to
4335 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4338 2018-04-17 Jim Wilson <jimw@sifive.com>
4341 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4342 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4343 Set arg_pointer_offset after using pretend_args_size.
4345 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4347 PR rtl-optimization/85431
4348 * dse.c (record_store): Ignore zero width stores.
4351 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4352 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4353 __builtin_stack_restore rather than after it.
4354 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4355 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4356 argument instead of virtual_dynamic_stack_rtx.
4358 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4360 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4362 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4363 Add note to error message to explain internal mapping of overloaded
4364 built-in function name to non-overloaded built-in function name.
4365 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4368 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4371 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4372 where the inputs overlap with the output.
4374 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4376 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4377 (=v, v) alternative and explicit "memory" attribute.
4378 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4379 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4381 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4382 "sselog1" type instead of "sselog".
4383 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4384 "sselog". Remove explicit "memory" attribute.
4385 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4386 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4388 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4389 "sselog1" type instead of "sselog". Remove explicit "memory"
4391 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4392 use "sselog1" type instead of "sselog". Remove explicit "memory"
4394 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4395 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4397 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4398 "sselog1" type instead of "sselog". Remove explicit "memory"
4400 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4401 use "sselog1" type instead of "sselog". Remove explicit "memory"
4405 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4408 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4411 2018-04-17 Martin Jambor <mjambor@suse.cz>
4414 * ipa-cp.c (create_specialized_node): Call
4415 expand_all_artificial_thunks if necessary.
4417 2018-04-17 Martin Liska <mliska@suse.cz>
4420 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4421 in message, remote space in between '_G' and '('.
4423 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4426 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4427 avx512f_vmcmp<mode>3<round_saeonly_name>,
4428 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4429 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4430 avx512f_rndscale<mode><round_saeonly_name>,
4431 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4432 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4433 Use %<iptr>2 instead of %2 for -masm=intel.
4434 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4435 avx512f_vcvttss2usi<round_saeonly_name>,
4436 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4438 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4439 avx512f_vcvttsd2usi<round_saeonly_name>,
4440 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4441 Use %q1 instead of %1 for -masm=intel.
4442 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4443 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4444 of %3 for -masm=intel.
4445 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4447 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4449 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4451 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4452 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4454 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4456 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4457 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4458 %g1 and one with %0 and %1.
4459 (avx512er_vmrcp28<mode><round_saeonly_name>,
4460 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4462 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4463 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4464 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4465 of %0 and %{%4%} for -masm=intel.
4466 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4467 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4468 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4469 order of %0 and %{%5%}%{z%} for -masm=intel.
4471 2018-04-17 Jan Hubicka <jh@suse.cz>
4474 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4476 2018-04-17 Martin Liska <mliska@suse.cz>
4479 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4480 for target_clone error message. Make default implementation
4481 clone to be a local declaration.
4482 (separate_attrs): Add new argument and check for an empty
4484 (expand_target_clones): Handle it.
4485 (ipa_target_clone): Make redirection just for target_clones
4488 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4489 Tom de Vries <tom@codesourcery.com>
4492 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4493 tiled basic blocks without omp continue statements.
4495 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4498 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4499 vec_extract expression as having side effects to make sure it gets
4502 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4505 * config/i386/i386.c (get_builtin_code_for_version): Check
4508 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4511 * gcc/config.gcc: Support "skylake".
4512 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4514 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4515 (processor_target_table): Add "skylake".
4516 (ix86_option_override_internal): Add "skylake".
4517 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4518 PROCESSOR_CANNONLAKE.
4519 (get_builtin_code_for_version): Fix priority for
4520 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4521 PROCESSOR_SKYLAKE-AVX512.
4522 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4523 (processor_type): Add PROCESSOR_SKYLAKE.
4525 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4526 Jason Merrill <jason@redhat.com>
4529 * convert.c (convert_to_integer_1): Use direct recursion for
4530 enumeral types and types with a precision less than the number
4531 of bits in their mode.
4533 2018-04-16 Julia Koval <julia.koval@intel.com>
4536 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4537 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4539 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4542 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4543 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4544 and -mno-direct-move.
4546 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4549 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4550 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4551 Ensure negative shifts result in {0}.
4553 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4555 PR rtl-optimization/79916
4556 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4557 regs (if any) to define how to gnerate SD moves when LRA is in
4560 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4562 PR rtl-optimization/85393
4563 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4564 * except.c (expand_dw2_landing_pad_for_region): Make static.
4565 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4566 a label and unconditional jump to old_bb, rather than
4567 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4570 PR rtl-optimization/85376
4571 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4572 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4573 instead of a specific value.
4575 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4576 Bin Cheng <bin.cheng@arm.com>
4578 PR tree-optimization/82965
4579 PR tree-optimization/83991
4580 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4581 by_profile_only parameter.
4582 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4583 information if the loop was predicted to iterate too many times.
4584 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4586 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4589 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4592 2018-04-13 Martin Liska <mliska@suse.cz>
4593 Jakub Jelinek <jakub@redhat.com>
4596 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4597 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4598 * builtins.c (expand_builtin_memory_copy_args): Use
4599 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4600 handle dest_addr == pc_rtx.
4602 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4605 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4606 asked to not generate direct moves.
4607 (fix_trunc<mode>si2_stfiwx): Similar.
4608 (fix_trunc<mode>si2_internal): Similar.
4610 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4613 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4614 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4615 lookup if dest in some wider mode is known to be const0_rtx and
4616 if so, record permanent equivalence for it to be ZERO_EXTEND of
4617 the narrower mode destination.
4619 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4621 * lto-streamer-out.c (output_function): Revert 259346.
4622 * omp-expand.c (expand_oacc_for): Likewise.
4624 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4626 PR rtl-optimization/85354
4627 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4628 * sel-sched.c (sel_global_init): ... here.
4630 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4633 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4634 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4635 mode for PE-COFF targets.
4636 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4637 (i386_pe_asm_lto_end): Likewise.
4638 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4639 (TARGET_ASM_LTO_END): Likewise.
4640 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4641 (i386_pe_asm_lto_start): New function.
4642 (i386_pe_asm_lto_end): Likewise.
4644 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4645 Richard Biener <rguenther@suse.de>
4648 * lto-streamer-out.c (output_function): Fix CFG loop state before
4650 * omp-expand.c (expand_oacc_for): Handle calls to internal
4651 functions like regular functions.
4653 2018-04-12 Richard Biener <rguenther@suse.de>
4656 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4657 for the early LTO debug to properly generate references to it
4658 during DIE emission. Do not re-use that for the skeleton for
4660 (dwarf2out_early_finish): Likewise.
4662 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4665 * config/i386/sse.md
4666 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4667 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4668 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4669 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4670 and output is a reg, avoid creating invalid lowpart subreg, but
4671 instead split into a 512-bit move. Don't split if not AVX512VL,
4672 input is xmm16+ reg and output is a mem.
4673 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4674 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4675 xmm16+ reg and output is a mem.
4677 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4679 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4680 also for flag_dwarf2_cfi_asm.
4682 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4684 PR rtl-optimization/85342
4685 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4686 a bool scalar var inside of the loop instead. Don't try to update
4687 recog_data.operand after failed apply_change_group.
4689 2018-04-12 Tom de Vries <tom@codesourcery.com>
4692 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4693 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4694 array with flexible array member as array without given dimension.
4695 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4696 argument for undefined param to true.
4698 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4701 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4702 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4703 from PowerPC section.
4704 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4705 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4706 help text that is too long.
4707 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4708 help text that is too long.
4709 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4710 help text that is too long.
4712 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4714 * config/alpha/alpha.md (stack_probe_internal): Rename
4715 from "probe_stack". Update all callers.
4717 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4719 PR rtl-optimization/84566
4720 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4721 sched_macro_fuse_insns.
4723 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4726 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4727 (compute_block_dependences): ... from here.
4729 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4731 PR tree-optimization/85331
4732 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4733 from int to HOST_WIDE_INT.
4735 2018-04-11 Martin Jambor <mjambor@suse.cz>
4738 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4739 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4740 not the same as the source val.
4741 (cgraph_edge_brings_value_p): New parameter.
4742 (gather_edges_for_value): Pass destination value to
4743 cgraph_edge_brings_value_p.
4744 (perhaps_add_new_callers): Likewise.
4745 (get_info_about_necessary_edges): Likewise and exclude values brought
4746 only by self-recursive edges.
4747 (create_specialized_node): Redirect only clones of self-calling edges.
4748 (+self_recursive_pass_through_p): New function.
4749 (find_more_scalar_values_for_callers_subset): Use it.
4750 (find_aggregate_values_for_callers_subset): Likewise.
4751 (known_aggs_to_agg_replacement_list): Removed.
4752 (decide_whether_version_node): Re-calculate known constants for all
4753 remaining context clones.
4755 2018-04-11 Richard Biener <rguenther@suse.de>
4758 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4759 from early DWARF output.
4760 (dwarf2out_early_finish): Output line info unconditionally into
4761 early DWARF and add reference to it.
4763 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4766 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4767 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4768 other than V2DFmode using iptr mode attribute.
4769 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4771 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4773 PR rtl-optimization/84659
4774 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4776 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4779 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4781 (output_loc_list): Pass address of a dummy size variable even in the
4782 locview handling loop.
4783 (index_location_lists): Add comment on why skip_loc_list_entry can't
4786 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4789 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4792 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4795 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4796 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4797 and -mstring-compare-inline-limit.
4799 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4802 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4803 for stack clash protection in a register whenever we need it to be in
4806 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4808 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4809 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4811 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4814 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4816 (mlong-double-): Ditto.
4817 * config/rs6000/sysv4.opt (msdata=): Ditto.
4818 (mtls-size=): Ditto.
4820 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4822 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4823 erroneous entries for
4824 "vector int vec_ldl (int, long int *)", and
4825 "vector unsigned int vec_ldl (int, unsigned long int *)".
4826 Add comments and entries for
4827 "vector bool char vec_ldl (int, bool char *)",
4828 "vector bool short vec_ldl (int, bool short *)",
4829 "vector bool int vec_ldl (int, bool int *)",
4830 "vector bool long long vec_ldl (int, bool long long *)",
4831 "vector pixel vec_ldl (int, pixel *)",
4832 "vector long long vec_ldl (int, long long *)",
4833 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4834 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4835 type tree bool_long_long_type_node and correct definition of
4836 bool_V2DI_type_node to make reference to this new type tree.
4837 (rs6000_mangle_type): Replace erroneous reference to
4838 bool_long_type_node with bool_long_long_type_node.
4839 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4840 comments to emphasize sign distinctions for char and int types and
4841 replace RS6000_BTI_bool_long constant with
4842 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4843 use of RS6000_BTI_pixel.
4844 (bool_long_type_node): Remove this macro definition.
4845 (bool_long_long_type_node): New macro definition
4847 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4849 PR rtl-optimization/85300
4850 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4851 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4852 simplify_unary_operation fails.
4854 2018-04-10 Martin Liska <mliska@suse.cz>
4856 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4857 cgraph_edge and ipa_ref.
4859 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4863 * config/i386/sse.md
4864 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4865 computation of the VEC_MERGE selector from mask.
4866 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4867 Fix decoding of the VEC_MERGE selector into mask.
4869 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4871 PR tree-optimization/85286
4872 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4874 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4876 * final.c (final_1): Set insn_last_address as well as
4877 insn_current_address.
4879 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4882 * explow.c (emit_stack_probe): Call validize_mem on memory location
4883 before passing it to gen_probe_stack. Create address operand and
4884 legitimize it for the probe_stack_address case.
4886 2018-04-09 Jan Hubicka <jh@suse.cz>
4889 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4890 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4891 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4893 (free_lang_data): Rebuild type inheritance graph.
4895 2018-04-09 Martin Sebor <msebor@redhat.com>
4897 * invoke.texi (-finline-small-functions): Mention other optimization
4899 (-findirect-inlining, -fpartial-inlining): Same.
4900 (-finline-functions-called-once): Same.
4901 (-freorder-blocks-and-partition): Same.
4903 2018-04-09 Jan Hubicka <jh@suse.cz>
4906 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4907 jumps; choose last target that matches the criteria (i.e.
4908 no partition changes for non-crossing jumps).
4909 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4910 support for redirecting crossing jumps to non-crossing.
4912 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4914 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4915 also for naked functions.
4917 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4919 * config/arc/arc.md (add_shift): New pattern.
4920 (add_shift2): Likewise.
4921 (sub_shift): Likewise.
4922 (sub_shift_cmp0_noout): Likewise.
4923 (compare_si_ashiftsi): Likewise.
4924 (xbfu_cmp0_noout): New combine pattern.
4925 (xbfu_cmp0"): Likewise.
4926 (movsi_set_cc_insn): Place the predicable variant first.
4927 (commutative_binary_cmp0_noout): Remove clobber.
4928 (commutative_binary_cmp0): New pattern.
4929 (noncommutative_binary_cmp0): Likewise.
4930 (noncommutative_binary_cmp0_noout): Likewise.
4931 (noncommutative_binary_comparison_result_used): Removed.
4932 (rsub_cmp0): New pattern.
4933 (rsub_cmp0_noout): Likewise.
4934 (extzvsi): Changed, keep only meaningful variants.
4935 (SQH, SEZ): New iterators.
4936 (SQH_postfix): New mode attribute.
4937 (SEZ_prefix): New code attribute.
4938 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4939 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4940 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4942 (noncommutative_operator): Check the availability of barrel
4945 2018-04-09 Richard Biener <rguenther@suse.de>
4947 PR tree-optimization/85284
4948 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4949 Only use the niter constraining form of simple_iv when the exit
4952 2018-04-09 Tom de Vries <tom@codesourcery.com>
4955 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4956 (define_expand "*memory_barrier"): New define_expand.
4957 (define_insn "memory_barrier"): New insn.
4959 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4961 PR rtl-optimization/80463
4962 PR rtl-optimization/83972
4963 PR rtl-optimization/83480
4965 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4966 correct producer for the insn.
4967 (tidy_control_flow): Fixup seqnos in case of debug insns.
4969 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4971 PR rtl-optimization/83913
4973 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4974 different sched-times when merging exprs.
4976 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4978 PR rtl-optimization/83962
4980 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4981 tidy_fallthru_edge and tidy_control_flow.
4983 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4985 PR rtl-optimization/83530
4987 * sel-sched.c (force_next_insn): New global variable.
4988 (remove_insn_for_debug): When force_next_insn is true, also leave only
4989 next insn in the ready list.
4990 (sel_sched_region): When the region wasn't scheduled, make another pass
4991 over it with force_next_insn set to 1.
4993 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4995 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4997 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4998 for interrupt control.
4999 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5000 functions for interrupt control.
5001 * config/nds32/nds32-intrinsic.md: Likewise.
5002 * config/nds32/nds32_intrinsic.h: Likewise.
5003 * config/nds32/nds32.h (nds32_builtins): Likewise.
5005 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5007 * config/nds32/nds32.c (nds32_init_machine_status,
5008 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
5009 strict_aligned_p field.
5010 (nds32_expand_to_rtl_hook): New function.
5011 (TARGET_EXPAND_TO_RTL_HOOK): Define.
5012 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
5014 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5015 Chung-Ju Wu <jasonwucj@gmail.com>
5017 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
5018 * config/nds32/nds32-n7.md: New file.
5019 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
5020 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
5022 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
5023 * config/nds32/nds32.md (pipeline_model): Add n7.
5024 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
5025 * config/nds32/pipelines.md: Include n7 settings.
5027 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5028 Chung-Ju Wu <jasonwucj@gmail.com>
5030 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
5031 * config/nds32/nds32-e8.md: New file.
5032 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
5033 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
5035 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
5036 * config/nds32/nds32.md (pipeline_model): Add e8.
5037 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5038 * config/nds32/pipelines.md: Include e8 settings.
5040 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5041 Chung-Ju Wu <jasonwucj@gmail.com>
5043 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5044 * config/nds32/nds32-n8.md: New file.
5045 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5046 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5048 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5049 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5050 * config/nds32/nds32.md (pipeline_model): Add n8.
5051 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5052 * config/nds32/pipelines.md: Include n8 settings.
5054 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5055 Chung-Ju Wu <jasonwucj@gmail.com>
5057 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5058 * config/nds32/nds32-n9-2r1w.md: New file.
5059 * config/nds32/nds32-n9-3r2w.md: New file.
5060 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5061 nds32_register_ports): New or modify for cpu n9.
5062 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5064 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5065 * config/nds32/nds32-utils.c: New file.
5066 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5067 TARGET_MUL_SLOW): Define.
5068 * config/nds32/nds32.md (pipeline_model): New attribute.
5069 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5070 New options that support cpu n9.
5071 * config/nds32/pipelines.md: Include n9 settings.
5072 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5074 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5076 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5077 information if necessary.
5078 (output_cond_branch_compare_zero): Likewise.
5079 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5080 (nds32_target_alignment): Refine for alignment.
5081 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5082 (FUNCTION_BOUNDARY): Modify.
5083 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5085 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5087 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5089 * config/nds32/constants.md (unspec_volatile_element): Add values for
5090 TLB operation and data prefetch.
5091 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5092 functions for TLB operation and data prefetch.
5093 * config/nds32/nds32-intrinsic.md: Likewise.
5094 * config/nds32/nds32_intrinsic.h: Likewise.
5095 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5096 (nds32_print_operand): Likewise.
5097 * config/nds32/nds32.h (nds32_builtins): Likewise.
5099 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5100 Andrew Pinski <pinsika@gcc.gnu.org>
5103 * match.pd: Use constant_boolean_node of correct type instead of
5104 boolean_true_node or boolean_false_node for simplifying
5105 pointer comparisons to zero.
5107 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5109 PR tree-optimization/80021
5110 * tree.c (verify_type_variant): Make error call in verify_variant_match
5111 translatable and remove final full stop.
5113 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5115 * config/nds32/constants.md (unspec_volatile_element): Add
5116 UNSPEC_VOLATILE_EH_RETURN.
5117 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5118 nds32_output_stack_pop): Support dwarf exception handling process.
5119 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5120 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5121 exception handling process.
5122 (nds32_compute_stack_frame): Likewise.
5123 (nds32_return_addr_rtx): Likewise.
5124 (nds32_initial_elimination_offset): Likewise.
5125 (nds32_expand_prologue): Likewise.
5126 (nds32_expand_epilogue): Likewise.
5127 (nds32_dynamic_chain_address): New function.
5128 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5130 (DYNAMIC_CHAIN_ADDRESS): Define.
5131 (EH_RETURN_DATA_REGNO): Define.
5132 (EH_RETURN_STACKADJ_RTX): Define.
5133 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5134 patterns for dwarf exception handling.
5136 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5138 * config/nds32/nds32.h: Clean up obsolete macros.
5140 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5142 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5143 Add enum values for particular instructions.
5144 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5145 particular intrinsic functions.
5146 * config/nds32/nds32-intrinsic.md: Likewise.
5147 * config/nds32/nds32_intrinsic.h: Likewise.
5148 * config/nds32/nds32.h (nds32_builtins): Likewise.
5149 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5150 (btst, ave): New patterns for particular instructions.
5152 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5154 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5155 Add enum values for atomic load/store and memory sync.
5156 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5158 * config/nds32/nds32-intrinsic.md: Likewise.
5159 * config/nds32/nds32_intrinsic.h: Likewise.
5160 * config/nds32/nds32.h (nds32_builtins): Likewise.
5162 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5164 PR tree-optimization/85257
5165 * fold-const.c (native_encode_vector): If not all elts could fit
5166 and off is -1, return 0 rather than offset.
5167 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5168 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5169 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5170 adjust buffer in native_interpret_expr call.
5172 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5174 * config/nds32/constants.md (unspec_volatile_element): Add cache
5175 control enum values.
5176 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5177 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5178 * config/nds32/nds32.c (nds32_cctl_names): New.
5179 (nds32_print_operand): Handle cache control register names.
5180 * config/nds32/nds32.h (nds32_builtins): New enum values.
5181 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5183 * config/nds32/nds32.md (type): Add mmu.
5184 * config/nds32/pipelines.md (simple_insn): Add mmu.
5186 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5188 * config/nds32/nds32.md (type): Remove call.
5189 * config/nds32/pipelines.md (simple_insn): Likewise.
5191 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5193 * config/nds32/constants.md (unspec_volatile_element): Add
5194 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5195 UNSPEC_VOLATILE_FMFCFG.
5196 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5197 description for fmfcfg and fmfcsr.
5198 (bdesc_1arg): Add fmtcsr.
5199 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5200 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5201 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5202 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5203 unspec_fmfcfg): New patterns.
5204 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5205 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5206 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5207 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5208 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5209 __nds32__fmfcfg): Define.
5211 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5213 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5214 intrinsic register names.
5215 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5216 intrinsic register enum values and macros.
5218 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5220 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5221 for load/store addressing form.
5222 (nds32_print_operand_address): Likewise.
5224 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5227 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5228 based on LABEL_REF. Remove useless assertion.
5229 (pic_address_needs_scratch): Fix formatting.
5230 (sparc_legitimize_pic_address): Minor tweaks.
5231 (sparc_delegitimize_address): Adjust assertion accordingly.
5232 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5233 into symbolic_operand.
5234 (movsi_high_pic_label_ref): Likewise.
5235 (movsi_lo_sum_pic_label_ref): Likewise.
5236 (movdi_pic_label_ref): Likewise.
5237 (movdi_high_pic_label_ref): Likewise.
5238 (movdi_lo_sum_pic_label_ref): Likewise.
5240 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5242 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5243 custom LIB_SPEC setup.
5245 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5246 Kito Cheng <kito.cheng@gmail.com>
5248 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5249 * config/riscv/freebsd.h: New.
5251 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5253 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5254 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5257 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5258 Kito Cheng <kito.cheng@gmail.com>
5260 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5261 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5262 * config/nds32/nds32-protos.h (nds32_output_call,
5263 nds32_output_return): Declare.
5264 * config/nds32/nds32.md: Refine all the call and return patterns.
5266 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5269 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5270 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5272 PR rtl-optimization/84872
5273 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5274 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5277 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5279 * expr.c (copy_blkmode_to_reg): Revert 254862.
5280 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5282 2018-04-06 Richard Biener <rguenther@suse.de>
5285 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5286 after seeing a component reference with an adjacent field. Treat
5287 refs to arrays at struct end of external decls similar to
5288 refs to unconstrained commons.
5290 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5293 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5294 look through SAVE_EXPRs with non-side-effects argument. Adjust
5296 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5299 2018-04-06 Richard Biener <rguenther@suse.de>
5302 * alias.c (find_base_term): New wrapper around find_base_term
5303 unwinding CSELIB_VAL_PTR changes.
5304 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5307 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5309 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5311 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5312 constant definitions.
5313 ("nop"): lr 0,0 -> nopr r0
5314 ("nop_lr0", "nop_lr1"): New insn definitions.
5316 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5318 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5319 NDS32_V3PUSH_AVAILABLE_P macro.
5321 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5322 Chung-Ju Wu <jasonwucj@gmail.com>
5324 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5325 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5326 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5327 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5328 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5329 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5330 * config/nds32/constraints.md: New constraints and checking for hard
5331 float configuration.
5332 * config/nds32/iterators.md: New mode iterator and attribute for hard
5333 float configuration.
5334 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5336 * config/nds32/nds32-fpu.md: New file.
5337 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5338 deal with hard float code generation.
5339 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5341 (abi_type, float_reg_number): New enum type.
5342 * config/nds32/nds32-predicates.c: New predicates for hard float.
5343 * config/nds32/nds32-protos.h: Declare functions for hard float.
5344 * config/nds32/nds32.c: Implementation for hard float configuration.
5345 * config/nds32/nds32.h: Definitions for hard float configuration.
5346 * config/nds32/nds32.md: Include hard float machine description and
5347 modify patterns for hard float configuration.
5348 * config/nds32/nds32.opt: New options for hard float configuration.
5349 * config/nds32/predicates.md: New predicates for hard float
5352 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5354 * common/config/nds32/nds32-common.c
5355 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5357 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5360 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5361 CONSTRUCTOR_ELT (ctor, ...)->value.
5363 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5366 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5368 2018-04-05 Tom de Vries <tom@codesourcery.com>
5371 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5374 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5375 Kito Cheng <kito.cheng@gmail.com>
5377 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5378 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5379 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5380 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5382 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5383 Kito Cheng <kito.cheng@gmail.com>
5385 * config/nds32/constraints.md (Ufe): New memory constraint.
5386 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5387 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5388 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5390 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5391 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5393 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5395 * config/nds32/nds32.md: Use optimize_size in the condition for
5396 alu-shift instructions.
5398 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5400 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5402 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5404 * config/nds32/nds32.md (negsi2): Refine pattern.
5406 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5407 Chung-Ju Wu <jasonwucj@gmail.com>
5409 * config/nds32/iterators.md (shift_rotate): New code iterator.
5410 (shift): New code attribute.
5411 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5412 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5413 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5414 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5415 bit-wise operations.
5416 (andsi3, *andsi3): Ditto.
5417 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5418 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5419 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5420 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5421 nds32_ior_operand, nds32_xor_operand): New predicates.
5423 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5425 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5426 (addsi3, subsi3): ... this.
5428 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5430 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5432 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5434 * config/nds32/nds32.md: Adjust indention.
5436 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5438 * config/nds32/nds32.md (feature): New attribute.
5440 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5442 * config/nds32/nds32.md (subtype): New attribute.
5444 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5447 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5448 expansion to perform a bitwise AND of the argument followed by a
5449 boolean negation of the result.
5451 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5453 PR rtl-optimization/84878
5454 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5455 the basic block. Assert the use reference is not artificial and that
5456 it has an associated insn.
5458 2018-04-04 Michael Matz <matz@suse.de>
5460 * builtins.c (compute_objsize): Pass correct operand
5461 to array_at_struct_end_p.
5463 2018-04-04 Richard Biener <rguenther@suse.de>
5466 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5467 from contexts for DINFO_LEVEL_TERSE and below.
5469 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5471 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5472 resiter_operand condition.
5473 * config/nds32/nds32.md (*move<mode>): Ditto.
5475 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5476 Monk Chiang <sh.chiang04@gmail.com>
5478 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5480 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5482 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5484 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5485 Kito Cheng <kito.cheng@gmail.com>
5487 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5488 nds32_cond_code_str, output_cond_branch,
5489 output_cond_branch_compare_zero, 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: New functions.
5495 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5496 nds32_expand_cstore, nds32_expand_movcc,
5497 nds32_output_cbranchsi4_equality_zero,
5498 nds32_output_cbranchsi4_equality_reg,
5499 nds32_output_cbranchsi4_equality_reg_or_const_int,
5500 nds32_output_cbranchsi4_greater_less_zero): Declare.
5501 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5502 nds32_rimm11s_operand): New predicates.
5503 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5504 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5507 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5509 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5510 * config/nds32/nds32.md: Ditto.
5511 * config/nds32/pipelines.md: Ditto.
5513 2018-04-04 Richard Biener <rguenther@suse.de>
5515 PR tree-optimization/85168
5516 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5517 propagating abnormals.
5519 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5521 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5523 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5524 Kito Cheng <kito.cheng@gmail.com>
5526 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5527 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5528 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5529 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5530 * config/nds32/nds32.md (sibcall_internal): New.
5531 (sibcall_register): Remove.
5532 (sibcall_immediate): Remove.
5533 (sibcall_value_internal): New.
5534 (sibcall_value_register): Remove.
5535 (sibcall_value_immediate): Remove.
5536 * config/nds32/predicates.md (nds32_general_register_operand): New.
5537 (nds32_call_address_operand): New.
5539 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5541 PR rtl-optimization/85167
5542 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5543 bb_defs if *split_p, instead preinitialize it to NULL.
5545 PR tree-optimization/85156
5546 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5547 evaluating the argument multiple times.
5549 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5551 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5553 (_mm_cvtpd_ps): Likewise.
5554 (_mm_cvttpd_epi32): Likewise.
5555 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5556 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5557 vector, pixel, and bool following altivec.h include.
5559 2018-04-03 Martin Sebor <msebor@redhat.com>
5561 * doc/extend.texi (Common Function Attributes): Clarify.
5562 (const attribute): Likewise.
5563 (pure attribute): Likewise.
5565 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5568 * config/i386/i386.c (ix86_expand_vector_set): Use
5569 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5571 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5573 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5574 instructions when changing rounding bits to preserve precision bits
5575 in the x87 control word.
5577 2018-04-03 Martin Liska <mliska@suse.cz>
5579 PR tree-optimization/82491
5580 * rtl.h (strip_offset_and_add): Replace += suboffset with
5581 poly_uint64 () + suboffset.
5583 2018-03-29 Martin Liska <mliska@suse.cz>
5584 Martin Jambor <mjambor@suse.cz>
5587 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5588 param_type is not an integral or pointer type.
5590 2018-04-03 Richard Biener <rguenther@suse.de>
5592 * sese.h (recompute_all_dominators): Remove.
5594 2018-04-02 Martin Sebor <msebor@redhat.com>
5596 * doc/invoke.texi (-Wrestrict): Fix typos.
5598 2018-04-02 Jim Wilson <jimw@sifive.com>
5600 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5601 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5602 (<optab>di3, <optab>si3_extend): Likewise.
5603 (<optab>si3_mask, <optab>si3_mask_1): New.
5604 (<optab>di3_mask, <optab>di3_mask_1): New.
5605 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5606 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5607 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5609 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5611 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5614 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5616 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5617 (nds32_canonicalize_comparison): New function.
5619 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5620 Kito Cheng <kito.cheng@gmail.com>
5621 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5623 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5624 * config/nds32/constants.md (unspec_volatile_element): Add
5625 UNSPEC_VOLATILE_RELAX_GROUP.
5626 * config/nds32/nds32-relax-opt.c: New file.
5627 * config/nds32/nds32-predicates.c
5628 (nds32_symbol_load_store_p): New function.
5629 * config/nds32/nds32-protos.h
5630 (nds32_symbol_load_store_p): Declare function.
5631 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5632 * config/nds32/nds32.c
5633 (nds32_register_pass): New function to register pass.
5634 (nds32_register_passes): New function to register passes.
5635 * config/nds32/nds32.md (relax_group): New pattern.
5636 * config/nds32/nds32.opt (mrelax-hint): New option.
5637 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5639 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5641 * config/nds32/t-nds32: Modify files dependency.
5643 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5645 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5646 (PROFILE_HOOK): Define its implementation.
5648 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5650 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5651 type and 32-bit size.
5653 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5656 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5657 (V_128_256): New mode iterator.
5658 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5659 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5660 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5662 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5663 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5665 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5668 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5669 NaN inputs correctly.
5671 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5674 * config/rs6000/vsx.md (??r): New mode attribute.
5675 (*vsx_mov<mode>_64bit): Use it.
5676 (*vsx_mov<mode>_32bit): Likewise.
5678 2018-03-30 Martin Sebor <msebor@redhat.com>
5680 PR tree-optimization/84818
5681 * builtins.c (check_access): Use warning_n.
5683 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5686 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5688 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5691 2018-03-30 Julia Koval <julia.koval@intel.com>
5694 * x86-tune.def (movx, partial_reg_dependency): Enable for
5697 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5700 * lra-constraints.c (process_alt_operands): Move setting
5701 this_alternative_matches below.
5703 2018-03-29 Martin Liska <mliska@suse.cz>
5706 * doc/invoke.texi: Document how LTO works with debug info.
5707 Describe auto-load support of binutils. Mention 'x86-64'
5708 as valid option value of -march option.
5710 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5712 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5715 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5716 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5717 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5720 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5723 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5724 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5725 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5726 for RS6000_BTM_POWERPC64.
5727 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5728 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5729 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5734 2018-03-28 Carl Love <cel@us.ibm.com>
5737 2017-09-27 Carl Love <cel@us.ibm.com>
5739 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5740 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5741 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5744 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5746 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5747 instead of __vector bool.
5748 (_mm_max_pu8): Likewise.
5749 (_mm_min_pi16): Likewise.
5751 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5754 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5755 (DIVWEUO): Likewise.
5757 (DIVDEUO): Likewise.
5758 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5759 DIVWEUO and DIVDEUO.
5760 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5761 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5762 (div_extend): Likewise.
5763 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5765 (__builtin_divweuo): Likewise.
5766 (__builtin_divdeo): Likewise.
5767 (__builtin_divdeuo): Likewise.
5769 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5772 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5773 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5775 PR tree-optimization/82004
5776 * gimple-match-head.c (optimize_pow_to_exp): New function.
5777 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5778 Don't fold to exp if optimize_pow_to_exp is false.
5780 2018-03-28 Martin Liska <mliska@suse.cz>
5783 * calls.c (initialize_argument_information): Fix trailing space.
5784 * common.opt: Fix typo and provide better explanation for
5785 -fsanitize-coverage option.
5786 * config/i386/i386.opt: Fix typo.
5788 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5789 Martin Liska <mliska@suse.cz>
5792 * gimplify.c (asan_poison_variable): Don't do the check for
5793 gimplify_omp_ctxp here.
5794 (gimplify_decl_expr): Do it here.
5795 (gimplify_target_expr): Likewise.
5797 2018-03-28 Martin Liska <mliska@suse.cz>
5800 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5801 chkp_type_bounds_count if MPX is not enabled.
5803 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5805 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5807 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5810 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5811 function to create the function decl for complex long double
5812 multiply and divide for -mabi=ieeelongdouble.
5813 (init_float128_ieee): Call it.
5815 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5818 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5819 -fcf-protection=branch -mibt.
5820 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5822 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5825 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5827 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5830 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5831 extern array declarations.
5833 2018-03-27 Richard Biener <rguenther@suse.de>
5836 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5837 explicit single_use checks.
5839 2018-03-27 Richard Biener <rguenther@suse.de>
5841 PR tree-optimization/85082
5842 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5845 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5847 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5848 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5849 Turn on fasynchronous-unwind-tables and funwind-tables.
5851 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5854 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5855 (*bmi_blsr_<mode>_ccz): Ditto.
5857 2018-03-26 Tom de Vries <tom@codesourcery.com>
5859 PR tree-optimization/85063
5860 * omp-general.c (offloading_function_p): New function. Factor out
5862 * omp-offload.c (pass_omp_target_link::gate): ... here.
5863 * omp-general.h (offloading_function_p): Declare.
5864 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5865 with attribute omp declare target for offloading functions.
5867 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5869 PR tree-optimization/84005
5870 * tree-data-ref.h (get_base_for_alignment): Declare.
5871 * tree-data-ref.c (get_base_for_alignment_1): New function.
5872 (get_base_for_alignment): Likewise.
5873 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5874 get_base_for_alignment to find a suitable base object, instead
5875 of always using drb->base_address.
5877 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5880 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5881 known size by default.
5883 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5886 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5887 and non BLKmode operands.
5889 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5892 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5893 Clean up attributes.
5895 2018-03-23 Richard Biener <rguenther@suse.de>
5898 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5899 we are going to emit early debug for LTO.
5901 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5904 * function.c (match_asm_constraints_1): Don't optimize if input
5905 doesn't satisfy general_operand predicate for output's mode.
5908 * alias.c (write_dependence_p): Don't require for x_canonicalized
5909 non-VOIDmode if x has VOIDmode.
5912 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5913 just don't try to optimize it rather than assert it never happens.
5915 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5917 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5918 macro expansions for definition of ST_INTERNAL_<mode> and
5919 LD_INTERNAL_<mode> builtins.
5920 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5922 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5924 (altivec_expand_st_builtin): Likewise.
5925 (altivec_expand_builtin): Remove calls to deleted functions.
5926 (rs6000_address_for_altivec): Delete this function.
5927 * config/rs6000/vector.md: Remove expands for
5928 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5930 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5933 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5934 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5935 re-computing once computed.
5936 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5937 (arm_init_machine_status): Initialize
5938 machine->static_chain_stack_bytes.
5940 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5943 * doc/extend.texi: Add four new prototypes for vec_ld.
5944 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5945 definitions for more logical presentation.
5946 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5947 entries for V1TI variants of __builtin_altivec_ld builtin.
5948 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5949 handling of V1TI variant of LVX icode pattern.
5950 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5951 (rs6000_gimple_fold_builtin): Likewise.
5952 (altivec_init_builtins): Add code to define
5953 __builtin_altivec_lvx_v1ti function.
5955 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5958 * function.c (match_asm_constraints_1): Don't do the optimization
5959 if input isn't a REG, SUBREG, MEM or constant.
5961 2018-03-22 Tom de Vries <tom@codesourcery.com>
5963 PR tree-optimization/84956
5964 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5965 bb_has_abnormal_pred.
5967 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5970 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5971 DECL_INITIAL (decl) to decl at the end.
5972 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5975 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5977 * doc/extend.texi (__builtin_tgmath): Document when complex
5978 integer types are treated as _Complex _Float64.
5980 2018-03-21 Tom de Vries <tom@codesourcery.com>
5982 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5984 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5986 PR tree-optimization/84960
5987 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5988 if it is ENTRY block, move them into single succ of ENTRY in that case.
5990 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5992 PR tree-optimization/84811
5993 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5994 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5995 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5996 (known_subrange_p): Update accordingly. Cast each value involved
5997 in the size comparison, rather than casting the result of the
6000 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6002 PR tree-optimization/84982
6003 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
6004 by flipping the least significant bit rather than all bits from
6005 bitpos to bitpos + bitsize - 1.
6007 2018-03-21 Nathan Sidwell <nathan@acm.org>
6009 * doc/extend.texi (Deprecated Features): Remove mention of
6010 long-deleted deprecations.
6012 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6015 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
6016 * configure: Regenerate.
6018 2018-03-21 Tom de Vries <tom@codesourcery.com>
6020 PR tree-optimization/83126
6021 * tree-parloops.c (num_phis): New function.
6022 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
6024 2018-03-21 Nathan Sidwell <nathan@acm.org>
6026 * doc/extend.texi (Deprecated Features): Update deprecated flags,
6027 mention anon-struct/union members and trailing attributes.
6029 2018-03-21 Bin Cheng <bin.cheng@arm.com>
6031 PR tree-optimization/84969
6032 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
6033 builtin memset partitions if they set different rhs values.
6035 2018-03-21 Jakub Jelinek <jakub@redhat.com>
6037 PR rtl-optimization/84989
6038 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6039 VEC_DUPLICATE with scalar result mode.
6041 2018-03-21 Martin Liska <mliska@suse.cz>
6044 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6045 not intended return statement.
6047 2018-03-21 Martin Liska <mliska@suse.cz>
6050 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6051 (chkp_find_bound_slots_1): Limit number of iterations.
6053 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6056 * Minor grammar fixes for x86 options.
6058 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6061 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6062 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6064 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6067 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6068 (altivec_lvx_<mode>_1op): Likewise.
6069 (altivec_stvx_<mode>_2op): Likewise.
6070 (altivec_stvx_<mode>_1op): Likewise.
6071 (altivec_lvx_<VM2:mode>): New define_expand.
6072 (altivec_stvx_<VM2:mode>): Likewise.
6073 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6074 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6075 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6076 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6077 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6078 (rs6000_gen_lvx): Likewise.
6079 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6080 (altivec_expand_stv_builtin): Likewise.
6081 (altivec_expand_builtin): Likewise.
6082 * config/rs6000/vector.md: Likewise.
6084 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6087 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6090 2018-03-20 Richard Biener <rguenther@suse.de>
6093 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6094 sign-conversions as zero, fall back to standard scalar_stmt
6097 2018-03-20 Martin Liska <mliska@suse.cz>
6100 * predict.c (rebuild_frequencies): Handle case when we have
6101 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6103 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6106 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6107 flag_section_anchors.
6108 * varasm.c (use_blocks_for_decl_p): Remove hack for
6109 dw2_force_const_mem.
6112 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6114 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6115 be created, use lowpart_subreg of operands[0] rather than operands[0]
6117 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6118 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6119 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6120 and n constraint instead of aarch64_shift_imm_di and Usd.
6121 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6122 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6124 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6127 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6128 to favor GPR over NEON registers.
6129 (<shift>di3_neon): Likewise.
6131 2018-03-20 Tom de Vries <tom@codesourcery.com>
6134 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6135 (nvptx_process_pars): Emit bar.sync asap and alap.
6137 2018-03-20 Tom de Vries <tom@codesourcery.com>
6140 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6141 seen_label if seen_label is already set.
6143 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6146 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6147 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6148 Use 1U instead of 1. Formatting fixes.
6151 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6152 instead of TREE_TYPE (s1) for the return value.
6154 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6156 PR tree-optimization/84946
6157 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6158 bitsize + bitsize in poly_uint64 rather than poly_int64.
6161 * dwarf2asm.c: Include fold-const.c.
6162 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6163 of decl rather than decl itself.
6165 PR rtl-optimization/84643
6166 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6168 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6171 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6172 calling assemble_variable.
6174 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6177 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6178 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6180 2018-03-19 Jim Wilson <jimw@sifive.com>
6183 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6184 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6185 (riscv_first_stack_step): Likewise.
6186 (riscv_option_override): Use STACK_BOUNDARY instead of
6188 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6190 (BIGGEST_ALIGNMENT): Set to 128.
6191 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6192 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6195 2018-03-19 Richard Biener <rguenther@suse.de>
6197 PR tree-optimization/84933
6198 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6199 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6201 2018-03-19 Richard Biener <rguenther@suse.de>
6203 PR tree-optimization/84859
6204 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6205 (cond_if_else_store_replacement): Perform sinking operation on
6206 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6207 Generalize what a BB with a single eligible store is.
6209 2018-03-19 Richard Biener <rguenther@suse.de>
6211 PR tree-optimization/84929
6212 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6213 chrec_is_positive against non-chrec arg.
6215 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6218 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6220 2018-03-18 Martin Liska <mliska@suse.cz>
6222 PR rtl-optimization/84635
6223 * regrename.c (build_def_use): Use matches_mode only when
6226 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6228 PR tree-optimization/84913
6229 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6230 vectorize chains of COND_EXPRs.
6232 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6234 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6236 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6238 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6240 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6242 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6244 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6245 Kito Cheng <kito.cheng@gmail.com>
6247 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6248 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6249 (nds32_adjust_reg_alloc_order): New function.
6250 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6252 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6254 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6255 nds32_print_operand, nds32_print_operand_address): Use
6256 HOST_WIDE_INT_PRINT_DEC instead.
6258 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6260 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6262 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6265 * config/i386/i386.c (initial_ix86_tune_features,
6266 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6268 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6269 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6270 rather than 1u << ix86_tune. Formatting fix.
6271 (ix86_option_override_internal): Change ix86_arch_mask from
6272 unsigned int to unsigned HOST_WIDE_INT, initialize to
6273 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6274 (ix86_function_specific_restore): Likewise.
6276 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6279 * postreload.c (reload_combine_recognize_pattern): Perform
6280 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6281 truncate_int_for_mode the result for the destination's mode.
6284 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6285 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6287 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6290 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6293 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6294 regno_allocno_class_array and sorted_pseudos.
6295 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6296 insns where regno is used.
6298 2018-03-16 Martin Liska <mliska@suse.cz>
6301 * multiple_target.c (create_dispatcher_calls): Redirect
6302 reference in the symbol table.
6304 2018-03-16 Martin Liska <mliska@suse.cz>
6307 * multiple_target.c (create_dispatcher_calls): Redirect also
6310 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6314 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6315 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6316 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6318 2018-03-16 Julia Koval <julia.koval@intel.com>
6320 * doc/invoke.texi (Skylake Server): Add CLWB.
6321 Cannonlake): Remove CLWB.
6323 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6325 PR tree-optimization/84841
6326 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6328 (FLOAT_ONE_CONST_TYPE): Define.
6329 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6330 (sort_by_operand_rank): Put entries with higher constant_type last
6331 rather than first to match comments.
6333 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6335 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6338 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6341 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6342 check and use address of target always.
6344 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6347 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6348 (indirect_thunk_bnd_needed): Likewise.
6349 (indirect_thunks_used): Likewise.
6350 (indirect_thunks_bnd_used): Likewise.
6351 (indirect_return_needed): New.
6352 (indirect_return_bnd_needed): Likewise.
6353 (output_indirect_thunk_function): Add a bool argument for
6355 (output_indirect_thunk_function): Don't generate alias for
6356 function return thunk.
6357 (ix86_code_end): Call output_indirect_thunk_function to generate
6358 function return thunks.
6359 (ix86_output_function_return): Set indirect_return_bnd_needed
6360 and indirect_return_needed instead of indirect_thunk_bnd_needed
6361 and indirect_thunk_needed.
6363 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6365 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6366 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6367 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6369 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6370 Paul Hua <paul.hua.gm@gmail.com>
6373 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6375 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6377 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6378 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6381 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6384 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6385 instead of GET_MODE_SIZE when comparing Units.
6387 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6390 * varasm.c (hash_section): Return an unchangeble hash value
6391 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6392 Return !aarch64_can_use_per_function_literal_pools_p ().
6394 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6397 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6398 rather than address of cmode as last argument to prepare_cmp_insn.
6400 2018-03-15 Julia Koval <julia.koval@intel.com>
6402 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6403 F_AVX512VNNI, F_AVX512BITALG): New.
6405 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6408 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6409 insn for floating-point loads and stores.
6411 2018-03-14 Carl Love <cel@us.ibm.com>
6413 * config/rs6000/rs6000-c.c: Add macro definitions for
6414 ALTIVEC_BUILTIN_VEC_PERMXOR.
6415 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6416 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6417 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6418 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6420 * config/doc/extend.texi: Add prototypes for vec_permxor.
6422 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6425 * diagnostic-show-locus.c (class layout_point): Convert m_line
6426 from int to linenum_type.
6427 (line_span::comparator): Use linenum "compare" function when
6428 comparing line numbers.
6429 (test_line_span): New function.
6430 (layout_range::contains_point): Convert param "row" from int to
6432 (layout_range::intersects_line_p): Likewise.
6433 (layout::will_show_line_p): Likewise.
6434 (layout::print_source_line): Likewise.
6435 (layout::should_print_annotation_line_p): Likewise.
6436 (layout::print_annotation_line): Likewise.
6437 (layout::print_leading_fixits): Likewise.
6438 (layout::annotation_line_showed_range_p): Likewise.
6439 (struct line_corrections): Likewise for field m_row.
6440 (line_corrections::line_corrections): Likewise for param "row".
6441 (layout::print_trailing_fixits): Likewise.
6442 (layout::get_state_at_point): Likewise.
6443 (layout::get_x_bound_for_row): Likewise.
6444 (layout::print_line): Likewise.
6445 (diagnostic_show_locus): Likewise for locals "last_line" and
6447 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6448 * input.c (selftest::test_linenum_comparisons): New function.
6449 (selftest::input_c_tests): Call it.
6450 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6451 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6452 * selftest.h (ASSERT_GT): New macro.
6453 (ASSERT_GT_AT): New macro.
6454 (ASSERT_LT): New macro.
6455 (ASSERT_LT_AT): New macro.
6457 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6459 PR rtl-optimization/84780
6460 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6462 2018-03-14 Martin Liska <mliska@suse.cz>
6464 * tree.c (record_node_allocation_statistics): Use
6465 get_stats_node_kind.
6466 (get_stats_node_kind): New function extracted from
6467 record_node_allocation_statistics.
6468 (free_node): Use get_stats_node_kind.
6470 2018-03-14 Richard Biener <rguenther@suse.de>
6472 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6473 that the value-set of ANTIC_IN doesn't grow.
6476 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6478 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6479 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6481 2018-03-14 Julia Koval <julia.koval@intel.com>
6483 * config.gcc (icelake-client, icelake-server): New.
6485 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6486 (initial_ix86_arch_features): Ditto.
6487 (PTA_SKYLAKE): Add SGX.
6488 (PTA_ICELAKE): Remove.
6489 (PTA_ICELAKE_CLIENT): New.
6490 (PTA_ICELAKE_SERVER): New.
6491 (ix86_option_override_internal): Split up icelake on icelake client and
6493 (get_builtin_code_for_version): Ditto.
6494 (fold_builtin_cpu): Ditto.
6495 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6496 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6497 * config/i386/i386.h (processor_type): Ditto.
6498 * doc/invoke.texi: Ditto.
6500 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6503 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6504 INTEGER_CST offset, add it together with bitpos / 8 and
6505 sign extend based on POINTER_SIZE.
6509 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6512 * config/i386/constraints.md (Yc): New register constraint.
6513 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6514 Use Yc constraint for alternative 2 of operand 0. Remove
6515 preferred_for_speed attribute.
6517 2018-03-14 Richard Biener <rguenther@suse.de>
6519 PR tree-optimization/84830
6520 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6521 with the old one to avoid oscillations.
6523 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6526 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6528 (assign_by_spills): Return a flag of reload assignment failure.
6529 Do not process the reload assignment failures. Do not spill other
6530 reload pseudos if they has the same reg class. Update n if
6532 (lra_assign): Add a return arg. Set up from the result of
6533 assign_by_spills call.
6534 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6535 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6536 usage_insns if it is not NULL.
6537 (spill_hard_reg_in_range): New function.
6538 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6539 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6540 function prototypes.
6541 (lra_assign): Change prototype.
6542 * lra.c (lra): Add code to deal with fails by splitting hard reg
6545 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6547 * config/riscv/riscv.opt (mrelax): New option.
6548 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6549 "norelax" when riscv_mrelax is disabled.
6550 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6552 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6555 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6556 reassociation for int modes.
6558 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6560 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6561 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6563 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6564 * config/aarch64/aarch64-sve.md
6565 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6566 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6567 (*extend<mode><Vwide>2): Rename to...
6568 (aarch64_sve_extend<mode><Vwide>2): ...this.
6569 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6570 renaming the old pattern to...
6571 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6573 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6574 define_expand, renaming the old pattern to...
6575 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6576 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6577 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6578 account when deciding which SVE instruction the optab should use.
6579 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6581 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6583 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6584 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6585 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6586 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6587 (tlsdesc_small_<mode>): Turn a define_expand and use
6588 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6589 (tlsdesc_small_advsimd_<mode>): ...this.
6590 (tlsdesc_small_sve_<mode>): New pattern.
6592 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6594 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6595 (UNSPEC_UMUL_HIGHPART): New constants.
6596 (MUL_HIGHPART): New int iteraor.
6597 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6598 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6600 (*<su>mul<mode>3_highpart): New define_insn.
6602 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6605 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6608 2018-03-13 Martin Liska <mliska@suse.cz>
6611 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6613 (sem_item_optimizer::~sem_item_optimizer): Release it.
6614 (sem_item_optimizer::merge_classes): Register variable aliases.
6615 (sem_item_optimizer::fixup_pt_set): New function.
6616 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6617 * ipa-icf.h: Declare new variables and functions.
6619 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6622 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6623 integer_pow2p@2 and test integer_pow2p in condition.
6624 (A < 0 ? C : 0): Similarly for @1.
6627 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6628 characters starting at p contain '\0' character, don't look beyond
6632 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6633 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6636 * reg-stack.c (change_stack): Change update_end var from int to
6637 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6638 also call set_block_for_insn on the newly added insns and rescan.
6641 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6642 on the last operand.
6645 * tree.c (stabilize_reference_1): Return save_expr (e) for
6646 STATEMENT_LIST even if it doesn't have side-effects.
6648 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6650 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6652 2018-03-12 Renlin Li <renlin.li@arm.com>
6654 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6655 aarch64_output_scalar_simd_mov_immediate.
6657 2018-03-12 Martin Sebor <msebor@redhat.com>
6659 PR tree-optimization/83456
6660 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6661 for perfectly overlapping calls to memcpy.
6662 (gimple_fold_builtin_memory_chk): Same.
6663 (gimple_fold_builtin_strcpy): Handle no-warning.
6664 (gimple_fold_builtin_stxcpy_chk): Same.
6665 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6667 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6669 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6670 parameter. Use it for SFmode.
6671 (rs6000_function_arg_advance_1): Adjust.
6672 (rs6000_function_arg): Adjust.
6673 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6675 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6677 PR rtl-optimization/84169
6678 PR rtl-optimization/84780
6679 * combine.c (can_combine_p): Check for a 2-insn combination whether
6680 the destination register is used between the two insns, too.
6682 2018-03-12 Richard Biener <rguenther@suse.de>
6684 PR tree-optimization/84803
6685 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6686 for refs DR analysis didn't process.
6688 2018-03-12 Richard Biener <rguenther@suse.de>
6690 PR tree-optimization/84777
6691 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6692 force-vectorize loops ignore whether we are optimizing for size.
6694 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6696 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6697 (TARGET_MD_ASM_ADJUST): Define.
6699 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6700 Kito Cheng <kito.cheng@gmail.com>
6701 Chung-Ju Wu <jasonwucj@gmail.com>
6703 * config/nds32/nds32.c (nds32_compute_stack_frame,
6704 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6705 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6706 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6707 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6708 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6709 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6710 * config/nds32/nds32.md (prologue, epilogue): Use macro
6711 NDS32_V3PUSH_AVAILABLE_P to do checking.
6713 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6716 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6717 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6718 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6719 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6720 addition of most attributes on !orig_type_die or the attribute not
6721 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6723 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6724 Chung-Ju Wu <jasonwucj@gmail.com>
6726 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6728 * config/nds32/nds32.opt (mvh): New option.
6730 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6731 Chung-Ju Wu <jasonwucj@gmail.com>
6733 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6735 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6736 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6739 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6740 Chung-Ju Wu <jasonwucj@gmail.com>
6742 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6744 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6745 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6747 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6748 Kito Cheng <kito.cheng@gmail.com>
6749 Chung-Ju Wu <jasonwucj@gmail.com>
6751 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6752 UNSPEC_FFMISM and UNSPEC_FLMISM.
6753 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6754 for ffb, ffmism and flmism.
6755 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6756 (unspec_ffmism): Ditto.
6757 (unspec_flmism): Ditto.
6758 (nds32_expand_builtin_impl): Check if string extension is available.
6759 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6760 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6762 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6765 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6768 * lra-assigns.c (assign_by_spills): Return a flag of reload
6769 assignment failure. Do not process the reload assignment
6770 failures. Do not spill other reload pseudos if they has the same
6772 (lra_assign): Add a return arg. Set up from the result of
6773 assign_by_spills call.
6774 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6775 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6776 usage_insns if it is not NULL.
6777 (spill_hard_reg_in_range): New function.
6778 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6779 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6780 function prototypes.
6781 (lra_assign): Change prototype.
6782 * lra.c (lra): Add code to deal with fails by splitting hard reg
6785 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6788 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6790 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6793 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6794 (dw_val_node): Add val_symbolic_view.
6795 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6796 (symview_upper_bound): New.
6797 (new_line_info_table): Initialize symviews_since_reset.
6798 (dwarf2out_source_line): Count symviews_since_reset and set
6799 symview_upper_bound.
6800 (dw_val_equal_p): Handle symview.
6801 (add_AT_symview): New.
6802 (print_dw_val): Handle symview.
6803 (attr_checksum, attr_checksum_ordered): Likewise.
6804 (same_dw_val_p, size_of_die): Likewise.
6805 (value_format, output_die): Likewise.
6806 (add_high_low_attributes): Use add_AT_symview for entry_view.
6807 (dwarf2out_finish): Reset symview_upper_bound, clear
6810 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6813 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6814 Add strict argument and use it.
6815 (rs6000_split_multireg_move): Update for new strict argument.
6816 (mem_operand_gpr): Disallow all non-offsettable addresses.
6817 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6819 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6822 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6823 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6824 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6827 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6828 decl, use remap_type if we want to use the type.
6830 2018-03-09 Martin Sebor <msebor@redhat.com>
6832 PR tree-optimization/84526
6833 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6835 (builtin_access::generic_overlap): Be prepared to handle non-array
6838 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6840 PR rtl-optimization/84682
6841 * lra-constraints.c (process_address_1): Check is_address flag
6842 for address constraints.
6843 (process_alt_operands): Likewise.
6844 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6845 preprocess_constraints.
6846 * recog.h (preprocess_constraints): Add oploc parameter.
6848 * recog.c (preprocess_constraints): Test address_operand for
6849 CT_ADDRESS constraints.
6851 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6854 * lra-assigns.c (assign_by_spills): Return a flag of reload
6855 assignment failure. Do not process the reload assignment
6856 failures. Do not spill other reload pseudos if they has the same
6858 (lra_assign): Add a return arg. Set up from the result of
6859 assign_by_spills call.
6860 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6861 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6862 usage_insns if it is not NULL.
6863 (spill_hard_reg_in_range): New function.
6864 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6865 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6866 function prototypes.
6867 (lra_assign): Change prototype.
6868 * lra.c (lra): Add code to deal with fails by splitting hard reg
6871 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6874 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6875 Accept complain bool parameter. Only emit errors if it is true.
6876 (arm_parse_cpu_option_name): Likewise.
6877 (arm_target_thumb_only): Adjust callers of the above.
6878 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6879 prototype to take a default true bool parameter.
6880 (arm_parse_arch_option_name): Likewise.
6882 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6883 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6887 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6888 * configure: Regenerate.
6889 * configure.ac ("linker --version-script option"): New.
6890 ("linker soname option"): New.
6892 2018-03-09 Richard Biener <rguenther@suse.de>
6894 PR tree-optimization/84775
6895 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6896 immediate uses of predicate stmts and mark them modified.
6899 PR tree-optimization/84178
6900 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6902 (version_loop_for_if_conversion): Delay update_ssa call.
6903 (tree_if_conversion): Delay update_ssa until after predicate
6906 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6909 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6910 when the function accesses prior frames.
6912 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6915 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6916 gen_llsym, otherwise call maybe_gen_llsym.
6919 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6920 has ',' character inside of it.
6922 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6925 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6926 as clobbering CC_REGNUM.
6928 2018-03-08 Richard Biener <rguenther@suse.de>
6931 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6932 (follow_copies_to_constant): Do not follow SSA names registered
6935 2018-03-08 Richard Biener <rguenther@suse.de>
6937 PR tree-optimization/84178
6938 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6940 (version_loop_for_if_conversion): Delay update_ssa call.
6941 (tree_if_conversion): Delay update_ssa until after predicate
6944 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6946 PR tree-optimization/84178
6947 * tree-if-conv.c (release_bb_predicate): Remove the
6948 the assertion that the stmts have NULL use_ops.
6949 Discard the statements, asserting that they haven't
6950 yet been added to a BB.
6952 2018-03-08 Richard Biener <rguenther@suse.de>
6954 PR tree-optimization/84746
6955 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6956 (phi_translate): Pass in destination ANTIC_OUT set.
6957 (phi_translate_1): Likewise. For a simplified result lookup
6958 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6959 (phi_translate_set): Adjust.
6960 (do_pre_regular_insertion): Likewise.
6961 (do_pre_partial_partial_insertion): Likewise.
6963 2018-03-08 Martin Liska <mliska@suse.cz>
6965 PR gcov-profile/84735
6966 * doc/gcov.texi: Document usage of profile files.
6967 * gcov-io.h: Document changes in the format.
6969 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6973 * dwarf2out.c (struct dw_line_info_table): Update comments for
6975 (FORCE_RESET_NEXT_VIEW): New.
6976 (FORCE_RESETTING_VIEW_P): New.
6977 (RESETTING_VIEW_P): Check for -1 too.
6978 (ZERO_VIEW_P): Likewise.
6979 (new_line_info_table): Force-reset next view.
6980 (dwarf2out_begin_function): Likewise.
6981 (dwarf2out_source_line): Simplify zero_view_p initialization.
6982 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6983 view directly. Omit view when omitting .loc at line 0.
6985 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6987 PR tree-optimization/84740
6988 * tree-switch-conversion.c (process_switch): Call build_constructors
6989 only if info.phi_count is non-zero.
6991 PR tree-optimization/84739
6992 * tree-tailcall.c (find_tail_calls): Check call arguments against
6993 DECL_ARGUMENTS (current_function_decl) rather than
6994 DECL_ARGUMENTS (func) when checking for tail recursion.
6996 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6998 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6999 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
7000 Volker Reichelt's entry and add entries for people that perform
7001 GCC fuzzy testing and report numerous bugs.
7003 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
7006 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
7007 readonly data in sdata, if that is disabled.
7008 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
7009 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
7010 -mreadonly-in-sdata option.
7012 2018-03-07 Martin Sebor <msebor@redhat.com>
7014 PR tree-optimization/84468
7015 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
7016 basic block when looking for nul assignment.
7018 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
7021 * except.h (output_function_exception_table): Adjust prototype.
7022 * except.c (output_function_exception_table): Remove FNNAME parameter
7023 and add SECTION parameter. Ouput one part of the table at a time.
7024 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
7025 the first part of the exception table and emit unwind directives.
7026 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
7027 (i386_pe_seh_cold_init): Likewise.
7028 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
7029 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
7030 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
7031 (ix86_output_call_insn): Emit a nop in one more case for SEH.
7032 * config/i386/winnt.c: Include except.h.
7033 (struct seh_frame_state): Add reg_offset, after_prologue and
7034 in_cold_section fields.
7035 (i386_pe_seh_end_prologue): Set seh->after_prologue.
7036 (i386_pe_seh_cold_init): New function.
7037 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7038 to seh->in_cold_section.
7039 (seh_emit_push): Record the offset of the push.
7040 (seh_emit_save): Record the offet of the save.
7041 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7042 Test seh->after_prologue to disregard the epilogue.
7043 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7044 (i386_pe_end_cold_function): New function.
7046 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7049 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7050 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7053 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7055 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7056 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7059 * multiple_target.c: Include tree-inline.h and intl.h.
7060 (expand_target_clones): Diagnose and fail if node->definition and
7061 !tree_versionable_function_p (node->decl).
7063 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7065 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7067 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7068 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7069 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7071 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7074 * combine.c (try_combine): Use reg_or_subregno instead of handling
7075 just paradoxical SUBREGs and REGs.
7077 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7079 * config/arc/arc.c (arc_finalize_pic): Remove function.
7080 (arc_must_save_register): We use single base PIC register, remove
7081 checks to save/restore the PIC register.
7082 (arc_expand_prologue): Likewise.
7083 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7085 (arc_verify_short): Likewise.
7086 (arc_attr_type): Likewise.
7087 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7088 (walk_stores): Likewise.
7089 (arc_address_cost): Make it static.
7090 (arc_verify_short): Likewise.
7091 (branch_dest): Likewise.
7092 (arc_attr_type): Likewise.
7093 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7094 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7095 (arc_final_prescan_insn): Remove inserting the nops due to
7096 hardware hazards. It is done in reorg step.
7097 (insn_length_variant_t): Remove.
7098 (insn_length_parameters_t): Likewise.
7099 (arc_insn_length_parameters): Likewise.
7100 (arc_get_insn_variants): Likewise.
7101 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7103 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7106 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7109 PR tree-optimization/84687
7110 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7112 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7114 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7116 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7117 Rename to ppc_speculation_barrier.
7118 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7119 __builtin_ppc_speculation_barrier.
7121 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7124 * combine.c (combine_simplify_rtx): Don't try to simplify if
7125 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7128 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7130 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7131 to 32 bytes when compiling for POWER9.
7133 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7136 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7137 regparm >= 3 with no arg reg available also for calls with
7138 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7141 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7143 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7145 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7148 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7150 2018-03-05 Richard Biener <rguenther@suse.de>
7152 PR tree-optimization/84486
7153 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7154 When inserting a __builtin_assume_aligned call set the LHS
7155 SSA name alignment info accordingly.
7157 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7159 PR tree-optimization/84114
7160 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7161 Avoid reassociation of FLOAT_MODE addition.
7163 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7165 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7166 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7167 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7168 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7169 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7170 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7171 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7173 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7174 __builtin_ia32_wbinvd): New builtins.
7175 (SPECIAL_ARGS2): New.
7176 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7177 (SPECIAL_ARGS2): New.
7178 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7179 (ix86_valid_target_attribute_inner_p): Ditto.
7180 (ix86_init_mmx_sse_builtins): Add special_args2.
7181 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7182 TARGET_WBNOINVD_P): New.
7183 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7184 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7185 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7186 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7187 * config/i386/pconfigintrin.h: New file.
7188 * config/i386/wbnoinvdintrin.h: Ditto.
7189 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7190 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7192 2018-03-05 Richard Biener <rguenther@suse.de>
7194 PR tree-optimization/84670
7195 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7197 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7198 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7199 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7200 doesn't grow if all successors (recursively) were visited at least
7203 2018-03-05 Richard Biener <rguenther@suse.de>
7205 PR tree-optimization/84650
7206 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7207 if executed in the loop pipeline.
7209 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7211 * doc/configfiles.texi (Configuration Files): Move info about
7212 conditionalizing $target-protos.h to...
7213 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7214 differs from $target-protos.h.
7216 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7217 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7220 * config/nds32/nds32-multiple.md (setmemsi): Define.
7221 * config/nds32/nds32-memory-manipulation.c
7222 (nds32_gen_dup_4_byte_to_word_value): New.
7223 (emit_setmem_word_loop): New.
7224 (emit_setmem_byte_loop): New.
7225 (nds32_expand_setmem_loop): New.
7226 (nds32_expand_setmem_loop_v3m): New.
7227 (nds32_expand_setmem_unroll): New.
7228 (nds32_expand_setmem): New.
7230 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7231 Chung-Ju Wu <jasonwucj@gmail.com>
7233 * config/nds32/nds32-memory-manipulation.c
7234 (nds32_emit_load_store): New.
7235 (nds32_emit_post_inc_load_store): New.
7236 (nds32_emit_mem_move): New.
7237 (nds32_emit_mem_move_block): New.
7238 (nds32_expand_movmemsi_loop_unknown_size): New.
7239 (nds32_expand_movmemsi_loop_known_size): New.
7240 (nds32_expand_movmemsi_loop): New.
7241 (nds32_expand_movmemsi_unroll): New.
7242 (nds32_expand_movmemqi): Rename ...
7243 (nds32_expand_movmemsi): ... to this.
7244 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7245 (movmemsi): ... to this.
7246 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7247 (nds32_expand_movmemsi): ... to this.
7249 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7250 Monk Chiang <sh.chiang04@gmail.com>
7251 Chung-Ju Wu <jasonwucj@gmail.com>
7253 * config/nds32/nds32-protos.h
7254 (nds32_expand_load_multiple): New arguments.
7255 (nds32_expand_store_multiple): Ditto.
7256 (nds32_valid_multiple_load_store): Rename ...
7257 (nds32_valid_multiple_load_store_p): ... to this.
7258 * config/nds32/nds32-memory-manipulation.c
7259 (nds32_expand_load_multiple): Refine implementation.
7260 (nds32_expand_store_multiple): Ditto.
7261 * config/nds32/nds32-multiple.md
7262 (load_multiple): Update nds32_expand_load_multiple interface.
7263 (store_multiple): Update nds32_expand_store_multiple interface.
7264 * config/nds32/nds32-predicates.c
7265 (nds32_valid_multiple_load_store): Rename ...
7266 (nds32_valid_multiple_load_store_p): ... to this and refine
7268 * config/nds32/predicates.md
7269 (nds32_load_multiple_and_update_address_operation): New predicate.
7270 (nds32_store_multiple_and_update_address_operation): New predicate.
7272 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7273 Chung-Ju Wu <jasonwucj@gmail.com>
7275 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7276 (combo): New attribute.
7277 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7279 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7281 * config/nds32/nds32.opt: Change -mcmodel= default value.
7283 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7284 Monk Chiang <sh.chiang04@gmail.com>
7285 Chung-Ju Wu <jasonwucj@gmail.com>
7287 * config/nds32/constants.md (unspec_element): New enum.
7288 * config/nds32/constraints.md (Umw): New constraint.
7289 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7290 * config/nds32/nds32-intrinsic.md: Likewise.
7291 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7292 (nds32_valid_smw_lwm_base_p): New.
7293 (nds32_output_smw_single_word): New.
7294 (nds32_output_lmw_single_word): New.
7295 (nds32_expand_unaligned_load): New.
7296 (nds32_expand_unaligned_store): New.
7297 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7298 (nds32_output_smw_single_word): Declare.
7299 (nds32_output_lmw_single_word): Declare.
7300 (nds32_expand_unaligned_load): Declare.
7301 (nds32_expand_unaligned_store): Declare.
7302 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7303 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7304 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7305 NDS32_BUILTIN_UASTORE_DW.
7306 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7309 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7310 Kito Cheng <kito.cheng@gmail.com>
7311 Chung-Ju Wu <jasonwucj@gmail.com>
7313 * config/nds32/nds32-intrinsic.c
7314 (nds32_expand_builtin_null_ftype_reg): Delete.
7315 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7316 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7317 (nds32_read_argument): New.
7318 (nds32_legitimize_target): Ditto.
7319 (nds32_legitimize_argument): Ditto.
7320 (nds32_check_constant_argument): Ditto.
7321 (nds32_expand_unop_builtin): Ditto.
7322 (nds32_expand_unopimm_builtin): Ditto.
7323 (nds32_expand_binop_builtin): Ditto.
7324 (nds32_builtin_decl_impl): Ditto.
7325 (builtin_description): Ditto.
7326 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7327 (nds32_init_builtins_impl): Ditto.
7328 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7329 (nds32_builtin_decl): New.
7330 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7331 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7333 2018-03-02 Jeff Law <law@redhat.com>
7335 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7336 (redundant_insn, fill_simple_delay_slots): Likewise.
7337 (fill_slots_from_thread): Likewise.
7338 * resource.c (mark_referenced_resources): Likewise.
7339 (mark_set_resources, find_dead_or_set_registers): Likewise.
7341 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7343 * substring-locations.h (format_warning_va): Formatting fix for
7345 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7347 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7348 * substring-locations.c: Include intl.h.
7349 (format_warning_va): Turned into small wrapper around
7350 format_warning_n_va, renamed to ...
7351 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7352 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7354 (format_warning_at_substring_n): New function.
7355 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7356 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7357 format_warning_at_substring with just a shorter name instead of
7358 const function pointer.
7359 (fmtwarn_n): New function.
7360 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7361 appropriate, get rid of all the fmtstr temporaries, move conditionals
7362 with G_() wrapped string literals directly into fmtwarn arguments,
7363 cast dir.len to (int), formatting fixes.
7365 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7367 * doc/invoke.texi: Remove "Cilk Plus" references.
7369 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7370 Richard Biener <rguenther@suse.de>
7373 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7374 for error or warning attributes if CALL_FROM_THUNK_P is set.
7377 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7380 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7381 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7384 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7385 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7387 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7388 instead of -1U in last predictors element's probability member.
7390 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7393 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7394 arguments if they are comparable.
7396 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7398 PR tree-optimization/84634
7399 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7400 masks and masked_loop_p with a single loop_masks, making sure it's
7401 null for bb vectorization.
7403 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7405 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7406 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7407 loop->force_vectorize to check whether there is no alias.
7409 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7412 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7414 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7416 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7417 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7418 instead of a loop around prev_real_insn.
7419 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7423 * config/i386/i386.c (ix86_print_operand): Use conditional
7424 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7427 2018-03-02 Richard Biener <rguenther@suse.de>
7429 PR tree-optimization/84427
7430 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7431 (bitmap_set_subtract_values): Rewrite to handle multiple
7434 (prune_clobbered_mems): Likewise.
7435 (phi_translate): Take edge instead of pred/phiblock.
7436 (phi_translate_1): Likewise.
7437 (phi_translate_set): Likewise. Insert all translated
7438 exprs for a value into the set, keeping possibly multiple
7439 expressions per value.
7440 (compute_antic_aux): Adjust for phi_translate changes.
7441 When intersecting union the expressions and prune those
7442 not in the final value set, keeping possibly multiple
7443 expressions per value. Do not use value-insertion
7444 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7445 all expressions. Add verification that the value-sets
7446 only shrink during iteration.
7447 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7448 (do_pre_regular_insertion): Likewise.
7449 (do_pre_partial_partial_insertion): Likewise.
7451 2018-03-02 Richard Biener <rguenther@suse.de>
7454 * config/darwin.c (saved_debug_info_level): New static global.
7455 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7456 (darwin_asm_lto_end): Restore debug info generation settings.
7458 2018-03-01 Martin Liska <mliska@suse.cz>
7461 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7464 2018-03-01 Richard Biener <rguenther@suse.de>
7467 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7468 type annotation with LTO when debug was disabled at compile-time.
7470 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7472 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7474 (mips_final_postscan_insn): Likewise.
7476 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7478 PR rtl-optimization/84528
7479 * alias.c (init_alias_target): Add commentary.
7480 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7481 a unique base value if the frame pointer is not eliminated
7482 to the stack pointer.
7484 2018-03-01 Tom de Vries <tom@codesourcery.com>
7486 PR rtl-optimization/83327
7487 * lra-int.h (hard_regs_spilled_into): Declare.
7488 * lra.c (hard_regs_spilled_into): Define.
7489 (init_reg_info): Init hard_regs_spilled_into.
7490 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7491 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7492 (process_bb_lives): Handle hard_regs_spilled_into.
7493 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7494 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7496 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7498 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7499 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7500 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7501 * config/rs6000/aix72.h: New file.
7503 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7505 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7506 instead of warning_at with conditional singular and plural messages
7510 * stor-layout.c (update_alignment_for_field): For
7511 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7512 && !DECL_PACKED (field), do the alignment update, just use
7513 only desired_align instead of MAX (type_align, desired_align)
7515 (place_field): Don't do known_align < desired_align handling
7516 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7517 is non-NULL, instead do it after rli->prev_field handling and
7518 only if not within a bitfield word. For DECL_PACKED (field)
7519 use type_align of BITS_PER_UNIT.
7521 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7523 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7524 superfluous parentheses and trailing spaces.
7526 2018-02-28 Richard Biener <rguenther@suse.de>
7528 PR tree-optimization/84584
7529 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7530 SCoPs with fake exit edge.
7532 2018-02-28 Martin Liska <mliska@suse.cz>
7535 * timevar.c (timer::print): Fix format to properly print 100%
7538 2018-02-28 Richard Biener <rguenther@suse.de>
7541 * genmatch.c (capture_info::walk_match): Do not mark
7542 captured expressions without operands as expr_p given
7543 they act more like predicates and should be subject to
7544 "lost tail" side-effect preserving.
7546 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7548 PR rtl-optimization/81611
7549 * auto-inc-dec.c (attempt_change): Move dead note from
7550 mem_insn if it's the next use of regno
7551 (find_address): Take address use of reg holding
7552 non-incremented value. Add parm to limit search to the named
7554 (merge_in_block): Attempt to use a mem insn that is the next
7555 use of the original regno.
7557 2018-02-27 Martin Sebor <msebor@redhat.com>
7560 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7561 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7563 2018-02-27 Martin Sebor <msebor@redhat.com>
7565 PR translation/84207
7566 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7567 n argument to unsigned HOST_WIDE_INT.
7568 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7569 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7570 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7571 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7573 2018-02-27 Richard Biener <rguenther@suse.de>
7575 PR tree-optimization/84512
7576 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7577 Do not use the estimate returned from record_stmt_cost for
7578 the scalar iteration cost but sum properly using add_stmt_cost.
7580 2018-02-27 Richard Biener <rguenther@suse.de>
7582 PR tree-optimization/84466
7583 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7584 Adjust last change to less strictly validate use operands.
7586 2018-02-27 Martin Liska <mliska@suse.cz>
7588 PR gcov-profile/84548
7589 * gcov.c (process_file): Allow partial overlap and consider it
7590 also as group functions.
7591 (output_lines): Properly calculate range of lines for a group.
7593 2018-02-27 Martin Liska <mliska@suse.cz>
7595 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7596 'ggc' suffixes. Change first column width.
7597 (timer::print): Fix formatting of the column.
7599 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7601 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7602 preserve inline entry blocks for the sake of debug inline
7603 entry point markers alone.
7604 (remove_unused_locals): Suggest in comments a better place to
7605 force the preservation of inline entry blocks that are
7606 otherwise unused, but do not preserve them.
7608 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7610 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7612 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7615 * config/i386/constraints.md (Bs): Replace
7616 ix86_indirect_branch_register with
7617 TARGET_INDIRECT_BRANCH_REGISTER.
7619 * config/i386/i386.md (indirect_jump): Likewise.
7620 (tablejump): Likewise.
7621 (*sibcall_memory): Likewise.
7622 (*sibcall_value_memory): Likewise.
7623 Peepholes of indirect call and jump via memory: Likewise.
7624 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7625 (*sibcall_value_GOT_32): Likewise.
7626 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7627 (GOT_memory_operand): Likewise.
7628 (call_insn_operand): Likewise.
7629 (sibcall_insn_operand): Likewise.
7630 (GOT32_symbol_operand): Likewise.
7631 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7633 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7635 PR rtl-optimization/83496
7636 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7637 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7638 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7639 redundant insn, if any.
7640 (relax_delay_slots): Likewise.
7641 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7643 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7645 PR tree-optimization/83965
7646 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7647 that grouped statements are part of a reduction chain. Return
7648 true if the statement is not marked as a reduction itself but
7650 (vect_recog_dot_prod_pattern): Don't check whether the statement
7651 is part of a group here.
7652 (vect_recog_sad_pattern): Likewise.
7653 (vect_recog_widen_sum_pattern): Likewise.
7655 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7658 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7660 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7663 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7665 (ix86_output_indirect_function_return): New prototype.
7666 (ix86_split_simple_return_pop_internal): Likewise.
7667 * config/i386/i386.c (indirect_return_via_cx): New.
7668 (indirect_return_via_cx_bnd): Likewise.
7669 (indirect_thunk_name): Handle return va CX_REG.
7670 (output_indirect_thunk_function): Create alias for
7671 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7672 (ix86_output_indirect_jmp): Remove the bool argument.
7673 (ix86_output_indirect_function_return): New function.
7674 (ix86_split_simple_return_pop_internal): Likewise.
7675 * config/i386/i386.md (*indirect_jump): Don't pass false
7676 to ix86_output_indirect_jmp.
7677 (*tablejump_1): Likewise.
7678 (simple_return_pop_internal): Change it to define_insn_and_split.
7679 Call ix86_split_simple_return_pop_internal to split it for
7681 (simple_return_indirect_internal): Call
7682 ix86_output_indirect_function_return instead of
7683 ix86_output_indirect_jmp.
7685 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7688 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7689 memset and value initialization afterwards.
7691 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7693 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7695 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7698 * common/config/aarch64/aarch64-common.c
7699 (aarch_option_optimization_table[]): Switch
7700 off fomit-frame-pointer
7702 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7703 Chung-Ju Wu <jasonwucj@gmail.com>
7705 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7707 (store_multiple): Ditto.
7709 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7711 * config.gcc: Add --with-cpu support for nds32 target.
7712 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7713 * config/nds32/nds32.opt: Add -mcpu= option.
7715 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7717 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7718 isel=yes): Warn for these deprecated options.
7720 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7722 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7723 ISA_2_5_MASKS_EMBEDDED.
7725 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7727 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7728 p->max as pointers rather than using iterative_hash_expr.
7730 2018-02-23 Carl Love <cel@us.ibm.com>
7732 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7733 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7735 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7736 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7737 P8V_BUILTIN_VEC_VUNSIGNED2.
7739 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7742 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7743 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7745 * lra-constraints.c (curr_insn_transform): Set up
7746 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7748 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7749 * lra-eliminations.c (spill_pseudos): Ditto.
7750 (process_insn_for_elimination): Ditto.
7751 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7752 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7753 LRA_NON_CLOBBERED_ALT.
7755 2018-02-22 Martin Sebor <msebor@redhat.com>
7757 PR tree-optimization/84480
7758 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7759 to maybe_diag_stxncpy_trunc. Call it.
7760 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7761 from gimple_fold_builtin_strcpy. Print inlining stack.
7762 (handle_builtin_stxncpy): Print inlining stack.
7763 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7765 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7768 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7769 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7770 and -fcheck-pointer-bounds are used together.
7771 (indirect_thunk_prefix): New enum.
7772 (indirect_thunk_need_prefix): New function.
7773 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7774 "_nt" instead of "_bnd" for NOTRACK prefix.
7775 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7776 (output_indirect_thunk_function): Likewise.
7778 (ix86_code_end): Update output_indirect_thunk_function calls.
7779 (ix86_output_indirect_branch_via_reg): Replace
7780 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7781 (ix86_output_indirect_branch_via_push): Likewise.
7782 (ix86_output_function_return): Likewise.
7783 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7784 incompatible with -fcf-protection=branch and
7785 -fcheck-pointer-bounds.
7787 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7790 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7791 Change gcc_assert call to output_operand_lossage.
7793 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7795 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7797 2018-02-22 DJ Delorie <dj@redhat.com>
7798 Sebastian Perta <sebastian.perta@renesas.com>
7799 Oleg Endo <olegendo@gcc.gnu.org>
7801 * config/rx/rx.c (rx_rtx_costs): New function.
7802 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7804 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7806 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7808 2018-02-22 Martin Liska <mliska@suse.cz>
7811 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7812 Add "native" as a possible value.
7814 2018-02-22 Martin Liska <mliska@suse.cz>
7817 * config/i386/i386.c (ix86_option_override_internal):
7818 Add "native" as a possible value for -march and -mtune.
7820 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7823 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7824 to all type variants.
7826 PR tree-optimization/84503
7827 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7828 width as info->bitpos + info->bitsize - start.
7829 (merged_store_group::merge_overlapping): Simplify width computation.
7830 (check_no_overlap): New function.
7831 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7832 start + width and last_order of the group, fail if check_no_overlap
7834 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7835 to group if check_no_overlap fails.
7837 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7839 * config/rs6000/altivec.md: Delete contraint arguments to
7840 define_expand, define_split, and define_peephole2, and in
7841 define_insn_and_split if always unused.
7842 * config/rs6000/darwin.md: Ditto.
7843 * config/rs6000/dfp.md: Ditto.
7844 * config/rs6000/rs6000.md: Ditto.
7845 * config/rs6000/sync.md: Ditto.
7846 * config/rs6000/vector.md: Ditto.
7847 * config/rs6000/vsx.md: Ditto.
7849 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7851 * config/rs6000/altivec.md: Write output control strings as braced
7852 blocks instead of double-quoted strings.
7853 * config/rs6000/darwin.md: Ditto.
7854 * config/rs6000/rs6000.md: Ditto.
7855 * config/rs6000/vector.md: Ditto.
7856 * config/rs6000/vsx.md: Ditto.
7858 2018-02-21 Jason Merrill <jason@redhat.com>
7860 PR c++/84314 - ICE with templates and fastcall attribute.
7861 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7863 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7865 * ipa-cp.c (determine_versionability): Fix comment typos.
7867 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7870 * ipa-cp.c (determine_versionability): Do not version functions caling
7873 2018-02-21 Martin Liska <mliska@suse.cz>
7876 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7877 Add "native" as a possible value.
7878 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7879 the macro when native cpu detection is available.
7881 2018-02-21 Martin Liska <mliska@suse.cz>
7884 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7885 Add "native" as a possible value.
7886 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7887 when native cpu detection is available.
7889 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7890 Martin Sebor <msebor@redhat.com>
7892 PR tree-optimization/84478
7893 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7895 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7896 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7897 support which is conservatively correct, for 2 only stay conservative
7898 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7899 argument to the 2 argument get_range_strlen, adjust 6 arg
7900 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7902 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7903 (gimple_fold_builtin_strlen): Pass true as last argument to
7906 2018-02-20 Martin Sebor <msebor@redhat.com>
7909 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7910 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7911 (builtin_memref::builtin_memref): Factor out parts into
7912 set_base_and_offset and call it.
7914 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7917 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7918 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7919 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7920 search at the associated MODE_INT.
7922 2018-02-20 Jeff Law <law@redhat.com>
7925 PR tree-optimization/81592
7927 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7928 for range data rather than using global data.
7929 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7930 range data rather than using global data.
7931 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7932 pass it to children as needed.
7933 (struct directive::fmtresult): Similarly.
7934 (struct directive::set_width): Similarly.
7935 (struct directive::set_precision): Similarly.
7936 (format_integer, format_directive, parse_directive): Similarly.
7937 (format_none): Accept unnamed vr_values parameter.
7938 (format_percent, format_floating, format_character): Similarly.
7939 (format_string, format_plain): Similarly.
7940 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7941 the EVRP range analyzer for range data rather than using global data.
7942 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7943 gimple-ssa-evrp-analyze.h
7944 (class sprintf_dom_walker): Add after_dom_children member function.
7945 Add evrp_range_analyzer member.
7946 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7947 range analyzer as needed.
7948 (sprintf_dom_walker::after_dom_children): New member function.
7949 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7951 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7952 (evrp_range_analyzer::pop_to_marker): Likewise.
7954 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7956 PR tree-optimization/84419
7957 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7958 with the required type if its current type is compatible but
7961 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7964 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7965 after vectorization.
7967 2018-02-20 Martin Liska <mliska@suse.cz>
7970 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7971 possible values if we don't have a hint.
7973 2018-02-20 Martin Liska <mliska@suse.cz>
7977 * final.c (shorten_branches): Build align_tab array with one
7979 * opts.c (finish_options): Add alignment option limit check.
7980 (MAX_CODE_ALIGN): Likewise.
7981 (MAX_CODE_ALIGN_VALUE): Likewise.
7982 * doc/invoke.texi: Document maximum allowed option value for
7983 all -falign-* options.
7985 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7988 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7989 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7990 * var-tracking.c (emit_note_insn_var_location): Remove all references
7991 to NOTE_INSN_CALL_ARG_LOCATION.
7992 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7993 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7994 Use copy_rtx_if_shared.
7995 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7996 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7997 (dwarf2out_var_location): Remove handling of
7998 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
8000 * final.c (final_scan_insn): Remove all references to
8001 NOTE_INSN_CALL_ARG_LOCATION.
8002 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
8003 before dumping final insns.
8004 * except.c (emit_note_eh_region_end): Remove all references to
8005 NOTE_INSN_CALL_ARG_LOCATION.
8006 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
8007 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
8008 * config/arc/arc.c (hwloop_optimize): Likewise.
8009 * config/arm/arm.c (create_fix_barrier): Likewise.
8010 * config/s390/s390.c (s390_chunkify_start): Likewise.
8011 * config/sh/sh.c (find_barrier): Likewise.
8012 * config/i386/i386.c (rest_of_insert_endbranch,
8013 ix86_seh_fixup_eh_fallthru): Likewise.
8014 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
8015 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
8016 * config/frv/frv.c (frv_function_prologue): Likewise.
8017 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
8019 (note_outside_basic_block_p): Remove all references to
8020 NOTE_INSN_CALL_ARG_LOCATION.
8021 * gengtype.c (adjust_field_rtx_def): Likewise.
8022 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
8024 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
8025 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
8028 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
8031 PR tree-optimization/84452
8032 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
8033 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
8036 2018-02-19 Martin Liska <mliska@suse.cz>
8039 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8041 2018-02-19 Martin Liska <mliska@suse.cz>
8042 Richard Sandiford <richard.sandiford@linaro.org>
8044 PR tree-optimization/82491
8045 * gimple-fold.c (get_base_constructor): Make earlier bail out
8048 2018-02-19 Carl Love <cel@us.ibm.com>
8050 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8051 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8053 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8054 P8V_BUILTIN_VEC_NEG.
8056 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8058 * config/rl78/rl78.md (movdf): New define expand.
8060 2018-02-19 Martin Liska <mliska@suse.cz>
8063 * doc/invoke.texi: Fix typo.
8064 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8066 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8068 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8069 handle rs6000_single_float and rs6000_double_float specially for
8072 2018-02-16 Jeff Law <law@redhat.com>
8074 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8077 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8080 * ipa-inline.c (inline_small_functions): Fix a typo.
8082 2018-02-16 Nathan Sidwell <nathan@acm.org>
8084 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8086 2018-02-16 Carl Love <cel@us.ibm.com>
8088 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8089 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8090 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8091 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8092 expansion to P8V_BUILTIN_VEC_FLOAT2.
8094 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8096 PR rtl-optimization/70023
8097 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8098 src_regno into account.
8100 2018-02-16 Carl Love <cel@us.ibm.com>
8102 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8103 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8104 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8105 * config/rs6000/rs6000.c: Remove case statements for
8106 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8107 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8108 and P9V_BUILTIN_VEC_VINSERT4B.
8109 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8110 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8111 * config/rs6000/vsx.md:
8112 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8115 2018-02-16 Carl Love <cel@us.ibm.com>
8117 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8119 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8121 * config/rs6000/rs6000-c.c: Add the definitions for
8122 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8123 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8124 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8125 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8126 definition for insert4b and define insn *insert3b_internal.
8127 * doc/extend.texi: Add documentation for vec_extract4b.
8129 2018-02-16 Nathan Sidwell <nathan@acm.org>
8131 * doc/extend.texi (Backwards Compatibility): Mention friend
8132 injection. Note for-scope is deprecated.
8133 * doc/invoke.texi (-ffriend-injection): Deprecate.
8135 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8137 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8138 that moved to I2, also allow destinations that are a paradoxical
8139 subreg (instead of a normal reg).
8141 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8144 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8147 2018-02-16 Richard Biener <rguenther@suse.de>
8149 PR tree-optimization/84037
8150 PR tree-optimization/84016
8152 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8153 Adjust vec_construct for the fact we need additional higher latency
8154 128bit inserts for AVX256 and AVX512 vector builds.
8155 (ix86_add_stmt_cost): Scale vector construction cost for
8158 2018-02-16 Richard Biener <rguenther@suse.de>
8160 PR tree-optimization/84417
8161 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8162 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8163 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8165 2018-02-16 Martin Liska <mliska@suse.cz>
8168 * internal-fn.def (ASAN_CHECK): Set proper flags.
8169 (ASAN_MARK): Likewise.
8171 2018-02-16 Julia Koval <julia.koval@intel.com>
8173 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8174 from PTA_CANNONLAKE.
8176 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8179 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8180 Use ++iter rather than iter++ for std::list iterators.
8181 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8182 defer deleting them until all nodes in the forest are processed. Do
8183 free even leaf nodes. Change to_process into auto_vec.
8186 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8187 * vec.h (vec_default_construct): Use memset instead of placement new
8188 if BROKEN_VALUE_INITIALIZATION is defined.
8189 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8190 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8193 PR rtl-optimization/83723
8194 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8195 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8196 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8198 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8200 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8202 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8204 PR rtl-optimization/81443
8205 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8206 from inner REGs to paradoxical SUBREGs.
8208 2018-02-16 Richard Biener <rguenther@suse.de>
8210 PR tree-optimization/84399
8211 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8212 For operands we can analyze at their definition make sure we can
8213 analyze them at each use as well.
8215 2018-02-16 Richard Biener <rguenther@suse.de>
8217 PR tree-optimization/84190
8218 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8219 volatile accesses if the decl isn't volatile.
8221 2018-02-15 Jason Merrill <jason@redhat.com>
8223 PR c++/84314 - ICE with templates and fastcall attribute.
8224 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8225 TYPE_CANONICAL on an existing type.
8227 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8229 PR tree-optimization/84383
8230 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8231 dstoff nor call operand_equal_p if dstbase is NULL.
8233 PR tree-optimization/84334
8234 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8235 also a CONSTANT_CLASS_P, punt.
8237 2018-02-14 Jim Wilson <jimw@sifive.com>
8239 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8240 first SMALL_OPERAND check. New local min_second_step. Move assert
8241 to where locals are set. Add TARGET_RVC support.
8242 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8244 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8246 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8248 2018-02-14 Martin Sebor <msebor@redhat.com>
8250 PR tree-optimization/83698
8251 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8252 arrays constrain the offset range to their bounds.
8253 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8254 (builtin_access::overlap): Avoid setting the size of overlap if it's
8256 (maybe_diag_overlap): Also consider arrays when deciding what values
8257 of offsets to include in diagnostics.
8259 2018-02-14 Martin Sebor <msebor@redhat.com>
8262 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8263 that correspond to the kind of a declaration.
8265 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8268 * config/pa/pa.md: Load address of PIC label using the linkage table
8269 if the label is nonlocal.
8271 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8273 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8274 warning message if user requests -maltivec=be.
8275 * doc/invoke.texi: Document deprecation of -maltivec=be.
8277 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8280 * config/rs6000/rs6000-c.c: Update definitions for
8281 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8282 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8284 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8287 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8288 add _get_ssp intrinsics. Remove argument from
8289 __builtin_ia32_rdssp[d|q].
8290 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8291 * config/i386/i386-builtin.def: Remove argument from
8292 __builtin_ia32_rdssp[d|q].
8293 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8294 ix86_expand_special_args_builtin for _rdssp[d|q].
8295 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8296 Clear register before usage.
8297 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8298 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8300 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8302 PR tree-optimization/84357
8303 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8304 operand 1 of an ARRAY_REF too.
8306 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8309 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8310 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8312 (set_of_reg): New struct.
8313 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8314 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8315 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8317 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8318 Split into bitclr, bitset, bitinvert patterns if appropriate.
8319 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8320 use rx_fuse_in_memory_bitop.
8321 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8322 to named insn, correct maximum insn length.
8324 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8327 * machmode.def: Define a complex mode for PARTIAL_INT.
8328 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8330 * doc/rtl.texi: Document CSPImode.
8331 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8333 (msp430_hard_regno_nregs_with_padding): Likewise.
8335 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8338 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8340 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8342 PR rtl-optimization/84169
8343 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8344 we generated a parallel as new i3 and we split that to new i2 and i3
8345 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8346 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8347 those to i2, not i1. Partially rewrite this scan code.
8349 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8352 * stor-layout.c (place_field): For variable length fields, adjust
8353 offset_align afterwards not just based on the field's alignment,
8354 but also on the size.
8357 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8358 of exps and logs in the use_exp2 case.
8360 2018-02-13 Jeff Law <law@redhat.com>
8362 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8365 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8368 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8372 * common.opt (gas-loc-support, gas-locview-support): New.
8373 (ginline-points, ginternal-reset-location-views): New.
8374 * doc/invoke.texi: Document them. Use @itemx where intended.
8375 (gvariable-location-views): Adjust.
8376 * target.def (reset_location_view): New.
8377 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8378 (TARGET_RESET_LOCATION_VIEW): New.
8379 * doc/tm.texi: Rebuilt.
8380 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8381 (dwarf2out_default_as_locview_support): New.
8382 (output_asm_line_debug_info): Use option variables.
8383 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8384 (output_loc_list): Likewise.
8385 (add_high_low_attributes): Check option variables.
8386 Don't output entry view attribute in strict mode.
8387 (gen_inlined_subroutine_die): Check option variables.
8388 (dwarf2out_inline_entry): Likewise.
8389 (init_sections_and_labels): Likewise.
8390 (dwarf2out_early_finish): Likewise.
8391 (maybe_reset_location_view): New, from...
8392 (dwarf2out_var_location): ... here. Call it.
8393 * debug.h (dwarf2out_default_as_loc_support): Declare.
8394 (dwarf2out_default_as_locview_support): Declare.
8395 * hooks.c (hook_int_rtx_insn_0): New.
8396 * hooks.h (hook_int_rtx_insn_0): Declare.
8397 * toplev.c (process_options): Take -gas-loc-support and
8398 -gas-locview-support from dwarf2out. Enable
8399 -gvariable-location-views by default only with locview
8400 assembler support. Enable -ginternal-reset-location-views by
8401 default only if the target defines the corresponding hook.
8402 Enable -ginline-points by default if location views are
8403 enabled; force it disabled if statement frontiers are
8405 * tree-inline.c (expand_call_inline): Check option variables.
8406 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8408 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8410 PR tree-optimization/84321
8411 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8412 handling. Also check whether the anti-range contains any values
8413 that satisfy the mask; switch to a VR_RANGE if not.
8415 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8418 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8420 2018-02-13 Martin Jambor <mjambor@suse.cz>
8423 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8424 of call statements, also set location of a load to a temporary.
8426 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8428 * config/rl78/rl78.c (add_vector_labels): New function.
8429 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8430 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8431 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8432 which checks that no arguments are passed.
8433 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8434 * doc/extend.texi: Documentation for the new attribute.
8436 2018-02-13 Andreas Schwab <schwab@suse.de>
8438 * config/riscv/linux.h (CPP_SPEC): Define.
8440 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8443 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8444 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8445 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8446 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8447 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8448 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8449 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8450 temporarily for AES and PCLMUL builtins.
8452 PR tree-optimization/84339
8453 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8454 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8458 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8459 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8460 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8462 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8464 * omp-simd-clone.h: New file.
8465 * omp-simd-clone.c: Include omp-simd-clone.h.
8466 (expand_simd_clones): No longer static.
8467 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8468 cgraph.h and omp-simd-clone.h.
8469 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8470 (vect_recog_widen_shift_pattern): Formatting fix.
8471 (vect_pattern_recog_1): Don't check optab for calls.
8474 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8475 operands[2] into a REG before using gen_lowpart on it.
8477 2018-02-12 Jeff Law <law@redhat.com>
8480 * config/sh/sh.c (find_barrier): Consider a sibling call
8483 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8484 successfully back substituting a reg.
8486 2018-02-12 Richard Biener <rguenther@suse.de>
8488 PR tree-optimization/84037
8489 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8490 parameter, move visited init to caller.
8491 (vect_slp_analyze_operations): Separate cost from validity
8492 check, initialize visited once for all instances.
8493 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8495 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8497 (vect_model_promotion_demotion_cost): Likewise.
8498 (vectorizable_bswap): Guard cost modeling with !slp_node
8499 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8501 (vectorizable_call): Likewise.
8502 (vectorizable_conversion): Likewise.
8503 (vectorizable_assignment): Likewise.
8504 (vectorizable_shift): Likewise.
8505 (vectorizable_operation): Likewise.
8506 (vectorizable_store): Likewise.
8507 (vectorizable_load): Likewise.
8508 (vectorizable_condition): Likewise.
8509 (vectorizable_comparison): Likewise.
8511 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8514 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8515 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8516 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8518 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8521 * params.def (inline-min-speedup): Increase from 8 to 15.
8522 (max-inline-insns-auto): Decrease from 40 to 30.
8523 * ipa-split.c (consider_split): Add some buffer for function to
8524 be considered inlining candidate.
8525 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8528 2018-02-12 Richard Biener <rguenther@suse.de>
8530 PR tree-optimization/84037
8531 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8532 matched stmts if we cannot swap the non-matched ones.
8534 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8536 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8537 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8538 _mm_maskz_scalef_round_ss): New intrinsics.
8539 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8540 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8541 __builtin_ia32_scalefss_round): Remove.
8542 (__builtin_ia32_scalefsd_mask_round,
8543 __builtin_ia32_scalefss_mask_round): New intrinsics.
8544 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8545 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8546 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8547 "<round_constraint>")): Changed to ...
8548 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8549 "<round_scalar_constraint>")): ... this.
8550 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8551 %0, %1, %2<round_op3>}"): Changed to ...
8552 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8553 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8554 %2<round_scalar_mask_op3>}"): ... this.
8555 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8557 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8559 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8560 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8561 (_mm_maskz_sqrt_round_ss): New intrinsics.
8562 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8563 (__builtin_ia32_sqrtsd_mask_round)
8564 (__builtin_ia32_sqrtss_mask_round): New builtins.
8565 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8566 (__builtin_ia32_sqrtss_round): Remove.
8567 (__builtin_ia32_sqrtsd_mask_round)
8568 (__builtin_ia32_sqrtss_mask_round): New builtins.
8569 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8570 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8571 ((match_operand:VF_128 1 "vector_operand"
8572 "xBm,<round_constraint>")): Changed to ...
8573 ((match_operand:VF_128 1 "vector_operand"
8574 "xBm,<round_scalar_constraint>")): ... this.
8575 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8576 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8577 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8578 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8579 %<iptr>1<round_scalar_mask_op3>}): ... this.
8580 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8581 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8583 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8586 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8587 Cast vec_cmpeq result to correct type.
8588 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8589 Cast vec_cmpgt result to correct type.
8591 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8593 * final.c (final_scan_insn_1): Renamed from...
8594 (final_scan_insn): ... this. New wrapper, to recover
8595 seen from the outermost call in recursive ones.
8596 * config/sparc/sparc.c (output_return): Drop seen from call.
8597 (output_sibcall): Likewise.
8598 * config/visium/visium.c (output_branch): Likewise.
8600 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8602 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8605 2018-02-10 Alan Modra <amodra@gmail.com>
8608 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8609 Specify LR as an input.
8611 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8614 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8615 remove_member_access_dummy_vars): New functions.
8616 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8617 lower_omp_1, execute_lower_omp): Use them.
8619 PR rtl-optimization/84308
8620 * shrink-wrap.c (spread_components): Release todo vector.
8622 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8624 PR rtl-optimization/57193
8625 * ira-color.c (struct allocno_color_data): Add member
8626 conflict_allocno_hard_prefs.
8627 (update_conflict_allocno_hard_prefs): New.
8628 (bucket_allocno_compare_func): Add a preference based on
8629 conflict_allocno_hard_prefs.
8630 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8631 (color_allocnos): Remove a dead code. Initiate
8632 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8634 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8637 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8638 constraint from =wa to wa. Avoid a subreg on the output operand,
8639 instead use a pseudo and subreg it in a move.
8640 (p9_xxbrd_<mode>): Changed to ...
8641 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8642 (p9_xxbrd_v2df): New expander.
8643 (p9_xxbrw_<mode>): Changed to ...
8644 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8645 (p9_xxbrw_v4sf): New expander.
8647 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8649 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8651 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8654 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8655 multiply in 32-bit mode.
8656 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8657 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8660 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8662 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8663 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8664 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8665 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8667 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8670 * dwarf2out.c (is_trivial_indirect_ref): New function.
8671 (dwarf2out_late_global_decl): Do not generate a location
8672 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8673 and that are not defined in the current unit.
8675 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8677 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8678 instead of a libcall for UNORDERED.
8680 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8683 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8684 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8686 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8689 * config/s390/s390.c (s390_set_current_function): Invoke
8690 s390_indirect_branch_settings also if fndecl didn't change.
8692 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8694 * config/rs6000/rs6000.md (blockage): Set length to zero.
8696 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8698 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8700 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8703 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8704 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8708 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8709 PARALLEL incoming that failed vt_get_decl_and_offset check.
8712 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8713 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8715 (get_variable_section): For decls in named .bss* sections pass true as
8716 second argument to bss_initializer_p.
8718 2018-02-09 Marek Polacek <polacek@redhat.com>
8719 Jakub Jelinek <jakub@redhat.com>
8722 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8723 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8724 Sync some changes from cxx_fold_indirect_ref.
8726 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8728 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8730 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8731 (BLOCK_INLINE_ENTRY_LABEL): New.
8732 (dwarf2out_var_location): Disregard inline entry markers.
8733 (inline_entry_data): New struct.
8734 (inline_entry_data_hasher): New hashtable type.
8735 (inline_entry_data_hasher::hash): New.
8736 (inline_entry_data_hasher::equal): New.
8737 (inline_entry_data_table): New variable.
8738 (add_high_low_attributes): Add DW_AT_entry_pc and
8739 DW_AT_GNU_entry_view attributes if a pending entry is found
8740 in inline_entry_data_table. Add old entry_pc attribute only
8741 if debug nonbinding markers are disabled.
8742 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8743 markers are enabled.
8744 (block_within_block_p, dwarf2out_inline_entry): New.
8745 (dwarf2out_finish): Check that no entries remained in
8746 inline_entry_data_table.
8747 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8748 (final_scan_insn, notice_source_line): Likewise.
8749 (rest_of_clean_state): Skip inline entry markers.
8750 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8752 * gimple.c (gimple_build_debug_inline_entry): New.
8753 * gimple.h (enum gimple_debug_subcode): Add
8754 GIMPLE_DEBUG_INLINE_ENTRY.
8755 (gimple_build_debug_inline_entry): Declare.
8756 (gimple_debug_inline_entry_p): New.
8757 (gimple_debug_nonbind_marker_p): Adjust.
8758 * insn-notes.def (INLINE_ENTRY): New.
8759 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8760 inline entry marker notes.
8761 (print_insn): Likewise.
8762 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8763 (INSN_DEBUG_MARKER_KIND): Likewise.
8764 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8765 * tree-inline.c (expand_call_inline): Build and insert
8766 debug_inline_entry stmt.
8767 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8768 inline entry blocks early, if nonbind markers are enabled.
8769 (dump_scope_block): Dump fragment info.
8770 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8771 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8772 (gimple_build_debug_inline_entry): New.
8773 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8774 Enable/disable inline entry points too.
8775 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8776 (DEBUG_INSN): Describe inline entry markers.
8778 * common.opt (gvariable-location-views): New.
8779 (gvariable-location-views=incompat5): New.
8780 * config.in: Rebuilt.
8781 * configure: Rebuilt.
8782 * configure.ac: Test assembler for view support.
8783 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8784 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8785 * dwarf2out.c (var_loc_view): New typedef.
8786 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8787 (dwarf2out_locviews_in_attribute): New.
8788 (dwarf2out_locviews_in_loclist): New.
8789 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8790 (enum dw_line_info_opcode): Add LI_adv_address.
8791 (struct dw_line_info_table): Add view.
8792 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8793 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8794 (zero_view_p): New variable.
8795 (ZERO_VIEW_P): New macro.
8796 (output_asm_line_debug_info): New.
8797 (struct var_loc_node): Add view.
8798 (add_AT_view_list, AT_loc_list): New.
8799 (add_var_loc_to_decl): Add view param. Test it against last.
8800 (new_loc_list): Add view params. Record them.
8801 (AT_loc_list_ptr): Handle loc and view lists.
8802 (view_list_to_loc_list_val_node): New.
8803 (print_dw_val): Handle dw_val_class_view_list.
8804 (size_of_die): Likewise.
8805 (value_format): Likewise.
8806 (loc_list_has_views): New.
8807 (gen_llsym): Set vl_symbol too.
8808 (maybe_gen_llsym, skip_loc_list_entry): New.
8809 (dwarf2out_maybe_output_loclist_view_pair): New.
8810 (output_loc_list): Output view list or entries too.
8811 (output_view_list_offset): New.
8812 (output_die): Handle dw_val_class_view_list.
8813 (output_dwarf_version): New.
8814 (output_compilation_unit_header): Use it.
8815 (output_skeleton_debug_sections): Likewise.
8816 (output_rnglists, output_line_info): Likewise.
8817 (output_pubnames, output_aranges): Update version comments.
8818 (output_one_line_info_table): Output view numbers in asm comments.
8819 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8820 Call maybe_gen_llsym.
8821 (loc_list_from_tree_1): Adjust.
8822 (add_AT_location_description): Create view list attribute if
8823 needed, check it's absent otherwise.
8824 (convert_cfa_to_fb_loc_list): Adjust.
8825 (maybe_emit_file): Call output_asm_line_debug_info for test.
8826 (dwarf2out_var_location): Reset views as needed. Precompute
8827 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8828 attribute. Set view.
8829 (new_line_info_table): Reset next view.
8830 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8831 (dwarf2out_source_line): Likewise. Output view resets and labels to
8832 the assembler, or select appropriate line info opcodes.
8833 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8834 (optimize_string_length): Catch it. Adjust.
8835 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8836 dw_val_class_view_list, and remove it if no longer needed.
8837 (hash_loc_list): Hash view numbers.
8838 (loc_list_hasher::equal): Compare them.
8839 (optimize_location_lists): Check whether a view list symbol is
8840 needed, and whether the locview attribute is present, and
8841 whether they match. Remove the locview attribute if no longer
8843 (index_location_lists): Call skip_loc_list_entry for test.
8844 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8845 Use output_dwarf_version.
8846 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8847 (struct dw_val_node): Add val_view_list.
8848 * final.c (SEEN_NEXT_VIEW): New.
8849 (set_next_view_needed): New.
8850 (clear_next_view_needed): New.
8851 (maybe_output_next_view): New.
8852 (final_start_function): Rename to...
8853 (final_start_function_1): ... this. Take pointer to FIRST,
8854 add SEEN parameter. Emit param bindings in the initial view.
8855 (final_start_function): Reintroduce SEEN-less interface.
8856 (final): Rename to...
8857 (final_1): ... this. Take SEEN parameter. Output final pending
8858 next view at the end.
8859 (final): Reintroduce seen-less interface.
8860 (final_scan_insn): Output pending next view before switching
8861 sections or ending a block. Mark the next view as needed when
8862 outputting variable locations. Notify debug backend of section
8863 changes, and of location view changes.
8864 (rest_of_handle_final): Adjust.
8865 * toplev.c (process_options): Autodetect value for debug variable
8866 location views option. Warn on incompat5 without -gdwarf-5.
8867 * doc/invoke.texi (gvariable-location-views): New.
8868 (gvariable-location-views=incompat5): New.
8869 (gno-variable-location-views): New.
8871 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8873 PR tree-optimization/84136
8874 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8875 that the result of find_edge is non-NULL.
8877 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8880 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8881 storing integer register in SImode. Fix cost of 256 and 512
8882 byte aligned SSE register store.
8884 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8886 * config/i386/i386.c (ix86_multiplication_cost): Fix
8887 multiplication cost for TARGET_AVX512DQ.
8889 2018-02-08 Marek Polacek <polacek@redhat.com>
8891 PR tree-optimization/84238
8892 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8895 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8897 PR tree-optimization/84265
8898 * tree-vect-stmts.c (vectorizable_store): Don't treat
8899 VMAT_CONTIGUOUS accesses as grouped.
8900 (vectorizable_load): Likewise.
8902 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8904 PR tree-optimization/81635
8905 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8906 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8907 (test_round_for_mask): New functions.
8908 (wide_int_cc_tests): Call test_round_for_mask.
8909 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8910 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8911 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8912 range returned by get_range_info.
8914 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8917 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8918 * symtab.c: Include builtins.h
8919 (symtab_node::output_to_lto_symbol_table_p): Move here
8920 from lto-streamer-out.c:output_symbol_p.
8921 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8922 (output_symbol_p): Move all logic to symtab.c
8923 (produce_symtab): Update.
8925 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8927 * config/s390/s390-opts.h (enum indirect_branch): Define.
8928 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8929 (s390_indirect_branch_via_thunk)
8930 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8931 (enum s390_indirect_branch_type): Define.
8932 * config/s390/s390.c (struct s390_frame_layout, struct
8933 machine_function): Remove.
8934 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8935 (indirect_branch_table_label_no, indirect_branch_table_name):
8937 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8938 (enum s390_indirect_branch_option): Define.
8939 (s390_return_addr_from_memory): New function.
8940 (s390_handle_string_attribute): New function.
8941 (s390_attribute_table): Add new attribute handler.
8942 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8943 (s390_indirect_branch_via_thunk): New function.
8944 (s390_indirect_branch_via_inline_thunk): New function.
8945 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8946 sibling call optimization for non z10 compiles.
8947 (s390_emit_call): Force indirect branch target to be a single
8948 register. Add r1 clobber for non-z10 compiles.
8949 (s390_emit_epilogue): Emit return jump via return_use expander.
8950 (s390_reorg): Handle JUMP_INSNs as execute targets.
8951 (s390_option_override_internal): Perform validity checks for the
8952 new command line options.
8953 (s390_indirect_branch_attrvalue): New function.
8954 (s390_indirect_branch_settings): New function.
8955 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8956 (s390_output_indirect_thunk_function): New function.
8957 (s390_code_end): Implement target hook.
8958 (s390_case_values_threshold): Implement target hook.
8959 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8961 * config/s390/s390.h (struct s390_frame_layout)
8962 (struct machine_function): Move here from s390.c.
8963 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8964 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8965 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8966 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8967 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8968 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8969 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8970 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8971 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8972 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8973 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8974 (mnemonic attribute): Add values which aren't recognized
8976 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8977 pattern for branch conversion. Fix mnemonic attribute.
8978 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8979 indirect branch via thunk if requested.
8980 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8981 ("*indirect_jump"): Disable for branch conversion using out of
8983 ("indirect_jump_via_thunk<mode>_z10")
8984 ("indirect_jump_via_thunk<mode>")
8985 ("indirect_jump_via_inlinethunk<mode>_z10")
8986 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8987 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8988 ("casesi_jump_via_inlinethunk<mode>_z10")
8989 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8990 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8991 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8992 ("*indirect2_jump"): Disable for branch conversion.
8993 ("casesi_jump"): Turn into expander and expand patterns for branch
8995 ("return_use"): New expander.
8996 ("*return"): Emit return via thunk and rename it to ...
8997 ("*return<mode>"): ... this one.
8998 * config/s390/s390.opt: Add new options and and enum for the
9001 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
9003 * lra-constraints.c (match_reload): Unconditionally use
9004 gen_lowpart_SUBREG, rather than selecting between that
9005 and equivalent gen_rtx_SUBREG code.
9007 2018-02-08 Richard Biener <rguenther@suse.de>
9009 PR tree-optimization/84233
9010 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
9011 changed flag instead of boguously re-using phi_inserted.
9013 2018-02-08 Martin Jambor <mjambor@suse.cz>
9015 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
9016 static local variables.
9018 2018-02-08 Richard Biener <rguenther@suse.de>
9020 PR tree-optimization/84278
9021 * tree-vect-stmts.c (vectorizable_store): When looking for
9022 smaller vector types to perform grouped strided loads/stores
9023 make sure the mode is supported by the target.
9024 (vectorizable_load): Likewise.
9026 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9028 * config/aarch64/aarch64.c (aarch64_components_for_bb):
9029 Increase LDP/STP opportunities by adding adjacent callee-saves.
9031 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
9033 PR rtl-optimization/84068
9034 PR rtl-optimization/83459
9035 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
9037 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9039 PR tree-optimization/84224
9040 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9041 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9044 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9047 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9048 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9049 count, remove one USE.
9051 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9053 * doc/install.texi (Configuration): Document the
9054 --with-long-double-format={ibm,ieee} PowerPC configuration
9058 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9059 Convert from define_expand to be define_insn_and_split. Rework
9060 float/double/_Float128 conversions to QI/HI/SImode to work with
9061 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9062 conversions to QI/HImode types did a store and then a load to
9063 truncate the value. For conversions to VSX registers, don't split
9064 the insn, instead emit the code directly. Use the code iterator
9065 any_fix to combine signed and unsigned conversions.
9066 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9067 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9068 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9069 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9070 (fix_<mode>di2_hw): Likewise.
9071 (fixuns_<mode>di2_hw): Likewise.
9072 (fix_<mode>si2_hw): Likewise.
9073 (fixuns_<mode>si2_hw): Likewise.
9074 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9075 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9076 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9077 fix<uns>_trunc<SFDF:mode>si2_p8.
9078 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9080 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9081 (fix<uns>_<mode>_mem): Likewise.
9082 (fctiw<u>z_<mode>_mem): Likewise.
9083 (fix<uns>_<mode>_mem): Likewise.
9084 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9085 the register allocator from doing a direct move to the GPRs to do
9086 a store, and instead use the ISA 3.0 store byte/half-word from
9087 vector register instruction. For IEEE 128-bit floating point,
9088 also optimize stores of 32-bit ints.
9089 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9091 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9093 * genextract.c (push_pathstr_operand): New function to support
9095 (walk_rtx): Call push_pathstr_operand.
9096 (print_path): Support [a-zA-Z].
9098 2018-02-07 Richard Biener <rguenther@suse.de>
9100 PR tree-optimization/84037
9101 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9102 (cse_and_gimplify_to_preheader): Declare.
9103 (vect_get_place_in_interleaving_chain): Likewise.
9104 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9106 (_loop_vec_info::~_loop_vec_info): Delete it.
9107 (cse_and_gimplify_to_preheader): New function.
9108 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9109 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9110 (vectorizable_load): Likewise. For grouped stores always base
9111 the IV on the first element.
9112 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9113 condition before gimplifying.
9115 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9117 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9118 *DIV_EXPR and *MOD_EXPR.
9120 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9123 * config/i386/i386.c (ix86_option_override_internal): Mask out
9124 the CF_SET bit when checking -fcf-protection.
9126 2018-02-07 Tom de Vries <tom@codesourcery.com>
9129 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9132 2018-02-07 Richard Biener <rguenther@suse.de>
9134 PR tree-optimization/84204
9135 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9138 PR tree-optimization/84205
9139 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9140 special-case isl_ast_op_zdiv_r.
9142 PR tree-optimization/84223
9143 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9144 Only add conditions from within the region.
9145 (gather_bbs::after_dom_children): Adjust.
9147 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9150 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9151 * config/avr/avr.md: Only post-reload split REG-REG moves if
9152 either register is GENERAL_REG_P.
9154 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9156 PR tree-optimization/84235
9157 * tree-ssa-scopedtables.c
9158 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9159 if the subtraction is performed in floating point type where NaNs are
9160 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9161 build 1. Formatting fix.
9163 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9166 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9167 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9168 and skip it regardless of bb boundaries. Use CALL_P macro,
9169 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9170 unnecessarily, formatting fix.
9172 2018-02-06 Michael Collison <michael.collison@arm.com>
9174 * config/arm/thumb2.md:
9175 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9176 (*thumb_mov_notscc): Ditto.
9178 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9181 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9182 unsigned_fix, not "s".
9184 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9186 * configure.ac (gcc_fn_eh_frame_ro): New function.
9187 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9188 correct .eh_frame permissions.
9189 * configure: Regenerate.
9191 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9193 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9196 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9198 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9199 Display warning message for -mno-speculate-indirect-jumps.
9201 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9203 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9205 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9207 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9209 PR tree-optimization/84225
9210 * tree-eh.c (find_trapping_overflow): Only call
9211 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9213 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9216 * config/i386/i386.c: Reimplement the check of possible options
9217 -mibt/-mshstk conbination. Change error messages.
9218 * doc/invoke.texi: Fix a typo: remove extra '='.
9220 2018-02-06 Marek Polacek <polacek@redhat.com>
9222 PR tree-optimization/84228
9223 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9225 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9228 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9229 emitted arch directives.
9230 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9231 __ARM_FEATURE_COPROC before changing architectures.
9233 2018-02-06 Richard Biener <rguenther@suse.de>
9235 * config/i386/i386.c (print_reg): Fix typo.
9236 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9238 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9240 * configure: Regenerate.
9242 2018-02-05 Martin Sebor <msebor@redhat.com>
9244 PR tree-optimization/83369
9245 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9248 2018-02-05 Martin Liska <mliska@suse.cz>
9250 * doc/invoke.texi: Cherry-pick upstream r323995.
9252 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9254 * ira.c (ira_init_register_move_cost): Adjust comment.
9256 2018-02-05 Martin Liska <mliska@suse.cz>
9258 PR gcov-profile/84137
9259 * doc/gcov.texi: Fix typo in documentation.
9261 2018-02-05 Martin Liska <mliska@suse.cz>
9263 PR gcov-profile/83879
9264 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9265 using dlopen functionality.
9267 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9269 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9270 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9271 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9272 _mm_maskz_range_round_ss): New intrinsics.
9273 (__builtin_ia32_rangesd128_round)
9274 (__builtin_ia32_rangess128_round): Remove.
9275 (__builtin_ia32_rangesd128_mask_round,
9276 __builtin_ia32_rangess128_mask_round): New builtins.
9277 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9278 __builtin_ia32_rangess128_round): Remove.
9279 (__builtin_ia32_rangesd128_mask_round,
9280 __builtin_ia32_rangess128_mask_round): New builtins.
9281 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9282 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9283 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9284 "<round_saeonly_constraint>")): Changed to ...
9285 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9286 "<round_saeonly_scalar_constraint>")): ... this.
9287 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9288 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9289 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9290 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9291 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9293 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9295 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9297 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9298 Remove all values except native, 8540 and 8548.
9300 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9302 * config/i386/i386.c (ix86_output_function_return): Pass
9303 INVALID_REGNUM, instead of -1, as invalid register number to
9304 indirect_thunk_name and output_indirect_thunk.
9306 2018-02-02 Julia Koval <julia.koval@intel.com>
9308 * config.gcc: Add -march=icelake.
9309 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9310 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9311 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9312 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9313 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9314 (processor_target_table): Add icelake.
9315 (ix86_option_override_internal): Handle new PTAs.
9316 (get_builtin_code_for_version): Handle icelake.
9317 (M_INTEL_COREI7_ICELAKE): New.
9318 (fold_builtin_cpu): Handle icelake.
9319 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9320 * doc/invoke.texi: Add -march=icelake.
9322 2018-02-02 Julia Koval <julia.koval@intel.com>
9324 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9325 to wide_int_bitmask.
9326 * wide-int-bitmask.h: New.
9328 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9331 * config/i386/i386.md: Replace Pmode with word_mode in
9332 builtin_setjmp_setup and builtin_longjmp to support x32.
9334 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9338 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9340 (rs6000_supported_cpu_names): New static variable.
9341 (linux_cpu_translation_table): Likewise.
9342 (elf_platform) <cpu>: Define new static variable and use it.
9343 Translate kernel AT_PLATFORM name to canonical name if needed.
9344 Error if platform name is unknown.
9346 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9349 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9351 2018-02-01 Jeff Law <law@redhat.com>
9354 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9355 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9356 the scratch if RELEASE_VIA_POP is false.
9357 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9358 If we have to save a temporary register, decrement SIZE appropriately.
9359 Pass new arguments to release_scratch_register_on_entry.
9360 (ix86_adjust_stack_and_probe): Likewise.
9361 (ix86_emit_probe_stack_range): Pass new arguments to
9362 release_scratch_register_on_entry.
9364 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9366 PR rtl-optimization/84157
9367 * combine.c (change_zero_ext): Use REG_P predicate in
9368 front of HARD_REGISTER_P predicate.
9370 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9372 * config/avr/avr.c (avr_option_override): Move disabling of
9373 -fdelete-null-pointer-checks to...
9374 * common/config/avr/avr-common.c (avr_option_optimization_table):
9377 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9379 PR tree-optimization/81635
9380 * tree-data-ref.c (split_constant_offset_1): For types that
9381 wrap on overflow, try to use range info to prove that wrapping
9384 2018-02-01 Renlin Li <renlin.li@arm.com>
9387 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9389 (aarch64_register_move_cost): Likewise.
9390 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9392 (REG_CLASS_NAMES): Likewise.
9393 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9394 TAILCALL_ADDR_REGS. Remove IP registers.
9395 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9397 2018-02-01 Richard Biener <rguenther@suse.de>
9399 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9400 for specifying RPO order and allow NULL for that.
9401 * domwalk.c (dom_walker::dom_walker): Likewise.
9402 (dom_walker::walk): Handle NULL RPO order.
9403 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9405 (rewrite_update_dom_walker): Likewise.
9406 (mark_def_dom_walker): Likewise.
9408 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9410 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9411 (aarch64_maybe_expand_sve_subreg_move): Declare.
9412 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9413 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9415 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9416 that are semantically a reverse operation.
9417 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9418 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9419 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9421 (aarch64_can_change_mode_class): For big-endian, forbid changes
9422 between two SVE modes if they have different element sizes.
9424 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9426 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9427 the TImode handling for big-endian targets.
9429 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9431 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9432 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9434 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9435 Remove BSWAP handing for big-endian targets and use the form of
9436 LD1RQ appropariate for the mode.
9438 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9440 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9441 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9444 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9447 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9448 check for operands that need to go through aarch64_sve_reload_be.
9450 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9452 PR tree-optimization/81661
9453 PR tree-optimization/84117
9454 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9455 * tree-eh.c: Include gimplify.h.
9456 (find_trapping_overflow, replace_trapping_overflow,
9457 rewrite_to_non_trapping_overflow): New functions.
9458 * tree-vect-loop.c: Include tree-eh.h.
9459 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9460 * tree-data-ref.c: Include tree-eh.h.
9461 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9463 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9465 PR rtl-optimization/84123
9466 * combine.c (change_zero_ext): Check if hard register satisfies
9467 can_change_dest_mode before calling gen_lowpart_SUBREG.
9469 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9472 * ira.c (ira_init_register_move_cost): Remove assert.
9474 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9476 PR rtl-optimization/84071
9477 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9478 * doc/tm.texi: Regenerate.
9480 2018-01-31 Richard Biener <rguenther@suse.de>
9482 PR tree-optimization/84132
9483 * tree-data-ref.c (analyze_miv_subscript): Properly
9484 check whether evolution_function_is_affine_multivariate_p
9485 before calling gcd_of_steps_may_divide_p.
9487 2018-01-31 Julia Koval <julia.koval@intel.com>
9490 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9491 * config/i386/i386.md (rdpid_rex64) New.
9492 (rdpid): Make 32bit only.
9494 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9497 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9498 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9500 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9503 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9505 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9507 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9509 PR rtl-optimization/84071
9510 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9511 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9513 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9515 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9516 (arc_attribute_table): Add 'aux' attribute.
9517 (arc_in_small_data_p): Consider aux like variables.
9518 (arc_is_aux_reg_p): New function.
9519 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9520 (arc_get_aux_arg): New function.
9521 (prepare_move_operands): Handle aux-register access.
9522 (arc_handle_aux_attribute): New function.
9523 * doc/extend.texi (ARC Variable attributes): Add subsection.
9525 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9527 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9528 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9529 (arc_attribute_table): Add 'uncached' attribute.
9530 (arc_print_operand): Print '.di' flag for uncached memory
9532 (arc_in_small_data_p): Do not consider for small data the uncached
9534 (arc_is_uncached_mem_p): New function.
9535 * config/arc/predicates.md (compact_store_memory_operand): Check
9536 for uncached memory accesses.
9537 (nonvol_nonimm_operand): Likewise.
9538 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9540 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9543 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9544 falign-loops=): Add Optimization flag.
9546 2018-01-30 Jeff Law <law@redhat.com>
9549 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9550 INT_REGISTERS_SAVED. Check it prior to calling
9551 get_scratch_register_on_entry.
9552 (ix86_adjust_stack_and_probe): Similarly.
9553 (ix86_emit_probe_stack_range): Similarly.
9554 (ix86_expand_prologue): Corresponding changes.
9556 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9559 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9560 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9562 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9565 * lra-constraints.c (curr_insn_transform): Process AND in the
9568 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9570 PR rtl-optimization/83986
9571 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9572 dependence against last_pending_memory_flush in addition to
9575 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9577 PR tree-optimization/81611
9578 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9581 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9584 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9587 2018-01-30 Richard Biener <rguenther@suse.de>
9588 Jakub Jelinek <jakub@redhat.com>
9590 PR tree-optimization/84111
9591 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9592 inner loops added during recursion, as they don't have up-to-date
9595 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9598 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9599 (can_inline_edge_by_limits_p): ... here.
9600 (can_early_inline_edge_p, check_callers,
9601 update_caller_keys, update_callee_keys, recursive_inlining,
9602 add_new_edges_to_heap, speculation_useful_p,
9603 inline_small_functions,
9604 inline_small_functions, flatten_function,
9605 inline_to_all_callers_1): Update.
9607 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9609 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9612 2018-01-30 Richard Biener <rguenther@suse.de>
9614 PR tree-optimization/83008
9615 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9616 invariant and constant vector uses in stmts when they need
9619 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9622 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9623 * configure: Regenerate.
9625 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9627 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9629 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9630 Use gen_rtx_REG rather than gen_lowpart.
9632 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9634 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9635 rather than 0 when creating partial subregs.
9637 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9639 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9642 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9645 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9646 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9647 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9648 flags. This restores the settings used before the 2017-07-24.
9649 Turning off pre increment/decrement/modify allows IVOPTS to
9650 optimize DF/SF loops where the index is an int.
9652 2018-01-29 Richard Biener <rguenther@suse.de>
9653 Kelvin Nilsen <kelvin@gcc.gnu.org>
9656 * tree-vect-stmts.c (vectorizable_call): Don't call
9657 targetm.vectorize_builtin_md_vectorized_function if callee is
9660 2018-01-22 Carl Love <cel@us.ibm.com>
9662 * doc/extend.tex: Fix typo in second arg in
9663 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9665 2018-01-29 Richard Biener <rguenther@suse.de>
9667 PR tree-optimization/84086
9668 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9669 (flush_ssaname_freelist): When SSA names were released reset
9670 the SCEV hash table.
9672 2018-01-29 Richard Biener <rguenther@suse.de>
9674 PR tree-optimization/84057
9675 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9676 removed paths when removing edges.
9678 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9680 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9681 -mfunction-return=@var{choice}.
9683 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9686 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9688 (layout::print_source_line): Likewise.
9689 (test_get_line_width_without_trailing_whitespace): Add test cases.
9691 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9694 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9697 2018-01-26 Jim Wilson <jimw@sifive.com>
9699 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9701 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9704 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9706 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9707 and CMP + SUB-immediate -> SUBS.
9709 2018-01-26 Martin Sebor <msebor@redhat.com>
9711 PR tree-optimization/83896
9712 * tree-ssa-strlen.c (get_string_len): Rename...
9713 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9714 Avoid assuming length is constant.
9715 (handle_char_store): Use HOST_WIDE_INT for string length.
9717 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9720 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9721 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9723 2018-01-26 Richard Biener <rguenther@suse.de>
9725 PR rtl-optimization/84003
9726 * dse.c (record_store): Only record redundant stores when
9727 the earlier store aliases at least all accesses the later one does.
9729 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9731 PR rtl-optimization/83985
9732 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9733 REG_CFA_RESTORE insns.
9734 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9735 REG_CFA_RESTORE insns here.
9738 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9739 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9741 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9743 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9744 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9745 (arc_init): Likewise.
9746 (arc_override_options): Likewise.
9747 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9749 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9751 * config/arc/arc.h (TARGET_DBNZ): Define.
9752 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9753 properly set the tune attribute.
9754 (dbnz): Use TARGET_DBNZ guard.
9755 * config/arc/arc.opt (mtune): Add core3 option.
9757 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9759 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9760 recognize new pic like addresses.
9761 (arc_delegitimize_address): Clean up.
9763 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9765 * config/arc/arc-arches.def: Option mrf16 valid for all
9767 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9768 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9769 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9770 * config/arc/arc-tables.opt: Regenerate.
9771 * config/arc/arc.c (arc_conditional_register_usage): Handle
9772 reduced register file case.
9773 (arc_file_start): Set must have build attributes.
9774 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9776 * config/arc/arc.opt (mrf16): Add new option.
9777 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9778 * config/arc/genmultilib.awk: Handle new mrf16 option.
9779 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9780 * config/arc/t-multilib: Regenerate.
9781 * doc/invoke.texi (ARC Options): Document mrf16 option.
9783 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9785 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9786 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9787 (arc_attribute_table): Add 'secure_call' attribute.
9788 (arc_print_operand): Print secure call operand.
9789 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9791 (arc_is_secure_call_p): New function. * config/arc/arc.md
9792 (call_i): Add support for sjli instruction.
9793 (call_value_i): Likewise.
9794 * config/arc/constraints.md (Csc): New constraint.
9796 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9797 John Eric Martin <John.Martin@emmicro-us.com>
9799 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9800 * config/arc/arc.c (_arc_jli_section): New struct.
9801 (arc_jli_section): New type.
9802 (rc_jli_sections): New static variable.
9803 (arc_handle_jli_attribute): New function.
9804 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9805 (arc_file_end): New function.
9806 (TARGET_ASM_FILE_END): Define.
9807 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9808 (arc_add_jli_section): New function.
9809 (jli_call_scan): Likewise.
9810 (arc_reorg): Call jli_call_scan.
9811 (arc_output_addsi): Remove 'S' from printing asm operand.
9812 (arc_is_jli_call_p): New function.
9813 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9815 (movhi_insn): Likewise.
9816 (movsi_insn): Likewise.
9817 (movsi_set_cc_insn): Likewise.
9818 (loadqi_update): Likewise.
9819 (load_zeroextendqisi_update): Likewise.
9820 (load_signextendqisi_update): Likewise.
9821 (loadhi_update): Likewise.
9822 (load_zeroextendhisi_update): Likewise.
9823 (load_signextendhisi_update): Likewise.
9824 (loadsi_update): Likewise.
9825 (loadsf_update): Likewise.
9826 (movsicc_insn): Likewise.
9827 (bset_insn): Likewise.
9828 (bxor_insn): Likewise.
9829 (bclr_insn): Likewise.
9830 (bmsk_insn): Likewise.
9831 (bicsi3_insn): Likewise.
9832 (cmpsi_cc_c_insn): Likewise.
9833 (movsi_ne): Likewise.
9834 (movsi_cond_exec): Likewise.
9835 (clrsbsi2): Likewise.
9846 (call_i): Remove 'S' asm letter, add jli instruction.
9847 (call_value_i): Likewise.
9848 * config/arc/arc.op (mjli-always): New option.
9849 * config/arc/constraints.md (Cji): New constraint.
9850 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9852 (subsf3_fpx): Likewise.
9853 (mulsf3_fpx): Likewise.
9854 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9856 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9858 * doc/invoke.texi (ARC): Document mjli-always option.
9860 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9862 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9863 avoid addition with 0 and use incw and decw where possible.
9865 2018-01-26 Richard Biener <rguenther@suse.de>
9867 PR tree-optimization/81082
9868 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9869 association if it requires casting to unsigned.
9870 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9871 from fold_plusminus_mult_expr to catch important cases late when
9872 range info is available.
9874 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9876 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9877 * configure.ac (hidden_linkonce): New test.
9878 * configure: Regenerate.
9879 * config.in: Regenerate.
9881 2018-01-26 Julia Koval <julia.koval@intel.com>
9883 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9884 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9885 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9886 _mm_mask_bitshuffle_epi64_mask): Fix type.
9887 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9888 USI_FTYPE_V4DI_V4DI_USI): Remove.
9889 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9890 __builtin_ia32_vpshufbitqmb256_mask,
9891 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9892 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9893 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9895 2018-01-26 Alan Modra <amodra@gmail.com>
9898 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9899 UNSPEC_VBPERMQ. Sort other unspecs.
9901 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9903 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9905 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9908 * predict.c (drop_profile): Do not push/pop cfun; update also
9910 (handle_missing_profiles): Fix logic looking for zero profiles.
9912 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9915 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9916 on functions with #pragma omp declare simd or functions with simd
9918 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9919 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9920 Remove trailing \n from warning_at calls.
9922 2018-01-25 Tom de Vries <tom@codesourcery.com>
9925 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9926 for neutered workers.
9928 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9931 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9932 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9934 2018-01-24 Jeff Law <law@redhat.com>
9937 * i386.c (get_probe_interval): Move to earlier point.
9938 (ix86_compute_frame_layout): If -fstack-clash-protection and
9939 the frame is larger than the probe interval, then use pushes
9940 to save registers rather than reg->mem moves.
9941 (ix86_expand_prologue): Remove conditional for int_registers_saved
9944 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9947 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9948 min/max for never referenced object.
9950 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9953 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9955 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9956 attributes from DECL_ATTRIBUTES (decl) without affecting
9957 DECL_ATTRIBUTES (current_function_decl).
9958 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9959 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9961 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9963 PR tree-optimization/83979
9964 * fold-const.c (fold_comparison): Use constant_boolean_node
9965 instead of boolean_{true,false}_node.
9967 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9969 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9972 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9974 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9975 Simplify the clause that sets the length attribute.
9976 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9977 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9978 clause that sets the length attribute.
9979 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9981 2018-01-24 Tom de Vries <tom@codesourcery.com>
9984 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9985 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9986 Add strict parameter.
9987 (prevent_branch_around_nothing): Insert dummy insn between branch to
9988 label and label with no ptx insn inbetween.
9989 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9991 2018-01-24 Tom de Vries <tom@codesourcery.com>
9994 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9995 for neutered threads in warp.
9996 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9998 2018-01-24 Richard Biener <rguenther@suse.de>
10000 PR tree-optimization/83176
10001 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
10004 2018-01-24 Richard Biener <rguenther@suse.de>
10006 PR tree-optimization/82819
10007 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
10008 code generating pluses that are no-ops in the target precision.
10010 2018-01-24 Richard Biener <rguenther@suse.de>
10012 PR middle-end/84000
10013 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
10015 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10017 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
10018 to merge probabilities.
10019 * predict.c (probably_never_executed): Also mark as cold functions
10020 with global 0 profile and guessed local profile.
10021 * profile-count.c (profile_probability::combine_with_count): New
10023 * profile-count.h (profile_probability::operator*,
10024 profile_probability::operator*=, profile_probability::operator/,
10025 profile_probability::operator/=): Reduce precision to adjusted
10026 and set value to guessed on contradictory divisions.
10027 (profile_probability::combine_with_freq): Remove.
10028 (profile_probability::combine_wiht_count): Declare.
10029 (profile_count::force_nonzero):: Set to adjusted.
10030 (profile_count::probability_in):: Set quality to adjusted.
10031 * tree-ssa-tail-merge.c (replace_block_by): Use
10032 combine_with_count.
10034 2018-01-23 Andrew Waterman <andrew@sifive.com>
10035 Jim Wilson <jimw@sifive.com>
10037 * config/riscv/riscv.c (riscv_stack_boundary): New.
10038 (riscv_option_override): Set riscv_stack_boundary. Handle
10039 riscv_preferred_stack_boundary_arg.
10040 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10041 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10042 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10043 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10044 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10045 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10047 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10050 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10051 of struct ix86_frame.
10052 (ix86_expand_epilogue): Likewise. Add a local variable for
10053 the reg_save_offset field in struct ix86_frame.
10055 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10057 PR tree-optimization/82604
10058 * tree-loop-distribution.c (enum partition_kind): New enum item
10059 PKIND_PARTIAL_MEMSET.
10060 (partition_builtin_p): Support above new enum item.
10061 (generate_code_for_partition): Ditto.
10062 (compute_access_range): Differentiate cases that equality can be
10063 proven at all loops, the innermost loops or no loops.
10064 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10065 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10066 (finalize_partitions, distribute_loop): Don't fuse partition of
10067 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10068 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10069 parloop is enabled.
10071 2018-01-23 Martin Liska <mliska@suse.cz>
10073 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10074 order to ignore the predictor.
10075 (PRED_POLYMORPHIC_CALL): Likewise.
10076 (PRED_RECURSIVE_CALL): Likewise.
10078 2018-01-23 Martin Liska <mliska@suse.cz>
10080 * tree-profile.c (tree_profiling): Print function header to
10081 aware reader which function we are working on.
10082 * value-prof.c (gimple_find_values_to_profile): Do not print
10083 not interesting value histograms.
10085 2018-01-23 Martin Liska <mliska@suse.cz>
10087 * profile-count.h (enum profile_quality): Add
10088 profile_uninitialized as the first value. Do not number values
10089 as they are zero based.
10090 (profile_count::verify): Update sanity check.
10091 (profile_probability::verify): Likewise.
10093 2018-01-23 Nathan Sidwell <nathan@acm.org>
10095 * doc/invoke.texi (ffor-scope): Deprecate.
10097 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10099 PR tree-optimization/83510
10100 * domwalk.c (set_all_edges_as_executable): New function.
10101 (dom_walker::dom_walker): Convert bool param
10102 "skip_unreachable_blocks" to enum reachability. Move setup of
10103 edge flags to set_all_edges_as_executable and only do it when
10104 reachability is REACHABLE_BLOCKS.
10105 * domwalk.h (enum dom_walker::reachability): New enum.
10106 (dom_walker::dom_walker): Convert bool param
10107 "skip_unreachable_blocks" to enum reachability.
10108 (set_all_edges_as_executable): New decl.
10109 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10110 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10112 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10113 but converting true to REACHABLE_BLOCKS.
10114 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10116 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10117 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10118 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10120 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10121 if check_all_array_refs will be called.
10123 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10125 * tree.c (selftest::test_location_wrappers): Add more test
10128 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10130 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10131 (selftest::test_bit_in_range): Likewise.
10133 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10136 * doc/sourcebuild.texi (vect_float): Say that the selector
10137 only describes the situation when -funsafe-math-optimizations is on.
10138 (vect_float_strict): Document.
10140 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10142 PR tree-optimization/83965
10143 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10144 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10145 instead of checking only for a reduction.
10146 (vect_recog_widen_sum_pattern): Likewise.
10148 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10150 * predict.c (probably_never_executed): Only use precise profile info.
10151 (compute_function_frequency): Skip after inlining hack since we now
10152 have quality checking.
10154 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10156 * profile-count.h (profile_probability::very_unlikely,
10157 profile_probability::unlikely, profile_probability::even): Set
10158 precision to guessed.
10160 2018-01-23 Richard Biener <rguenther@suse.de>
10162 PR tree-optimization/83963
10163 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10164 Properly terminate dominator walk when crossing the exit edge not
10165 when visiting its source block.
10167 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10170 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10171 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10173 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10175 PR tree-optimization/83957
10176 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10177 semicolon after for body surrounded by braces.
10179 PR tree-optimization/83081
10180 * profile-count.h (profile_probability::split): New method.
10181 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10182 Use profile_probability::split.
10183 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10184 when splitting a single conditional jump into 2.
10186 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10188 PR tree-optimization/69452
10189 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10192 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10194 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10195 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10196 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10198 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10200 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10201 * config/rl78/rl78.md (movdi): New define_expand.
10202 * config/rl78/rl78.c (rl78_split_movdi): New function.
10204 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10207 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10209 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10210 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10211 128-bit to produce an UNSPEC move to get the double word with the
10212 signbit and then a shift directly to do signbit.
10213 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10214 implementation with a new version that just does either a direct
10215 move or a regular move. Move memory interface to separate insns.
10216 Move insns so they are next to the expander.
10217 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10218 with signbit move. Split big and little endian case.
10219 (signbit<mode>2_dm_mem_le): Likewise.
10220 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10221 (signbit<mode>2_dm2): Likewise.
10223 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10225 * config/rl78/rl78.md (anddi3): New define_expand.
10227 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10229 * config/rl78/rl78.md (umindi3): New define_expand.
10231 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10233 * config/rl78/rl78.md (smindi3): New define_expand.
10235 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10237 * config/rl78/rl78.md (smaxdi3): New define_expand.
10239 2018-01-22 Carl Love <cel@us.ibm.com>
10241 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10242 LVX_V1TI): Add macro expansion.
10243 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10244 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10245 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10246 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10247 Change check to determine if the instruction is a byte reversing
10248 entry. Fix typo in comment.
10249 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10250 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10251 Add def_builtin calls for new builtins.
10252 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10253 Add define_insn expansion.
10255 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10257 * config/rl78/rl78.md (umaxdi3): New define_expand.
10259 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10261 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10262 for non-QImode registers.
10264 2018-01-22 Richard Biener <rguenther@suse.de>
10266 PR tree-optimization/83963
10267 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10268 including the loop exit block.
10269 (scop_detection::merge_sese): Likewise.
10270 (scop_detection::add_scop): Do it here instead.
10272 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10274 * doc/sourcebuild.texi (arm_softfloat): Document.
10276 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10279 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10280 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10281 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10283 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10284 David Edelsohn <dje.gcc@gmail.com>
10287 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10288 Change "crset eq" to "crset 2".
10289 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10290 (*call_indirect_aix<mode>_nospec): Likewise.
10291 (*call_value_indirect_aix<mode>_nospec): Likewise.
10292 (*call_indirect_elfv2<mode>_nospec): Likewise.
10293 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10294 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10295 change assembly output from . to $.
10296 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10297 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10298 (*tablejump<mode>_internal1_nospec): Likewise.
10300 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10303 * config/sh/sh_optimize_sett_clrt.cc:
10304 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10306 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10308 PR tree-optimization/83940
10309 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10310 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10311 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10312 use it to pass back the definition type.
10313 (vect_check_store_rhs): Likewise rhs_dt_out.
10314 (vect_build_gather_load_calls): Add a mask_dt argument and use
10315 it instead of a call to vect_is_simple_use.
10316 (vectorizable_store): Update calls to vect_check_load_store_mask
10317 and vect_check_store_rhs. Use the dt returned by the latter instead
10318 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10319 instead of calls to vect_is_simple_use. Pass the scalar rather
10320 than the vector operand to vect_is_simple_use when handling
10321 second and subsequent copies of an rhs value.
10322 (vectorizable_load): Update calls to vect_check_load_store_mask
10323 and vect_build_gather_load_calls. Use the cached mask_dt and
10324 gs_info.offset_dt instead of calls to vect_is_simple_use.
10326 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10328 PR middle-end/83945
10329 * tree-emutls.c: Include gimplify.h.
10330 (lower_emutls_2): New function.
10331 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10332 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10333 it before further processing.
10336 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10337 UINTVAL (trueop1) instead of INTVAL (op1).
10339 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10343 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10344 INCOMING_FRAME_SP_OFFSET if not defined.
10345 (scan_trace): Add ENTRY argument. If true and
10346 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10347 emit a note to adjust the CFA offset.
10348 (create_cfi_notes): Adjust scan_trace callers.
10349 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10350 INCOMING_FRAME_SP_OFFSET in the CIE.
10351 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10352 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10354 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10355 * doc/tm.texi: Regenerated.
10357 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10359 PR rtl-optimization/83147
10360 * lra-constraints.c (remove_inheritance_pseudos): Use
10361 lra_substitute_pseudo_within_insn.
10363 2018-01-19 Tom de Vries <tom@codesourcery.com>
10364 Cesar Philippidis <cesar@codesourcery.com>
10367 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10369 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10372 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10373 spaces for function labels.
10375 2018-01-19 Martin Liska <mliska@suse.cz>
10377 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10378 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10379 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10380 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10381 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10382 (PRED_CONST_RETURN): Change from 69 to 65.
10383 (PRED_NULL_RETURN): Change from 91 to 71.
10384 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10385 (PRED_LOOP_GUARD): Change from 66 to 73.
10387 2018-01-19 Martin Liska <mliska@suse.cz>
10389 * predict.c (predict_insn_def): Add new assert.
10390 (struct branch_predictor): Change type to signed integer.
10391 (test_prediction_value_range): Amend test to cover
10392 PROB_UNINITIALIZED.
10393 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10394 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10395 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10396 (PRED_LOOP_IV_COMPARE): Likewise.
10397 * predict.h (PROB_UNINITIALIZED): Define new constant.
10399 2018-01-19 Martin Liska <mliska@suse.cz>
10401 * predict.c (dump_prediction): Add new format for
10402 analyze_brprob.py script which is enabled with -details
10404 * profile-count.h (precise_p): New function.
10406 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10408 PR tree-optimization/83922
10409 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10410 there are no statements that need masking.
10411 (vect_active_double_reduction_p): New function.
10412 (vect_analyze_loop_operations): Use it when handling phis that
10413 are not in the loop header.
10415 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10417 PR tree-optimization/83914
10418 * tree-vect-loop.c (vectorizable_induction): Don't convert
10419 init_expr or apply the peeling adjustment for inductions
10420 that are nested within the vectorized loop.
10422 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10424 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10427 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10431 * function.h (gimplify_parameters): Add gimple_seq * argument.
10432 * function.c: Include gimple.h and options.h.
10433 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10434 for the added local temporaries if needed.
10435 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10436 if there are any parameter cleanups, wrap whole body into a
10437 try/finally with the cleanups.
10439 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10442 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10443 Use GET_MODE_CLASS for scalar floating point.
10445 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10449 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10450 Fix call of call_cgraph_insertion_hooks.
10452 2018-01-18 Martin Sebor <msebor@redhat.com>
10454 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10456 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10459 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10462 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10465 * common.opt: (-ffile-prefix-map): New option.
10466 * opts.c (common_handle_option): Defer it.
10467 * opts-global.c (handle_common_deferred_options): Handle it.
10468 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10469 * file-prefix-map.h: New file.
10470 (remap_debug_filename, add_debug_prefix_map): ...here.
10471 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10472 * final.c (debug_prefix_map, add_debug_prefix_map
10473 remap_debug_filename): Move to...
10474 * file-prefix-map.c: New file.
10475 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10476 generalize, get rid of alloca(), use strrchr() instead of strchr().
10477 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10478 Implement in terms of add_prefix_map().
10479 (remap_macro_filename, remap_debug_filename): Implement in term of
10481 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10482 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10483 * dbxout.c: Include file-prefix-map.h.
10484 * varasm.c: Likewise.
10485 * vmsdbgout.c: Likewise.
10486 * xcoffout.c: Likewise.
10487 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10488 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10489 * doc/invoke.texi (-ffile-prefix-map): Document.
10490 (-fdebug-prefix-map): Update description.
10492 2018-01-18 Martin Liska <mliska@suse.cz>
10494 * config/i386/i386.c (indirect_thunk_name): Document that also
10496 (output_indirect_thunk): Document why both instructions
10497 (pause and lfence) are generated.
10499 2018-01-18 Richard Biener <rguenther@suse.de>
10501 PR tree-optimization/83887
10502 * graphite-scop-detection.c
10503 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10504 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10505 (scop_detection::merge_sese): Re-implement with a flood-fill
10506 algorithm that properly finds a SESE region if it exists.
10508 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10511 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10512 pointer_diff optimizations use view_convert instead of convert.
10514 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10516 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10517 Generate different code for -mno-speculate-indirect-jumps.
10518 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10519 (*call_indirect_aix<mode>): Disable for
10520 -mno-speculate-indirect-jumps.
10521 (*call_indirect_aix<mode>_nospec): New define_insn.
10522 (*call_value_indirect_aix<mode>): Disable for
10523 -mno-speculate-indirect-jumps.
10524 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10525 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10526 -mno-speculate-indirect-jumps.
10527 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10529 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10531 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10532 long double type, set the flags for noting the default long double
10533 type, even if we don't pass or return a long double type.
10535 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10538 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10541 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10543 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10544 support for merge[hl].
10545 (fold_mergehl_helper): New helper function.
10546 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10547 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10548 (altivec_vmrglw_direct): Add xxmrglw insn.
10550 2018-01-17 Andrew Waterman <andrew@sifive.com>
10552 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10553 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10555 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10558 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10559 call the lto_location_cache before reading the
10560 DECL_SOURCE_LOCATION of the types.
10562 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10563 Richard Sandiford <richard.sandiford@linaro.org>
10565 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10566 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10567 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10568 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10569 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10571 * config/aarch64/constraints.md (aarch64_movti_operand):
10573 * config/aarch64/predicates.md (Uti): Add new constraint.
10575 2018-01-17 Carl Love <cel@us.ibm.com>
10577 * config/rs6000/vsx.md (define_expand xl_len_r,
10578 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10579 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10581 (define_expand, define_insn): Move the shift left from the
10582 define_insn to the define_expand for lxvl and stxvl instructions.
10583 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10584 and XL_LEN_R definitions to PURE.
10586 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10588 * config/i386/i386.c (indirect_thunk_name): Declare regno
10589 as unsigned int. Compare regno with INVALID_REGNUM.
10590 (output_indirect_thunk): Ditto.
10591 (output_indirect_thunk_function): Ditto.
10592 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10593 in the call to output_indirect_thunk_function.
10595 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10597 PR middle-end/83884
10598 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10599 rather than the size of inner_type to determine the stack slot size
10600 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10602 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10605 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10608 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10610 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10611 endian Linux systems to optionally enable multilibs for selecting
10612 the long double type if the user configured an explicit type.
10613 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10614 have no long double multilibs if not defined.
10615 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10616 warn if the user used -mabi={ieee,ibm}longdouble and we built
10617 multilibs for long double.
10618 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10619 appropriate multilib option.
10620 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10622 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10623 for building long double multilibs.
10624 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10626 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10628 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10631 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10633 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10636 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10639 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10642 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10644 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10645 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10647 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10649 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10650 different rtl trees depending on TARGET_64BIT.
10651 (rs6000_gen_lvx): Likewise.
10653 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10655 * config/visium/visium.md (nop): Tweak comment.
10656 (hazard_nop): Likewise.
10658 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10660 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10661 -mspeculate-indirect-jumps.
10662 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10663 for -mno-speculate-indirect-jumps.
10664 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10665 (*call_value_indirect_elfv2<mode>): Disable for
10666 -mno-speculate-indirect-jumps.
10667 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10668 (indirect_jump): Emit different RTL for
10669 -mno-speculate-indirect-jumps.
10670 (*indirect_jump<mode>): Disable for
10671 -mno-speculate-indirect-jumps.
10672 (*indirect_jump<mode>_nospec): New define_insn.
10673 (tablejump): Emit different RTL for
10674 -mno-speculate-indirect-jumps.
10675 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10676 (tablejumpsi_nospec): New define_expand.
10677 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10678 (tablejumpdi_nospec): New define_expand.
10679 (*tablejump<mode>_internal1): Disable for
10680 -mno-speculate-indirect-jumps.
10681 (*tablejump<mode>_internal1_nospec): New define_insn.
10682 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10685 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10687 * caller-save.c (insert_save): Drop unnecessary parameter. All
10690 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10691 Richard Biener <rguenth@suse.de>
10694 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10695 return early, inline manually is_gimple_sizepos. Make sure if we
10696 call gimplify_expr we don't end up with a gimple constant.
10697 * tree.c (variably_modified_type_p): Don't return true for
10698 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10699 * gimplify.h (is_gimple_sizepos): Remove.
10701 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10703 PR tree-optimization/83857
10704 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10705 vectorizable_live_operation for pure SLP statements.
10706 (vectorizable_live_operation): Handle PHIs.
10708 2018-01-16 Richard Biener <rguenther@suse.de>
10710 PR tree-optimization/83867
10711 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10712 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10714 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10717 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10718 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10719 If off is not INTEGER_CST, issue a may not be aligned warning
10720 rather than isn't aligned. Use isn%'t rather than isn't.
10721 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10723 <case MULT_EXPR>: Improve the case when bottom and one of the
10724 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10725 operand, in that case check if the other operand is multiple of
10726 bottom divided by the INTEGER_CST operand.
10728 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10731 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10732 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10733 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10734 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10735 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10736 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10737 (pa_function_arg_size): New function.
10739 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10741 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10742 in a separate statement.
10744 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10746 PR tree-optimization/83847
10747 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10748 group gathers and scatters.
10750 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10752 PR rtl-optimization/86620
10753 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10755 PR rtl-optimization/83213
10756 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10757 to last if both are JUMP_INSNs.
10759 PR tree-optimization/83843
10760 * gimple-ssa-store-merging.c
10761 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10762 store_immediate_info for bswap/nop orig_stores.
10764 2018-01-15 Andrew Waterman <andrew@sifive.com>
10766 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10768 <UDIV>: Increase cost if !TARGET_DIV.
10770 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10772 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10773 (define_attr "cr_logical_3op"): New.
10774 (cceq_ior_compare): Adjust.
10775 (cceq_ior_compare_complement): Adjust.
10776 (*cceq_rev_compare): Adjust.
10777 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10778 (is_cracked_insn): Adjust.
10779 (insn_must_be_first_in_group): Adjust.
10780 * config/rs6000/40x.md: Adjust.
10781 * config/rs6000/440.md: Adjust.
10782 * config/rs6000/476.md: Adjust.
10783 * config/rs6000/601.md: Adjust.
10784 * config/rs6000/603.md: Adjust.
10785 * config/rs6000/6xx.md: Adjust.
10786 * config/rs6000/7450.md: Adjust.
10787 * config/rs6000/7xx.md: Adjust.
10788 * config/rs6000/8540.md: Adjust.
10789 * config/rs6000/cell.md: Adjust.
10790 * config/rs6000/e300c2c3.md: Adjust.
10791 * config/rs6000/e500mc.md: Adjust.
10792 * config/rs6000/e500mc64.md: Adjust.
10793 * config/rs6000/e5500.md: Adjust.
10794 * config/rs6000/e6500.md: Adjust.
10795 * config/rs6000/mpc.md: Adjust.
10796 * config/rs6000/power4.md: Adjust.
10797 * config/rs6000/power5.md: Adjust.
10798 * config/rs6000/power6.md: Adjust.
10799 * config/rs6000/power7.md: Adjust.
10800 * config/rs6000/power8.md: Adjust.
10801 * config/rs6000/power9.md: Adjust.
10802 * config/rs6000/rs64.md: Adjust.
10803 * config/rs6000/titan.md: Adjust.
10805 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10807 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10808 ix86_indirect_branch_register logic.
10810 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10812 * config/i386/constraints.md (Bs): Update
10813 ix86_indirect_branch_register check. Don't check
10814 ix86_indirect_branch_register with GOT_memory_operand.
10816 * config/i386/predicates.md (GOT_memory_operand): Don't check
10817 ix86_indirect_branch_register here.
10818 (GOT32_symbol_operand): Likewise.
10820 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10822 * config/i386/predicates.md (constant_call_address_operand):
10823 Rewrite ix86_indirect_branch_register logic.
10824 (sibcall_insn_operand): Likewise.
10826 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10828 * config/i386/constraints.md (Bs): Replace
10829 ix86_indirect_branch_thunk_register with
10830 ix86_indirect_branch_register.
10832 * config/i386/i386.md (indirect_jump): Likewise.
10833 (tablejump): Likewise.
10834 (*sibcall_memory): Likewise.
10835 (*sibcall_value_memory): Likewise.
10836 Peepholes of indirect call and jump via memory: Likewise.
10837 * config/i386/i386.opt: Likewise.
10838 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10839 (GOT_memory_operand): Likewise.
10840 (call_insn_operand): Likewise.
10841 (sibcall_insn_operand): Likewise.
10842 (GOT32_symbol_operand): Likewise.
10844 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10846 PR middle-end/83837
10847 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10848 type rather than type addr's type points to.
10849 (expand_omp_atomic_mutex): Likewise.
10850 (expand_omp_atomic): Likewise.
10852 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10855 * config/i386/i386.c (output_indirect_thunk_function): Use
10856 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10857 for __x86_return_thunk.
10859 2018-01-15 Richard Biener <rguenther@suse.de>
10861 PR middle-end/83850
10862 * expmed.c (extract_bit_field_1): Fix typo.
10864 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10867 * config/arm/iterators.md (VF): New mode iterator.
10868 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10869 Remove integer-related logic from pattern.
10870 (neon_vabd<mode>_3): Likewise.
10872 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10874 PR middle-end/82694
10875 * common.opt (fstrict-overflow): No longer an alias.
10876 (fwrapv-pointer): New option.
10877 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10878 also for pointer types based on flag_wrapv_pointer.
10879 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10880 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10881 opts->x_flag_wrapv got set.
10882 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10883 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10884 POINTER_TYPE_OVERFLOW_UNDEFINED.
10885 * match.pd: Likewise in address comparison pattern.
10886 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10888 2018-01-15 Richard Biener <rguenther@suse.de>
10891 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10892 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10893 Reset type names to their identifier if their TYPE_DECL doesn't
10894 have linkage (and thus is used for ODR and devirt).
10895 (save_debug_info_for_decl): Remove.
10896 (save_debug_info_for_type): Likewise.
10897 (add_tree_to_fld_list): Adjust.
10898 * tree-pretty-print.c (dump_generic_node): Make dumping of
10899 type names more robust.
10901 2018-01-15 Richard Biener <rguenther@suse.de>
10903 * BASE-VER: Bump to 8.0.1.
10905 2018-01-14 Martin Sebor <msebor@redhat.com>
10908 * builtins.c (check_access): Avoid warning when the no-warning bit
10911 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10913 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10914 * ira-color (allocno_hard_regs_compare): Likewise.
10916 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10919 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10920 Use .pushsection/.popsection.
10922 2018-01-14 Martin Sebor <msebor@redhat.com>
10925 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10927 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10929 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10930 entry from extra_headers.
10931 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10932 extra_headers, make the list bitwise identical to the i?86-*-* one.
10934 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10936 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10937 -mcmodel=large with -mindirect-branch=thunk,
10938 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10939 -mfunction-return=thunk-extern.
10940 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10941 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10942 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10944 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10946 * config/i386/i386.c (print_reg): Print the name of the full
10947 integer register without '%'.
10948 (ix86_print_operand): Handle 'V'.
10949 * doc/extend.texi: Document 'V' modifier.
10951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10953 * config/i386/constraints.md (Bs): Disallow memory operand for
10954 -mindirect-branch-register.
10956 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10957 (GOT_memory_operand): Likewise.
10958 (call_insn_operand): Likewise.
10959 (sibcall_insn_operand): Likewise.
10960 (GOT32_symbol_operand): Likewise.
10961 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10962 for -mindirect-branch-register.
10963 (tablejump): Likewise.
10964 (*sibcall_memory): Likewise.
10965 (*sibcall_value_memory): Likewise.
10966 Disallow peepholes of indirect call and jump via memory for
10967 -mindirect-branch-register.
10968 (*call_pop): Replace m with Bw.
10969 (*call_value_pop): Likewise.
10970 (*sibcall_pop_memory): Replace m with Bs.
10971 * config/i386/i386.opt (mindirect-branch-register): New option.
10972 * doc/invoke.texi: Document -mindirect-branch-register option.
10974 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10976 * config/i386/i386-protos.h (ix86_output_function_return): New.
10977 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10978 set function_return_type.
10979 (indirect_thunk_name): Add ret_p to indicate thunk for function
10981 (output_indirect_thunk_function): Pass false to
10982 indirect_thunk_name.
10983 (ix86_output_indirect_branch_via_reg): Likewise.
10984 (ix86_output_indirect_branch_via_push): Likewise.
10985 (output_indirect_thunk_function): Create alias for function
10986 return thunk if regno < 0.
10987 (ix86_output_function_return): New function.
10988 (ix86_handle_fndecl_attribute): Handle function_return.
10989 (ix86_attribute_table): Add function_return.
10990 * config/i386/i386.h (machine_function): Add
10991 function_return_type.
10992 * config/i386/i386.md (simple_return_internal): Use
10993 ix86_output_function_return.
10994 (simple_return_internal_long): Likewise.
10995 * config/i386/i386.opt (mfunction-return=): New option.
10996 (indirect_branch): Mention -mfunction-return=.
10997 * doc/extend.texi: Document function_return function attribute.
10998 * doc/invoke.texi: Document -mfunction-return= option.
11000 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
11002 * config/i386/i386-opts.h (indirect_branch): New.
11003 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
11004 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
11005 with local indirect jump when converting indirect call and jump.
11006 (ix86_set_indirect_branch_type): New.
11007 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
11008 (indirectlabelno): New.
11009 (indirect_thunk_needed): Likewise.
11010 (indirect_thunk_bnd_needed): Likewise.
11011 (indirect_thunks_used): Likewise.
11012 (indirect_thunks_bnd_used): Likewise.
11013 (INDIRECT_LABEL): Likewise.
11014 (indirect_thunk_name): Likewise.
11015 (output_indirect_thunk): Likewise.
11016 (output_indirect_thunk_function): Likewise.
11017 (ix86_output_indirect_branch_via_reg): Likewise.
11018 (ix86_output_indirect_branch_via_push): Likewise.
11019 (ix86_output_indirect_branch): Likewise.
11020 (ix86_output_indirect_jmp): Likewise.
11021 (ix86_code_end): Call output_indirect_thunk_function if needed.
11022 (ix86_output_call_insn): Call ix86_output_indirect_branch if
11024 (ix86_handle_fndecl_attribute): Handle indirect_branch.
11025 (ix86_attribute_table): Add indirect_branch.
11026 * config/i386/i386.h (machine_function): Add indirect_branch_type
11027 and has_local_indirect_jump.
11028 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
11030 (tablejump): Likewise.
11031 (*indirect_jump): Use ix86_output_indirect_jmp.
11032 (*tablejump_1): Likewise.
11033 (simple_return_indirect_internal): Likewise.
11034 * config/i386/i386.opt (mindirect-branch=): New option.
11035 (indirect_branch): New.
11038 (thunk-inline): Likewise.
11039 (thunk-extern): Likewise.
11040 * doc/extend.texi: Document indirect_branch function attribute.
11041 * doc/invoke.texi: Document -mindirect-branch= option.
11043 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11046 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11048 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11050 * ipa-inline.c (want_inline_small_function_p): Return false if
11051 inlining has already failed with CIF_FINAL_ERROR.
11052 (update_caller_keys): Call want_inline_small_function_p before
11054 (update_callee_keys): Likewise.
11056 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11058 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11060 (rs6000_quadword_masked_address_p): Likewise.
11061 (quad_aligned_load_p): Likewise.
11062 (quad_aligned_store_p): Likewise.
11063 (const_load_sequence_p): Add comment to describe the outer-most loop.
11064 (mimic_memory_attributes_and_flags): New function.
11065 (rs6000_gen_stvx): Likewise.
11066 (replace_swapped_aligned_store): Likewise.
11067 (rs6000_gen_lvx): Likewise.
11068 (replace_swapped_aligned_load): Likewise.
11069 (replace_swapped_load_constant): Capitalize argument name in
11070 comment describing this function.
11071 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11072 and stores that access quad-word aligned addresses and replace
11073 with stvx or lvx instructions when appropriate.
11074 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11075 New function prototype.
11076 (rs6000_quadword_masked_address_p): Likewise.
11077 (rs6000_gen_lvx): Likewise.
11078 (rs6000_gen_stvx): Likewise.
11079 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11080 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11081 when memory address is aligned.
11082 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11083 this split to select lvx instruction when memory address is aligned.
11084 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11085 instruction when memory address is aligned.
11086 (*vsx_le_perm_load_v16qi): Likewise.
11087 (four unnamed splitters): Modify to select the stvx instruction
11088 when memory is aligned.
11090 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11092 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11093 which appears in the queue multiple times.
11095 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11096 Alan Hayward <alan.hayward@arm.com>
11097 David Sherwood <david.sherwood@arm.com>
11099 * tree-vectorizer.h (vec_lower_bound): New structure.
11100 (_loop_vec_info): Add check_nonzero and lower_bounds.
11101 (LOOP_VINFO_CHECK_NONZERO): New macro.
11102 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11103 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11104 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11105 fields. Make seg_len the distance travelled, not including the
11107 (dr_direction_indicator): Declare.
11108 (dr_zero_step_indicator): Likewise.
11109 (dr_known_forward_stride_p): Likewise.
11110 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11112 (runtime_alias_check_p): Allow runtime alias checks with
11114 (operator ==): Compare access_size and align.
11115 (prune_runtime_alias_test_list): Rework for new distinction between
11116 the access_size and seg_len.
11117 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11119 (get_segment_min_max): New function.
11120 (create_intersect_range_checks): Use it.
11121 (dr_step_indicator): New function.
11122 (dr_direction_indicator): Likewise.
11123 (dr_zero_step_indicator): Likewise.
11124 (dr_known_forward_stride_p): Likewise.
11125 * tree-loop-distribution.c (data_ref_segment_size): Return
11126 DR_STEP * (niters - 1).
11127 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11129 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11130 (vect_preserves_scalar_order_p): New function, split out from...
11131 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11132 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11133 (vect_vfa_access_size): New function.
11134 (vect_vfa_align): Likewise.
11135 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11136 (dump_lower_bound): New function.
11137 (vect_check_lower_bound): Likewise.
11138 (vect_small_gap_p): Likewise.
11139 (vectorizable_with_step_bound_p): Likewise.
11140 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11141 depencies if the vectorization factor is 1. Convert the checks
11142 for nonzero steps into checks on the bounds of DR_STEP. Try using
11143 a bunds check for variable steps if the minimum required step is
11144 relatively small. Update calls to the dr_with_seg_len
11145 constructor and to vect_compile_time_alias.
11146 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11148 (vect_loop_versioning): Call it.
11149 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11151 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11154 Alan Hayward <alan.hayward@arm.com>
11155 David Sherwood <david.sherwood@arm.com>
11157 * doc/sourcebuild.texi (vect_scatter_store): Document.
11158 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11160 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11162 * genopinit.c (main): Add supports_vec_scatter_store and
11163 supports_vec_scatter_store_cached to target_optabs.
11164 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11165 IFN_MASK_SCATTER_STORE.
11166 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11168 * internal-fn.h (internal_store_fn_p): Declare.
11169 (internal_fn_stored_value_index): Likewise.
11170 * internal-fn.c (scatter_store_direct): New macro.
11171 (expand_scatter_store_optab_fn): New function.
11172 (direct_scatter_store_optab_supported_p): New macro.
11173 (internal_store_fn_p): New function.
11174 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11175 IFN_MASK_SCATTER_STORE.
11176 (internal_fn_mask_index): Likewise.
11177 (internal_fn_stored_value_index): New function.
11178 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11179 for scatter stores.
11180 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11181 * optabs-query.c (supports_vec_scatter_store_p): New function.
11182 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11183 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11184 true for scatter stores.
11185 (vect_gather_scatter_fn_p): Handle scatter stores too.
11186 (vect_check_gather_scatter): Consider using scatter stores if
11187 supports_vec_scatter_store_p.
11188 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11189 scatter stores too.
11190 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11191 internal_fn_stored_value_index.
11192 (check_load_store_masking): Handle scatter stores too.
11193 (vect_get_store_rhs): Make public.
11194 (vectorizable_call): Use internal_store_fn_p.
11195 (vectorizable_store): Handle scatter store internal functions.
11196 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11197 when deciding whether the end of the group has been reached.
11198 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11199 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11200 (mask_scatter_store<mode>): New insns.
11202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11203 Alan Hayward <alan.hayward@arm.com>
11204 David Sherwood <david.sherwood@arm.com>
11206 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11207 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11208 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11210 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11211 Use vect_truncate_gather_scatter_offset if we can't treat the
11212 operation as a normal gather load or scatter store.
11213 (get_group_load_store_type): Take the gather_scatter_info
11214 as argument. Try using a gather load or scatter store for
11215 single-element groups.
11216 (get_load_store_type): Update calls to get_group_load_store_type
11217 and vect_use_strided_gather_scatters_p.
11219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11220 Alan Hayward <alan.hayward@arm.com>
11221 David Sherwood <david.sherwood@arm.com>
11223 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11224 optional tree argument.
11225 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11227 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11228 but continue to use the current value as a fallback.
11229 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11230 to compare the updates.
11231 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11232 (get_load_store_type): Use it when handling a strided access.
11233 (vect_get_strided_load_store_ops): New function.
11234 (vect_get_data_ptr_increment): Likewise.
11235 (vectorizable_load): Handle strided gather loads. Always pass
11236 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11239 Alan Hayward <alan.hayward@arm.com>
11240 David Sherwood <david.sherwood@arm.com>
11242 * doc/md.texi (gather_load@var{m}): Document.
11243 (mask_gather_load@var{m}): Likewise.
11244 * genopinit.c (main): Add supports_vec_gather_load and
11245 supports_vec_gather_load_cached to target_optabs.
11246 * optabs-tree.c (init_tree_optimization_optabs): Use
11247 ggc_cleared_alloc to allocate target_optabs.
11248 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11249 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11251 * internal-fn.h (internal_load_fn_p): Declare.
11252 (internal_gather_scatter_fn_p): Likewise.
11253 (internal_fn_mask_index): Likewise.
11254 (internal_gather_scatter_fn_supported_p): Likewise.
11255 * internal-fn.c (gather_load_direct): New macro.
11256 (expand_gather_load_optab_fn): New function.
11257 (direct_gather_load_optab_supported_p): New macro.
11258 (direct_internal_fn_optab): New function.
11259 (internal_load_fn_p): Likewise.
11260 (internal_gather_scatter_fn_p): Likewise.
11261 (internal_fn_mask_index): Likewise.
11262 (internal_gather_scatter_fn_supported_p): Likewise.
11263 * optabs-query.c (supports_at_least_one_mode_p): New function.
11264 (supports_vec_gather_load_p): Likewise.
11265 * optabs-query.h (supports_vec_gather_load_p): Declare.
11266 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11267 and memory_type field.
11268 (NUM_PATTERNS): Bump to 15.
11269 * tree-vect-data-refs.c: Include internal-fn.h.
11270 (vect_gather_scatter_fn_p): New function.
11271 (vect_describe_gather_scatter_call): Likewise.
11272 (vect_check_gather_scatter): Try using internal functions for
11273 gather loads. Recognize existing calls to a gather load function.
11274 (vect_analyze_data_refs): Consider using gather loads if
11275 supports_vec_gather_load_p.
11276 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11277 (vect_get_gather_scatter_offset_type): Likewise.
11278 (vect_convert_mask_for_vectype): Likewise.
11279 (vect_add_conversion_to_patterm): Likewise.
11280 (vect_try_gather_scatter_pattern): Likewise.
11281 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11282 (vect_vect_recog_func_ptrs): Add it.
11283 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11284 internal_fn_mask_index and internal_gather_scatter_fn_p.
11285 (check_load_store_masking): Take the gather_scatter_info as an
11286 argument and handle gather loads.
11287 (vect_get_gather_scatter_ops): New function.
11288 (vectorizable_call): Check internal_load_fn_p.
11289 (vectorizable_load): Likewise. Handle gather load internal
11291 (vectorizable_store): Update call to check_load_store_masking.
11292 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11293 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11294 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11295 (aarch64_gather_scale_operand_d): New predicates.
11296 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11297 (mask_gather_load<mode>): New insns.
11299 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11300 Alan Hayward <alan.hayward@arm.com>
11301 David Sherwood <david.sherwood@arm.com>
11303 * optabs.def (fold_left_plus_optab): New optab.
11304 * doc/md.texi (fold_left_plus_@var{m}): Document.
11305 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11306 * internal-fn.c (fold_left_direct): Define.
11307 (expand_fold_left_optab_fn): Likewise.
11308 (direct_fold_left_optab_supported_p): Likewise.
11309 * fold-const-call.c (fold_const_fold_left): New function.
11310 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11311 * tree-parloops.c (valid_reduction_p): New function.
11312 (gather_scalar_reductions): Use it.
11313 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11314 (vect_finish_replace_stmt): Declare.
11315 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11316 (needs_fold_left_reduction_p): New function, split out from...
11317 (vect_is_simple_reduction): ...here. Accept reductions that
11318 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11319 (vect_force_simple_reduction): Also store the reduction type in
11320 the assignment's STMT_VINFO_REDUC_TYPE.
11321 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11322 (merge_with_identity): New function.
11323 (vect_expand_fold_left): Likewise.
11324 (vectorize_fold_left_reduction): Likewise.
11325 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11326 scalar phi in place for it. Check for target support and reject
11327 cases that would reassociate the operation. Defer the transform
11328 phase to vectorize_fold_left_reduction.
11329 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11330 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11331 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11335 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11336 call to ifc_temp_var.
11338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11339 Alan Hayward <alan.hayward@arm.com>
11340 David Sherwood <david.sherwood@arm.com>
11342 * target.def (legitimize_address_displacement): Take the original
11343 offset as a poly_int.
11344 * targhooks.h (default_legitimize_address_displacement): Update
11346 * targhooks.c (default_legitimize_address_displacement): Likewise.
11347 * doc/tm.texi: Regenerate.
11348 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11349 as an argument, moving assert of ad->disp == ad->disp_term to...
11350 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11351 Try calling targetm.legitimize_address_displacement before expanding
11352 the address rather than afterwards, and adjust for the new interface.
11353 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11354 Match the new hook interface. Handle SVE addresses.
11355 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11356 new hook interface.
11358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11360 * Makefile.in (OBJS): Add early-remat.o.
11361 * target.def (select_early_remat_modes): New hook.
11362 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11363 * doc/tm.texi: Regenerate.
11364 * targhooks.h (default_select_early_remat_modes): Declare.
11365 * targhooks.c (default_select_early_remat_modes): New function.
11366 * timevar.def (TV_EARLY_REMAT): New timevar.
11367 * passes.def (pass_early_remat): New pass.
11368 * tree-pass.h (make_pass_early_remat): Declare.
11369 * early-remat.c: New file.
11370 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11372 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 Alan Hayward <alan.hayward@arm.com>
11376 David Sherwood <david.sherwood@arm.com>
11378 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11379 vfm1 with a bound_epilog parameter.
11380 (vect_do_peeling): Update calls accordingly, and move the prologue
11381 call earlier in the function. Treat the base bound_epilog as 0 for
11382 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11383 this base when peeling for gaps.
11384 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11385 with fully-masked loops.
11386 (vect_estimate_min_profitable_iters): Handle the single peeled
11387 iteration in that case.
11389 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11390 Alan Hayward <alan.hayward@arm.com>
11391 David Sherwood <david.sherwood@arm.com>
11393 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11394 single-element interleaving even if the size is not a power of 2.
11395 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11396 accesses for single-element interleaving if the group size is
11399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11400 Alan Hayward <alan.hayward@arm.com>
11401 David Sherwood <david.sherwood@arm.com>
11403 * doc/md.texi (fold_extract_last_@var{m}): Document.
11404 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11405 * optabs.def (fold_extract_last_optab): New optab.
11406 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11407 * internal-fn.c (fold_extract_direct): New macro.
11408 (expand_fold_extract_optab_fn): Likewise.
11409 (direct_fold_extract_optab_supported_p): Likewise.
11410 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11411 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11412 EXTRACT_LAST_REDUCTION.
11413 (get_initial_def_for_reduction): Do not create an initial vector
11414 for EXTRACT_LAST_REDUCTION reductions.
11415 (vectorizable_reduction): Leave the scalar phi in place for
11416 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11417 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11418 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11419 transform phase to vectorizable_condition.
11420 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11422 (vect_finish_stmt_generation): ...here.
11423 (vect_finish_replace_stmt): New function.
11424 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11425 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11427 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11430 Alan Hayward <alan.hayward@arm.com>
11431 David Sherwood <david.sherwood@arm.com>
11433 * doc/md.texi (extract_last_@var{m}): Document.
11434 * optabs.def (extract_last_optab): New optab.
11435 * internal-fn.def (EXTRACT_LAST): New internal function.
11436 * internal-fn.c (cond_unary_direct): New macro.
11437 (expand_cond_unary_optab_fn): Likewise.
11438 (direct_cond_unary_optab_supported_p): Likewise.
11439 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11440 loops using EXTRACT_LAST.
11441 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11442 (extract_last_<mode>): ...this optab.
11443 (vec_extract<mode><Vel>): Update accordingly.
11445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11446 Alan Hayward <alan.hayward@arm.com>
11447 David Sherwood <david.sherwood@arm.com>
11449 * target.def (empty_mask_is_expensive): New hook.
11450 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11451 * doc/tm.texi: Regenerate.
11452 * targhooks.h (default_empty_mask_is_expensive): Declare.
11453 * targhooks.c (default_empty_mask_is_expensive): New function.
11454 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11455 if the target says that empty masks are expensive.
11456 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11458 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11460 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11461 Alan Hayward <alan.hayward@arm.com>
11462 David Sherwood <david.sherwood@arm.com>
11464 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11465 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11466 (vect_use_loop_mask_for_alignment_p): New function.
11467 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11468 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11469 niters_skip argument. Make sure that the first niters_skip elements
11470 of the first iteration are inactive.
11471 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11472 Update call to vect_set_loop_masks_directly.
11473 (get_misalign_in_elems): New function, split out from...
11474 (vect_gen_prolog_loop_niters): ...here.
11475 (vect_update_init_of_dr): Take a code argument that specifies whether
11476 the adjustment should be added or subtracted.
11477 (vect_update_init_of_drs): Likewise.
11478 (vect_prepare_for_masked_peels): New function.
11479 (vect_do_peeling): Skip prologue peeling if we're using a mask
11480 instead. Update call to vect_update_inits_of_drs.
11481 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11483 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11484 alignment. Do not include the number of peeled iterations in
11485 the minimum threshold in that case.
11486 (vectorizable_induction): Adjust the start value down by
11487 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11488 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11489 Take the number of skipped iterations into account when calculating
11491 * tree-vect-stmts.c (vect_gen_while_not): New function.
11493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11494 Alan Hayward <alan.hayward@arm.com>
11495 David Sherwood <david.sherwood@arm.com>
11497 * doc/sourcebuild.texi (vect_fully_masked): Document.
11498 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11499 default value to 0.
11500 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11502 (vect_analyze_loop_2): ...here. Don't check the vectorization
11503 factor against the number of loop iterations if the loop is
11506 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11507 Alan Hayward <alan.hayward@arm.com>
11508 David Sherwood <david.sherwood@arm.com>
11510 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11511 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11512 (dump_groups): Update accordingly.
11513 (iv_use::mem_type): New member variable.
11514 (address_p): New function.
11515 (record_use): Add a mem_type argument and initialize the new
11517 (record_group_use): Add a mem_type argument. Use address_p.
11518 Remove obsolete null checks of base_object. Update call to record_use.
11519 (find_interesting_uses_op): Update call to record_group_use.
11520 (find_interesting_uses_cond): Likewise.
11521 (find_interesting_uses_address): Likewise.
11522 (get_mem_type_for_internal_fn): New function.
11523 (find_address_like_use): Likewise.
11524 (find_interesting_uses_stmt): Try find_address_like_use before
11525 calling find_interesting_uses_op.
11526 (addr_offset_valid_p): Use the iv mem_type field as the type
11527 of the addressed memory.
11528 (add_autoinc_candidates): Likewise.
11529 (get_address_cost): Likewise.
11530 (split_small_address_groups_p): Use address_p.
11531 (split_address_groups): Likewise.
11532 (add_iv_candidate_for_use): Likewise.
11533 (autoinc_possible_for_pair): Likewise.
11534 (rewrite_groups): Likewise.
11535 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11536 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11537 (get_alias_ptr_type_for_ptr_address): New function.
11538 (rewrite_use_address): Rewrite address uses in calls that were
11539 identified by find_address_like_use.
11541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11542 Alan Hayward <alan.hayward@arm.com>
11543 David Sherwood <david.sherwood@arm.com>
11545 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11547 * gimple-expr.h (is_gimple_addressable: Likewise.
11548 * gimple-expr.c (is_gimple_address): Likewise.
11549 * internal-fn.c (expand_call_mem_ref): New function.
11550 (expand_mask_load_optab_fn): Use it.
11551 (expand_mask_store_optab_fn): Likewise.
11553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11554 Alan Hayward <alan.hayward@arm.com>
11555 David Sherwood <david.sherwood@arm.com>
11557 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11558 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11559 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11560 (cond_umax@var{mode}): Document.
11561 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11562 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11563 (cond_umin_optab, cond_umax_optab): New optabs.
11564 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11565 (COND_IOR, COND_XOR): New internal functions.
11566 * internal-fn.h (get_conditional_internal_fn): Declare.
11567 * internal-fn.c (cond_binary_direct): New macro.
11568 (expand_cond_binary_optab_fn): Likewise.
11569 (direct_cond_binary_optab_supported_p): Likewise.
11570 (get_conditional_internal_fn): New function.
11571 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11572 Cope with reduction statements that are vectorized as calls rather
11574 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11575 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11576 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11577 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11578 (UNSPEC_COND_EOR): New unspecs.
11579 (optab): Add mappings for them.
11580 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11581 (sve_int_op, sve_fp_op): New int attributes.
11583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11584 Alan Hayward <alan.hayward@arm.com>
11585 David Sherwood <david.sherwood@arm.com>
11587 * optabs.def (while_ult_optab): New optab.
11588 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11589 * internal-fn.def (WHILE_ULT): New internal function.
11590 * internal-fn.h (direct_internal_fn_supported_p): New override
11591 that takes two types as argument.
11592 * internal-fn.c (while_direct): New macro.
11593 (expand_while_optab_fn): New function.
11594 (convert_optab_supported_p): Likewise.
11595 (direct_while_optab_supported_p): New macro.
11596 * wide-int.h (wi::udiv_ceil): New function.
11597 * tree-vectorizer.h (rgroup_masks): New structure.
11598 (vec_loop_masks): New typedef.
11599 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11600 and fully_masked_p.
11601 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11602 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11603 (vect_max_vf): New function.
11604 (slpeel_make_loop_iterate_ntimes): Delete.
11605 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11606 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11607 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11608 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11609 internal-fn.h, stor-layout.h and optabs-query.h.
11610 (vect_set_loop_mask): New function.
11611 (add_preheader_seq): Likewise.
11612 (add_header_seq): Likewise.
11613 (interleave_supported_p): Likewise.
11614 (vect_maybe_permute_loop_masks): Likewise.
11615 (vect_set_loop_masks_directly): Likewise.
11616 (vect_set_loop_condition_masked): Likewise.
11617 (vect_set_loop_condition_unmasked): New function, split out from
11618 slpeel_make_loop_iterate_ntimes.
11619 (slpeel_make_loop_iterate_ntimes): Rename to..
11620 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11621 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11622 (vect_do_peeling): Update call accordingly.
11623 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11625 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11626 mask_compare_type, can_fully_mask_p and fully_masked_p.
11627 (release_vec_loop_masks): New function.
11628 (_loop_vec_info): Use it to free the loop masks.
11629 (can_produce_all_loop_masks_p): New function.
11630 (vect_get_max_nscalars_per_iter): Likewise.
11631 (vect_verify_full_masking): Likewise.
11632 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11633 retries, and free the mask rgroups before retrying. Check loop-wide
11634 reasons for disallowing fully-masked loops. Make the final decision
11635 about whether use a fully-masked loop or not.
11636 (vect_estimate_min_profitable_iters): Do not assume that peeling
11637 for the number of iterations will be needed for fully-masked loops.
11638 (vectorizable_reduction): Disable fully-masked loops.
11639 (vectorizable_live_operation): Likewise.
11640 (vect_halve_mask_nunits): New function.
11641 (vect_double_mask_nunits): Likewise.
11642 (vect_record_loop_mask): Likewise.
11643 (vect_get_loop_mask): Likewise.
11644 (vect_transform_loop): Handle the case in which the final loop
11645 iteration might handle a partial vector. Call vect_set_loop_condition
11646 instead of slpeel_make_loop_iterate_ntimes.
11647 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11648 (check_load_store_masking): New function.
11649 (prepare_load_store_mask): Likewise.
11650 (vectorizable_store): Handle fully-masked loops.
11651 (vectorizable_load): Likewise.
11652 (supportable_widening_operation): Use vect_halve_mask_nunits for
11654 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11655 (vect_gen_while): New function.
11656 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11657 (aarch64_uqdec<mode>): New insn.
11659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11660 Alan Hayward <alan.hayward@arm.com>
11661 David Sherwood <david.sherwood@arm.com>
11663 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11664 (reduc_xor_scal_optab): New optabs.
11665 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11666 (reduc_xor_scal_@var{m}): Document.
11667 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11668 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11669 internal functions.
11670 * fold-const-call.c (fold_const_call): Handle them.
11671 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11672 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11673 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11674 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11675 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11676 (UNSPEC_XORV): New unspecs.
11677 (optab): Add entries for them.
11678 (BITWISEV): New int iterator.
11679 (bit_reduc_op): New int attributes.
11681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11682 Alan Hayward <alan.hayward@arm.com>
11683 David Sherwood <david.sherwood@arm.com>
11685 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11686 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11687 * optabs.def (vec_shl_insert_optab): New optab.
11688 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11689 (duplicate_and_interleave): Likewise.
11690 * tree-vect-loop.c: Include internal-fn.h.
11691 (neutral_op_for_slp_reduction): New function, split out from
11692 get_initial_defs_for_reduction.
11693 (get_initial_def_for_reduction): Handle option 2 for variable-length
11694 vectors by loading the neutral value into a vector and then shifting
11695 the initial value into element 0.
11696 (get_initial_defs_for_reduction): Replace the code argument with
11697 the neutral value calculated by neutral_op_for_slp_reduction.
11698 Use gimple_build_vector for constant-length vectors.
11699 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11700 but the first group_size elements have a neutral value.
11701 Use duplicate_and_interleave otherwise.
11702 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11703 Update call to get_initial_defs_for_reduction. Handle SLP
11704 reductions for variable-length vectors by creating one vector
11705 result for each scalar result, with the elements associated
11706 with other scalar results stubbed out with the neutral value.
11707 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11708 Require IFN_VEC_SHL_INSERT for double reductions on
11709 variable-length vectors, or SLP reductions that have
11710 a neutral value. Require can_duplicate_and_interleave_p
11711 support for variable-length unchained SLP reductions if there
11712 is no neutral value, such as for MIN/MAX reductions. Also require
11713 the number of vector elements to be a multiple of the number of
11714 SLP statements when doing variable-length unchained SLP reductions.
11715 Update call to vect_create_epilog_for_reduction.
11716 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11717 and remove initial values.
11718 (duplicate_and_interleave): Make public.
11719 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11720 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11723 Alan Hayward <alan.hayward@arm.com>
11724 David Sherwood <david.sherwood@arm.com>
11726 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11727 (can_duplicate_and_interleave_p): New function.
11728 (vect_get_and_check_slp_defs): Take the vector of statements
11729 rather than just the current one. Remove excess parentheses.
11730 Restriction rejectinon of vect_constant_def and vect_external_def
11731 for variable-length vectors to boolean types, or types for which
11732 can_duplicate_and_interleave_p is false.
11733 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11734 (duplicate_and_interleave): New function.
11735 (vect_get_constant_vectors): Use gimple_build_vector for
11736 constant-length vectors and suitable variable-length constant
11737 vectors. Use duplicate_and_interleave for other variable-length
11738 vectors. Don't defer the update when inserting new statements.
11740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11741 Alan Hayward <alan.hayward@arm.com>
11742 David Sherwood <david.sherwood@arm.com>
11744 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11745 min_profitable_iters doesn't go negative.
11747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11748 Alan Hayward <alan.hayward@arm.com>
11749 David Sherwood <david.sherwood@arm.com>
11751 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11752 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11753 * optabs.def (vec_mask_load_lanes_optab): New optab.
11754 (vec_mask_store_lanes_optab): Likewise.
11755 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11756 (MASK_STORE_LANES): Likewise.
11757 * internal-fn.c (mask_load_lanes_direct): New macro.
11758 (mask_store_lanes_direct): Likewise.
11759 (expand_mask_load_optab_fn): Handle masked operations.
11760 (expand_mask_load_lanes_optab_fn): New macro.
11761 (expand_mask_store_optab_fn): Handle masked operations.
11762 (expand_mask_store_lanes_optab_fn): New macro.
11763 (direct_mask_load_lanes_optab_supported_p): Likewise.
11764 (direct_mask_store_lanes_optab_supported_p): Likewise.
11765 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11767 (vect_load_lanes_supported): Likewise.
11768 * tree-vect-data-refs.c (strip_conversion): New function.
11769 (can_group_stmts_p): Likewise.
11770 (vect_analyze_data_ref_accesses): Use it instead of checking
11771 for a pair of assignments.
11772 (vect_store_lanes_supported): Take a masked_p parameter.
11773 (vect_load_lanes_supported): Likewise.
11774 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11775 vect_store_lanes_supported and vect_load_lanes_supported.
11776 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11777 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11778 parameter. Don't allow gaps for masked accesses.
11779 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11780 and vect_load_lanes_supported.
11781 (get_load_store_type): Take a masked_p parameter and update
11782 call to get_group_load_store_type.
11783 (vectorizable_store): Update call to get_load_store_type.
11784 Handle IFN_MASK_STORE_LANES.
11785 (vectorizable_load): Update call to get_load_store_type.
11786 Handle IFN_MASK_LOAD_LANES.
11788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11789 Alan Hayward <alan.hayward@arm.com>
11790 David Sherwood <david.sherwood@arm.com>
11792 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11794 * config/aarch64/aarch64-protos.h
11795 (aarch64_sve_struct_memory_operand_p): Declare.
11796 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11797 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11798 (VPRED, vpred): Handle SVE structure modes.
11799 * config/aarch64/constraints.md (Utx): New constraint.
11800 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11801 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11802 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11803 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11804 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11805 structure modes. Split into pieces after RA.
11806 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11807 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11809 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11810 SVE structure modes.
11811 (aarch64_classify_address): Likewise.
11812 (sizetochar): Move earlier in file.
11813 (aarch64_print_operand): Handle SVE register lists.
11814 (aarch64_array_mode): New function.
11815 (aarch64_sve_struct_memory_operand_p): Likewise.
11816 (TARGET_ARRAY_MODE): Redefine.
11818 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11819 Alan Hayward <alan.hayward@arm.com>
11820 David Sherwood <david.sherwood@arm.com>
11822 * target.def (array_mode): New target hook.
11823 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11824 * doc/tm.texi: Regenerate.
11825 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11826 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11827 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11828 targetm.array_mode.
11829 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11833 Alan Hayward <alan.hayward@arm.com>
11834 David Sherwood <david.sherwood@arm.com>
11836 * fold-const.c (fold_binary_loc): Check the argument types
11837 rather than the result type when testing for a vector operation.
11839 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11841 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11842 * doc/tm.texi: Regenerate.
11844 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11845 Alan Hayward <alan.hayward@arm.com>
11846 David Sherwood <david.sherwood@arm.com>
11848 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11849 (sve): Document new AArch64 extension.
11850 * doc/md.texi (w): Extend the description of the AArch64
11851 constraint to include SVE vectors.
11852 (Upl, Upa): Document new AArch64 predicate constraints.
11853 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11855 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11856 (msve-vector-bits=): New option.
11857 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11858 SVE when these are disabled.
11859 (sve): New extension.
11860 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11861 modes. Adjust their number of units based on aarch64_sve_vg.
11862 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11863 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11864 aarch64_addr_query_type.
11865 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11866 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11867 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11868 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11869 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11870 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11871 (aarch64_simd_imm_zero_p): Delete.
11872 (aarch64_check_zero_based_sve_index_immediate): Declare.
11873 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11874 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11875 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11876 (aarch64_sve_float_mul_immediate_p): Likewise.
11877 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11878 rather than an rtx.
11879 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11880 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11881 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11882 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11883 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11884 (aarch64_regmode_natural_size): Likewise.
11885 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11886 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11888 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11889 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11890 for VG and the SVE predicate registers.
11891 (V_ALIASES): Add a "z"-prefixed alias.
11892 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11893 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11894 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11895 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11896 (REG_CLASS_NAMES): Add entries for them.
11897 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11898 and the predicate registers.
11899 (aarch64_sve_vg): Declare.
11900 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11901 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11902 (REGMODE_NATURAL_SIZE): Define.
11903 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11905 * config/aarch64/aarch64.c: Include cfgrtl.h.
11906 (simd_immediate_info): Add a constructor for series vectors,
11907 and an associated step field.
11908 (aarch64_sve_vg): New variable.
11909 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11910 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11911 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11912 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11913 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11914 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11915 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11916 (aarch64_get_mask_mode): New functions.
11917 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11918 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11919 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11920 predicate modes and predicate registers. Explicitly restrict
11921 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11922 to store a vector mode if it is recognized by
11923 aarch64_classify_vector_mode.
11924 (aarch64_regmode_natural_size): New function.
11925 (aarch64_hard_regno_caller_save_mode): Return the original mode
11927 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11928 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11929 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11930 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11932 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11933 does not overlap dest if the function is frame-related. Handle
11935 (aarch64_split_add_offset): New function.
11936 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11937 them aarch64_add_offset.
11938 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11939 and update call to aarch64_sub_sp.
11940 (aarch64_add_cfa_expression): New function.
11941 (aarch64_expand_prologue): Pass extra temporary registers to the
11942 functions above. Handle the case in which we need to emit new
11943 DW_CFA_expressions for registers that were originally saved
11944 relative to the stack pointer, but now have to be expressed
11945 relative to the frame pointer.
11946 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11948 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11949 IP0 and IP1 values for SVE frames.
11950 (aarch64_expand_vec_series): New function.
11951 (aarch64_expand_sve_widened_duplicate): Likewise.
11952 (aarch64_expand_sve_const_vector): Likewise.
11953 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11954 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11955 into the register, rather than emitting a SET directly.
11956 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11957 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11958 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11959 (offset_9bit_signed_scaled_p): New functions.
11960 (aarch64_replicate_bitmask_imm): New function.
11961 (aarch64_bitmask_imm): Use it.
11962 (aarch64_cannot_force_const_mem): Reject expressions involving
11963 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11964 (aarch64_classify_index): Handle SVE indices, by requiring
11965 a plain register index with a scale that matches the element size.
11966 (aarch64_classify_address): Handle SVE addresses. Assert that
11967 the mode of the address is VOIDmode or an integer mode.
11968 Update call to aarch64_classify_symbol.
11969 (aarch64_classify_symbolic_expression): Update call to
11970 aarch64_classify_symbol.
11971 (aarch64_const_vec_all_in_range_p): New function.
11972 (aarch64_print_vector_float_operand): Likewise.
11973 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11974 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11975 and the FP immediates 1.0 and 0.5.
11976 (aarch64_print_address_internal): Handle SVE addresses.
11977 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11978 (aarch64_regno_regclass): Handle predicate registers.
11979 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11981 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11982 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11983 (aarch64_convert_sve_vector_bits): New function.
11984 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11985 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11986 rather than an rtx.
11987 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11988 Handle SVE vector and predicate modes. Accept VL-based constants
11989 that need only one temporary register, and VL offsets that require
11990 no temporary registers.
11991 (aarch64_conditional_register_usage): Mark the predicate registers
11992 as fixed if SVE isn't available.
11993 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11994 Return true for SVE vector and predicate modes.
11995 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11996 rather than an unsigned int. Handle SVE modes.
11997 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11999 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
12001 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
12002 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
12003 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
12004 (aarch64_sve_float_mul_immediate_p): New functions.
12005 (aarch64_sve_valid_immediate): New function.
12006 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
12007 Explicitly reject structure modes. Check for INDEX constants.
12008 Handle PTRUE and PFALSE constants.
12009 (aarch64_check_zero_based_sve_index_immediate): New function.
12010 (aarch64_simd_imm_zero_p): Delete.
12011 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
12012 vector modes. Accept constants in the range of CNT[BHWD].
12013 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
12014 ask for an Advanced SIMD mode.
12015 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
12016 (aarch64_simd_vector_alignment): Handle SVE predicates.
12017 (aarch64_vectorize_preferred_vector_alignment): New function.
12018 (aarch64_simd_vector_alignment_reachable): Use it instead of
12020 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
12021 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
12023 (MAX_VECT_LEN): Delete.
12024 (expand_vec_perm_d): Add a vec_flags field.
12025 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
12026 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
12027 (aarch64_evpc_ext): Don't apply a big-endian lane correction
12029 (aarch64_evpc_rev): Rename to...
12030 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
12031 (aarch64_evpc_rev_global): New function.
12032 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
12033 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
12035 (aarch64_evpc_sve_tbl): New function.
12036 (aarch64_expand_vec_perm_const_1): Update after rename of
12037 aarch64_evpc_rev. Handle SVE permutes too, trying
12038 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12039 than aarch64_evpc_tbl.
12040 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12041 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12042 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12043 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12044 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12045 (aarch64_expand_sve_vcond): New functions.
12046 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12047 of aarch64_vector_mode_p.
12048 (aarch64_dwarf_poly_indeterminate_value): New function.
12049 (aarch64_compute_pressure_classes): Likewise.
12050 (aarch64_can_change_mode_class): Likewise.
12051 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12052 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12053 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12054 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12055 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12056 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12057 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12058 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12060 (Dn, Dl, Dr): Accept const as well as const_vector.
12061 (Dz): Likewise. Compare against CONST0_RTX.
12062 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12063 of "vector" where appropriate.
12064 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12065 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12066 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12067 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12068 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12069 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12070 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12071 (v_int_equiv): Extend to SVE modes.
12072 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12074 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12075 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12076 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12077 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12078 (SVE_COND_FP_CMP): New int iterators.
12079 (perm_hilo): Handle the new unpack unspecs.
12080 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12082 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12083 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12084 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12085 (aarch64_equality_operator, aarch64_constant_vector_operand)
12086 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12087 (aarch64_sve_nonimmediate_operand): Likewise.
12088 (aarch64_sve_general_operand): Likewise.
12089 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12090 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12091 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12092 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12093 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12094 (aarch64_sve_float_arith_immediate): Likewise.
12095 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12096 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12097 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12098 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12099 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12100 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12101 (aarch64_sve_float_arith_operand): Likewise.
12102 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12103 (aarch64_sve_float_mul_operand): Likewise.
12104 (aarch64_sve_vec_perm_operand): Likewise.
12105 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12106 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12107 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12108 as well as const_vector.
12109 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12110 in file. Use CONST0_RTX and CONSTM1_RTX.
12111 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12112 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12113 Use aarch64_simd_imm_zero.
12114 * config/aarch64/aarch64-sve.md: New file.
12115 * config/aarch64/aarch64.md: Include it.
12116 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12117 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12118 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12119 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12120 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12121 (sve): New attribute.
12122 (enabled): Disable instructions with the sve attribute unless
12124 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12125 aarch64_expand_mov_immediate.
12126 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12127 CNT[BHSD] immediates.
12128 (movti): Split CONST_POLY_INT moves into two halves.
12129 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12130 Split additions that need a temporary here if the destination
12131 is the stack pointer.
12132 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12133 (*add<mode>3_poly_1): New instruction.
12134 (set_clobber_cc): New expander.
12136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12138 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12139 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12140 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12141 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12142 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12143 Change innermode from fixed_mode_size to machine_mode.
12144 (simplify_subreg): Update call accordingly. Handle a constant-sized
12145 subreg of a variable-length CONST_VECTOR.
12147 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12148 Alan Hayward <alan.hayward@arm.com>
12149 David Sherwood <david.sherwood@arm.com>
12151 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12152 (add_offset_to_base): New function, split out from...
12153 (create_mem_ref): ...here. When handling a scale other than 1,
12154 check first whether the address is valid without the offset.
12155 Add it into the base if so, leaving the index and scale as-is.
12157 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12160 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12161 fold_for_warn before checking if arg2 is INTEGER_CST.
12163 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12165 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12166 (store_multiple_operation): Delete.
12167 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12168 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12169 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12170 guarded by TARGET_STRING.
12171 (rs6000_output_load_multiple): Delete.
12172 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12173 OPTION_MASK_STRING / TARGET_STRING handling.
12174 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12175 (const rs6000_opt_masks) <"string">: Change mask to 0.
12176 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12177 (MASK_STRING): Delete.
12178 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12180 (load_multiple): Delete.
12187 (store_multiple): Delete.
12194 (movmemsi_8reg): Delete.
12195 (corresponding unnamed define_insn): Delete.
12196 (movmemsi_6reg): Delete.
12197 (corresponding unnamed define_insn): Delete.
12198 (movmemsi_4reg): Delete.
12199 (corresponding unnamed define_insn): Delete.
12200 (movmemsi_2reg): Delete.
12201 (corresponding unnamed define_insn): Delete.
12202 (movmemsi_1reg): Delete.
12203 (corresponding unnamed define_insn): Delete.
12204 * config/rs6000/rs6000.opt (mno-string): New.
12205 (mstring): Replace by deprecation warning stub.
12206 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12208 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12210 * regrename.c (regrename_do_replace): If replacing the same
12211 reg multiple times, try to reuse last created gen_raw_REG.
12214 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12215 main to workaround a bug in GDB.
12217 2018-01-12 Tom de Vries <tom@codesourcery.com>
12220 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12222 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12224 PR rtl-optimization/80481
12225 * ira-color.c (get_cap_member): New function.
12226 (allocnos_conflict_by_live_ranges_p): Use it.
12227 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12228 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12230 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12233 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12234 (*saddl_se_1): Ditto.
12235 (*ssubsi_1): Ditto.
12236 (*ssubl_se_1): Ditto.
12238 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12240 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12241 rather than wi::to_widest for DR_INITs.
12242 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12243 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12244 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12246 (vect_analyze_group_access_1): Note that here.
12248 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12250 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12251 polynomial type sizes.
12253 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12255 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12256 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12257 (gimple_add_tmp_var): Likewise.
12259 2018-01-12 Martin Liska <mliska@suse.cz>
12261 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12262 (gimple_alloc_sizes): Likewise.
12263 (dump_gimple_statistics): Use PRIu64 in printf format.
12264 * gimple.h: Change uint64_t to int.
12266 2018-01-12 Martin Liska <mliska@suse.cz>
12268 * tree-core.h: Use uint64_t instead of int.
12269 * tree.c (tree_node_counts): Likewise.
12270 (tree_node_sizes): Likewise.
12271 (dump_tree_statistics): Use PRIu64 in printf format.
12273 2018-01-12 Martin Liska <mliska@suse.cz>
12275 * Makefile.in: As qsort_chk is implemented in vec.c, add
12276 vec.o to linkage of gencfn-macros.
12277 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12278 passing the info to record_node_allocation_statistics.
12279 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12281 * ggc-common.c (struct ggc_usage): Add operator== and use
12282 it in operator< and compare function.
12283 * mem-stats.h (struct mem_usage): Likewise.
12284 * vec.c (struct vec_usage): Remove operator< and compare
12285 function. Can be simply inherited.
12287 2018-01-12 Martin Jambor <mjambor@suse.cz>
12290 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12291 * tree-ssa-math-opts.c: Include domwalk.h.
12292 (convert_mult_to_fma_1): New function.
12293 (fma_transformation_info): New type.
12294 (fma_deferring_state): Likewise.
12295 (cancel_fma_deferring): New function.
12296 (result_of_phi): Likewise.
12297 (last_fma_candidate_feeds_initial_phi): Likewise.
12298 (convert_mult_to_fma): Added deferring logic, split actual
12299 transformation to convert_mult_to_fma_1.
12300 (math_opts_dom_walker): New type.
12301 (math_opts_dom_walker::after_dom_children): New method, body moved
12302 here from pass_optimize_widening_mul::execute, added deferring logic
12304 (pass_optimize_widening_mul::execute): Moved most of code to
12305 math_opts_dom_walker::after_dom_children.
12306 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12307 * config/i386/i386.c (ix86_option_override_internal): Added
12308 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12310 2018-01-12 Richard Biener <rguenther@suse.de>
12313 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12314 inline instance vars.
12316 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12319 * config/rx/rx.c (rx_is_restricted_memory_address):
12320 Handle SUBREG case.
12322 2018-01-12 Richard Biener <rguenther@suse.de>
12324 PR tree-optimization/80846
12325 * target.def (split_reduction): New target hook.
12326 * targhooks.c (default_split_reduction): New function.
12327 * targhooks.h (default_split_reduction): Declare.
12328 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12329 target requests first reduce vectors by combining low and high
12331 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12332 (get_vectype_for_scalar_type_and_size): Export.
12333 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12334 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12335 * doc/tm.texi: Regenerate.
12336 * config/i386/i386.c (ix86_split_reduction): Implement
12337 TARGET_VECTORIZE_SPLIT_REDUCTION.
12339 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12342 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12343 in PIC mode except for TARGET_VXWORKS_RTP.
12344 * config/sparc/sparc.c: Include cfgrtl.h.
12345 (TARGET_INIT_PIC_REG): Define.
12346 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12347 (sparc_pic_register_p): New predicate.
12348 (sparc_legitimate_address_p): Use it.
12349 (sparc_legitimize_pic_address): Likewise.
12350 (sparc_delegitimize_address): Likewise.
12351 (sparc_mode_dependent_address_p): Likewise.
12352 (gen_load_pcrel_sym): Remove 4th parameter.
12353 (load_got_register): Adjust call to above. Remove obsolete stuff.
12354 (sparc_expand_prologue): Do not call load_got_register here.
12355 (sparc_flat_expand_prologue): Likewise.
12356 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12357 (sparc_use_pseudo_pic_reg): New function.
12358 (sparc_init_pic_reg): Likewise.
12359 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12360 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12362 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12364 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12365 Add item for branch_cost.
12367 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12369 PR rtl-optimization/83565
12370 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12371 not extend the result to a larger mode for rotate operations.
12372 (num_sign_bit_copies1): Likewise.
12374 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12377 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12379 Use values-Xc.o for -pedantic.
12380 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12382 2018-01-12 Martin Liska <mliska@suse.cz>
12385 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12387 (possible_polymorphic_call_targets): Use it.
12388 (ipa_devirt): Likewise.
12390 2018-01-12 Martin Liska <mliska@suse.cz>
12392 * profile-count.h (enum profile_quality): Use 0 as invalid
12393 enum value of profile_quality.
12395 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12397 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12398 -mext-string options.
12400 2018-01-12 Richard Biener <rguenther@suse.de>
12402 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12403 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12404 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12406 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12408 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12410 * configure.ac (--with-long-double-format): Add support for the
12411 configuration option to change the default long double format on
12413 * config.gcc (powerpc*-linux*-*): Likewise.
12414 * configure: Regenerate.
12415 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12416 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12417 used without modification.
12419 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12421 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12422 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12423 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12424 MISC_BUILTIN_SPEC_BARRIER.
12425 (rs6000_init_builtins): Likewise.
12426 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12428 (speculation_barrier): New define_insn.
12429 * doc/extend.texi: Document __builtin_speculation_barrier.
12431 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12434 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12435 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12436 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12438 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12439 integral modes instead of "ss" and "sd".
12440 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12441 vectors with 32-bit and 64-bit elements.
12442 (vecdupssescalarmodesuffix): New mode attribute.
12443 (vec_dup<mode>): Use it.
12445 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12448 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12449 frame if argument is passed on stack.
12451 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12454 * ree.c (combine_reaching_defs): Optimize also
12455 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12456 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12458 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12460 PR middle-end/83189
12461 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12463 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12465 PR middle-end/83718
12466 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12467 after they are computed.
12469 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12471 PR tree-optimization/83695
12472 * gimple-loop-linterchange.cc
12473 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12474 reset cached scev information after interchange.
12475 (pass_linterchange::execute): Remove call to scev_reset_htab.
12477 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12479 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12480 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12481 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12482 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12483 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12484 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12485 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12486 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12487 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12488 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12489 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12490 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12491 (V_lane_reg): Likewise.
12492 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12494 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12495 (vfmal_lane_low<mode>_intrinsic,
12496 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12497 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12498 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12499 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12500 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12501 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12503 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12505 * config/arm/arm-cpus.in (fp16fml): New feature.
12506 (ALL_SIMD): Add fp16fml.
12507 (armv8.2-a): Add fp16fml as an option.
12508 (armv8.3-a): Likewise.
12509 (armv8.4-a): Add fp16fml as part of fp16.
12510 * config/arm/arm.h (TARGET_FP16FML): Define.
12511 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12513 * config/arm/arm-modes.def (V2HF): Define.
12514 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12515 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12516 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12517 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12518 vfmsl_low, vfmsl_high): New set of builtins.
12519 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12520 (vfml_op): New code attribute.
12521 (VFMLHALVES): New int iterator.
12522 (VFML, VFMLSEL): New mode attributes.
12523 (V_reg): Define mapping for V2HF.
12524 (V_hi, V_lo): New mode attributes.
12525 (VF_constraint): Likewise.
12526 (vfml_half, vfml_half_selector): New int attributes.
12527 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12529 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12530 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12532 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12533 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12534 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12535 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12537 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12538 Document new effective target and option set.
12540 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12542 * config/arm/arm-cpus.in (armv8_4): New feature.
12543 (ARMv8_4a): New fgroup.
12544 (armv8.4-a): New arch.
12545 * config/arm/arm-tables.opt: Regenerate.
12546 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12547 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12548 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12549 Add matching rules for -march=armv8.4-a and extensions.
12550 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12552 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12555 * config/rx/rx.md (BW): New mode attribute.
12556 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12558 2018-01-11 Richard Biener <rguenther@suse.de>
12560 PR tree-optimization/83435
12561 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12562 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12563 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12565 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12571 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12572 (aarch64_print_address_internal): Use it to check for a zero offset.
12574 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12575 Alan Hayward <alan.hayward@arm.com>
12576 David Sherwood <david.sherwood@arm.com>
12578 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12579 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12580 Return a poly_int64 rather than a HOST_WIDE_INT.
12581 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12582 rather than a HOST_WIDE_INT.
12583 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12584 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12585 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12586 final_offset from HOST_WIDE_INT to poly_int64.
12587 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12588 to_constant when getting the number of units in an Advanced SIMD
12590 (aarch64_builtin_vectorized_function): Check for a constant number
12592 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12594 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12595 attribute instead of GET_MODE_NUNITS.
12596 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12597 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12598 GET_MODE_SIZE for fixed-size registers.
12599 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12600 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12601 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12602 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12603 (aarch64_print_operand, aarch64_print_address_internal)
12604 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12605 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12606 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12607 Handle polynomial GET_MODE_SIZE.
12608 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12609 wider than SImode without modification.
12610 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12611 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12612 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12613 passing and returning SVE modes.
12614 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12615 rather than GEN_INT.
12616 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12617 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12618 (aarch64_allocate_and_probe_stack_space): Likewise.
12619 (aarch64_layout_frame): Cope with polynomial offsets.
12620 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12621 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12622 polynomial offsets.
12623 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12624 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12625 poly_int64 rather than a HOST_WIDE_INT.
12626 (aarch64_get_separate_components, aarch64_process_components)
12627 (aarch64_expand_prologue, aarch64_expand_epilogue)
12628 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12629 (aarch64_anchor_offset): New function, split out from...
12630 (aarch64_legitimize_address): ...here.
12631 (aarch64_builtin_vectorization_cost): Handle polynomial
12632 TYPE_VECTOR_SUBPARTS.
12633 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12635 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12636 number of elements from the PARALLEL rather than the mode.
12637 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12638 rather than GET_MODE_BITSIZE.
12639 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12640 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12641 (aarch64_expand_vec_perm_const_1): Handle polynomial
12642 d->perm.length () and d->perm elements.
12643 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12644 Apply to_constant to d->perm elements.
12645 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12646 polynomial CONST_VECTOR_NUNITS.
12647 (aarch64_move_pointer): Take amount as a poly_int64 rather
12649 (aarch64_progress_pointer): Avoid temporary variable.
12650 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12651 the mode attribute instead of GET_MODE.
12653 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12654 Alan Hayward <alan.hayward@arm.com>
12655 David Sherwood <david.sherwood@arm.com>
12657 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12658 x exists before using it.
12659 (aarch64_add_constant_internal): Rename to...
12660 (aarch64_add_offset_1): ...this. Replace regnum with separate
12661 src and dest rtxes. Handle the case in which they're different,
12662 including when the offset is zero. Replace scratchreg with an rtx.
12663 Use 2 additions if there is no spare register into which we can
12664 move a 16-bit constant.
12665 (aarch64_add_constant): Delete.
12666 (aarch64_add_offset): Replace reg with separate src and dest
12667 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12668 Use aarch64_add_offset_1.
12669 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12670 an rtx rather than an int. Take the delta as a poly_int64
12671 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12672 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12673 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12674 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12675 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12676 and aarch64_add_sp.
12677 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12678 aarch64_add_constant.
12680 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12682 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12683 Use scalar_float_mode.
12685 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12687 * config/aarch64/aarch64-simd.md
12688 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12689 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12690 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12691 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12692 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12693 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12694 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12695 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12696 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12697 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12699 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12702 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12703 targ_options->x_arm_arch_string is non NULL.
12705 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12707 * config/aarch64/aarch64.h
12708 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12710 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12713 * expmed.c (emit_store_flag_force): Swap if const op0
12714 and change VOIDmode to mode of op0.
12716 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12718 PR rtl-optimization/83761
12719 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12720 than bytes to mode_for_size.
12722 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12724 PR middle-end/83189
12725 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12726 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12729 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12731 PR middle-end/83575
12732 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12733 when in layout mode.
12734 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12735 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12738 2018-01-10 Michael Collison <michael.collison@arm.com>
12740 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12741 * config/aarch64/aarch64-option-extension.def: Add
12742 AARCH64_OPT_EXTENSION of 'fp16fml'.
12743 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12744 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12745 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12746 * config/aarch64/constraints.md (Ui7): New constraint.
12747 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12748 (VFMLA_SEL_W): Ditto.
12751 (VFMLA16_LOW): New int iterator.
12752 (VFMLA16_HIGH): Ditto.
12753 (UNSPEC_FMLAL): New unspec.
12754 (UNSPEC_FMLSL): Ditto.
12755 (UNSPEC_FMLAL2): Ditto.
12756 (UNSPEC_FMLSL2): Ditto.
12757 (f16mac): New code attribute.
12758 * config/aarch64/aarch64-simd-builtins.def
12759 (aarch64_fmlal_lowv2sf): Ditto.
12760 (aarch64_fmlsl_lowv2sf): Ditto.
12761 (aarch64_fmlalq_lowv4sf): Ditto.
12762 (aarch64_fmlslq_lowv4sf): Ditto.
12763 (aarch64_fmlal_highv2sf): Ditto.
12764 (aarch64_fmlsl_highv2sf): Ditto.
12765 (aarch64_fmlalq_highv4sf): Ditto.
12766 (aarch64_fmlslq_highv4sf): Ditto.
12767 (aarch64_fmlal_lane_lowv2sf): Ditto.
12768 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12769 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12770 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12771 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12772 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12773 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12774 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12775 (aarch64_fmlal_lane_highv2sf): Ditto.
12776 (aarch64_fmlsl_lane_highv2sf): Ditto.
12777 (aarch64_fmlal_laneq_highv2sf): Ditto.
12778 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12779 (aarch64_fmlalq_lane_highv4sf): Ditto.
12780 (aarch64_fmlsl_lane_highv4sf): Ditto.
12781 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12782 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12783 * config/aarch64/aarch64-simd.md:
12784 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12785 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12786 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12787 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12788 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12789 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12790 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12791 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12792 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12793 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12794 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12795 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12796 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12797 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12798 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12799 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12800 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12801 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12802 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12803 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12804 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12805 (vfmlsl_low_u32): Ditto.
12806 (vfmlalq_low_u32): Ditto.
12807 (vfmlslq_low_u32): Ditto.
12808 (vfmlal_high_u32): Ditto.
12809 (vfmlsl_high_u32): Ditto.
12810 (vfmlalq_high_u32): Ditto.
12811 (vfmlslq_high_u32): Ditto.
12812 (vfmlal_lane_low_u32): Ditto.
12813 (vfmlsl_lane_low_u32): Ditto.
12814 (vfmlal_laneq_low_u32): Ditto.
12815 (vfmlsl_laneq_low_u32): Ditto.
12816 (vfmlalq_lane_low_u32): Ditto.
12817 (vfmlslq_lane_low_u32): Ditto.
12818 (vfmlalq_laneq_low_u32): Ditto.
12819 (vfmlslq_laneq_low_u32): Ditto.
12820 (vfmlal_lane_high_u32): Ditto.
12821 (vfmlsl_lane_high_u32): Ditto.
12822 (vfmlal_laneq_high_u32): Ditto.
12823 (vfmlsl_laneq_high_u32): Ditto.
12824 (vfmlalq_lane_high_u32): Ditto.
12825 (vfmlslq_lane_high_u32): Ditto.
12826 (vfmlalq_laneq_high_u32): Ditto.
12827 (vfmlslq_laneq_high_u32): Ditto.
12828 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12829 (AARCH64_FL_FOR_ARCH8_4): New.
12830 (AARCH64_ISA_F16FML): New ISA flag.
12831 (TARGET_F16FML): New feature flag for fp16fml.
12832 (doc/invoke.texi): Document new fp16fml option.
12834 2018-01-10 Michael Collison <michael.collison@arm.com>
12836 * config/aarch64/aarch64-builtins.c:
12837 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12838 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12839 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12840 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12841 (AARCH64_ISA_SHA3): New ISA flag.
12842 (TARGET_SHA3): New feature flag for sha3.
12843 * config/aarch64/iterators.md (sha512_op): New int attribute.
12844 (CRYPTO_SHA512): New int iterator.
12845 (UNSPEC_SHA512H): New unspec.
12846 (UNSPEC_SHA512H2): Ditto.
12847 (UNSPEC_SHA512SU0): Ditto.
12848 (UNSPEC_SHA512SU1): Ditto.
12849 * config/aarch64/aarch64-simd-builtins.def
12850 (aarch64_crypto_sha512hqv2di): New builtin.
12851 (aarch64_crypto_sha512h2qv2di): Ditto.
12852 (aarch64_crypto_sha512su0qv2di): Ditto.
12853 (aarch64_crypto_sha512su1qv2di): Ditto.
12854 (aarch64_eor3qv8hi): Ditto.
12855 (aarch64_rax1qv2di): Ditto.
12856 (aarch64_xarqv2di): Ditto.
12857 (aarch64_bcaxqv8hi): Ditto.
12858 * config/aarch64/aarch64-simd.md:
12859 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12860 (aarch64_crypto_sha512su0qv2di): Ditto.
12861 (aarch64_crypto_sha512su1qv2di): Ditto.
12862 (aarch64_eor3qv8hi): Ditto.
12863 (aarch64_rax1qv2di): Ditto.
12864 (aarch64_xarqv2di): Ditto.
12865 (aarch64_bcaxqv8hi): Ditto.
12866 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12867 (vsha512h2q_u64): Ditto.
12868 (vsha512su0q_u64): Ditto.
12869 (vsha512su1q_u64): Ditto.
12870 (veor3q_u16): Ditto.
12871 (vrax1q_u64): Ditto.
12872 (vxarq_u64): Ditto.
12873 (vbcaxq_u16): Ditto.
12874 * config/arm/types.md (crypto_sha512): New type attribute.
12875 (crypto_sha3): Ditto.
12876 (doc/invoke.texi): Document new sha3 option.
12878 2018-01-10 Michael Collison <michael.collison@arm.com>
12880 * config/aarch64/aarch64-builtins.c:
12881 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12882 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12883 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12884 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12885 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12886 (AARCH64_ISA_SM4): New ISA flag.
12887 (TARGET_SM4): New feature flag for sm4.
12888 * config/aarch64/aarch64-simd-builtins.def
12889 (aarch64_sm3ss1qv4si): Ditto.
12890 (aarch64_sm3tt1aq4si): Ditto.
12891 (aarch64_sm3tt1bq4si): Ditto.
12892 (aarch64_sm3tt2aq4si): Ditto.
12893 (aarch64_sm3tt2bq4si): Ditto.
12894 (aarch64_sm3partw1qv4si): Ditto.
12895 (aarch64_sm3partw2qv4si): Ditto.
12896 (aarch64_sm4eqv4si): Ditto.
12897 (aarch64_sm4ekeyqv4si): Ditto.
12898 * config/aarch64/aarch64-simd.md:
12899 (aarch64_sm3ss1qv4si): Ditto.
12900 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12901 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12902 (aarch64_sm4eqv4si): Ditto.
12903 (aarch64_sm4ekeyqv4si): Ditto.
12904 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12905 (sm3part_op): Ditto.
12906 (CRYPTO_SM3TT): Ditto.
12907 (CRYPTO_SM3PART): Ditto.
12908 (UNSPEC_SM3SS1): New unspec.
12909 (UNSPEC_SM3TT1A): Ditto.
12910 (UNSPEC_SM3TT1B): Ditto.
12911 (UNSPEC_SM3TT2A): Ditto.
12912 (UNSPEC_SM3TT2B): Ditto.
12913 (UNSPEC_SM3PARTW1): Ditto.
12914 (UNSPEC_SM3PARTW2): Ditto.
12915 (UNSPEC_SM4E): Ditto.
12916 (UNSPEC_SM4EKEY): Ditto.
12917 * config/aarch64/constraints.md (Ui2): New constraint.
12918 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12919 * config/arm/types.md (crypto_sm3): New type attribute.
12920 (crypto_sm4): Ditto.
12921 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12922 (vsm3tt1aq_u32): Ditto.
12923 (vsm3tt1bq_u32): Ditto.
12924 (vsm3tt2aq_u32): Ditto.
12925 (vsm3tt2bq_u32): Ditto.
12926 (vsm3partw1q_u32): Ditto.
12927 (vsm3partw2q_u32): Ditto.
12928 (vsm4eq_u32): Ditto.
12929 (vsm4ekeyq_u32): Ditto.
12930 (doc/invoke.texi): Document new sm4 option.
12932 2018-01-10 Michael Collison <michael.collison@arm.com>
12934 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12935 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12936 (AARCH64_FL_FOR_ARCH8_4): New.
12937 (AARCH64_FL_V8_4): New flag.
12938 (doc/invoke.texi): Document new armv8.4-a option.
12940 2018-01-10 Michael Collison <michael.collison@arm.com>
12942 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12943 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12944 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12945 * config/aarch64/aarch64-option-extension.def: Add
12946 AARCH64_OPT_EXTENSION of 'sha2'.
12947 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12948 (crypto): Disable sha2 and aes if crypto disabled.
12949 (crypto): Enable aes and sha2 if enabled.
12950 (simd): Disable sha2 and aes if simd disabled.
12951 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12953 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12954 (TARGET_SHA2): New feature flag for sha2.
12955 (TARGET_AES): New feature flag for aes.
12956 * config/aarch64/aarch64-simd.md:
12957 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12958 conditional on TARGET_AES.
12959 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12960 (aarch64_crypto_sha1hsi): Make pattern conditional
12962 (aarch64_crypto_sha1hv4si): Ditto.
12963 (aarch64_be_crypto_sha1hv4si): Ditto.
12964 (aarch64_crypto_sha1su1v4si): Ditto.
12965 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12966 (aarch64_crypto_sha1su0v4si): Ditto.
12967 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12968 (aarch64_crypto_sha256su0v4si): Ditto.
12969 (aarch64_crypto_sha256su1v4si): Ditto.
12970 (doc/invoke.texi): Document new aes and sha2 options.
12972 2018-01-10 Martin Sebor <msebor@redhat.com>
12974 PR tree-optimization/83781
12975 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12978 2018-01-11 Martin Sebor <msebor@gmail.com>
12979 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12981 PR tree-optimization/83501
12982 PR tree-optimization/81703
12984 * tree-ssa-strlen.c (get_string_cst): Rename...
12985 (get_string_len): ...to this. Handle global constants.
12986 (handle_char_store): Adjust.
12988 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12989 Jim Wilson <jimw@sifive.com>
12991 * config/riscv/riscv-protos.h (riscv_output_return): New.
12992 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12993 (riscv_attribute_table, riscv_output_return),
12994 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12995 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12996 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12998 (riscv_expand_prologue): Add early return for naked function.
12999 (riscv_expand_epilogue): Likewise.
13000 (riscv_function_ok_for_sibcall): Return false for naked function.
13001 (riscv_set_current_function): New.
13002 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
13003 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
13004 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
13005 * doc/extend.texi (RISC-V Function Attributes): New.
13007 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
13009 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
13010 check for 128-bit long double before checking TCmode.
13011 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
13012 128-bit long doubles before checking TFmode or TCmode.
13013 (FLOAT128_IBM_P): Likewise.
13015 2018-01-10 Martin Sebor <msebor@redhat.com>
13017 PR tree-optimization/83671
13018 * builtins.c (c_strlen): Unconditionally return zero for the empty
13020 Use -Warray-bounds for warnings.
13021 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
13022 for non-constant array indices with COMPONENT_REF, arrays of
13023 arrays, and pointers to arrays.
13024 (gimple_fold_builtin_strlen): Determine and set length range for
13025 non-constant character arrays.
13027 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
13029 PR middle-end/81897
13030 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
13033 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
13035 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
13037 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13040 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13041 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13042 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13043 indexed_or_indirect_operand predicate.
13044 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13045 (*vsx_le_perm_load_v8hi): Likewise.
13046 (*vsx_le_perm_load_v16qi): Likewise.
13047 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13048 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13049 (*vsx_le_perm_store_v8hi): Likewise.
13050 (*vsx_le_perm_store_v16qi): Likewise.
13051 (eight unnamed splitters): Likewise.
13053 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13055 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13056 * config/rs6000/emmintrin.h: Likewise.
13057 * config/rs6000/mmintrin.h: Likewise.
13058 * config/rs6000/xmmintrin.h: Likewise.
13060 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13063 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13065 * tree.c (tree_nop_conversion): Return true for location wrapper
13067 (maybe_wrap_with_location): New function.
13068 (selftest::check_strip_nops): New function.
13069 (selftest::test_location_wrappers): New function.
13070 (selftest::tree_c_tests): Call it.
13071 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13072 (maybe_wrap_with_location): New decl.
13073 (EXPR_LOCATION_WRAPPER_P): New macro.
13074 (location_wrapper_p): New inline function.
13075 (tree_strip_any_location_wrapper): New inline function.
13077 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13080 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13081 stack_realign_offset for the largest alignment of stack slot
13083 (ix86_find_max_used_stack_alignment): New function.
13084 (ix86_finalize_stack_frame_flags): Use it. Set
13085 max_used_stack_alignment if we don't realign stack.
13086 * config/i386/i386.h (machine_function): Add
13087 max_used_stack_alignment.
13089 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13091 * config/arm/arm.opt (-mbranch-cost): New option.
13092 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13095 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13098 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13099 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13101 2018-01-10 Richard Biener <rguenther@suse.de>
13104 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13105 early out so it also covers the case where we have a non-NULL
13108 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13110 PR tree-optimization/83753
13111 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13112 for non-strided grouped accesses if the number of elements is 1.
13114 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13117 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13118 * i386.h (TARGET_USE_GATHER): Define.
13119 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13121 2018-01-10 Martin Liska <mliska@suse.cz>
13124 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13125 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13127 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13128 CLEANUP_NO_PARTITIONING is not set.
13130 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13132 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13133 for vectors, as a partial revert of r254296.
13134 * rtl.h (const_vec_p): Delete.
13135 (const_vec_duplicate_p): Don't test for vector CONSTs.
13136 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13137 * expmed.c (make_tree): Likewise.
13140 * common.md (E, F): Use CONSTANT_P instead of checking for
13142 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13143 checking for CONST_VECTOR.
13145 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13147 PR middle-end/83575
13148 * predict.c (force_edge_cold): Handle in more sane way edges
13149 with no prediction.
13151 2018-01-09 Carl Love <cel@us.ibm.com>
13153 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13155 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13156 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13157 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13158 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13159 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13160 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13161 * config/rs6000/rs6000-protos.h: Add extern defition for
13162 rs6000_generate_float2_double_code.
13163 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13165 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13166 (float2_v2df): Add define_expand.
13168 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13171 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13172 op_mode in the force_to_mode call.
13174 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13176 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13177 instead of checking each element individually.
13178 (aarch64_evpc_uzp): Likewise.
13179 (aarch64_evpc_zip): Likewise.
13180 (aarch64_evpc_ext): Likewise.
13181 (aarch64_evpc_rev): Likewise.
13182 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13183 instead of checking each element individually. Return true without
13185 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13186 whether all selected elements come from the same input, instead of
13187 checking each element individually. Remove calls to gen_rtx_REG,
13188 start_sequence and end_sequence and instead assert that no rtl is
13191 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13193 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13194 order of HIGH and CONST checks.
13196 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13198 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13199 if the destination isn't an SSA_NAME.
13201 2018-01-09 Richard Biener <rguenther@suse.de>
13203 PR tree-optimization/83668
13204 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13206 (canonicalize_loop_form): ... here, renamed from ...
13207 (canonicalize_loop_closed_ssa_form): ... this and amended to
13208 swap successor edges for loop exit blocks to make us use
13209 the RPO order we need for initial schedule generation.
13211 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13213 PR tree-optimization/64811
13214 * match.pd: When optimizing comparisons with Inf, avoid
13215 introducing or losing exceptions from comparisons with NaN.
13217 2018-01-09 Martin Liska <mliska@suse.cz>
13220 * asan.c (shadow_mem_size): Add gcc_assert.
13222 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13224 Don't save registers in main().
13227 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13228 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13229 * config/avr/avr.c (avr_set_current_function): Don't error if
13230 naked, OS_task or OS_main are specified at the same time.
13231 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13233 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13235 * common/config/avr/avr-common.c (avr_option_optimization_table):
13236 Switch on -mmain-is-OS_task for optimizing compilations.
13238 2018-01-09 Richard Biener <rguenther@suse.de>
13240 PR tree-optimization/83572
13241 * graphite.c: Include cfganal.h.
13242 (graphite_transform_loops): Connect infinite loops to exit
13243 and remove fake edges at the end.
13245 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13247 * ipa-inline.c (edge_badness): Revert accidental checkin.
13249 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13252 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13253 symbols; not inline clones.
13255 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13258 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13259 hard registers. Formatting fixes.
13261 PR preprocessor/83722
13262 * gcc.c (try_generate_repro): Pass
13263 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13264 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13267 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13268 Kito Cheng <kito.cheng@gmail.com>
13270 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13271 (riscv_leaf_function_p): Delete.
13272 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13274 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13276 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13278 (do_ifelse): New function.
13279 (do_isel): New function.
13280 (do_sub3): New function.
13281 (do_add3): New function.
13282 (do_load_mask_compare): New function.
13283 (do_overlap_load_compare): New function.
13284 (expand_compare_loop): New function.
13285 (expand_block_compare): Call expand_compare_loop() when appropriate.
13286 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13287 option description.
13288 (-mblock-compare-inline-loop-limit): New option.
13290 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13293 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13294 Reverse order of second and third operands in first alternative.
13295 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13296 of first and second elements in UNSPEC_VPERMR vector.
13297 (altivec_expand_vec_perm_le): Likewise.
13299 2018-01-08 Jeff Law <law@redhat.com>
13301 PR rtl-optimizatin/81308
13302 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13303 (process_switch): If group_case_labels makes a change, then set
13305 (pass_convert_switch::execute): If a switch is converted, then
13306 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13308 PR rtl-optimization/81308
13309 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13312 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13314 PR target/83663 - Revert r255946
13315 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13316 generation for cases where splatting a value is not useful.
13317 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13318 across a vec_duplicate and a paradoxical subreg forming a vector
13319 mode to a vec_concat.
13321 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13323 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13324 -march=armv8.3-a variants.
13325 * config/arm/t-multilib: Likewise.
13326 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13328 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13330 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13332 (cceq_ior_compare_complement): Give it a name so I can use it, and
13333 change boolean_or_operator predicate to boolean_operator so it can
13334 be used to generate a crand.
13335 (eqne): New code iterator.
13336 (bd/bd_neg): New code_attrs.
13337 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13338 a single define_insn.
13339 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13340 decrement (bdnzt/bdnzf/bdzt/bdzf).
13341 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13342 with the new names of the branch decrement patterns, and added the
13343 names of the branch decrement conditional patterns.
13345 2018-01-08 Richard Biener <rguenther@suse.de>
13347 PR tree-optimization/83563
13348 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13351 2018-01-08 Richard Biener <rguenther@suse.de>
13353 PR middle-end/83713
13354 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13356 2018-01-08 Richard Biener <rguenther@suse.de>
13358 PR tree-optimization/83685
13359 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13360 references to abnormals.
13362 2018-01-08 Richard Biener <rguenther@suse.de>
13365 * dwarf2out.c (output_indirect_strings): Handle empty
13366 skeleton_debug_str_hash.
13367 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13369 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13371 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13372 (emit_store_direct): Likewise.
13373 (arc_trampoline_adjust_address): Likewise.
13374 (arc_asm_trampoline_template): New function.
13375 (arc_initialize_trampoline): Use asm_trampoline_template.
13376 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13377 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13378 * config/arc/arc.md (flush_icache): Delete pattern.
13380 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13382 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13383 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13386 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13389 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13390 by not USED_FOR_TARGET.
13391 (make_pass_resolve_sw_modes): Likewise.
13393 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13395 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13398 2018-01-08 Richard Biener <rguenther@suse.de>
13400 PR middle-end/83580
13401 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13403 2018-01-08 Richard Biener <rguenther@suse.de>
13405 PR middle-end/83517
13406 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13408 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13410 PR middle-end/81897
13411 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13412 basic blocks with a small number of successors.
13413 (convert_control_dep_chain_into_preds): Improve handling of
13415 (dump_predicates): Split apart into...
13416 (dump_pred_chain): ...here...
13417 (dump_pred_info): ...and here.
13418 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13419 (can_chain_union_be_invalidated_p): Improve check for invalidation
13421 (uninit_uses_cannot_happen): Avoid unnecessary if
13422 convert_control_dep_chain_into_preds yielded nothing.
13424 2018-01-06 Martin Sebor <msebor@redhat.com>
13426 PR tree-optimization/83640
13427 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13428 subtracting negative offset from size.
13429 (builtin_access::overlap): Adjust offset bounds of the access to fall
13430 within the size of the object if possible.
13432 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13434 PR rtl-optimization/83699
13435 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13436 extract_bit_field_as_subreg to cases in which the extracted
13437 value is also a vector.
13439 * lra-constraints.c (process_alt_operands): Test for the equivalence
13440 substitutions when detecting a possible reload cycle.
13442 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13445 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13446 by default if flag_selective_schedling{,2}. Formatting fixes.
13448 PR rtl-optimization/83682
13449 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13450 if it has non-VECTOR_MODE element mode.
13451 (vec_duplicate_p): Likewise.
13453 PR middle-end/83694
13454 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13455 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13457 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13460 * config/i386/i386-builtin.def
13461 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13462 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13463 Require also OPTION_MASK_ISA_AVX512F in addition to
13464 OPTION_MASK_ISA_GFNI.
13465 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13466 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13467 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13468 to OPTION_MASK_ISA_GFNI.
13469 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13470 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13471 OPTION_MASK_ISA_AVX512BW.
13472 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13473 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13474 addition to OPTION_MASK_ISA_GFNI.
13475 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13476 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13477 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13478 to OPTION_MASK_ISA_GFNI.
13479 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13480 a requirement for all ISAs rather than any of them with a few
13482 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13484 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13485 bitmasks to be enabled with 3 exceptions, instead of requiring any
13486 enabled ISA with lots of exceptions.
13487 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13488 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13489 Change avx512bw in isa attribute to avx512f.
13490 * config/i386/sgxintrin.h: Add license boilerplate.
13491 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13492 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13493 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13494 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13496 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13497 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13498 temporarily sse2 rather than sse if not enabled already.
13501 * config/i386/sse.md (VI248_VLBW): Rename to ...
13502 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13503 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13504 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13505 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13506 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13507 mode iterator instead of VI248_VLBW.
13509 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13511 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13512 (record_modified): Skip clobbers; add debug output.
13513 (param_change_prob): Use sreal frequencies.
13515 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13517 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13518 punt for user-aligned variables.
13520 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13522 * tree-chrec.c (chrec_contains_symbols): Return true for
13525 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13528 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13529 of (x|y) == x for BICS pattern.
13531 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13533 PR tree-optimization/83605
13534 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13535 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13538 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13540 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13541 * config/epiphany/rtems.h: New file.
13543 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13544 Uros Bizjak <ubizjak@gmail.com>
13547 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13548 QIreg_operand instead of register_operand predicate.
13549 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13550 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13551 comments instead of -fmitigate[-_]rop.
13553 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13556 * cgraphunit.c (symbol_table::compile): Switch to text_section
13557 before calling assembly_start debug hook.
13558 * run-rtl-passes.c (run_rtl_passes): Likewise.
13561 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13563 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13564 range_int_cst_p rather than !symbolic_range_p before calling
13565 extract_range_from_multiplicative_op_1.
13567 2018-01-04 Jeff Law <law@redhat.com>
13569 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13570 redundant test in assertion.
13572 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13574 * doc/rtl.texi: Document machine_mode wrapper classes.
13576 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13578 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13579 using tree_to_uhwi.
13581 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13583 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13584 the VEC_PERM_EXPR fold to fail.
13586 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13589 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13590 to switched_sections.
13592 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13595 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13596 test for d.testing.
13598 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13601 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13602 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13604 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13607 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13608 is BLKmode and bitpos not zero or mode change is needed.
13610 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13613 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13616 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13619 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13620 instead of MULT rtx. Update all corresponding splitters.
13621 (*saddl_se): Ditto.
13622 (*ssub<modesuffix>): Ditto.
13623 (*ssubl_se): Ditto.
13624 (*cmp_sadd_di): Update split patterns.
13625 (*cmp_sadd_si): Ditto.
13626 (*cmp_sadd_sidi): Ditto.
13627 (*cmp_ssub_di): Ditto.
13628 (*cmp_ssub_si): Ditto.
13629 (*cmp_ssub_sidi): Ditto.
13630 * config/alpha/predicates.md (const23_operand): New predicate.
13631 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13632 Look for ASHIFT, not MULT inner operand.
13633 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13635 2018-01-04 Martin Liska <mliska@suse.cz>
13637 PR gcov-profile/83669
13638 * gcov.c (output_intermediate_file): Add version to intermediate
13640 * doc/gcov.texi: Document new field 'version' in intermediate
13641 file format. Fix location of '-k' option of gcov command.
13643 2018-01-04 Martin Liska <mliska@suse.cz>
13646 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13648 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13650 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13652 2018-01-03 Martin Sebor <msebor@redhat.com>
13654 PR tree-optimization/83655
13655 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13656 checking calls with invalid arguments.
13658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13660 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13661 (vectorizable_mask_load_store): Delete.
13662 (vectorizable_call): Return false for masked loads and stores.
13663 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13664 instead of gimple_assign_rhs1.
13665 (vectorizable_load): Handle IFN_MASK_LOAD.
13666 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13670 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13672 (vectorizable_mask_load_store): ...here.
13673 (vectorizable_load): ...and here.
13675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13677 * tree-vect-stmts.c (vect_build_all_ones_mask)
13678 (vect_build_zero_merge_argument): New functions, split out from...
13679 (vectorizable_load): ...here.
13681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13683 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13685 (vectorizable_mask_load_store): ...here.
13686 (vectorizable_store): ...and here.
13688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13690 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13692 (vectorizable_mask_load_store): ...here.
13694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13696 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13697 (vect_model_store_cost): Take a vec_load_store_type instead of a
13699 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13700 (vect_model_store_cost): Take a vec_load_store_type instead of a
13702 (vectorizable_mask_load_store): Update accordingly.
13703 (vectorizable_store): Likewise.
13704 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13708 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13709 IFN_MASK_LOAD calls here rather than...
13710 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13712 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13713 Alan Hayward <alan.hayward@arm.com>
13714 David Sherwood <david.sherwood@arm.com>
13716 * expmed.c (extract_bit_field_1): For vector extracts,
13717 fall back to extract_bit_field_as_subreg if vec_extract
13720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13721 Alan Hayward <alan.hayward@arm.com>
13722 David Sherwood <david.sherwood@arm.com>
13724 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13725 they are variable or constant sized.
13726 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13727 slots for constant-sized data.
13729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13730 Alan Hayward <alan.hayward@arm.com>
13731 David Sherwood <david.sherwood@arm.com>
13733 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13734 handling COND_EXPRs with boolean comparisons, try to find a better
13735 basis for the mask type than the boolean itself.
13737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13739 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13740 is calculated and how it can be overridden.
13741 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13742 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13744 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13748 Alan Hayward <alan.hayward@arm.com>
13749 David Sherwood <david.sherwood@arm.com>
13751 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13752 Remove the mode argument.
13753 (aarch64_simd_valid_immediate): Remove the mode and inverse
13755 * config/aarch64/iterators.md (bitsize): New iterator.
13756 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13757 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13758 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13759 aarch64_simd_valid_immediate.
13760 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13761 (aarch64_reg_or_bic_imm): Likewise.
13762 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13763 with an insn_type enum and msl with a modifier_type enum.
13764 Replace element_width with a scalar_mode. Change the shift
13765 to unsigned int. Add constructors for scalar_float_mode and
13766 scalar_int_mode elements.
13767 (aarch64_vect_float_const_representable_p): Delete.
13768 (aarch64_can_const_movi_rtx_p)
13769 (aarch64_simd_scalar_immediate_valid_for_move)
13770 (aarch64_simd_make_constant): Update call to
13771 aarch64_simd_valid_immediate.
13772 (aarch64_advsimd_valid_immediate_hs): New function.
13773 (aarch64_advsimd_valid_immediate): Likewise.
13774 (aarch64_simd_valid_immediate): Remove mode and inverse
13775 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13776 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13777 and aarch64_float_const_representable_p on the result.
13778 (aarch64_output_simd_mov_immediate): Remove mode argument.
13779 Update call to aarch64_simd_valid_immediate and use of
13780 simd_immediate_info.
13781 (aarch64_output_scalar_simd_mov_immediate): Update call
13784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13785 Alan Hayward <alan.hayward@arm.com>
13786 David Sherwood <david.sherwood@arm.com>
13788 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13789 (mode_nunits): Likewise CONST_MODE_NUNITS.
13790 * machmode.def (ADJUST_NUNITS): Document.
13791 * genmodes.c (mode_data::need_nunits_adj): New field.
13792 (blank_mode): Update accordingly.
13793 (adj_nunits): New variable.
13794 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13796 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13797 listed in adj_nunits.
13798 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13799 listed in adj_nunits. Don't emit case statements for such modes.
13800 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13801 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13802 nothing if adj_nunits is nonnull.
13803 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13804 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13805 (emit_mode_fbit): Update use of print_maybe_const_decl.
13806 (emit_move_size): Likewise. Treat the array as non-const
13808 (emit_mode_adjustments): Handle adj_nunits.
13810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13812 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13813 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13814 (VECTOR_MODES): Use it.
13815 (make_vector_modes): Take the prefix as an argument.
13817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13818 Alan Hayward <alan.hayward@arm.com>
13819 David Sherwood <david.sherwood@arm.com>
13821 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13822 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13823 for MODE_VECTOR_BOOL.
13824 * machmode.def (VECTOR_BOOL_MODE): Document.
13825 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13826 (make_vector_bool_mode): New function.
13827 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13829 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13830 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13832 * stor-layout.c (int_mode_for_mode): Likewise.
13833 * tree.c (build_vector_type_for_mode): Likewise.
13834 * varasm.c (output_constant_pool_2): Likewise.
13835 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13836 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13837 for MODE_VECTOR_BOOL.
13838 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13839 of mode class checks.
13840 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13841 instead of a list of mode class checks.
13842 (expand_vector_scalar_condition): Likewise.
13843 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13845 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13846 Alan Hayward <alan.hayward@arm.com>
13847 David Sherwood <david.sherwood@arm.com>
13849 * machmode.h (mode_size): Change from unsigned short to
13851 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13852 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13853 or if measurement_type is not polynomial.
13854 (fixed_size_mode::includes_p): Check for constant-sized modes.
13855 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13856 return a poly_uint16 rather than an unsigned short.
13857 (emit_mode_size): Change the type of mode_size from unsigned short
13858 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13859 (emit_mode_adjustments): Cope with polynomial vector sizes.
13860 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13862 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13864 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13865 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13866 * caller-save.c (setup_save_areas): Likewise.
13867 (replace_reg_with_saved_mem): Likewise.
13868 * calls.c (emit_library_call_value_1): Likewise.
13869 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13870 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13871 (gen_lowpart_for_combine): Likewise.
13872 * convert.c (convert_to_integer_1): Likewise.
13873 * cse.c (equiv_constant, cse_insn): Likewise.
13874 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13875 (cselib_subst_to_values): Likewise.
13876 * dce.c (word_dce_process_block): Likewise.
13877 * df-problems.c (df_word_lr_mark_ref): Likewise.
13878 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13879 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13880 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13881 (rtl_for_decl_location): Likewise.
13882 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13883 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13884 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13885 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13886 (expand_expr_real_1): Likewise.
13887 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13888 (pad_below): Likewise.
13889 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13890 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13891 * ira.c (get_subreg_tracking_sizes): Likewise.
13892 * ira-build.c (ira_create_allocno_objects): Likewise.
13893 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13894 (ira_sort_regnos_for_alter_reg): Likewise.
13895 * ira-costs.c (record_operand_costs): Likewise.
13896 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13897 (resolve_simple_move): Likewise.
13898 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13899 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13900 (lra_constraints): Likewise.
13901 (CONST_POOL_OK_P): Reject variable-sized modes.
13902 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13903 (add_pseudo_to_slot, lra_spill): Likewise.
13904 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13905 * optabs-query.c (get_best_extraction_insn): Likewise.
13906 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13907 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13908 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13909 * recog.c (offsettable_address_addr_space_p): Likewise.
13910 * regcprop.c (maybe_mode_change): Likewise.
13911 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13912 * regrename.c (build_def_use): Likewise.
13913 * regstat.c (dump_reg_info): Likewise.
13914 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13915 (find_reloads, find_reloads_subreg_address): Likewise.
13916 * reload1.c (eliminate_regs_1): Likewise.
13917 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13918 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13919 (simplify_binary_operation_1, simplify_subreg): Likewise.
13920 * targhooks.c (default_function_arg_padding): Likewise.
13921 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13922 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13923 (verify_gimple_assign_ternary): Likewise.
13924 * tree-inline.c (estimate_move_cost): Likewise.
13925 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13926 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13927 (get_address_cost_ainc): Likewise.
13928 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13929 (vect_supportable_dr_alignment): Likewise.
13930 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13931 (vectorizable_reduction): Likewise.
13932 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13933 (vectorizable_operation, vectorizable_load): Likewise.
13934 * tree.c (build_same_sized_truth_vector_type): Likewise.
13935 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13936 * var-tracking.c (emit_note_insn_var_location): Likewise.
13937 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13938 (ADDR_VEC_ALIGN): Likewise.
13940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13941 Alan Hayward <alan.hayward@arm.com>
13942 David Sherwood <david.sherwood@arm.com>
13944 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13946 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13947 or if measurement_type is polynomial.
13948 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13949 * combine.c (make_extraction): Likewise.
13950 * dse.c (find_shift_sequence): Likewise.
13951 * dwarf2out.c (mem_loc_descriptor): Likewise.
13952 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13953 (extract_bit_field, extract_low_bits): Likewise.
13954 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13955 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13956 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13957 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13958 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13959 * reload.c (find_reloads): Likewise.
13960 * reload1.c (alter_reg): Likewise.
13961 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13962 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13963 * tree-if-conv.c (predicate_mem_writes): Likewise.
13964 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13965 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13966 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13967 * valtrack.c (dead_debug_insert_temp): Likewise.
13968 * varasm.c (mergeable_constant_section): Likewise.
13969 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13971 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13972 Alan Hayward <alan.hayward@arm.com>
13973 David Sherwood <david.sherwood@arm.com>
13975 * expr.c (expand_assignment): Cope with polynomial mode sizes
13976 when assigning to a CONCAT.
13978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13979 Alan Hayward <alan.hayward@arm.com>
13980 David Sherwood <david.sherwood@arm.com>
13982 * machmode.h (mode_precision): Change from unsigned short to
13984 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13986 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13987 or if measurement_type is not polynomial.
13988 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13989 in which the mode is already known to be a scalar_int_mode.
13990 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13991 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13993 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13994 for GET_MODE_PRECISION.
13995 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13996 for GET_MODE_PRECISION.
13997 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13999 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
14000 (expand_field_assignment, make_extraction): Likewise.
14001 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
14002 (get_last_value): Likewise.
14003 * convert.c (convert_to_integer_1): Likewise.
14004 * cse.c (cse_insn): Likewise.
14005 * expr.c (expand_expr_real_1): Likewise.
14006 * lra-constraints.c (simplify_operand_subreg): Likewise.
14007 * optabs-query.c (can_atomic_load_p): Likewise.
14008 * optabs.c (expand_atomic_load): Likewise.
14009 (expand_atomic_store): Likewise.
14010 * ree.c (combine_reaching_defs): Likewise.
14011 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
14012 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
14013 * tree.h (type_has_mode_precision_p): Likewise.
14014 * ubsan.c (instrument_si_overflow): Likewise.
14016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14017 Alan Hayward <alan.hayward@arm.com>
14018 David Sherwood <david.sherwood@arm.com>
14020 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
14021 polynomial numbers of units.
14022 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
14023 (valid_vector_subparts_p): New function.
14024 (build_vector_type): Remove temporary shim and take the number
14025 of units as a poly_uint64 rather than an int.
14026 (build_opaque_vector_type): Take the number of units as a
14027 poly_uint64 rather than an int.
14028 * tree.c (build_vector_from_ctor): Handle polynomial
14029 TYPE_VECTOR_SUBPARTS.
14030 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
14031 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
14032 (build_vector_from_val): If the number of units is variable,
14033 use build_vec_duplicate_cst for constant operands and
14034 VEC_DUPLICATE_EXPR otherwise.
14035 (make_vector_type): Remove temporary is_constant ().
14036 (build_vector_type, build_opaque_vector_type): Take the number of
14037 units as a poly_uint64 rather than an int.
14038 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14040 * cfgexpand.c (expand_debug_expr): Likewise.
14041 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14042 (store_constructor, expand_expr_real_1): Likewise.
14043 (const_scalar_mask_from_tree): Likewise.
14044 * fold-const-call.c (fold_const_reduction): Likewise.
14045 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14046 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14047 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14048 (fold_relational_const): Likewise.
14049 (native_interpret_vector): Likewise. Change the size from an
14050 int to an unsigned int.
14051 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14052 TYPE_VECTOR_SUBPARTS.
14053 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14054 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14055 duplicating a non-constant operand into a variable-length vector.
14056 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14057 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14058 * ipa-icf.c (sem_variable::equals): Likewise.
14059 * match.pd: Likewise.
14060 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14061 * print-tree.c (print_node): Likewise.
14062 * stor-layout.c (layout_type): Likewise.
14063 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14064 * tree-cfg.c (verify_gimple_comparison): Likewise.
14065 (verify_gimple_assign_binary): Likewise.
14066 (verify_gimple_assign_ternary): Likewise.
14067 (verify_gimple_assign_single): Likewise.
14068 * tree-pretty-print.c (dump_generic_node): Likewise.
14069 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14070 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14071 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14072 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14073 (vect_shift_permute_load_chain): Likewise.
14074 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14075 (expand_vector_condition, optimize_vector_constructor): Likewise.
14076 (lower_vec_perm, get_compute_type): Likewise.
14077 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14078 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14079 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14080 (vect_recog_mask_conversion_pattern): Likewise.
14081 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14082 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14083 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14084 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14085 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14086 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14087 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14088 (supportable_widening_operation): Likewise.
14089 (supportable_narrowing_operation): Likewise.
14090 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14092 * varasm.c (output_constant): Likewise.
14094 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14095 Alan Hayward <alan.hayward@arm.com>
14096 David Sherwood <david.sherwood@arm.com>
14098 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14099 so that both the length == 3 and length != 3 cases set up their
14100 own permute vectors. Add comments explaining why we know the
14101 number of elements is constant.
14102 (vect_permute_load_chain): Likewise.
14104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14105 Alan Hayward <alan.hayward@arm.com>
14106 David Sherwood <david.sherwood@arm.com>
14108 * machmode.h (mode_nunits): Change from unsigned char to
14110 (ONLY_FIXED_SIZE_MODES): New macro.
14111 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14112 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14113 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14115 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14116 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14117 or if measurement_type is not polynomial.
14118 * genmodes.c (ZERO_COEFFS): New macro.
14119 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14121 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14122 Use ZERO_COEFFS when emitting initializers.
14123 * data-streamer.h (bp_pack_poly_value): New function.
14124 (bp_unpack_poly_value): Likewise.
14125 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14126 for GET_MODE_NUNITS.
14127 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14128 for GET_MODE_NUNITS.
14129 * tree.c (make_vector_type): Remove temporary shim and make
14130 the real function take the number of units as a poly_uint64
14131 rather than an int.
14132 (build_vector_type_for_mode): Handle polynomial nunits.
14133 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14134 * emit-rtl.c (const_vec_series_p_1): Likewise.
14135 (gen_rtx_CONST_VECTOR): Likewise.
14136 * fold-const.c (test_vec_duplicate_folding): Likewise.
14137 * genrecog.c (validate_pattern): Likewise.
14138 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14139 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14140 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14141 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14142 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14143 * rtlanal.c (subreg_get_info): Likewise.
14144 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14145 (vect_grouped_load_supported): Likewise.
14146 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14147 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14148 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14149 (simplify_const_unary_operation, simplify_binary_operation_1)
14150 (simplify_const_binary_operation, simplify_ternary_operation)
14151 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14152 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14153 instead of CONST_VECTOR_NUNITS.
14154 * varasm.c (output_constant_pool_2): Likewise.
14155 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14156 explicit-encoded elements in the XVEC for variable-length vectors.
14158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14160 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14163 Alan Hayward <alan.hayward@arm.com>
14164 David Sherwood <david.sherwood@arm.com>
14166 * coretypes.h (fixed_size_mode): Declare.
14167 (fixed_size_mode_pod): New typedef.
14168 * builtins.h (target_builtins::x_apply_args_mode)
14169 (target_builtins::x_apply_result_mode): Change type to
14170 fixed_size_mode_pod.
14171 * builtins.c (apply_args_size, apply_result_size, result_vector)
14172 (expand_builtin_apply_args_1, expand_builtin_apply)
14173 (expand_builtin_return): Update accordingly.
14175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14177 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14178 * cselib.c (cselib_hash_rtx): Likewise.
14179 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14180 CONST_VECTOR encoding.
14182 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14183 Jeff Law <law@redhat.com>
14186 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14187 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14188 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14189 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14192 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14193 explicitly probe *sp in a noreturn function if there were any callee
14194 register saves or frame pointer is needed.
14196 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14199 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14200 BLKmode for ternary, binary or unary expressions.
14203 * var-tracking.c (delete_vta_debug_insn): New inline function.
14204 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14205 insns from get_insns () to NULL instead of each bb separately.
14206 Use delete_vta_debug_insn. No longer static.
14207 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14208 delete_vta_debug_insns callers.
14209 * rtl.h (delete_vta_debug_insns): Declare.
14210 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14211 instead of variable_tracking_main.
14213 2018-01-03 Martin Sebor <msebor@redhat.com>
14215 PR tree-optimization/83603
14216 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14217 arguments past the endof the argument list in functions declared
14218 without a prototype.
14219 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14220 Avoid checking when arguments are null.
14222 2018-01-03 Martin Sebor <msebor@redhat.com>
14225 * doc/extend.texi (attribute const): Fix a typo.
14226 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14227 issuing -Wsuggest-attribute for void functions.
14229 2018-01-03 Martin Sebor <msebor@redhat.com>
14231 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14232 offset_int::from instead of wide_int::to_shwi.
14233 (maybe_diag_overlap): Remove assertion.
14234 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14235 * gimple-ssa-sprintf.c (format_directive): Same.
14236 (parse_directive): Same.
14237 (sprintf_dom_walker::compute_format_length): Same.
14238 (try_substitute_return_value): Same.
14240 2018-01-03 Jeff Law <law@redhat.com>
14242 PR middle-end/83654
14243 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14244 non-constant residual for zero at runtime and avoid probing in
14245 that case. Reorganize code for trailing problem to mirror handling
14248 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14250 PR tree-optimization/83501
14251 * tree-ssa-strlen.c (get_string_cst): New.
14252 (handle_char_store): Call get_string_cst.
14254 2018-01-03 Martin Liska <mliska@suse.cz>
14256 PR tree-optimization/83593
14257 * tree-ssa-strlen.c: Include tree-cfg.h.
14258 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14259 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14260 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14262 (strlen_dom_walker::before_dom_children): Call
14263 gimple_purge_dead_eh_edges. Dump tranformation with details
14265 (strlen_dom_walker::before_dom_children): Update call by adding
14266 new argument cleanup_eh.
14267 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14269 2018-01-03 Martin Liska <mliska@suse.cz>
14272 * cif-code.def (VARIADIC_THUNK): New enum value.
14273 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14276 2018-01-03 Jan Beulich <jbeulich@suse.com>
14278 * sse.md (mov<mode>_internal): Tighten condition for when to use
14279 vmovdqu<ssescalarsize> for TI and OI modes.
14281 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14283 Update copyright years.
14285 2018-01-03 Martin Liska <mliska@suse.cz>
14288 * ipa-visibility.c (function_and_variable_visibility): Skip
14289 functions with noipa attribure.
14291 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14293 * gcc.c (process_command): Update copyright notice dates.
14294 * gcov-dump.c (print_version): Ditto.
14295 * gcov.c (print_version): Ditto.
14296 * gcov-tool.c (print_version): Ditto.
14297 * gengtype.c (create_file): Ditto.
14298 * doc/cpp.texi: Bump @copying's copyright year.
14299 * doc/cppinternals.texi: Ditto.
14300 * doc/gcc.texi: Ditto.
14301 * doc/gccint.texi: Ditto.
14302 * doc/gcov.texi: Ditto.
14303 * doc/install.texi: Ditto.
14304 * doc/invoke.texi: Ditto.
14306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14308 * vector-builder.h (vector_builder::m_full_nelts): Change from
14309 unsigned int to poly_uint64.
14310 (vector_builder::full_nelts): Update prototype accordingly.
14311 (vector_builder::new_vector): Likewise.
14312 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14313 (vector_builder::operator ==): Likewise.
14314 (vector_builder::finalize): Likewise.
14315 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14316 Take the number of elements as a poly_uint64 rather than an
14318 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14319 from unsigned int to poly_uint64.
14320 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14321 (vec_perm_indices::new_vector): Likewise.
14322 (vec_perm_indices::length): Likewise.
14323 (vec_perm_indices::nelts_per_input): Likewise.
14324 (vec_perm_indices::input_nelts): Likewise.
14325 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14326 number of elements per input as a poly_uint64 rather than an
14327 unsigned int. Use the original encoding for variable-length
14328 vectors, rather than clamping each individual element.
14329 For the second and subsequent elements in each pattern,
14330 clamp the step and base before clamping their sum.
14331 (vec_perm_indices::series_p): Handle polynomial element counts.
14332 (vec_perm_indices::all_in_range_p): Likewise.
14333 (vec_perm_indices_to_tree): Likewise.
14334 (vec_perm_indices_to_rtx): Likewise.
14335 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14336 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14337 (tree_vector_builder::new_binary_operation): Handle polynomial
14338 element counts. Return false if we need to know the number
14339 of elements at compile time.
14340 * fold-const.c (fold_vec_perm): Punt if the number of elements
14341 isn't known at compile time.
14343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14345 * vec-perm-indices.h (vec_perm_builder): Change element type
14346 from HOST_WIDE_INT to poly_int64.
14347 (vec_perm_indices::element_type): Update accordingly.
14348 (vec_perm_indices::clamp): Handle polynomial element_types.
14349 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14350 (vec_perm_indices::all_in_range_p): Likewise.
14351 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14353 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14354 polynomial vec_perm_indices element types.
14355 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14356 * fold-const.c (fold_vec_perm): Likewise.
14357 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14358 * tree-vect-generic.c (lower_vec_perm): Likewise.
14359 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14360 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14361 element type to HOST_WIDE_INT.
14363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14364 Alan Hayward <alan.hayward@arm.com>
14365 David Sherwood <david.sherwood@arm.com>
14367 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14368 rather than an int. Use plus_constant.
14369 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14370 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14373 Alan Hayward <alan.hayward@arm.com>
14374 David Sherwood <david.sherwood@arm.com>
14376 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14377 a HOST_WIDE_INT to a poly_int64.
14379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14380 Alan Hayward <alan.hayward@arm.com>
14381 David Sherwood <david.sherwood@arm.com>
14383 * calls.c (load_register_parameters): Cope with polynomial
14384 mode sizes. Require a constant size for BLKmode parameters
14385 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14386 forces a parameter to be padded at the lsb end in order to
14387 fill a complete number of words, require the parameter size
14388 to be ordered wrt UNITS_PER_WORD.
14390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14391 Alan Hayward <alan.hayward@arm.com>
14392 David Sherwood <david.sherwood@arm.com>
14394 * reload1.c (spill_stack_slot_width): Change element type
14395 from unsigned int to poly_uint64_pod.
14396 (alter_reg): Treat mode sizes as polynomial.
14398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14399 Alan Hayward <alan.hayward@arm.com>
14400 David Sherwood <david.sherwood@arm.com>
14402 * reload.c (complex_word_subreg_p): New function.
14403 (reload_inner_reg_of_subreg, push_reload): Use it.
14405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14406 Alan Hayward <alan.hayward@arm.com>
14407 David Sherwood <david.sherwood@arm.com>
14409 * lra-constraints.c (process_alt_operands): Reject matched
14410 operands whose sizes aren't ordered.
14411 (match_reload): Refer to this check here.
14413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14414 Alan Hayward <alan.hayward@arm.com>
14415 David Sherwood <david.sherwood@arm.com>
14417 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14418 that the mode size is in the set {1, 2, 4, 8, 16}.
14420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14421 Alan Hayward <alan.hayward@arm.com>
14422 David Sherwood <david.sherwood@arm.com>
14424 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14425 Use plus_constant instead of gen_rtx_PLUS.
14427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14428 Alan Hayward <alan.hayward@arm.com>
14429 David Sherwood <david.sherwood@arm.com>
14431 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14432 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14433 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14434 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14435 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14436 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14437 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14438 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14439 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14440 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14442 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14443 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14444 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14445 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14446 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14447 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14448 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14449 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14450 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14451 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14453 * expr.c (emit_move_resolve_push): Treat the input and result
14454 of PUSH_ROUNDING as a poly_int64.
14455 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14456 (emit_push_insn): Likewise.
14457 * lra-eliminations.c (mark_not_eliminable): Likewise.
14458 * recog.c (push_operand): Likewise.
14459 * reload1.c (elimination_effects): Likewise.
14460 * rtlanal.c (nonzero_bits1): Likewise.
14461 * calls.c (store_one_arg): Likewise. Require the padding to be
14462 known at compile time.
14464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14465 Alan Hayward <alan.hayward@arm.com>
14466 David Sherwood <david.sherwood@arm.com>
14468 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14469 Use plus_constant instead of gen_rtx_PLUS.
14471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14472 Alan Hayward <alan.hayward@arm.com>
14473 David Sherwood <david.sherwood@arm.com>
14475 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14476 rather than an int.
14478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14479 Alan Hayward <alan.hayward@arm.com>
14480 David Sherwood <david.sherwood@arm.com>
14482 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14483 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14484 via stack temporaries. Treat the mode size as polynomial too.
14486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14487 Alan Hayward <alan.hayward@arm.com>
14488 David Sherwood <david.sherwood@arm.com>
14490 * expr.c (expand_expr_real_2): When handling conversions involving
14491 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14492 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14493 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14502 Alan Hayward <alan.hayward@arm.com>
14503 David Sherwood <david.sherwood@arm.com>
14505 * combine.c (can_change_dest_mode): Handle polynomial
14506 REGMODE_NATURAL_SIZE.
14507 * expmed.c (store_bit_field_1): Likewise.
14508 * expr.c (store_constructor): Likewise.
14509 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14510 and polynomial REGMODE_NATURAL_SIZE.
14511 (gen_lowpart_common): Likewise.
14512 * reginfo.c (record_subregs_of_mode): Likewise.
14513 * rtlanal.c (read_modify_subreg_p): Likewise.
14515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14516 Alan Hayward <alan.hayward@arm.com>
14517 David Sherwood <david.sherwood@arm.com>
14519 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14520 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
14528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14529 Alan Hayward <alan.hayward@arm.com>
14530 David Sherwood <david.sherwood@arm.com>
14532 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14533 in a POINTER_PLUS_EXPR.
14535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14536 Alan Hayward <alan.hayward@arm.com>
14537 David Sherwood <david.sherwood@arm.com>
14539 * omp-simd-clone.c (simd_clone_subparts): New function.
14540 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14541 (ipa_simd_modify_function_body): Likewise.
14543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14544 Alan Hayward <alan.hayward@arm.com>
14545 David Sherwood <david.sherwood@arm.com>
14547 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14548 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14549 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14550 (expand_vector_condition, vector_element): Likewise.
14551 (subparts_gt): New function.
14552 (get_compute_type): Use subparts_gt.
14553 (count_type_subparts): Delete.
14554 (expand_vector_operations_1): Use subparts_gt instead of
14555 count_type_subparts.
14557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14558 Alan Hayward <alan.hayward@arm.com>
14559 David Sherwood <david.sherwood@arm.com>
14561 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14562 (vect_compile_time_alias): ...this new function. Do the calculation
14563 on poly_ints rather than trees.
14564 (vect_prune_runtime_alias_test_list): Update call accordingly.
14566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14567 Alan Hayward <alan.hayward@arm.com>
14568 David Sherwood <david.sherwood@arm.com>
14570 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14572 (vect_schedule_slp_instance): Likewise.
14574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14575 Alan Hayward <alan.hayward@arm.com>
14576 David Sherwood <david.sherwood@arm.com>
14578 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14579 constant and extern definitions for variable-length vectors.
14580 (vect_get_constant_vectors): Note that the number of units
14581 is known to be constant.
14583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14584 Alan Hayward <alan.hayward@arm.com>
14585 David Sherwood <david.sherwood@arm.com>
14587 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14588 of units as polynomial. Choose between WIDE and NARROW based
14591 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14592 Alan Hayward <alan.hayward@arm.com>
14593 David Sherwood <david.sherwood@arm.com>
14595 * tree-vect-stmts.c (simd_clone_subparts): New function.
14596 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14599 Alan Hayward <alan.hayward@arm.com>
14600 David Sherwood <david.sherwood@arm.com>
14602 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14603 vectors as polynomial. Use build_index_vector for
14604 IFN_GOMP_SIMD_LANE.
14606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14607 Alan Hayward <alan.hayward@arm.com>
14608 David Sherwood <david.sherwood@arm.com>
14610 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14611 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14612 for variable-length vectors.
14613 (vectorizable_mask_load_store): Treat the number of units as
14614 polynomial, asserting that it is constant if the condition has
14615 already been enforced.
14616 (vectorizable_store, vectorizable_load): Likewise.
14618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14619 Alan Hayward <alan.hayward@arm.com>
14620 David Sherwood <david.sherwood@arm.com>
14622 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14623 of units as polynomial. Punt if we can't tell at compile time
14624 which vector contains the final result.
14626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14627 Alan Hayward <alan.hayward@arm.com>
14628 David Sherwood <david.sherwood@arm.com>
14630 * tree-vect-loop.c (vectorizable_induction): Treat the number
14631 of units as polynomial. Punt on SLP inductions. Use an integer
14632 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14633 cast of such a series for variable-length floating-point
14636 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14637 Alan Hayward <alan.hayward@arm.com>
14638 David Sherwood <david.sherwood@arm.com>
14640 * tree.h (build_index_vector): Declare.
14641 * tree.c (build_index_vector): New function.
14642 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14643 of units as polynomial, forcibly converting it to a constant if
14644 vectorizable_reduction has already enforced the condition.
14645 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14646 to create a {1,2,3,...} vector.
14647 (vectorizable_reduction): Treat the number of units as polynomial.
14648 Choose vectype_in based on the largest scalar element size rather
14649 than the smallest number of units. Enforce the restrictions
14652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14653 Alan Hayward <alan.hayward@arm.com>
14654 David Sherwood <david.sherwood@arm.com>
14656 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14657 number of units as polynomial.
14659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14660 Alan Hayward <alan.hayward@arm.com>
14661 David Sherwood <david.sherwood@arm.com>
14663 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14664 * target.def (autovectorize_vector_sizes): Return the vector sizes
14665 by pointer, using vector_sizes rather than a bitmask.
14666 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14667 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14668 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14670 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14671 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14672 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14673 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14674 * omp-general.c (omp_max_vf): Likewise.
14675 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14676 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14677 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14678 * tree-vect-slp.c (vect_slp_bb): Likewise.
14679 * doc/tm.texi: Regenerate.
14680 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14682 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14683 the vector size as a poly_uint64 rather than an unsigned int.
14684 (current_vector_size): Change from an unsigned int to a poly_uint64.
14685 (get_vectype_for_scalar_type): Update accordingly.
14686 * tree.h (build_truth_vector_type): Take the size and number of
14687 units as a poly_uint64 rather than an unsigned int.
14688 (build_vector_type): Add a temporary overload that takes
14689 the number of units as a poly_uint64 rather than an unsigned int.
14690 * tree.c (make_vector_type): Likewise.
14691 (build_truth_vector_type): Take the number of units as a poly_uint64
14692 rather than an unsigned int.
14694 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14695 Alan Hayward <alan.hayward@arm.com>
14696 David Sherwood <david.sherwood@arm.com>
14698 * target.def (get_mask_mode): Take the number of units and length
14699 as poly_uint64s rather than unsigned ints.
14700 * targhooks.h (default_get_mask_mode): Update accordingly.
14701 * targhooks.c (default_get_mask_mode): Likewise.
14702 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14703 * doc/tm.texi: Regenerate.
14705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14706 Alan Hayward <alan.hayward@arm.com>
14707 David Sherwood <david.sherwood@arm.com>
14709 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14710 * omp-general.c (omp_max_vf): Likewise.
14711 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14712 (expand_omp_simd): Handle polynomial safelen.
14713 * omp-low.c (omplow_simd_context): Add a default constructor.
14714 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14715 (lower_rec_simd_input_clauses): Update accordingly.
14716 (lower_rec_input_clauses): Likewise.
14718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14719 Alan Hayward <alan.hayward@arm.com>
14720 David Sherwood <david.sherwood@arm.com>
14722 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14723 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14724 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14725 (vect_analyze_slp_cost): Likewise.
14726 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14727 (vect_model_load_cost): Likewise.
14729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14730 Alan Hayward <alan.hayward@arm.com>
14731 David Sherwood <david.sherwood@arm.com>
14733 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14734 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14735 from an unsigned int * to a poly_uint64_pod *.
14736 (calculate_unrolling_factor): New function.
14737 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14740 Alan Hayward <alan.hayward@arm.com>
14741 David Sherwood <david.sherwood@arm.com>
14743 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14744 from an unsigned int to a poly_uint64.
14745 (_loop_vec_info::slp_unrolling_factor): Likewise.
14746 (_loop_vec_info::vectorization_factor): Change from an int
14748 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14749 (vect_get_num_vectors): New function.
14750 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14751 (vect_get_num_copies): Use vect_get_num_vectors.
14752 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14753 to an unsigned int *.
14754 (vect_analyze_data_refs): Change min_vf from an int * to a
14756 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14757 than an unsigned HOST_WIDE_INT.
14758 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14759 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14760 to an unsigned int *.
14761 (vect_analyze_data_ref_dependences): Likewise.
14762 (vect_compute_data_ref_alignment): Handle polynomial vf.
14763 (vect_enhance_data_refs_alignment): Likewise.
14764 (vect_prune_runtime_alias_test_list): Likewise.
14765 (vect_shift_permute_load_chain): Likewise.
14766 (vect_supportable_dr_alignment): Likewise.
14767 (dependence_distance_ge_vf): Take the vectorization factor as a
14768 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14769 (vect_analyze_data_refs): Change min_vf from an int * to a
14771 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14772 vfm1 as a poly_uint64 rather than an int. Make the same change
14773 for the returned bound_scalar.
14774 (vect_gen_vector_loop_niters): Handle polynomial vf.
14775 (vect_do_peeling): Likewise. Update call to
14776 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14777 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14779 * tree-vect-loop.c (vect_determine_vectorization_factor)
14780 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14781 (vect_get_known_peeling_cost): Likewise.
14782 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14783 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14784 (vect_transform_loop): Likewise. Use the lowest possible VF when
14785 updating the upper bounds of the loop.
14786 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14787 rather than an int.
14788 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14789 polynomial unroll factors.
14790 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14791 (vect_make_slp_decision): Likewise.
14792 (vect_supported_load_permutation_p): Likewise, and polynomial
14794 (vect_analyze_slp_cost): Handle polynomial vf.
14795 (vect_slp_analyze_node_operations): Likewise.
14796 (vect_slp_analyze_bb_1): Likewise.
14797 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14798 than an unsigned HOST_WIDE_INT.
14799 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14800 (vectorizable_load): Handle polynomial vf.
14801 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14803 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14806 Alan Hayward <alan.hayward@arm.com>
14807 David Sherwood <david.sherwood@arm.com>
14809 * match.pd: Handle bit operations involving three constants
14810 and try to fold one pair.
14812 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14814 * tree-vect-loop-manip.c: Include gimple-fold.h.
14815 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14816 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14817 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14818 Add a path that uses a step of VF instead of 1, but disable it
14820 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14821 and niters_no_overflow parameters. Update calls to
14822 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14823 Create a new SSA name if the latter choses to use a ste other
14824 than zero, and return it via niters_vector_mult_vf_var.
14825 * tree-vect-loop.c (vect_transform_loop): Update calls to
14826 vect_do_peeling, vect_gen_vector_loop_niters and
14827 slpeel_make_loop_iterate_ntimes.
14828 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14829 (vect_gen_vector_loop_niters): Update declarations after above changes.
14831 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14833 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14834 128-bit round to integer instructions.
14835 (ceil<mode>2): Likewise.
14836 (btrunc<mode>2): Likewise.
14837 (round<mode>2): Likewise.
14839 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14841 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14842 unaligned VSX load/store on P8/P9.
14843 (expand_block_clear): Allow the use of unaligned VSX
14844 load/store on P8/P9.
14846 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14848 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14850 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14851 swap associated with both a load and a store.
14853 2018-01-02 Andrew Waterman <andrew@sifive.com>
14855 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14856 * config/riscv/riscv.md (clear_cache): Use it.
14858 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14860 * web.c: Remove out-of-date comment.
14862 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14864 * expr.c (fixup_args_size_notes): Check that any existing
14865 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14866 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14867 (emit_single_push_insn): ...here.
14869 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14871 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14872 (const_vector_encoded_nelts): New function.
14873 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14874 (const_vector_int_elt, const_vector_elt): Declare.
14875 * emit-rtl.c (const_vector_int_elt_1): New function.
14876 (const_vector_elt): Likewise.
14877 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14878 of CONST_VECTOR_ELT.
14880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14882 * expr.c: Include rtx-vector-builder.h.
14883 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14884 directly on the tree encoding.
14885 (const_vector_from_tree): Likewise.
14886 * optabs.c: Include rtx-vector-builder.h.
14887 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14888 sequence of "u" values.
14889 * vec-perm-indices.c: Include rtx-vector-builder.h.
14890 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14891 directly on the vec_perm_indices encoding.
14893 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14895 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14896 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14897 * rtx-vector-builder.h: New file.
14898 * rtx-vector-builder.c: Likewise.
14899 * rtl.h (rtx_def::u2): Add a const_vector field.
14900 (CONST_VECTOR_NPATTERNS): New macro.
14901 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14902 (CONST_VECTOR_DUPLICATE_P): Likewise.
14903 (CONST_VECTOR_STEPPED_P): Likewise.
14904 (CONST_VECTOR_ENCODED_ELT): Likewise.
14905 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14906 (unwrap_const_vec_duplicate): Likewise.
14907 (const_vec_series_p): Check for a non-duplicated vector encoding.
14908 Say that the function only returns true for integer vectors.
14909 * emit-rtl.c: Include rtx-vector-builder.h.
14910 (gen_const_vec_duplicate_1): Delete.
14911 (gen_const_vector): Call gen_const_vec_duplicate instead of
14912 gen_const_vec_duplicate_1.
14913 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14914 (gen_const_vec_duplicate): Use rtx_vector_builder.
14915 (gen_const_vec_series): Likewise.
14916 (gen_rtx_CONST_VECTOR): Likewise.
14917 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14918 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14919 Build a new vector rather than modifying a CONST_VECTOR in-place.
14920 (handle_special_swappables): Update call accordingly.
14921 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14922 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14923 Build a new vector rather than modifying a CONST_VECTOR in-place.
14924 (handle_special_swappables): Update call accordingly.
14926 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14928 * simplify-rtx.c (simplify_const_binary_operation): Use
14929 CONST_VECTOR_ELT instead of XVECEXP.
14931 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14933 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14934 the selector elements to be different from the data elements
14935 if the selector is a VECTOR_CST.
14936 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14937 ssizetype for the selector.
14939 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14941 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14942 before testing each element individually.
14943 * tree-vect-generic.c (lower_vec_perm): Likewise.
14945 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14947 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14948 * selftest-run-tests.c (selftest::run_tests): Call it.
14949 * vector-builder.h (vector_builder::operator ==): New function.
14950 (vector_builder::operator !=): Likewise.
14951 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14952 (vec_perm_indices::all_from_input_p): New function.
14953 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14954 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14955 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14956 instead of reading the VECTOR_CST directly. Detect whether both
14957 vector inputs are the same before constructing the vec_perm_indices,
14958 and update the number of inputs argument accordingly. Use the
14959 utility functions added above. Only construct sel2 if we need to.
14961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14963 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14964 the broadcast of the low byte.
14965 (expand_mult_highpart): Use an explicit encoding for the permutes.
14966 * optabs-query.c (can_mult_highpart_p): Likewise.
14967 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14968 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14969 (vectorizable_bswap): Likewise.
14970 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14971 explicit encoding for the power-of-2 permutes.
14972 (vect_permute_store_chain): Likewise.
14973 (vect_grouped_load_supported): Likewise.
14974 (vect_permute_load_chain): Likewise.
14976 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14978 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14979 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14980 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14981 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14982 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14983 (vect_gen_perm_mask_any): Likewise.
14985 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14987 * int-vector-builder.h: New file.
14988 * vec-perm-indices.h: Include int-vector-builder.h.
14989 (vec_perm_indices): Redefine as an int_vector_builder.
14990 (auto_vec_perm_indices): Delete.
14991 (vec_perm_builder): Redefine as a stand-alone class.
14992 (vec_perm_indices::vec_perm_indices): New function.
14993 (vec_perm_indices::clamp): Likewise.
14994 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14995 (vec_perm_indices::new_vector): New function.
14996 (vec_perm_indices::new_expanded_vector): Update for new
14997 vec_perm_indices class.
14998 (vec_perm_indices::rotate_inputs): New function.
14999 (vec_perm_indices::all_in_range_p): Operate directly on the
15000 encoded form, without computing elided elements.
15001 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
15002 encoding. Update for new vec_perm_indices class.
15003 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
15004 the given vec_perm_builder.
15005 (expand_vec_perm_var): Update vec_perm_builder constructor.
15006 (expand_mult_highpart): Use vec_perm_builder instead of
15007 auto_vec_perm_indices.
15008 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
15009 vec_perm_indices instead of auto_vec_perm_indices. Use a single
15010 or double series encoding as appropriate.
15011 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
15012 vec_perm_indices instead of auto_vec_perm_indices.
15013 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15014 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15015 (vect_permute_store_chain): Likewise.
15016 (vect_grouped_load_supported): Likewise.
15017 (vect_permute_load_chain): Likewise.
15018 (vect_shift_permute_load_chain): Likewise.
15019 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15020 (vect_transform_slp_perm_load): Likewise.
15021 (vect_schedule_slp_instance): Likewise.
15022 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15023 (vectorizable_mask_load_store): Likewise.
15024 (vectorizable_bswap): Likewise.
15025 (vectorizable_store): Likewise.
15026 (vectorizable_load): Likewise.
15027 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
15028 vec_perm_indices instead of auto_vec_perm_indices. Use
15029 tree_to_vec_perm_builder to read the vector from a tree.
15030 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
15031 vec_perm_builder instead of a vec_perm_indices.
15032 (have_whole_vector_shift): Use vec_perm_builder and
15033 vec_perm_indices instead of auto_vec_perm_indices. Leave the
15034 truncation to calc_vec_perm_mask_for_shift.
15035 (vect_create_epilog_for_reduction): Likewise.
15036 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
15037 from auto_vec_perm_indices to vec_perm_indices.
15038 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15039 instead of changing individual elements.
15040 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15041 the vector in d.perm.
15042 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15043 from auto_vec_perm_indices to vec_perm_indices.
15044 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15045 instead of changing individual elements.
15046 (arm_vectorize_vec_perm_const): Use new_vector to install
15047 the vector in d.perm.
15048 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15049 Update vec_perm_builder constructor.
15050 (rs6000_expand_interleave): Likewise.
15051 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15052 (rs6000_expand_interleave): Likewise.
15054 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15056 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15057 to qimode could truncate the indices.
15058 * optabs.c (expand_vec_perm_var): Likewise.
15060 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15062 * Makefile.in (OBJS): Add vec-perm-indices.o.
15063 * vec-perm-indices.h: New file.
15064 * vec-perm-indices.c: Likewise.
15065 * target.h (vec_perm_indices): Replace with a forward class
15067 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15068 * optabs.h: Include vec-perm-indices.h.
15069 (expand_vec_perm): Delete.
15070 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15071 (expand_vec_perm_const): Declare.
15072 * target.def (vec_perm_const_ok): Replace with...
15073 (vec_perm_const): ...this new hook.
15074 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15075 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15076 * doc/tm.texi: Regenerate.
15077 * optabs.def (vec_perm_const): Delete.
15078 * doc/md.texi (vec_perm_const): Likewise.
15079 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15080 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15081 expand_vec_perm for constant permutation vectors. Assert that
15082 the mode of variable permutation vectors is the integer equivalent
15083 of the mode that is being permuted.
15084 * optabs-query.h (selector_fits_mode_p): Declare.
15085 * optabs-query.c: Include vec-perm-indices.h.
15086 (selector_fits_mode_p): New function.
15087 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15088 is defined, instead of checking whether the vec_perm_const_optab
15089 exists. Use targetm.vectorize.vec_perm_const instead of
15090 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15091 fit in the vector mode before using a variable permute.
15092 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15093 vec_perm_indices instead of an rtx.
15094 (expand_vec_perm): Replace with...
15095 (expand_vec_perm_const): ...this new function. Take the selector
15096 as a vec_perm_indices rather than an rtx. Also take the mode of
15097 the selector. Update call to shift_amt_for_vec_perm_mask.
15098 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15099 Use vec_perm_indices::new_expanded_vector to expand the original
15100 selector into bytes. Check whether the indices fit in the vector
15101 mode before using a variable permute.
15102 (expand_vec_perm_var): Make global.
15103 (expand_mult_highpart): Use expand_vec_perm_const.
15104 * fold-const.c: Includes vec-perm-indices.h.
15105 * tree-ssa-forwprop.c: Likewise.
15106 * tree-vect-data-refs.c: Likewise.
15107 * tree-vect-generic.c: Likewise.
15108 * tree-vect-loop.c: Likewise.
15109 * tree-vect-slp.c: Likewise.
15110 * tree-vect-stmts.c: Likewise.
15111 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15113 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15114 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15115 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15116 (aarch64_vectorize_vec_perm_const): ...this new function.
15117 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15118 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15119 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15120 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15121 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15122 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15123 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15125 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15126 check for NEON modes.
15127 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15128 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15129 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15130 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15132 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15133 the old VEC_PERM_CONST conditions.
15134 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15135 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15136 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15137 (ia64_vectorize_vec_perm_const_ok): Merge into...
15138 (ia64_vectorize_vec_perm_const): ...this new function.
15139 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15140 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15141 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15142 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15143 * config/mips/mips.c (mips_expand_vec_perm_const)
15144 (mips_vectorize_vec_perm_const_ok): Merge into...
15145 (mips_vectorize_vec_perm_const): ...this new function.
15146 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15147 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15148 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15149 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15150 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15151 (rs6000_expand_vec_perm_const): Delete.
15152 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15154 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15155 (altivec_expand_vec_perm_const_le): Take each operand individually.
15156 Operate on constant selectors rather than rtxes.
15157 (altivec_expand_vec_perm_const): Likewise. Update call to
15158 altivec_expand_vec_perm_const_le.
15159 (rs6000_expand_vec_perm_const): Delete.
15160 (rs6000_vectorize_vec_perm_const_ok): Delete.
15161 (rs6000_vectorize_vec_perm_const): New function.
15162 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15163 an element count and rtx array.
15164 (rs6000_expand_extract_even): Update call accordingly.
15165 (rs6000_expand_interleave): Likewise.
15166 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15167 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15168 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15169 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15170 (rs6000_expand_vec_perm_const): Delete.
15171 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15172 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15173 (altivec_expand_vec_perm_const_le): Take each operand individually.
15174 Operate on constant selectors rather than rtxes.
15175 (altivec_expand_vec_perm_const): Likewise. Update call to
15176 altivec_expand_vec_perm_const_le.
15177 (rs6000_expand_vec_perm_const): Delete.
15178 (rs6000_vectorize_vec_perm_const_ok): Delete.
15179 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15180 reference to the SPE evmerge intructions.
15181 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15182 an element count and rtx array.
15183 (rs6000_expand_extract_even): Update call accordingly.
15184 (rs6000_expand_interleave): Likewise.
15185 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15186 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15188 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15192 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15193 vector mode and that that mode matches the mode of the data
15195 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15196 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15197 directly using expand_vec_perm_1 when forcing selectors into
15199 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15203 * optabs-query.h (can_vec_perm_p): Delete.
15204 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15205 * optabs-query.c (can_vec_perm_p): Split into...
15206 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15207 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15208 particular selector is valid.
15209 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15210 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15211 (vect_grouped_load_supported): Likewise.
15212 (vect_shift_permute_load_chain): Likewise.
15213 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15214 (vect_transform_slp_perm_load): Likewise.
15215 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15216 (vectorizable_bswap): Likewise.
15217 (vect_gen_perm_mask_checked): Likewise.
15218 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15219 implementations of variable permutation vectors into account
15220 when deciding which selector to use.
15221 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15222 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15223 with a false third argument.
15224 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15225 to test whether the constant selector is valid and can_vec_perm_var_p
15226 to test whether a variable selector is valid.
15228 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15230 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15231 * optabs-query.c (can_vec_perm_p): Likewise.
15232 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15233 instead of vec_perm_indices.
15234 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15235 (vect_gen_perm_mask_checked): Likewise,
15236 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15237 (vect_gen_perm_mask_checked): Likewise,
15239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15241 * optabs-query.h (qimode_for_vec_perm): Declare.
15242 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15243 (qimode_for_vec_perm): ...this new function.
15244 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15246 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15248 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15249 does not have a conditional at the top.
15251 2018-01-02 Richard Biener <rguenther@suse.de>
15253 * ipa-inline.c (big_speedup_p): Fix expression.
15255 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15258 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15261 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15265 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15266 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15267 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15268 cond_taken_branch_cost 3->4.
15270 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15272 PR tree-optimization/83581
15273 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15274 TODO_cleanup_cfg if any changes have been made.
15276 PR middle-end/83608
15277 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15278 convert_modes if target mode has the right side, but different mode
15281 PR middle-end/83609
15282 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15283 last argument when extracting from CONCAT. If either from_real or
15284 from_imag is NULL, use expansion through memory. If result is not
15285 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15286 the parts directly to inner mode, if even that fails, use expansion
15289 PR middle-end/83623
15290 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15291 check for bswap in mode rather than HImode and use that in expand_unop
15294 Copyright (C) 2018 Free Software Foundation, Inc.
15296 Copying and distribution of this file, with or without modification,
15297 are permitted in any medium without royalty provided the copyright
15298 notice and this notice are preserved.