1 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
5 qualifier_void_pointer and qualifier_const_void_pointer.
6 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
7 (arm_init_builtins): Handle the above.
8 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
9 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
13 2018-06-05 Martin Liska <mliska@suse.cz>
15 * auto-profile.c (read_autofdo_file): Do not use
16 gcov_ctr_summary struct.
17 (afdo_callsite_hot_enough_for_early_inline): Likewise.
18 * coverage.c (struct counts_entry): Likewise.
19 (read_counts_file): Read just single summary entry.
20 (get_coverage_counts): Use gcov_summary struct.
21 * coverage.h (get_coverage_counts): Likewise.
22 * gcov-dump.c (dump_working_sets): Likewise.
23 (tag_summary): Dump just single summary.
24 * gcov-io.c (gcov_write_summary): Write just histogram
26 (gcov_read_summary): Read just single summary.
27 (compute_working_sets): Use gcov_summary struct.
28 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
29 of GCOV_COUNTERS_SUMMABLE.
30 (GCOV_COUNTERS_SUMMABLE): Remove.
31 (GCOV_FIRST_VALUE_COUNTER): Replace with
32 GCOV_COUNTER_V_INTERVAL.
33 (struct gcov_ctr_summary): Remove.
34 (struct gcov_summary): Directly use fields of former
36 (compute_working_sets): Use gcov_summary struct.
37 * gcov.c (read_count_file): Do not use ctrs fields.
38 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
40 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
42 * profile.c: Likewise.
43 * profile.h: Likewise.
45 2018-06-05 Martin Liska <mliska@suse.cz>
48 * gcov.c (output_lines): Print working directory only
49 in intermediate format.
51 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
53 * config/s390/s390-builtin-types.def: Add void function type.
54 * config/s390/s390-builtins.def: Use the function type for the
57 2018-06-04 Jim Wilson <jimw@sifive.com>
59 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
61 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
62 and maybe_eh_return. Change regno to unsigned int. Use new args to
63 handle EH_RETURN_DATA_REGNO registers properly.
64 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
65 (riscv_expand_epilogue): Update comment. Change argument name and
66 type. Update code to use new name and type. Pass new args to
67 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
69 * config/riscv/riscv.md (NORMAL_RETURN): New.
70 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
71 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
72 (eh_return): Call gen_eh_return_internal and emit barrier.
73 (eh_return_internal): Call riscv_expand_epilogue.
75 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
77 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
78 bit_insertion field and declare can_be_merged_into method.
79 (merged_store_group::can_be_merged_into): New method.
80 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
81 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
82 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
84 2018-06-04 Richard Biener <rguenther@suse.de>
86 PR tree-optimization/85955
87 * builtins.c (fold_builtin_sincos): Convert pointers to
88 destination to appropriate type before dereferencing.
90 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
92 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
94 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
96 * expr.c (expand_expr_real_1): Force the operand into memory if
97 its TYPE_MODE is BLKmode and if there is no integer mode for
98 the number of bits being extracted.
100 2018-06-04 Jakub Jelinek <jakub@redhat.com>
104 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
105 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
107 2018-06-04 Richard Biener <rguenther@suse.de>
109 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
110 (cleanup_tree_cfg_noloop): ... single caller. Do
111 start_recording_case_labels later.
113 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
115 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
116 to _IMMINTRIN_H_INCLUDED.
117 * config/i386/pconfigintrin.h: Ditto.
118 * config/i386/waitpkgintrin.h: Ditto.
119 * config/i386/immintrin.h: Add includes for sgxintrin.h,
120 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
121 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
122 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
123 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
124 waitpkgintrin.h and cldemoteintrin.h.
126 2018-06-04 Richard Biener <rguenther@suse.de>
128 PR tree-optimization/86038
129 * tracer.c (find_best_successor): Check probability for
130 being initialized, bail out if not.
132 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
135 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
136 of bits to ignore when comparing architectures.
138 2018-06-04 Jakub Jelinek <jakub@redhat.com>
140 PR tree-optimization/69615
141 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
142 maximum or minimum of the type, try to merge it also as if
143 range1 is + [-, x - 1] or + [x + 1, -].
146 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
148 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
150 PR tree-optimization/86034
151 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
152 the unsigned bitfield type in a bit insertion sequence if it does not
153 have a larger precision than the bitfield size.
154 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
156 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
158 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
160 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
162 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
163 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
164 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
165 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
167 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
169 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
170 Disable -fdelete-null-pointer-checks for ELF toolchain.
172 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
173 Kito Cheng <kito.cheng@gmail.com>
175 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
176 (nds32le-*-*, nds32be-*-*): Integrate checking process.
177 (nds32*-*-*): Add glibc and uclibc conditions.
178 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
179 (TARGET_EXCEPT_UNWIND_INFO): Define.
180 * config/nds32/elf.h: New file.
181 * config/nds32/linux.h: New file.
182 * config/nds32/nds32-elf.opt: New file.
183 * config/nds32/nds32-linux.opt: New file.
184 * config/nds32/nds32-fp-as-gp.c
185 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
186 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
188 (nds32_asm_file_end): Ditto.
189 (nds32_print_operand): Ditto.
190 (nds32_insert_attributes): Ditto.
191 (nds32_init_libfuncs): New function.
192 (TARGET_HAVE_TLS): Define.
193 (TARGET_INIT_LIBFUNCS): Define.
194 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
196 (TARGET_ELF): Apply different mcmodel setting.
197 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
198 been migrated into elf.h and linux.h files.
199 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
200 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
201 (mcmodel): The content has been migrated into nds32-elf.opt and
202 nds32-linux.opt files.
203 * config/nds32/t-elf: New file.
204 * config/nds32/t-linux: New file.
206 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
207 Shiva Chen <shiva0217@gmail.com>
209 * config/nds32/constants.md (unspec_volatile_element): Add
210 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
211 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
213 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
214 (make_pass_nds32_fp_as_gp): Declare.
215 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
217 (nds32_asm_function_end_prologue): Remove unused asm output.
218 (nds32_asm_function_begin_epilogue): Remove unused asm output.
219 (nds32_asm_file_start): Output necessary fp_as_gp information.
220 (nds32_option_override): Adjust register usage.
221 (nds32_expand_prologue): Consider fp_as_gp situation.
222 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
223 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
226 (simple_return): Ditto.
227 (omit_fp_begin): Output special directive for fp_as_gp.
228 (omit_fp_end): Output special directive for fp_as_gp.
229 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
230 mforbid-fp-as-gp): New options.
232 2018-06-01 Mark Wielaard <mark@klomp.org>
234 * dwarf2out.c (dwarf2out_finish): Remove generation of
237 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
239 * gimple-ssa-store-merging.c: Include gimple-fold.h.
240 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
241 (struct merged_store_group): Add bit_insertion field.
242 (dump_char_array): Use standard hexadecimal format.
243 (merged_store_group::merged_store_group): Set bit_insertion to false.
244 (merged_store_group::apply_stores): Use optimal buffer size. Deal
245 with BIT_INSERT_EXPR stores. Move up code updating the mask and
246 also print the mask in the dump file.
247 (pass_store_merging::gate): Minor tweak.
248 (imm_store_chain_info::coalesce_immediate): Fix wrong association
249 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
250 stores with INTEGER_CST stores.
251 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
252 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
253 and use it throughout. Generate bit insertion sequences if need be.
254 (pass_store_merging::process_store): Remove redundant condition.
255 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
257 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
259 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
260 the 128-bit floating point types. Fix function comment.
262 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
264 * config/aarch64/aarch64-simd.md
265 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
267 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
270 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
272 PR tree-optimization/85989
273 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
275 (backprop::intersect_uses): Check it when deciding whether this
276 is a backedge reference.
277 (backprop::process_block): Add each phi to m_visited_phis
278 after visiting it, then clear it at the end.
280 2018-06-01 Richard Biener <rguenther@suse.de>
282 * tree-vectorizer.h (vect_dr_stmt): New function.
283 (vect_get_load_cost): Adjust.
284 (vect_get_store_cost): Likewise.
285 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
286 Use vect_dr_stmt instead of DR_SMTT.
287 (vect_record_base_alignments): Likewise.
288 (vect_calculate_target_alignment): Likewise.
289 (vect_compute_data_ref_alignment): Likewise and make static.
290 (vect_update_misalignment_for_peel): Likewise.
291 (vect_verify_datarefs_alignment): Likewise.
292 (vector_alignment_reachable_p): Likewise.
293 (vect_get_data_access_cost): Likewise. Pass down
294 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
295 (vect_get_peeling_costs_all_drs): Likewise.
296 (vect_peeling_hash_get_lowest_cost): Likewise.
297 (vect_enhance_data_refs_alignment): Likewise.
298 (vect_find_same_alignment_drs): Likewise.
299 (vect_analyze_data_refs_alignment): Likewise.
300 (vect_analyze_group_access_1): Likewise.
301 (vect_analyze_group_access): Likewise.
302 (vect_analyze_data_ref_access): Likewise.
303 (vect_analyze_data_ref_accesses): Likewise.
304 (vect_vfa_segment_size): Likewise.
305 (vect_small_gap_p): Likewise.
306 (vectorizable_with_step_bound_p): Likewise.
307 (vect_prune_runtime_alias_test_list): Likewise.
308 (vect_analyze_data_refs): Likewise.
309 (vect_supportable_dr_alignment): Likewise.
310 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
311 (vect_gen_prolog_loop_niters): Likewise.
312 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
313 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
315 (vect_recog_mask_conversion_pattern): Likewise.
316 (vect_try_gather_scatter_pattern): Likewise.
317 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
318 to vect_get_store_cost.
319 (vect_get_store_cost): Get stmt_info instead of DR.
320 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
321 (vect_get_load_cost): Get stmt_info instead of DR.
323 2018-06-01 Richard Biener <rguenther@suse.de>
326 * gimple-fold.c (var_decl_component_p): Also allow offsetted
327 vars wrapped in MEM_REFs.
329 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
331 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
332 Fix subreg tests so that we only return a choice between
333 GENERAL_REGS and FP_REGS if the original classes included both.
335 2018-06-01 Richard Biener <rguenther@suse.de>
338 * tree-ssa-structalias.c (get_function_part_constraint):
339 Handle NULL fi->decl.
340 (find_func_aliases_for_call): Properly handle indirect
342 (find_func_clobbers): Likewise.
343 (ipa_pta_execute): Likewise.
344 (create_variable_info_for): For functions that are ifunc_resolver
345 resolve to a varinfo that contains the result of the resolver
347 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
350 2018-05-31 Michael Collison <michael.collison@arm.com>
352 * config/aarch64/aarch64.md:
353 (*fix_to_zero_extenddfdi2): New pattern.
354 * gcc.target/aarch64/fix_extend1.c: New testcase.
356 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
360 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
361 and BUILT_IN_STRNCMP_EQ.
362 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
364 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
365 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
366 (gimple_fold_builtin): Likewise.
367 * tree-ssa-strlen.c (compute_string_length): New function.
368 (determine_min_obsize): New function.
369 (handle_builtin_string_cmp): New function to handle calls to
370 string compare functions.
371 (strlen_optimize_stmt): Add handling to builtin string compare
373 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
374 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
375 * tree.c (build_common_builtin_nodes): Add new defines of
376 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
378 2018-05-31 Jakub Jelinek <jakub@redhat.com>
381 * bb-reorder.c (pass_partition_blocks::gate): Return false for
382 functions with naked attribute.
384 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
386 * config/i386/sse.md (avx_vec_concat<mode>):
387 Substitute concat_tg_mode mode attribute with xtg_mode.
388 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
389 (concat_tg_mode): Remove mode attribute.
391 2018-05-31 Martin Sebor <msebor@redhat.com>
394 * calls.c (alloc_max_size): Correct a logic error/typo.
395 Treat excessive arguments as infinite. Warn for invalid arguments.
396 * doc/invoke.texi (-Walloc-size-larger-than): Update.
398 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
401 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
402 and movx for Haswell.
404 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
405 Cesar Philippidis <cesar@codesourcery.com>
408 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
409 when emitting error on private/firstprivate reductions.
410 * omp-low.c (lower_omp_target): Avoid reference-type processing
411 on pointers for firstprivate clause.
413 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
415 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
418 * config/aarch64/aarch64-simd.md
419 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
420 (aarch64_ld1_x3_<mode>): Likewise
421 (aarch64_st1x2<VALLDIF:mode>): Likewise
422 (aarch64_st1_x2_<mode>): Likewise
423 (aarch64_st1x3<VALLDIF:mode>): Likewise
424 (aarch64_st1_x3_<mode>): Likewise
425 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
426 (vld1_s8_x3): Likewise.
427 (vld1_u16_x3): Likewise.
428 (vld1_s16_x3): Likewise.
429 (vld1_u32_x3): Likewise.
430 (vld1_s32_x3): Likewise.
431 (vld1_u64_x3): Likewise.
432 (vld1_s64_x3): Likewise.
433 (vld1_f16_x3): Likewise.
434 (vld1_f32_x3): Likewise.
435 (vld1_f64_x3): Likewise.
436 (vld1_p8_x3): Likewise.
437 (vld1_p16_x3): Likewise.
438 (vld1_p64_x3): Likewise.
439 (vld1q_u8_x3): Likewise.
440 (vld1q_s8_x3): Likewise.
441 (vld1q_u16_x3): Likewise.
442 (vld1q_s16_x3): Likewise.
443 (vld1q_u32_x3): Likewise.
444 (vld1q_s32_x3): Likewise.
445 (vld1q_u64_x3): Likewise.
446 (vld1q_s64_x3): Likewise.
447 (vld1q_f16_x3): Likewise.
448 (vld1q_f32_x3): Likewise.
449 (vld1q_f64_x3): Likewise.
450 (vld1q_p8_x3): Likewise.
451 (vld1q_p16_x3): Likewise.
452 (vld1q_p64_x3): Likewise.
453 (vst1_s64_x2): Likewise.
454 (vst1_u64_x2): Likewise.
455 (vst1_f64_x2): Likewise.
456 (vst1_s8_x2): Likewise.
457 (vst1_p8_x2): Likewise.
458 (vst1_s16_x2): Likewise.
459 (vst1_p16_x2): Likewise.
460 (vst1_s32_x2): Likewise.
461 (vst1_u8_x2): Likewise.
462 (vst1_u16_x2): Likewise.
463 (vst1_u32_x2): Likewise.
464 (vst1_f16_x2): Likewise.
465 (vst1_f32_x2): Likewise.
466 (vst1_p64_x2): Likewise.
467 (vst1q_s8_x2): Likewise.
468 (vst1q_p8_x2): Likewise.
469 (vst1q_s16_x2): Likewise.
470 (vst1q_p16_x2): Likewise.
471 (vst1q_s32_x2): Likewise.
472 (vst1q_s64_x2): Likewise.
473 (vst1q_u8_x2): Likewise.
474 (vst1q_u16_x2): Likewise.
475 (vst1q_u32_x2): Likewise.
476 (vst1q_u64_x2): Likewise.
477 (vst1q_f16_x2): Likewise.
478 (vst1q_f32_x2): Likewise.
479 (vst1q_f64_x2): Likewise.
480 (vst1q_p64_x2): Likewise.
481 (vst1_s64_x3): Likewise.
482 (vst1_u64_x3): Likewise.
483 (vst1_f64_x3): Likewise.
484 (vst1_s8_x3): Likewise.
485 (vst1_p8_x3): Likewise.
486 (vst1_s16_x3): Likewise.
487 (vst1_p16_x3): Likewise.
488 (vst1_s32_x3): Likewise.
489 (vst1_u8_x3): Likewise.
490 (vst1_u16_x3): Likewise.
491 (vst1_u32_x3): Likewise.
492 (vst1_f16_x3): Likewise.
493 (vst1_f32_x3): Likewise.
494 (vst1_p64_x3): Likewise.
495 (vst1q_s8_x3): Likewise.
496 (vst1q_p8_x3): Likewise.
497 (vst1q_s16_x3): Likewise.
498 (vst1q_p16_x3): Likewise.
499 (vst1q_s32_x3): Likewise.
500 (vst1q_s64_x3): Likewise.
501 (vst1q_u8_x3): Likewise.
502 (vst1q_u16_x3): Likewise.
503 (vst1q_u32_x3): Likewise.
504 (vst1q_u64_x3): Likewise.
505 (vst1q_f16_x3): Likewise.
506 (vst1q_f32_x3): Likewise.
507 (vst1q_f64_x3): Likewise.
508 (vst1q_p64_x3): Likewise.
510 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
512 * config/msp430/msp430.c (msp430_output_labelref): Prepend
513 user_label_prefix to name.
515 * tree-core.h: Update comment about the format of NAME string
516 passed to handler in attribute_spec.
518 * config/msp430/msp430.md: Remove erroneous subreg expression from
519 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
520 zero_extend{q,h}isi2.
522 2018-05-30 Borislav Petkov <bp@suse.de>
524 * doc/extend.texi: Document some architecture specific
525 constraints and sort entries.
527 2018-05-30 Martin Sebor <msebor@redhat.com>
530 * builtins.c (expand_builtin_stpcpy_1): New function.
531 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
532 only if the former succeeds.
534 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
536 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
539 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
541 * doc/invoke.texi (-flinker-output): Document
543 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
545 * passes.c (ipa_write_summaries): Only modify statements if body
547 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
548 incrementally linking.
549 (ipa_passes): Likewise.
550 * lto-cgraph.c (lto_output_node): When incrementally linking do not
551 pass down resolution info.
552 * common.opt (flag_incremental_link): Update info.
553 * gcc.c (plugin specs): Turn flinker-output=* to
554 -plugin-opt=-linker-output-known
555 * toplev.c (compile_file): Also cut compilation when doing incremental
557 * flag-types. (enum lto_partition_model): Add
558 LTO_LINKER_OUTPUT_NOLTOREL.
559 (invoke.texi): Add -flinker-output docs.
560 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
561 link same way as WPA; do not stream in dead initializers.
563 * dwarf2out.c (dwarf2out_die_ref_for_decl,
564 darf2out_register_external_decl): Support incremental link.
566 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
568 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
570 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
572 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
573 it down to simple_object_copy_lto_debug_sections.
574 (run_gcc): Determine incremental LTO link time and configure
575 lto1 into non-wpa mode, disable renaming of debug sections.
577 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
579 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
580 descriptions of various incorrectly documented functions.
582 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
585 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
586 address check not strict.
588 2018-05-30 Richard Biener <rguenther@suse.de>
590 PR tree-optimization/85964
591 * tracer.c (better_p): Drop initialized count check, we only
592 call the function with initialized counts now.
593 (find_best_successor): Do find a best edge if one
594 has uninitialized count.
595 (find_best_predecessor): Likewise. Do BB frequency check only
596 if count is initialized.
598 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
600 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
601 (aarch64_ldrstr_offset_compare): New.
602 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
603 load/store orderings.
604 (aarch64_gen_adjusted_ldpstp): Likewise.
606 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
608 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
609 Check for subset of GENERAL_REGS and FP_REGS.
610 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
613 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
615 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
616 and wi::to_poly_offset. Add the current offset and then check
617 whether the sum fits, rather than using an unchecked addition of
618 a checked term. Check for a shwi rather than a uhwi.
619 * expr.c (get_bit_range): Use tree_to_poly_uint64.
620 (store_constructor): Use poly_int_tree_p.
621 (expand_expr_real_1): Likewise.
622 * function.c (assign_temp): Likewise.
623 * fold-const.c (const_binop): Use poly_int_tree_p and
625 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
627 * ipa-icf-gimple.c (func_checker::compare_operand): Use
628 to_poly_offset for MEM offsets.
629 * ipa-icf.c (sem_variable::equals): Likewise.
630 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
631 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
632 wi::to_poly_offset for BIT_FIELD_REF offsets.
633 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
635 * var-tracking.c (emit_note_insn_var_location): Use
638 2018-05-29 Jim Wilson <jimw@sifive.com>
640 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
642 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
645 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
646 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
648 (sse4_1_round<mode>2): Use nonimmediate_operand
649 for operand 1 predicate.
651 2018-05-29 Martin Sebor <msebor@redhat.com>
652 Richard Biener <rguenther@suse.de>
655 * calls.c (get_size_range): Call determine_value_range instead
657 * tree-vrp.h (determine_value_range): Declared new function.
658 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
660 2018-05-29 Richard Biener <rguenther@suse.de>
662 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
663 sure to use non-pattern stmts for get_earlier_stmt arguments.
664 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
665 called on pattern stmts.
666 (get_later_stmt): Likewise.
668 2018-05-29 Martin Liska <mliska@suse.cz>
670 PR gcov-profile/85759
671 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
674 2018-05-29 Jakub Jelinek <jakub@redhat.com>
676 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
678 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
682 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
683 VEC_PACK_FLOAT_EXPR): New tree codes.
684 * tree-pretty-print.c (op_code_prio): Handle
685 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
686 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
687 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
688 * tree-inline.c (estimate_operator_cost): Likewise.
689 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
690 * fold-const.c (const_binop): Likewise.
691 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
692 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
693 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
694 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
695 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
696 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
697 * expr.c (expand_expr_real_2): Likewise.
698 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
699 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
700 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
702 * optabs.c (expand_widen_pattern_expr): For
703 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
704 sign from result type rather than operand's type.
705 (expand_binop_directly): For vec_packu_float_optab and
706 vec_packs_float_optab allow result type to be different from operand's
708 * optabs-tree.c (optab_for_tree_code): Handle
709 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
710 VEC_PACK_FLOAT_EXPR. Formatting fixes.
711 * tree-vect-generic.c (expand_vector_operations_1): Handle
712 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
714 * tree-vect-stmts.c (supportable_widening_operation): Handle
716 (supportable_narrowing_operation): Handle FLOAT_EXPR.
717 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
718 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
719 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
720 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
722 (vec_pack<floatprefix>_float_<mode>): New expander.
723 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
725 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
726 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
727 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
728 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
729 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
731 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
732 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
733 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
734 VEC_PACK_FLOAT_EXPR): Document.
736 2018-05-29 Richard Biener <rguenther@suse.de>
738 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
740 (stmt_vec_info_vec): Make pointer.
741 (init_stmt_vec_info_vec): Remove.
742 (free_stmt_vec_info_vec): Likewise.
743 (set_stmt_vec_info_vec): New function.
744 (free_stmt_vec_infos): Likewise.
745 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
746 (set_vinfo_for_stmt): Likewise.
747 (get_earlier_stmt): Likewise.
748 (get_later_stmt): Likewise.
749 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
750 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
751 (vec_info::~vec_info): Free stmt_vec_infos.
752 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
753 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
754 (pass_slp_vectorize::execute): Likewise.
755 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
756 (free_stmt_vec_info_vec): Likewise.
757 (set_stmt_vec_info_vec): New function.
758 (free_stmt_vec_infos): Likewise.
759 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
760 the global stmt_vec_info_vec.
761 * tree-parloops.c (gather_scalar_reductions): Use
762 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
765 2018-05-29 Richard Biener <rguenther@suse.de>
767 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
769 2018-05-29 Martin Liska <mliska@suse.cz>
770 David Malcolm <dmalcolm@redhat.com>
772 * vec.c (test_reverse): New.
773 (vec_c_tests): Add new test.
774 * vec.h (vl_ptr>::reverse): New function.
776 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
778 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
780 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
783 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
785 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
787 2018-05-28 Richard Biener <rguenther@suse.de>
789 PR tree-optimization/85933
790 * tree-vect-data-refs.c (vect_record_base_alignments): Only
791 look at stmts marked as vectorizable.
793 2018-05-28 Richard Biener <rguenther@suse.de>
795 PR tree-optimization/85934
796 * tree-vect-generic.c (expand_vector_operations_1): Hoist
797 vector boolean check before scalar optimization.
799 2018-05-28 Jakub Jelinek <jakub@redhat.com>
801 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
804 2018-05-28 Mark Wielaard <mark@klomp.org>
806 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
807 if it is an expression containing a minus sign.
809 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
811 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
813 2018-05-27 Paul Koning <ni1d@arrl.net>
815 * config/pdp11/pdp11.md (truncsihi2): Remove.
817 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
818 Chung-Ju Wu <jasonwucj@gmail.com>
820 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
822 (unaligned_store_dw): Ditto.
823 * config/nds32/nds32-memory-manipulation.c
824 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
825 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
826 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
827 (emit_setmem_word_loop): Rename to ...
828 (emit_setmem_doubleword_loop): ... this.
829 (nds32_gen_dup_4_byte_to_word_value): New function.
830 (nds32_gen_dup_8_byte_to_double_word_value): New function.
831 (nds32_expand_setmem_loop): Refine implementation.
832 (nds32_expand_setmem_loop_v3m): Ditto.
833 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
836 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
838 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
840 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
842 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
843 (nds32_init_machine_status): Initialize machine->attr_naked_p and
844 machine->attr_no_prologue_p.
845 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
846 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
847 (nds32_expand_epilogue): Consider attr_naked_p.
848 (nds32_expand_epilogue_v3pop): Likewise.
849 (nds32_can_use_return_insn): Likewise.
850 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
851 attr_no_prologue_p fields.
852 * config/nds32/nds32.opt (mret-in-naked-func): New option.
854 2018-05-27 Jakub Jelinek <jakub@redhat.com>
857 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
860 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
862 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
864 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
866 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
868 (*<floatsuffix>floatv2div2sf2): Rename to ...
869 (*float<floatunssuffix>v2div2sf2): ... this.
870 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
871 (float<floatunssuffix>v2div2sf2_mask): ... this.
872 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
873 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
874 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
876 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
878 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
880 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
882 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
884 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
886 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
887 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
888 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
889 gen_ufix_truncv8dfv8si2.
890 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
891 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
892 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
893 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
894 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
895 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
896 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
897 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
899 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
903 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
905 2018-05-25 Jim Wilson <jimw@sifive.com>
907 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
908 * config/riscv/riscv.c (struct machine_function): Add
909 interrupt_handler_p and attribute_checked_p fields.
910 (riscv_attribute_table): Add interrupt.
911 (riscv_interrupt_type_p): New.
912 (riscv_save_reg_p): Save extra regs for interrupt handler.
913 (riscv_use_save_libcall): Return false for interrupt handler.
914 (riscv_first_stack_step): Add forward declaration.
915 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
916 for interrupt handler with large frame. Use it for saved reg list.
917 (riscv_expand_prologue): Move flag_stack_usage_info support to
918 eliminate duplication.
919 (riscv_expand_epilogue): Generate mret for interrupt handler.
920 (riscv_epilogue_uses): New.
921 (riscv_can_use_return_insn): Return false for interrupt handler.
922 (riscv_function_ok_for_sibcall): Likewise.
923 (riscv_set_current_function): Add interrupt handler support.
924 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
925 * config/riscv/riscv.md (UNSPECV_MRET): New.
927 (riscv_frflags, riscv_fsflags): Use tab after opcode.
929 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
931 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
933 PR tree-optimization/85712
934 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
935 this candidate has already been replaced in-situ by a copy.
937 2018-05-25 Jason Merrill <jason@redhat.com>
939 PR c++/80485 - inline function non-zero address.
940 * symtab.c (nonzero_address): Check DECL_COMDAT.
942 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
945 * config/alpha/alpha.md (ashlsi3): New insn pattern.
946 (*ashlsi_se): Rename from *ashldi_se. Define as sign
947 extension of SImode operation. Use const123_operand predicate.
949 (*saddl_se_1): Ditto.
951 (*ssubl_se_1): Ditto.
952 * config/alpha/predicates.md (const123_operand): New predicate.
953 * config/alpha/constraints.md (P): Use IN_RANGE.
955 2018-05-25 Richard Biener <rguenther@suse.de>
957 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
959 (ref_maybe_used_by_stmt_p): Likewise.
960 (stmt_may_clobber_ref_p): Likewise.
961 (stmt_may_clobber_ref_p_1): Likewise.
962 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
964 (ref_maybe_used_by_stmt_p): Likewise.
965 (stmt_may_clobber_ref_p): Likewise.
966 (stmt_may_clobber_ref_p_1): Likewise.
967 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
968 the alias oracle to disambiguate DRs with stmts DR analysis
970 (vect_analyze_data_refs): Do not give up on not analyzable
971 DRs for BB vectorization. Remove code truncating the dataref
974 2018-05-25 Jakub Jelinek <jakub@redhat.com>
977 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
978 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
979 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
981 2018-05-25 Richard Biener <rguenther@suse.de>
983 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
984 function, combining stmt data ref gathering and fatal analysis
986 (vect_analyze_data_refs): Remove now redudnant code and simplify.
987 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
988 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
989 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
990 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
992 2018-05-25 Bin Cheng <bin.cheng@arm.com>
994 PR tree-optimization/85720
995 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
996 SCC if all partitions are builtins.
997 (version_loop_by_alias_check): New parameter. Generate cancelable
998 runtime alias check if all partitions are builtins.
999 (distribute_loop): Update call to above function.
1001 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1003 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
1004 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
1005 (parm_default_def_partition_arg): Ditto.
1006 (set_parm_default_def_partition): Ditto.
1007 (get_parm_default_def_partitions): Ditto and make it static.
1008 (get_undefined_value_partitions): Ditto and make it static.
1009 (remove_ssa_form): Refactor call to init_var_map here.
1010 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
1011 computation for loop region.
1012 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
1013 (register_default_def): Delete.
1014 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
1015 (parm_default_def_partition_arg): Ditto.
1016 (set_parm_default_def_partition): Ditto.
1017 (get_parm_default_def_partitions): Ditto and make it static.
1018 (get_undefined_value_partitions): Ditto and make it static.
1019 (coalesce_with_default, coalesce_with_default): Update comment.
1020 (create_coalesce_list_for_region): New func factored out from
1021 create_outofssa_var_map.
1022 (populate_coalesce_list_for_outofssa): New func factored out from
1023 create_outofssa_var_map and coalesce_ssa_name.
1024 (create_outofssa_var_map): Delete.
1025 (coalesce_ssa_name): Refactor to support live range computation.
1026 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
1027 (get_parm_default_def_partitions): Delete.
1028 (get_undefined_value_partitions): Ditto.
1029 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
1030 computation for loop region.
1031 (new_tree_live_info, loe_visit_block): Ditto.
1032 (live_worklist, set_var_live_on_entry): Ditto.
1033 (calculate_live_on_exit, verify_live_on_entry): Ditto.
1034 * tree-ssa-live.h (struct _var_map): New fields.
1035 (init_var_map): Change decl.
1036 (region_contains_p): New.
1038 2018-05-25 Bin Cheng <bin.cheng@arm.com>
1040 * tree-ssa-live.h (live_merge_and_clear): Delete.
1042 2018-05-25 Richard Biener <rguenther@suse.de>
1045 * tree-dump.c (dequeue_and_dump): Remove access to removed
1046 operand 2 of a SWITCH_EXPR.
1048 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1050 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1051 multiplication and division.
1052 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1053 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1054 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1055 (cond_udiv_optab, cond_umod_optab): New optabs.
1056 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1057 (IFN_COND_RDIV): New internal functions.
1058 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1059 TRUNC_MOD_EXPR and RDIV_EXPR.
1060 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1061 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1063 (SVE_INT_BINARY): Include mult.
1064 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1065 (optab, sve_int_op): Handle mult.
1066 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1068 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1069 for SVE_INT_BINARY_SD.
1071 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1073 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1074 (optab, sve_int_op): Handle div and udiv.
1075 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1076 for SVE_INT_BINARY_SD.
1077 (*<optab><mode>3): New insn for the same.
1079 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1081 * tree-vect-patterns.c: Include predict.h.
1082 (vect_recog_divmod_pattern): Restrict check for division support
1083 to when optimizing for size.
1085 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1087 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1088 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1089 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1090 (gimple_match_op::set_op): Likewise.
1091 (gimple_resimplify4): Declare.
1092 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1093 (expr::gen_transform): Likewise.
1094 (decision_tree::gen): Generate a simplification routine for 4 operands.
1095 * gimple-match-head.c (gimple_simplify): Add an overload for
1096 4 operands. In the top-level function, handle up to 4 call
1097 arguments and call gimple_resimplify4.
1098 (gimple_resimplify4): New function.
1099 (build_call_internal): Pass a fourth operand.
1100 (maybe_push_to_seq): Likewise.
1101 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1102 Fold VEC_COND_EXPRs of an operation and a default value into
1103 an IFN_COND_* function if possible.
1104 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1106 (SVE_COND_FP_BINARY): Include them.
1107 (optab, sve_fp_op): Handle them.
1108 (SVE_INT_BINARY_REV): New code iterator.
1109 (SVE_COND_FP_BINARY_REV): New int iterator.
1110 (commutative): New int attribute.
1111 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1113 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1115 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1116 (*cond_<optab><mode>): New patterns for reversed operands.
1118 2018-05-25 Richard Biener <rguenther@suse.de>
1120 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1121 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1122 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1123 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1124 * tree-vect-data-refs.c (everywhere): Adjust users.
1125 * tree-vect-loop.c (everywhere): Likewise.
1126 * tree-vect-slp.c (everywhere): Likewise.
1127 * tree-vect-stmts.c (everywhere): Likewise.
1128 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1130 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1132 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1134 (gcc_cv_as_section_exclude): ... this.
1135 Try Solaris as #exclude syntax.
1136 * configure: Regenerate.
1137 * config.in: Regenerate.
1138 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1140 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1141 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1143 * varasm.c (default_elf_asm_named_section): Don't check if
1144 HAVE_GAS_SECTION_EXCLUDE is defined.
1146 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1148 * doc/md.texi: Update the documentation of the cond_* optabs
1149 to mention the new final operand. Fix GET_MODE_NUNITS call.
1150 Describe the scalar case too.
1151 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1152 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1154 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1155 (get_conditional_internal_fn): Update comment.
1156 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1157 accumulator value as a final argument to conditional functions.
1158 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1159 a define_expand and add an "else" operand. Assert for now that
1160 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1161 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1162 (*cond_<optab><mode>): New patterns.
1163 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1164 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1165 (UNSPEC_COND_EOR): Delete.
1166 (optab): Remove associated mappings.
1167 (SVE_INT_BINARY): New code iterator.
1168 (sve_int_op): Remove int attribute and add "minus" to the code
1170 (SVE_COND_INT_OP): Delete.
1171 (SVE_COND_FP_OP): Rename to...
1172 (SVE_COND_FP_BINARY): ...this.
1174 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1176 * optabs.c (can_reuse_operands_p): New function.
1177 (maybe_legitimize_operands): Try to reuse the results for
1180 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1182 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1183 Add {q} suffix to insn mnemonic.
1185 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1187 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1188 (msp430_warn_func_return): New.
1190 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1192 * fold-const.c (tree_nonzero_bits): New function.
1193 * fold-const.h (tree_nonzero_bits): Likewise.
1194 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1195 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1197 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1201 * varasm.c (assemble_alias): Check ifunc_resolver only on
1204 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1207 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1208 when memory input operand is handled.
1210 2018-05-24 Luis Machado <luis.machado@linaro.org>
1212 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1214 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1216 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1218 * match.pd: Delay FMA folds until after vectorization.
1220 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1223 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1224 address check not strict.
1226 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1228 * gimple-match.h (gimple_match_op): New class.
1229 (mprts_hook): Replace parameters with a gimple_match_op *.
1230 (maybe_build_generic_op): Likewise.
1231 (gimple_simplified_result_is_gimple_val): Replace parameters with
1232 a const gimple_match_op *.
1233 (gimple_simplify): Replace code_helper * and tree * parameters with
1234 a gimple_match_op * parameter.
1235 (gimple_resimplify1): Replace code_helper *, tree and tree *
1236 parameters with a gimple_match_op * parameter.
1237 (gimple_resimplify2): Likewise.
1238 (gimple_resimplify3): Likewise.
1239 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1240 parameters with a gimple_match_op * parameter.
1241 * gimple-match-head.c (gimple_simplify): Change prototypes of
1242 auto-generated functions to take a gimple_match_op * instead of
1243 separate code_helper * and tree * parameters. Make the same
1244 change in the top-level overload and update calls to the
1245 gimple_resimplify routines. Update calls to the auto-generated
1246 functions and to maybe_push_res_to_seq in the publicly-facing
1247 operation-specific gimple_simplify overloads.
1248 (gimple_match_op::MAX_NUM_OPS): Define.
1249 (gimple_resimplify1): Replace rcode and ops with a single res_op
1250 parameter. Update call to gimple_simplify.
1251 (gimple_resimplify2): Likewise.
1252 (gimple_resimplify3): Likewise.
1253 (mprts_hook): Replace parameters with a gimple_match_op *.
1254 (maybe_build_generic_op): Likewise.
1255 (build_call_internal): Replace type, nargs and ops with
1256 a gimple_match_op *.
1257 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1258 with a single gimple_match_op *. Update calls to mprts_hook,
1259 build_call_internal and gimple_simplified_result_is_gimple_val.
1260 Factor out code that is common to the tree_code and combined_fn cases.
1261 * genmatch.c (expr::gen_transform): Replace tem_code and
1262 tem_ops with a gimple_match_op called tem_op. Update calls
1263 to the gimple_resimplify functions and maybe_push_res_to_seq.
1264 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1265 res_ops. Update call to the gimple_resimplify functions.
1266 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1267 (decision_tree::gen): Make the functions take a gimple_match_op *
1268 called res_op instead of separate res_code and res_ops parameters.
1269 Update call accordingly.
1270 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1271 and ops with a single res_op parameter. Update calls to
1272 maybe_build_generic_op and maybe_push_res_to_seq.
1273 (fold_stmt_1): Update calls to gimple_simplify and
1274 replace_stmt_with_simplification.
1275 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1276 and gimple_simplified_result_is_gimple_val.
1277 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1279 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1280 with a gimple_match_op *.
1281 (vn_nary_build_or_lookup): Likewise. Update call to
1282 vn_nary_build_or_lookup_1.
1283 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1284 gimple_match_op *. Update calls to the gimple_resimplify routines
1285 and to gimple_simplified_result_is_gimple_val.
1286 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1287 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1288 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1289 (visit_nary_op): Likewise.
1290 (visit_reference_op_load): Likewise.
1292 2018-05-23 Luis Machado <luis.machado@linaro.org>
1294 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1295 modifier for printing the step amount.
1297 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1300 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1303 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1305 * doc/sourcebuild.texi (Endianness): New subsubsection.
1307 2018-05-23 Luis Machado <luis.machado@linaro.org>
1309 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1310 <prefetch_dynamic_strides>: New const bool field.
1311 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1312 prefetch_dynamic_strides.
1313 (exynosm1_prefetch_tune): Likewise.
1314 (thunderxt88_prefetch_tune): Likewise.
1315 (thunderx_prefetch_tune): Likewise.
1316 (thunderx2t99_prefetch_tune): Likewise.
1317 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1319 (aarch64_override_options_internal): Update to set
1320 PARAM_PREFETCH_DYNAMIC_STRIDES.
1321 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1322 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1323 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1324 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1325 prefetch-dynamic-strides setting.
1327 2018-05-23 Luis Machado <luis.machado@linaro.org>
1329 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1330 <minimum_stride>: New const int field.
1331 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1332 minimum_stride field defaulting to -1.
1333 (exynosm1_prefetch_tune): Likewise.
1334 (thunderxt88_prefetch_tune): Likewise.
1335 (thunderx_prefetch_tune): Likewise.
1336 (thunderx2t99_prefetch_tune): Likewise.
1337 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1338 <default_opt_level>: Set to 3.
1339 (aarch64_override_options_internal): Update to set
1340 PARAM_PREFETCH_MINIMUM_STRIDE.
1341 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1342 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1343 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1344 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1345 stride is constant and is below the minimum stride threshold.
1347 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1349 * config/arm/arm-cpus.in (mode26): Delete.
1350 (armv4): Delete mode26 reference.
1351 * config/arm/arm.c (arm_configure_build_target): Delete use of
1354 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1356 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1358 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1359 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1361 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1362 (floatunsdidf2): Ditto.
1364 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1366 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1367 (fixuns_trunc<mode>si2_avx512f): Ditto.
1368 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1369 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1370 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1372 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1374 PR rtl-optimization/79985
1375 * df-scan.c (df_insn_refs_collect): Remove special case for
1376 global registers and asm statements.
1378 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1380 * extend.texi (Global Register Variables): Rewrite the bullet list.
1381 Note that the register is available for allocation. Note that access
1382 via inline asm must use constraints. Add note about async-signal
1383 handlers. Remove paragraph about automagic register selection.
1385 2018-05-23 Richard Biener <rguenther@suse.de>
1387 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1388 of fixed offset from memset VN.
1390 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1392 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1394 (alloc_cand_and_find_basis): Initialize first_interp field.
1395 (slsr_process_mul): Modify first_interp field.
1396 (slsr_process_add): Likewise.
1397 (slsr_process_cast): Modify first_interp field for each new
1399 (slsr_process_copy): Likewise.
1400 (dump_candidate): Dump first_interp field.
1401 (replace_mult_candidate): Process all interpretations, not just
1403 (replace_rhs_if_not_dup): Likewise.
1404 (replace_one_candidate): Likewise.
1406 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1408 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1410 (aarch64_needs_frame_chain): New function.
1411 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1413 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1416 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1417 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1418 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1419 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1420 as true for strict-align.
1421 (aarch64_can_inline_p): Perform checks even when callee has no
1422 attributes to check for strict alignment.
1423 * doc/extend.texi (AArch64 Function Attributes): Document
1425 * doc/invoke.texi: (AArch64 Options): Likewise.
1427 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1429 PR tree-optimization/85853
1430 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1431 the handling of the root of the node to...
1432 (vect_slp_analyze_node_operations_1): ...this new function,
1433 and run the whole thing with the child nodes' def types
1434 set according to their SLP node's def type.
1436 2018-05-23 Richard Biener <rguenther@suse.de>
1439 * tree-data-ref.c (create_runtime_alias_checks): Defer
1440 and ignore overflow warnings.
1442 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1444 PR tree-optimization/85822
1445 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1448 2018-05-23 Richard Biener <rguenther@suse.de>
1450 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1451 memset constants via native_interpret_expr.
1453 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1456 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1458 (cgraph_node::create_alias): Likewise.
1459 (cgraph_node::get_availability): Check ifunc_resolver instead
1460 of looking up ifunc attribute.
1461 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1462 * varasm.c (do_assemble_alias): Likewise.
1463 (assemble_alias): Likewise.
1464 (default_binds_local_p_3): Likewise.
1465 * cgraph.h (cgraph_node): Add ifunc_resolver.
1466 (cgraph_node::only_called_directly_or_aliased_p): Return false
1468 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1470 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1471 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1472 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1473 instead of looking up ifunc attribute.
1475 2018-05-22 Luis Machado <luis.machado@linaro.org>
1477 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1479 2018-05-22 Martin Sebor <msebor@redhat.com>
1482 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1483 only when expasion succeeds.
1484 (expand_builtin_strcmp): Same.
1485 (expand_builtin_strncmp): Same.
1487 2018-05-22 Martin Sebor <msebor@redhat.com>
1489 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1491 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1492 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1494 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1495 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1497 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1498 Add check for legitimate_address.
1499 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1500 (aarch64_swap_ldrstr_operands): New.
1501 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1504 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1505 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1507 * config/aarch64/aarch64.md: New patterns to generate stp
1509 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1510 single words and double words.
1511 (load_pair_sw, load_pair_dw): Likewise.
1512 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1514 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1516 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1517 for different mode ldpstp and add peephole for merged zero stores.
1519 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1521 (aarch64_gen_store_pair): Rename calls to match new patterns.
1522 (aarch64_gen_load_pair): Rename calls to match new patterns.
1523 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1524 (load_pair<DREG:mode><DREG2:mode>): ... This.
1525 (store_pair<mode>): Rename to...
1526 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1527 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1529 (V_INT_EQUIV): Handle SImode.
1530 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1533 2018-05-22 Martin Sebor <msebor@redhat.com>
1536 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1537 or ajust the presumed bound on an operation to avoid unnecessary
1540 2018-05-22 Martin Sebor <msebor@redhat.com>
1542 PR tree-optimization/85826
1543 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1544 assuming that a DECL necesarily has a constant size.
1546 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1549 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1551 2018-05-22 Richard Biener <rguenther@suse.de>
1553 PR tree-optimization/85834
1554 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1555 non-constant and non-zero memset arguments.
1557 2018-05-22 Martin Liska <mliska@suse.cz>
1560 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1562 2018-05-22 Richard Biener <rguenther@suse.de>
1564 PR tree-optimization/85863
1565 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1566 comparisons when vectype is specified.
1567 (vectorizable_condition): Do not specify vectype for
1568 vect_is_simple_cond when SLP vectorizing.
1570 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1573 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1574 define __ibm128 as long double.
1575 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1576 as a distinct type when IEEE 128-bit support is enabled.
1577 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1578 128-bit types to use the correct functions.
1579 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1580 convert between 128-bit floating point types that have different
1581 modes but the same representation, instead of using gen_lowpart to
1583 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1585 (IFKF_reg): New attributes to give the register constraints for
1587 (extend<mode>tf2_internal): New insns to mark an explicit
1588 conversion between 128-bit floating point types that have a
1589 different mode but share the same representation.
1591 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1593 PR tree-optimization/85814
1594 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1595 a null return from get_strinfo when unsharing the next
1596 strinfo in the chain.
1598 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1601 * varasm.c (weak_finish): Clean up weak_decls.
1603 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1605 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1606 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1607 UNSPEC_UADALP values.
1608 * config/aarch64/iterators.md (ABAL): New int iterator.
1611 (sur): Add mappings for the above.
1612 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1614 (aarch64_<sur>abal<mode>_4): Likewise.
1615 (aarch64_<sur>adalp<mode>_3): Likewise.
1616 (<sur>sadv16qi): New define_expand.
1618 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1620 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1621 (*movdf_internal): Ditto.
1622 (*rcpsf2_sse): Ditto.
1623 (*rsqrtsf2_sse): Ditto.
1624 (*sqrt<mode>2_sse): Ditto.
1626 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1628 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1630 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1631 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1632 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1633 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1635 * config/aarch64/arm_neon.h: Likewise.
1636 * config/aarch64/iterators.md (VQ_I): New.
1638 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1640 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1642 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1645 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1647 * config/nds32/constraints.md (S): New constraint.
1648 * config/nds32/nds32.md (call_internal): Use constraint S.
1649 (call_value_internal): Likewise.
1650 (sibcall_internal): Likewise.
1651 (sibcall_value_internal): Likewise.
1653 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1654 Chung-Ju Wu <jasonwucj@gmail.com>
1656 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1659 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1660 Chung-Ju Wu <jasonwucj@gmail.com>
1662 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1663 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1664 (nds32_rtx_costs_impl): Simplify.
1665 (nds32_address_cost_impl): Simplify.
1666 (nds32_init_rtx_costs): New function.
1667 (nds32_rtx_costs_speed_prefer): Likewise.
1668 (nds32_rtx_costs_size_prefer): Likewise.
1669 (nds32_address_cost_speed_prefer): Likewise.
1670 (nds32_address_cost_speed_fwprop): Likewise.
1671 (nds32_address_cost_size_prefer): Likewise.
1672 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1673 * config/nds32/nds32.c (nds32_option_override): Use
1674 nds32_init_rtx_costs function.
1676 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1678 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1679 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1680 (TARGET_PIPELINE_N8): Likewise.
1681 (TARGET_PIPELINE_N10): Likewise.
1682 (TARGET_PIPELINE_N13): Likewise.
1683 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1685 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1687 * config/nds32/nds32-fpu.md: Update copyright year.
1689 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1691 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1693 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1695 * config/nds32/nds32.c
1696 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1697 * config/nds32/nds32.opt (minline-asm-r15): New option.
1699 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1701 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1703 * config/nds32/nds32.md (abssi2): New pattern.
1705 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1707 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1708 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1709 Merge insn pattern from sse_cvtsi2ss<round_name> and
1710 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1711 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1712 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1713 using SWI48 mode iterator.
1714 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1715 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1716 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1717 pattern from sse_cvttss2si<round_saeonly_name>
1718 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1719 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1720 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1721 using SWI48 mode iterator.
1722 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1723 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1724 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1725 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1726 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1727 using SWI48 mode iterator.
1728 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1729 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1730 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1731 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1732 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1733 SWI48 mode iterator.
1734 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1735 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1736 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1737 pattern from sse_cvttsd2si<round_saeonly_name>
1738 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1740 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1742 * config/nds32/nds32-md-auxiliary.c
1743 (nds32_valid_smw_lwm_base_p): Refine.
1744 (nds32_output_smw_single_word): Refine.
1745 (nds32_output_smw_double_word): New.
1746 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1748 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1750 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1751 (nds32_output_stack_pop): Refine.
1752 (nds32_expand_unaligned_load): Refine.
1753 (nds32_expand_unaligned_store): Refine.
1755 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1756 Chung-Ju Wu <jasonwucj@gmail.com>
1758 * config/nds32/constants.md: Add TP_REGNUM constant.
1759 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1760 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1762 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1763 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1764 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1765 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1766 and PIC code generation.
1767 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1769 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1771 * config/nds32/nds32.md: Support TLS and PIC.
1772 * config/nds32/nds32.c: Support TLS and PIC.
1773 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1774 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1777 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1779 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1780 mode with E_ prefix.
1782 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1783 Chung-Ju Wu <jasonwucj@gmail.com>
1785 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1786 * config/nds32/nds32-md-auxiliary.c
1787 (symbolic_reference_mentioned_p): New.
1788 (nds32_legitimize_ict_address): New.
1789 (nds32_expand_ict_move): New.
1790 (nds32_indirect_call_referenced_p): New.
1791 (nds32_symbol_binds_local_p): Delete.
1792 (nds32_long_call_p): Modify.
1793 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1794 * config/nds32/nds32-protos.h
1795 (symbolic_reference_mentioned_p): Declare.
1796 (nds32_legitimize_ict_address): Declare.
1797 (nds32_expand_ict_move): Declare.
1798 (nds32_indirect_call_referenced_p): Declare.
1799 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1800 (nds32_relax_group): Use nds32_ict_const_p as condition.
1801 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1802 (nds32_asm_file_start): Output ict_model directive in asm code.
1803 (nds32_legitimate_address_p): Consider indirect call.
1804 (nds32_print_operand): Consider indirect call.
1805 (nds32_print_operand_address): Consider indirect call.
1806 (nds32_insert_attributes): Handle "indirect_call" attribute.
1807 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1808 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1809 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1810 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1811 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1812 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1813 (TARGET_ICT_MODEL_SMALL): Define.
1814 (TARGET_ICT_MODEL_LARGE): Define.
1815 * config/nds32/nds32.md (movsi): Consider ict model.
1816 (call, call_value): Consider ict model.
1817 (sibcall, sibcall_value): Consider ict model.
1818 * config/nds32/nds32.opt (mict-model): New option.
1819 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1822 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1823 Monk Chiang <sh.chiang04@gmail.com>
1824 Jim Wilson <jimw@sifive.com>
1826 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1827 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1828 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1829 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1830 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1831 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1832 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1833 compute save_libcall_adjustment properly.
1834 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1835 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1836 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1837 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1838 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1839 (ABI_SPEC): Handle mabi=ilp32e.
1840 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1841 (RVE): Add RVE mask.
1842 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1843 <-march>: Add rv32e as an example.
1845 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1848 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1849 (intra_create_variable_infos): Handle C++ constructors.
1851 2018-05-18 Martin Liska <mliska@suse.cz>
1853 * passes.def: Remove a redundant pass.
1855 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1858 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1860 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1862 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1864 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1866 (armv2, armv2a, armv3, armv3m): Delete architectures.
1867 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1868 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1869 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1871 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1872 (*mulsidi3adddi): Likewise.
1873 (mulsidi3): Likewise.
1874 (*mulsidi3_nov6): Likewise.
1875 (umulsidi3): Likewise.
1876 (umulsidi3_nov6): Likewise.
1877 (umaddsidi4): Likewise.
1878 (*umulsidi3adddi): Likewise.
1879 (smulsi3_highpart): Likewise.
1880 (*smulsi3_highpart_nov6): Likewise.
1881 (umulsi3_highpart): Likewise.
1882 (*umulsi3_highpart_nov6): Likewise.
1883 * config/arm/arm.h (arm_arch3m): Delete.
1884 * config/arm/arm.c (arm_arch3m): Delete.
1885 (arm_option_override_internal): Update armv3-related comment.
1886 (arm_configure_build_target): Delete use of isa_bit_mode32.
1887 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1888 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1889 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1891 (mulusa3): Likewise.
1892 * config/arm/arm-protos.h (arm_arch3m): Delete.
1893 * config/arm/arm-tables.opt: Regenerate.
1894 * config/arm/arm-tune.md: Likewise.
1895 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1896 deleted architectures.
1898 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1900 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1901 (armv5t, armv5te): New features.
1902 (ARMv5, ARMv5e): Delete fgroups.
1903 (ARMv5t, ARMv5te): Adjust for above changes.
1905 (armv5, armv5e): Delete arches.
1906 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1908 (*call_reg_arm): Likewise.
1909 (*call_value_reg_armv5): Likewise.
1910 (*call_value_reg_arm): Likewise.
1911 (*call_symbol): Likewise.
1912 (*call_value_symbol): Likewise.
1913 (*sibcall_insn): Likewise.
1914 (*sibcall_value_insn): Likewise.
1916 (prefetch): Likewise.
1917 (define_split and define_peephole2 dependent on arm_arch5):
1919 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1921 (TARGET_ARM_QBIT): Likewise.
1922 (TARGET_DSP_MULTIPLY): Likewise.
1923 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1924 (arm_arch5, arm_arch5e): Delete.
1925 (arm_arch5t, arm_arch5te): Declare.
1926 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1927 (arm_arch5t): Declare.
1928 (arm_option_reconfigure_globals): Update for the above.
1929 (arm_options_perform_arch_sanity_checks): Update comment, replace
1930 use of arm_arch5 with arm_arch5t.
1931 (use_return_insn): Likewise.
1932 (arm_emit_call_insn): Likewise.
1933 (output_return_instruction): Likewise.
1934 (arm_final_prescan_insn): Likewise.
1935 (arm_coproc_builtin_available): Likewise.
1936 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1937 arm_arch5e with arm_arch5t and arm_arch5te.
1938 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1939 (arm_arch5t, arm_arch5te): Declare.
1940 * config/arm/arm-tables.opt: Regenerate.
1941 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1942 * config/arm/t-multilib: Likewise.
1943 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1944 instead of arm_arch5.
1945 (*call_reg_thumb1): Likewise.
1946 (*call_value_reg_thumb1_v5): Likewise.
1947 (*call_value_reg_thumb1): Likewise.
1948 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1950 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1952 2018-05-18 Martin Liska <mliska@suse.cz>
1954 PR gcov-profile/84846
1955 * doc/gcov.texi: Document -t option of gcov tool.
1957 2018-05-18 Martin Liska <mliska@suse.cz>
1959 PR gcov-profile/84846
1960 * gcov.c (print_usage): Add new -t option.
1961 (process_args): Handle the option.
1962 (generate_results): Use stdout as output when requested by
1965 2018-05-18 Martin Liska <mliska@suse.cz>
1967 PR gcov-profile/84846
1968 * coverage.c (coverage_init): Write PWD to .gcno file.
1969 * doc/gcov.texi: Document how working directory is printed.
1970 * gcov-dump.c (dump_gcov_file): Print PWD.
1971 * gcov.c (output_intermediate_file): Likewise.
1972 (read_graph_file): Read PWD string.
1973 (output_lines): Print PWD.
1975 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1978 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1979 for retval and return false if all args to phi are zero.
1981 2018-05-18 Richard Biener <rguenther@suse.de>
1983 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1985 (evrp_dom_walker::before_dom_children): Call it.
1987 2018-05-18 Richard Biener <rguenther@suse.de>
1989 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1990 results when processing array refs with variable index.
1992 2018-05-18 Toon Moene <toon@moene.org>
1994 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1995 directly after that of -floop-interchange. Indicate that both
1996 options are enabled by default when specifying -O3.
1998 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2000 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
2001 iterator. Delete separate integer-mode vec_set<mode> expander.
2002 (aarch64_simd_vec_setv2di): Delete.
2003 (vec_setv2di): Delete.
2004 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
2005 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
2006 the "w, r" alternative.
2008 2018-05-18 Martin Liska <mliska@suse.cz>
2010 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
2011 * tree-pass.h (make_pass_lower_switch_O0): New function.
2012 * tree-switch-conversion.c (node_has_low_bound): Remove.
2013 (node_has_high_bound): Likewise.
2014 (node_is_bounded): Likewise.
2015 (class pass_lower_switch): Make it a template type and create
2017 (pass_lower_switch::execute): Add template argument.
2018 (make_pass_lower_switch): New function.
2019 (make_pass_lower_switch_O0): New function.
2020 (do_jump_if_equal): Remove.
2021 (emit_case_nodes): Simplify to just handle all 3 cases and leave
2022 all the hard work to tree optimization passes.
2024 2018-05-18 Martin Liska <mliska@suse.cz>
2026 * dbgcnt.c (limit_low): Renamed from limit.
2027 (limit_high): New variable.
2028 (dbg_cnt_is_enabled): Check for upper limit.
2029 (dbg_cnt): Adjust dumping.
2030 (dbg_cnt_set_limit_by_index): Add new argument for high
2032 (dbg_cnt_set_limit_by_name): Likewise.
2033 (dbg_cnt_process_single_pair): Parse new format.
2034 (dbg_cnt_process_opt): Use strtok.
2035 (dbg_cnt_list_all_counters): Remove 'value' and add
2037 * doc/invoke.texi: Document changes.
2039 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
2041 * doc/sourcebuild.texi (scalar_all_fma): Document.
2042 * tree.def (FMA_EXPR): Delete.
2043 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
2044 * internal-fn.c (ternary_direct): New macro.
2045 (expand_ternary_optab_fn): Likewise.
2046 (direct_ternary_optab_supported_p): Likewise.
2047 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
2048 * builtins.c (fold_builtin_fma): Delete.
2049 (fold_builtin_3): Don't call it.
2050 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2051 * expr.c (expand_expr_real_2): Likewise.
2052 * fold-const.c (operand_equal_p): Likewise.
2053 (fold_ternary_loc): Likewise.
2054 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2055 * gimple.c (DEFTREECODE): Likewise.
2056 * gimplify.c (gimplify_expr): Likewise.
2057 * optabs-tree.c (optab_for_tree_code): Likewise.
2058 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2059 * tree-eh.c (operation_could_trap_p): Likewise.
2060 (stmt_could_throw_1_p): Likewise.
2061 * tree-inline.c (estimate_operator_cost): Likewise.
2062 * tree-pretty-print.c (dump_generic_node): Likewise.
2063 (op_code_prio): Likewise.
2064 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2065 * tree-ssa-operands.c (get_expr_operands): Likewise.
2066 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2067 * fold-const-call.h (fold_fma): Delete.
2068 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2069 CFN_FNMA and CFN_FNMS.
2071 * genmatch.c (combined_fn): New enum.
2072 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2073 (commutative_op): New function.
2074 (commutate): Use it. Handle more than 2 operands.
2075 (dt_operand::gen_gimple_expr): Use commutative_op.
2076 (parser::parse_expr): Allow :c to be used with non-binary
2077 operators if the commutative operand is known.
2078 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2079 CFN_FMS, CFN_FNMA and CFN_FNMS.
2080 (backprop::process_assign_use): Remove FMA_EXPR handling.
2081 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2082 (gen_hsa_fma): New function.
2083 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2084 IFN_FNMA and IFN_FNMS.
2085 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2086 * gimple-fold.h (follow_all_ssa_edges): Declare.
2087 * gimple-fold.c (follow_all_ssa_edges): New function.
2088 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2089 gimple_build interface and use follow_all_ssa_edges to fold the result.
2090 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2091 instead of checking for optabs directly.
2092 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2093 rather than FMA_EXPRs.
2094 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2095 call to IFN_FMA instead of an FMA_EXPR.
2097 2018-05-17 Jim Wilson <jimw@sifive.com>
2099 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2100 sign extended promoted subreg, and the range does not have the sign bit
2101 set, then do a sign extend.
2103 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2104 test, check for sign extended subreg and/or constant operands, and
2105 do a sign extend in that case.
2107 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2109 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2110 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2112 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2113 Change logics_shift_reg to logics_shift_imm.
2114 (thunderx2t99_fp_loadpair_basic): Delete.
2115 (thunderx2t99_fp_storepair_basic): Delete.
2116 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2117 (thunderx2t99_asimd_polynomial): Delete.
2118 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2119 and neon_fp_mul_d_scalar_q.
2120 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2121 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2122 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2123 (thunderx2t99_asimd_lut): Add missing tbl types.
2124 (thunderx2t99_asimd_ext): Delete.
2125 (thunderx2t99_asimd_load1_1_mult): Delete.
2126 (thunderx2t99_asimd_load1_2_mult): Delete.
2127 (thunderx2t99_asimd_load1_ldp): New.
2128 (thunderx2t99_asimd_load1): New.
2129 (thunderx2t99_asimd_load2): Add missing *load2* types.
2130 (thunderx2t99_asimd_load3): New.
2131 (thunderx2t99_asimd_load4): New.
2132 (thunderx2t99_asimd_store1_1_mult): Delete.
2133 (thunderx2t99_asimd_store1_2_mult): Delete.
2134 (thunderx2t99_asimd_store2_mult): Delete.
2135 (thunderx2t99_asimd_store2_onelane): Delete.
2136 (thunderx2t99_asimd_store_stp): New.
2137 (thunderx2t99_asimd_store1): New.
2138 (thunderx2t99_asimd_store2): New.
2139 (thunderx2t99_asimd_store3): New.
2140 (thunderx2t99_asimd_store4): New.
2142 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2144 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2145 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2147 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2148 Segher Boessenkool <segher@kernel.crashing.org>
2151 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2154 2018-05-17 Richard Biener <rguenther@suse.de>
2156 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2157 for pruning loop and prune defs feeding only already visited PHIs.
2159 2018-05-17 Richard Biener <rguenther@suse.de>
2161 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2163 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2164 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/85793
2167 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2168 for VMAT_ELEMENTWISE.
2170 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2172 * internal-fn.h (lookup_internal_fn): Declare
2173 * internal-fn.c (lookup_internal_fn): New function.
2174 * gimple.c (gimple_build_call_from_tree): Handle calls to
2176 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2177 internal function names.
2178 * tree-pretty-print.c (dump_generic_node): Likewise.
2179 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2181 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2183 * gimple-fold.h (gimple_build): Make the function forms take
2184 combined_fn rather than built_in_function.
2185 (gimple_simplify): Likewise.
2186 * gimple-match-head.c (gimple_simplify): Likewise.
2187 * gimple-fold.c (gimple_build): Likewise.
2188 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2189 rather than gimple_build_call_internal.
2190 (get_initial_defs_for_reduction): Likewise.
2191 (vect_create_epilog_for_reduction): Likewise.
2192 (vectorizable_live_operation): Likewise.
2194 2018-05-17 Martin Liska <mliska@suse.cz>
2196 * gimple-ssa-sprintf.c (format_directive): Do not use
2197 space in between 'G_' and '('.
2199 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2202 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2203 even if the mask is not all ones.
2206 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2208 (ix86_gimple_fold_builtin): Likewise.
2211 * config/i386/i386.c: Include tree-vector-builder.h.
2212 (ix86_vector_shift_count): New function.
2213 (ix86_fold_builtin): Fold shift builtins by scalar count.
2214 (ix86_gimple_fold_builtin): Likewise.
2216 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2217 _mm512_setzero): New intrinsics.
2219 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2220 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2222 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2223 code generation for cases where splatting a value is not useful.
2224 * simplify-rtx.c (simplify_ternary_operation): Simplify
2225 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2226 mode to a vec_concat.
2228 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2230 * config.gcc: Support "goldmont-plus".
2231 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2233 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2234 PROCESSOR_GOLDMONT_PLUS.
2235 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2236 (processor_target_table): Add "goldmont-plus".
2237 (PTA_GOLDMONT_PLUS): Define.
2238 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2239 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2240 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2241 (fold_builtin_cpu): Add "goldmont-plus".
2242 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2243 (ix86_option_override_internal): Add "goldmont-plus".
2244 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2245 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2246 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2247 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2249 2018-05-17 Richard Biener <rguenther@suse.de>
2251 PR tree-optimization/85757
2252 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2253 remove defs that only feed that PHI from further processing.
2255 2018-05-16 Jim Wilson <jimw@sifive.com>
2257 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2259 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2261 2018-05-16 Mark Wielaard <mark@klomp.org>
2263 * dwarf2out.c (count_index_strings): New function.
2264 (output_indirect_strings): Call count_index_strings and generate
2265 header for dwarf_version >= 5.
2267 2018-05-16 Mark Wielaard <mark@klomp.org>
2269 * dwarf2out.c (dwarf_FORM): New function.
2270 (set_indirect_string): Use dwarf_FORM.
2271 (reset_indirect_string): Likewise.
2272 (size_of_die): Likewise.
2273 (value_format): Likewise.
2274 (output_die): Likewise.
2275 (add_skeleton_AT_string): Likewise.
2276 (output_macinfo_op): Likewise.
2277 (index_string): Likewise.
2278 (output_index_string_offset): Likewise.
2279 (output_index_string): Likewise.
2280 (count_index_strings): Likewise.
2282 2018-05-16 Carl Love <cel@us.ibm.com>
2284 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2285 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2287 2018-05-16 Martin Jambor <mjambor@suse.cz>
2289 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2290 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2292 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2294 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2295 (fnma<mode>4): Likewise.
2296 (fms<mode>4): Likewise.
2297 (fnms<mode>4): Likewise.
2298 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2299 (aarch64_fnma<mode>4): Likewise.
2300 (aarch64_fms<mode>4): Likewise.
2301 (aarch64_fnms<mode>4): Likewise.
2302 (aarch64_fnmadd<mode>4): Likewise.
2304 2018-05-16 Jason Merrill <jason@redhat.com>
2306 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2308 2018-05-16 Richard Biener <rguenther@suse.de>
2310 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2311 (dump_stmt_cost): Declare.
2312 (add_stmt_cost): Dump cost we add.
2313 (add_stmt_costs): New function.
2314 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2316 (vect_analyze_stmt): Adjust prototype.
2317 (vectorizable_condition): Likewise.
2318 (vectorizable_live_operation): Likewise.
2319 (vectorizable_reduction): Likewise.
2320 (vectorizable_induction): Likewise.
2321 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2322 cost vector to pass to vectorizable_ and record afterwards.
2323 (vect_model_reduction_cost): Take cost vector argument and adjust.
2324 (vect_model_induction_cost): Likewise.
2325 (vectorizable_reduction): Likewise.
2326 (vectorizable_induction): Likewise.
2327 (vectorizable_live_operation): Likewise.
2328 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2329 SLP_TREE_NUMBER_OF_VEC_STMTS.
2330 (vect_analyze_slp_cost_1): Remove.
2331 (vect_analyze_slp_cost): Likewise.
2332 (vect_slp_analyze_node_operations): Take visited args and
2333 a target cost vector. Avoid processing already visited stmt sets.
2334 (vect_slp_analyze_operations): Use a local cost vector to gather
2335 costs and register those of non-discarded instances.
2336 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2337 (vect_schedule_slp_instance): Remove copying of
2338 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2340 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2341 adding cost. Record cost entry location.
2342 (vect_prologue_cost_for_slp_op): Function to compute cost of
2343 a constant or invariant generated for SLP vect in the prologue,
2344 split out from vect_analyze_slp_cost_1.
2345 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2346 (vect_model_promotion_demotion_cost): Likewise.
2347 (vect_model_store_cost): Likewise, make static.
2348 (vect_model_load_cost): Likewise.
2349 (vectorizable_bswap): Add cost vector arg and adjust.
2350 (vectorizable_call): Likewise.
2351 (vectorizable_simd_clone_call): Likewise.
2352 (vectorizable_conversion): Likewise.
2353 (vectorizable_assignment): Likewise.
2354 (vectorizable_shift): Likewise.
2355 (vectorizable_operation): Likewise.
2356 (vectorizable_store): Likewise.
2357 (vectorizable_load): Likewise.
2358 (vectorizable_condition): Likewise.
2359 (vectorizable_comparison): Likewise.
2360 (can_vectorize_live_stmts): Likewise.
2361 (vect_analyze_stmt): Likewise.
2362 (vect_transform_stmt): Adjust calls to vectorizable_*.
2363 * tree-vectorizer.c: Include gimple-pretty-print.h.
2364 (dump_stmt_cost): New function.
2366 2018-05-16 Richard Biener <rguenther@suse.de>
2368 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2369 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2370 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2371 (check_name): New callback.
2372 (dse_classify_store): Track cycles via a visited bitmap of PHI
2373 defs and simplify handling of in-loop and across loop dead stores
2374 and properly fail for loop-variant refs. Handle byte-tracking with
2375 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2378 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2380 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2381 (vect_get_mask_type_for_stmt): Likewise.
2382 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2384 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2385 to determine the statement's vector type and the vector type that
2386 should be used for calculating nunits. Deal with cases in which
2387 the type has to be deferred.
2388 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2389 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2390 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2391 (vect_determine_vf_for_stmt): New functions, split out from...
2392 (vect_determine_vectorization_factor): ...here.
2393 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2394 (vect_get_mask_type_for_stmt): New functions, split out from
2395 vect_determine_vectorization_factor.
2397 2018-05-16 Richard Biener <rguenther@suse.de>
2399 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2400 verify the [VEC_]COND_EXPR embedded comparison.
2402 2018-05-15 Martin Sebor <msebor@redhat.com>
2404 PR tree-optimization/85753
2405 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2406 RECORD_TYPE in addition to ARRAY_TYPE.
2408 2018-05-15 Martin Sebor <msebor@redhat.com>
2411 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2413 2018-05-15 Richard Biener <rguenther@suse.de>
2415 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2416 add by_clobber_p one. Change algorithm to collect all defs
2417 representing uses we need to walk and try reducing them to
2418 a single one before failing.
2419 (dse_dom_walker::dse_optimize_stmt): Adjust.
2421 2018-05-13 Mark Wielaard <mark@klomp.org>
2423 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2424 (size_of_loc_descr): Likewise.
2425 (output_loc_operands): Likewise.
2426 (output_loc_operands_raw): Likewise.
2427 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2428 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2429 (hash_loc_operands): Likewise.
2430 (compare_loc_operands): Likewise.
2432 2018-05-14 Mark Wielaard <mark@klomp.org>
2434 * dwarf2out.c (count_index_addrs): New function.
2435 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2437 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2439 PR tree-optimization/83648
2440 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2441 return value as malloc candidate.
2443 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2446 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2447 as true in call to suggest_attribute.
2449 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2451 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2452 -mreadonly-in-sdata.
2454 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2456 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2458 (aarch64_crypto_aesd_fused): Likewise.
2460 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2462 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2463 (movsi_aarch64): Likewise.
2464 (load_pairsi): Likewise.
2465 (load_pairdi): Likewise.
2466 (store_pairsi): Likewise.
2467 (store_pairdi): Likewise.
2468 (load_pairsf): Likewise.
2469 (load_pairdf): Likewise.
2470 (store_pairsf): Likewise.
2471 (store_pairdf): Likewise.
2472 (zero_extend): Likewise.
2473 (trunc): Swap alternatives.
2474 (fcvt_target): Add '?' to prefer w over r.
2476 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2479 * config/i386/i386.md: Disallow non-commutative arithmetics in
2480 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2481 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2482 in the peephole2 before it.
2484 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2486 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2487 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2488 (ix86_handle_option): Handle -mcldemote.
2489 * config.gcc: New header.
2490 * config/i386/cldemoteintrin.h: New file.
2491 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2492 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2494 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2495 OPTION_MASK_ISA_CLDEMOTE.
2496 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2497 (ix86_valid_target_attribute_inner_p): Ditto.
2498 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2499 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2500 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2501 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2502 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2504 * config/i386/i386.opt: Add -mcldemote.
2505 * config/i386/x86intrin.h: New header.
2506 * doc/invoke.texi: Add -mcldemote.
2508 2018-05-14 Richard Biener <rguenther@suse.de>
2510 * doc/match-and-simplify.texi: Adjust :s documentation.
2512 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2514 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2515 intended memcpy size.
2516 (REORDER_45): Likewise.
2518 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2520 * sort.cc: New file.
2521 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2522 * vec.c (qsort_chk): Use gcc_qsort.
2523 * Makefile.in (OBJS-libcommon): Add sort.o.
2524 (build/sort.o): New target. Use it...
2525 (BUILD_RTL): ... here, and...
2526 (build/gencfn-macros): ... here, and...
2527 (build/genmatch): ... here.
2529 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2530 Chung-Ju Wu <jasonwucj@gmail.com>
2532 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2533 * config/nds32/nds32-graywolf.md: New file.
2534 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2535 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2537 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2538 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2539 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2540 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2541 * config/nds32/pipelines.md: Include n15 settings.
2543 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2544 Chung-Ju Wu <jasonwucj@gmail.com>
2546 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2547 * config/nds32/nds32-n13.md: New file.
2548 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2549 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2551 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2552 * config/nds32/nds32.md (pipeline_model): Add n13.
2553 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2554 * config/nds32/pipelines.md: Include n13 settings.
2556 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2557 Chung-Ju Wu <jasonwucj@gmail.com>
2559 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2560 * config/nds32/nds32-n10.md: New file.
2561 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2562 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2564 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2565 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2566 * config/nds32/nds32.md (pipeline_model): Add n10.
2567 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2568 * config/nds32/pipelines.md: Include n10 settings.
2570 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2571 Kito Cheng <kito.cheng@gmail.com>
2572 Chung-Ju Wu <jasonwucj@gmail.com>
2574 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2575 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2576 Add enum values for DSP extension instructions.
2577 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2579 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2580 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2582 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2583 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2584 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2585 * config/nds32/nds32-intrinsic.md: Likewise.
2586 * config/nds32/nds32_intrinsic.h: Likewise.
2587 * config/nds32/nds32-md-auxiliary.c: Likewise.
2588 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2589 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2590 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2591 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2592 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2593 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2594 TYPE_DMAC in switch statement.
2595 * config/nds32/nds32.c: New checking and implementation for DSP
2596 extension instructions.
2597 * config/nds32/nds32.h: Likewise.
2598 * config/nds32/nds32.md: Likewise.
2599 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2600 * config/nds32/predicates.md: Implement new predicates for DSP
2603 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2605 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2606 Reformat alternatives and attributes so it is easier to identify
2607 which constraints/attributes go with which instruction.
2608 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2609 (mov<mode>_softfloat32, FMOVE64): Likewise.
2610 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2611 (mov<mode>_softfloat64, FMOVE64): Likewise.
2613 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2615 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2617 (Basic PowerPC Built-in Functions): The new name of the
2618 subsection previously known as "PowerPC Built-in Functions".
2619 (Basic PowerPC Built-in Functions Available on all Configurations):
2621 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2622 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2623 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2624 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2626 2018-05-11 Martin Jambor <mjambor@suse.cz>
2629 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2632 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2635 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2637 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2639 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2640 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2641 (ix86_handle_option): Handle -mwaitpkg.
2642 * config.gcc: New header.
2643 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2644 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2645 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2647 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2648 OPTION_MASK_ISA_WAITPKG.
2649 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2650 (ix86_option_override_internal): Add PTA_WAITPKG.
2651 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2652 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2653 IX86_BUILTIN_TPAUSE.
2654 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2655 __builtin_ia32_umwait and __builtin_ia32_tpause.
2656 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2657 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2658 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2659 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2660 UNSPECV_TPAUSE): New.
2661 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2662 * config/i386/i386.opt: Add -mwaitpkg.
2663 * config/i386/waitpkgintrin.h: New file.
2664 * config/i386/x86intrin.h: New header.
2665 * doc/invoke.texi: Add -mwaitpkg.
2667 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2670 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2672 (cortex-m0): Use armv6s-m isa.
2673 (cortex-m0plus): Likewise.
2674 (cortex-m1): Likewise.
2675 (cortex-m0.small-multiply): Likewise.
2676 (cortex-m0plus.small-multiply): Likewise.
2677 (cortex-m1.small-multiply): Likewise.
2679 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2680 Jakub Jelinek <jakub@redhat.com>
2682 PR tree-optimization/85692
2683 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2684 source permute as well.
2686 2018-05-11 Martin Liska <mliska@suse.cz>
2689 * doc/extend.texi: Document LLVM style format for no_sanitize
2692 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2694 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2695 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2696 (mode_supports_vsx_dform_quad): Likewise.
2697 (mode_supports_vmx_dform): Move these functions to be next to the
2698 other mode_supports functions.
2699 (mode_supports_dq_form): Likewise.
2700 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2701 mode_supports_dq_form.
2702 (reg_offset_addressing_ok_p): Likewise.
2703 (offsettable_ok_by_alignment): Likewise.
2704 (rs6000_legitimate_offset_address_p): Likewise.
2705 (legitimate_lo_sum_address_p): Likewise.
2706 (rs6000_legitimize_address): Likewise.
2707 (rs6000_legitimize_reload_address): Likewise.
2708 (rs6000_secondary_reload_inner): Likewise.
2709 (rs6000_preferred_reload_class): Likewise.
2710 (rs6000_output_move_128bit): Likewise.
2712 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2714 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2715 Generate SImode target register for null target.
2716 <case IX86_BUILTIN_XGETBV>: Ditto.
2717 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2718 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2720 2018-05-10 Carl Love <cel@us.ibm.com>
2722 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2723 dcbtt and dcbtstt if operands[2] is 0.
2725 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2728 * config/i386/sse.md (usadv64qi): New expander.
2730 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2732 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2733 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2734 -maltivec=be support.
2735 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2736 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2737 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2738 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2739 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2740 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2741 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2742 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2743 altivec_vsumsws): Adjust.
2744 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2745 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2746 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2748 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2749 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2750 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2751 (altivec_lve<VI_char>x): Delete expand.
2752 (*altivec_lve<VI_char>x_internal): Rename to...
2753 (altivec_lve<VI_char>x): ... this.
2754 (altivec_lvxl_<mode>): Delete expand.
2755 (*altivec_lvxl_<mode>_internal): Rename to ...
2756 (altivec_lvxl_<mode>): ... this.
2757 (altivec_stvxl_<mode>): Delete expand.
2758 (*altivec_stvxl_<mode>_internal): Rename to ...
2759 (altivec_stvxl_<mode>): ... this.
2760 (altivec_stve<VI_char>x): Delete expand.
2761 (*altivec_stve<VI_char>x_internal): Rename to ...
2762 (altivec_stve<VI_char>x): ... this.
2763 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2764 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2765 reduc_plus_scal_<mode>): Adjust.
2766 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2768 (rs6000_cpu_cpp_builtins): Adjust.
2769 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2770 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2771 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2772 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2773 -maltivec=be support.
2774 (rs6000_split_vec_extract_var): Adjust.
2775 (rs6000_split_v4si_init): Adjust.
2776 (swap_selector_for_mode): Delete.
2777 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2778 altivec_expand_stvex_be): Delete.
2779 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2780 -maltivec=be support.
2781 (rs6000_gimple_fold_builtin): Ditto.
2782 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2784 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2785 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2786 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2787 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2788 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2789 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2790 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2791 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2792 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2793 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2794 anonymous split): Adjust.
2795 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2796 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2798 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2800 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2801 when --with-gxx-include-dir is also specified.
2802 * configure: Regenerate.
2804 2018-05-09 Jim Wilson <jimw@sifive.com>
2807 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2808 * config/riscv/t-withmultilib: New.
2809 * config/riscv/withmultilib.h: New.
2810 * doc/install.texi: Document RISC-V --with-multilib-list support.
2812 2018-05-09 Richard Biener <rguenther@suse.de>
2814 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2816 (vect_bb_vectorization_profitable_p): Adjust. Compute
2817 actual scalar cost using the cost vector and the add_stmt_cost
2820 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2822 PR rtl-optimization/85645
2823 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2824 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2826 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2828 PR rtl-optimization/85645
2829 * shrink-wrap.c (spread_components): Return a boolean saying if
2830 anything was changed.
2831 (try_shrink_wrapping_separate): Iterate spread_components until
2832 nothing changes anymore.
2834 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2836 PR rtl-optimization/85645
2837 * regrename.c (build_def_use): Also kill the chains that include the
2838 destination of a REG_CFA_REGISTER note.
2840 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2842 PR rtl-optimization/85645
2843 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2844 insn that has a REG_CFA_REGISTER note.
2846 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2848 * cfgexpand.c (expand_clobber): New function.
2849 (expand_gimple_stmt_1): Use it.
2850 * tree-vect-stmts.c (vect_clobber_variable): New function,
2852 (vectorizable_simd_clone_call): ...here.
2853 (vectorizable_store): Emit a clobber either side of an
2854 IFN_STORE_LANES sequence.
2855 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2857 2018-05-09 Tom de Vries <tom@codesourcery.com>
2860 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2861 (define_insn "trap_if_false"): Add exit after trap.
2863 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2865 PR rtl-optimization/85638
2866 * bb-reorder.c: Include common/common-target.h.
2867 (create_forwarder_block): New function extracted from...
2868 (fix_up_crossing_landing_pad): ...here. Rename into...
2869 (dw2_fix_up_crossing_landing_pad): ...this.
2870 (sjlj_fix_up_crossing_landing_pad): New function.
2871 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2872 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2873 from both partitions and exit the loop after one iteration.
2875 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2878 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2880 (Basic PowerPC Built-in Functions): The new name of the
2881 subsection previously known as "PowerPC Built-in Functions".
2882 (Basic PowerPC Built-in Functions Available on all Configurations):
2884 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2886 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2888 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2890 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2893 2018-05-08 Jim Wilson <jimw@sifive.com>
2895 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2896 (LD_EMUL_SUFFIX): New.
2897 (LINK_SPEC): Use it.
2899 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2901 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2903 (Basic PowerPC Built-in Functions): The new name of the
2904 subsection previously known as "PowerPC Built-in Functions".
2905 (Basic PowerPC Built-in Functions Available on all Configurations):
2907 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2909 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2911 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2913 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2916 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2919 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2920 after cmpelim optimization.
2922 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2924 * config.gcc: Support "goldmont".
2925 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2926 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2928 * config/i386/i386.c (m_GOLDMONT): Define.
2929 (processor_target_table): Add "goldmont".
2930 (PTA_GOLDMONT): Define.
2931 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2932 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2933 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2934 (fold_builtin_cpu): Add "goldmont".
2935 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2936 (ix86_option_override_internal): Add "goldmont".
2937 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2938 (processor_type): Add PROCESSOR_GOLDMONT.
2939 * config/i386/i386.md: Add CPU "glm".
2940 * config/i386/glm.md: New file.
2941 * config/i386/x86-tune.def: Add m_GOLDMONT.
2942 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2944 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2947 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2949 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2950 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2951 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2954 * config/i386/i386.c (ix86_fold_builtin): Handle
2955 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2958 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2959 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2961 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2964 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2965 (check_arch): Likewise.
2966 (check_fpu): Return the result rather than printing it.
2967 (end arch): Fix operator precedence.
2968 (end cpu): Likewise.
2969 (END): Print the result from check_fpu.
2971 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2972 Alan Hayward <alan.hayward@arm.com>
2973 David Sherwood <david.sherwood@arm.com>
2975 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2976 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2977 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2978 (*fcmuo<mode>_and): New patterns.
2980 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2982 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2983 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2984 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2985 (cmp_op, sve_imm_con): New code attributes.
2986 (SVE_COND_INT_CMP, imm_con): Delete.
2987 (cmp_op): Remove above unspecs from int attribute.
2988 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2990 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2991 comparison-specific unspecs.
2992 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2993 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2994 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2995 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2996 (*vec_fcm<cmp_op><mode>): Rename to...
2997 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2998 (*vec_fcmuo<mode>): Rename to...
2999 (*fcmuo<mode>): ...this and adjust likewise.
3000 (*pred_fcm<cmp_op><mode>): New pattern.
3001 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
3002 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
3004 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
3006 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
3007 (aarch64_emit_sve_predicated_cond): New function.
3008 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
3009 (aarch64_emit_unspec_cond_or): Replace with...
3010 (aarch64_emit_sve_or_conds): ...this new function. Use
3011 aarch64_emit_sve_ptrue_op for the individual comparisons and
3012 aarch64_emit_binop to OR them together.
3013 (aarch64_emit_inverted_unspec_cond): Replace with...
3014 (aarch64_emit_sve_inverted_cond): ...this new function. Use
3015 aarch64_emit_sve_ptrue_op for the comparison and
3016 aarch64_emit_unop to invert the result.
3017 (aarch64_expand_sve_vec_cmp_float): Update after the above
3018 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
3020 2018-05-07 Nathan Sidwell <nathan@acm.org>
3022 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
3023 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
3024 (Backwards Compatibility): Likewise.
3026 2018-05-07 Luis Machado <luis.machado@linaro.org>
3030 2018-05-07 Luis Machado <luis.machado@linaro.org>
3032 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3033 <prefetch_dynamic_strides>: New const bool field.
3034 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3035 prefetch_dynamic_strides.
3036 (exynosm1_prefetch_tune): Likewise.
3037 (thunderxt88_prefetch_tune): Likewise.
3038 (thunderx_prefetch_tune): Likewise.
3039 (thunderx2t99_prefetch_tune): Likewise.
3040 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3041 (aarch64_override_options_internal): Update to set
3042 PARAM_PREFETCH_DYNAMIC_STRIDES.
3043 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3044 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3045 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3046 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3047 prefetch-dynamic-strides setting.
3049 2018-05-07 Luis Machado <luis.machado@linaro.org>
3051 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3052 <minimum_stride>: New const int field.
3053 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3054 minimum_stride field.
3055 (exynosm1_prefetch_tune): Likewise.
3056 (thunderxt88_prefetch_tune): Likewise.
3057 (thunderx_prefetch_tune): Likewise.
3058 (thunderx2t99_prefetch_tune): Likewise.
3059 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3060 (aarch64_override_options_internal): Update to set
3061 PARAM_PREFETCH_MINIMUM_STRIDE.
3062 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3063 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3064 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3065 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3066 stride is constant and is below the minimum stride threshold.
3068 2018-05-07 Luis Machado <luis.machado@linaro.org>
3070 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3073 2018-05-07 Luis Machado <luis.machado@linaro.org>
3075 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3076 <prefetch_dynamic_strides>: New const bool field.
3077 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3078 prefetch_dynamic_strides.
3079 (exynosm1_prefetch_tune): Likewise.
3080 (thunderxt88_prefetch_tune): Likewise.
3081 (thunderx_prefetch_tune): Likewise.
3082 (thunderx2t99_prefetch_tune): Likewise.
3083 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3084 (aarch64_override_options_internal): Update to set
3085 PARAM_PREFETCH_DYNAMIC_STRIDES.
3086 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3087 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3088 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3089 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3090 prefetch-dynamic-strides setting.
3092 2018-05-07 Luis Machado <luis.machado@linaro.org>
3094 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3095 <minimum_stride>: New const int field.
3096 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3097 minimum_stride field.
3098 (exynosm1_prefetch_tune): Likewise.
3099 (thunderxt88_prefetch_tune): Likewise.
3100 (thunderx_prefetch_tune): Likewise.
3101 (thunderx2t99_prefetch_tune): Likewise.
3102 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3103 (aarch64_override_options_internal): Update to set
3104 PARAM_PREFETCH_MINIMUM_STRIDE.
3105 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3106 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3107 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3108 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3109 stride is constant and is below the minimum stride threshold.
3111 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3114 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3115 the type is addressable. Don't force op into register if it has
3118 2018-05-05 Roland McGrath <mcgrathr@google.com>
3121 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3122 any section for which we don't know a specific type it should have,
3123 regardless of name. Previously this was done only for the exact
3124 names ".init_array", ".fini_array", and ".preinit_array".
3125 (default_elf_asm_named_section): Add comment about
3126 relationship with default_section_type_flags and SECTION_NOTYPE.
3127 (get_section): Don't consider it a type conflict if one side has
3128 SECTION_NOTYPE and the other doesn't, as long as neither has the
3129 SECTION_BSS et al used in the default_section_type_flags logic.
3131 2018-05-05 Tom de Vries <tom@codesourcery.com>
3134 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3135 (workaround_barsyncs): New function.
3136 (nvptx_reorg): Use workaround_barsyncs.
3137 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3138 (define_expand "nvptx_membar_cta"): New define_expand.
3139 (define_insn "*nvptx_membar_cta"): New insn.
3141 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3143 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3144 To improve optimization opportunities.
3145 * builtin-types.def: The new needed builtin types for the above.
3147 2018-05-04 Richard Biener <rguenther@suse.de>
3149 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3150 * gimple-ssa-store-merging.c
3151 (imm_store_chain_info::output_merged_store): Remove redundant create,
3152 release split_store vector contents on failure.
3153 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3154 scalar stmt vector on cache hit.
3156 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3158 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3160 * config.gcc (powerpc-xilinx-eabi*): Remove.
3161 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3163 (fusion_addis_mem_combo_load): Ditto.
3164 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3166 (rs6000_cpu_cpp_builtins): Ditto.
3167 * config/rs6000/rs6000-linux.c
3168 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3169 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3170 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3172 (rs6000_setup_reg_addr_masks): Ditto.
3173 (rs6000_init_hard_regno_mode_ok): Ditto.
3174 (rs6000_option_override_internal): Ditto.
3175 (legitimate_lo_sum_address_p): Ditto.
3176 (rs6000_legitimize_address): Ditto.
3177 (rs6000_legitimize_reload_address): Ditto.
3178 (rs6000_legitimate_address_p): Ditto.
3179 (abi_v4_pass_in_fpr): Ditto.
3180 (setup_incoming_varargs): Ditto.
3181 (rs6000_gimplify_va_arg): Ditto.
3182 (rs6000_split_multireg_move): Ditto.
3183 (rs6000_savres_strategy): Ditto.
3184 (rs6000_emit_prologue_components): Ditto.
3185 (rs6000_emit_epilogue_components): Ditto.
3186 (rs6000_emit_prologue): Ditto.
3187 (rs6000_emit_epilogue): Ditto.
3188 (rs6000_elf_file_end): Ditto.
3189 (rs6000_function_value): Ditto.
3190 (rs6000_libcall_value): Ditto.
3191 * config/rs6000/rs6000.h: Ditto.
3192 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3193 (TARGET_MINMAX): ... this. New.
3194 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3195 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3196 (*movsi_internal1_single): Delete.
3197 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3198 mfpu=, mxilinx-fpu): Delete.
3199 * config/rs6000/singlefp.h: Delete.
3200 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3201 * config/rs6000/t-rs6000: Ditto.
3202 * config/rs6000/t-xilinx: Delete.
3203 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3204 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3205 (VStype_simple): Delete.
3206 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3207 * config/rs6000/xfpu.h: Delete.
3208 * config/rs6000/xfpu.md: Delete.
3209 * config/rs6000/xilinx.h: Delete.
3210 * config/rs6000/xilinx.opt: Delete.
3211 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3212 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3214 2018-05-04 Tom de Vries <tom@codesourcery.com>
3217 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3220 2018-05-04 Richard Biener <rguenther@suse.de>
3223 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3224 (expand_complex_div_wide): Likewise.
3225 (expand_complex_operations_1): Likewise.
3226 (expand_complex_libcall): Preserve EH info of the original stmt.
3227 (tree_lower_complex): Handle removed blocks.
3228 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3229 on complex multiplication and division libcall builtins.
3231 2018-05-04 Richard Biener <rguenther@suse.de>
3234 * fold-const.c (negate_expr_p): Restrict negation of operand
3235 zero of a division to when we know that can happen without
3237 (fold_negate_expr_1): Likewise.
3239 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3242 * real.h (real_nextafter): Declare.
3243 * real.c (real_nextafter): New function.
3244 * fold-const-call.c (fold_const_nextafter): New function.
3245 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3246 CASE_CFN_NEXTTOWARD.
3247 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3248 even when arg1_mode is different from arg0_mode.
3250 2018-05-03 Nathan Sidwell <nathan@acm.org>
3252 * doc/extend.texi (Deprecated Features): Remove
3254 (Backwards Compatibility): Likewise.
3255 * doc/invoke.texi (C++ Language Options): Likewise.
3256 (C++ Dialect Options): Likewise.
3258 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3261 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3262 _mm512_mask_mullox_epi64): New intrinsics.
3264 2018-05-03 Tom de Vries <tom@codesourcery.com>
3267 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3268 dump files): Add offload-tree.
3270 2018-05-03 Richard Biener <rguenther@suse.de>
3272 PR tree-optimization/85615
3273 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3274 to loops not nested in BBs loop father to avoid creating multi-entry
3277 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3279 PR tree-optimization/70291
3280 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3281 arguments. Change return type to tree. Emit libcall as a new
3282 statement rather than replacing existing one when inplace_p is true.
3283 (expand_complex_multiplication_components): New function.
3284 (expand_complex_multiplication): Expand floating-point complex
3285 multiplication using the above.
3286 (expand_complex_division): Rename inner_type parameter to type.
3287 Update expand_complex_libcall call-site.
3288 (expand_complex_operations_1): Update expand_complex_multiplication
3289 and expand_complex_division call-sites.
3291 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3294 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3295 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3296 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3297 the highest significant bit of the shift count mask is clear. In
3298 check whether and[sq]i3 is needed verify that all significant bits
3299 of the shift count other than the highest are set.
3301 2018-05-02 Tom de Vries <tom@codesourcery.com>
3304 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3305 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3306 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3307 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3308 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3309 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3310 __builtin_goacc_parlevel_size.
3312 2018-05-02 Richard Biener <rguenther@suse.de>
3314 PR tree-optimization/85597
3315 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3316 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3319 2018-05-02 Tom de Vries <tom@codesourcery.com>
3322 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3323 dump files): Add ltrans-tree.
3325 2018-05-02 Tom de Vries <tom@codesourcery.com>
3328 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3329 dump files): Add wpa-ipa.
3331 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3333 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3334 powerpc*-*-linux*paired* target.
3335 * config/rs6000/750cl.h: Delete.
3336 * config/rs6000/paired.h: Delete.
3337 * config/rs6000/paired.md: Delete.
3338 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3340 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3341 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3342 comment. Remove paired float support.
3343 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3344 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3346 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3347 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3349 * config/rs6000/rs6000.c: Remove paired float support.
3350 (paired_expand_vector_init, paired_expand_vector_move,
3351 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3352 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3353 paired_expand_builtin, paired_expand_predicate_builtin,
3354 paired_init_builtins): Delete.
3355 * config/rs6000/rs6000.h: Remove paired float support.
3356 * config/rs6000/rs6000.md: Remove paired float support.
3357 (move_from_CR_ov_bit): Delete.
3358 * config/rs6000/rs6000.opt (mpaired): Delete.
3359 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3360 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3362 2018-05-02 Richard Biener <rguenther@suse.de>
3365 * gimplify.c (gimplify_save_expr): When in SSA form allow
3366 SAVE_EXPRs to compute to SSA vars.
3368 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3371 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3372 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3373 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3374 clobber operands[2], instead use a new pseudo. Formatting fixes.
3376 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3378 PR tree-optimization/85586
3379 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3380 exit early for statements in the same group if the accesses are
3383 2018-05-02 Tom de Vries <tom@codesourcery.com>
3386 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3389 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3391 PR tree-optimization/85143
3392 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3394 2018-05-01 Tom de Vries <tom@codesourcery.com>
3397 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3398 not found" error message.
3400 2018-05-01 Tom de Vries <tom@codesourcery.com>
3403 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3404 * vec.c (test_ordered_remove_if): New function.
3405 (vec_c_tests): Call test_ordered_remove_if.
3406 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3407 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3408 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3409 VEC_ORDERED_REMOVE_IF.
3411 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3413 PR tree-optimization/82665
3414 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3415 pointer subtraction where arguments come from a memchr call.
3417 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3419 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3420 --push-state --as-needed and --pop-state instead of --as-needed and
3421 --no-as-needed if ld supports it.
3422 * configure: Regenerated.
3425 * doc/install.texi2html: Replace _002d with - and _002a with * in
3426 generated html files using sed.
3428 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3431 * gcc-rich-location.c (blank_line_before_p): New function.
3432 (use_new_line): New function.
3433 (gcc_rich_location::add_fixit_insert_formatted): New function.
3434 * gcc-rich-location.h
3435 (gcc_rich_location::add_fixit_insert_formatted): New function.
3437 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3439 * selftest.c (assert_streq): Rename "expected" and "actual" to
3440 "val1" and "val2". Extend NULL-handling to cover both inputs
3441 symmetrically, while still requiring both to be non-NULL for a pass.
3442 * selftest.h (assert_streq): Rename "expected" and "actual" to
3444 (ASSERT_EQ): Likewise.
3445 (ASSERT_EQ_AT): Likewise.
3446 (ASSERT_KNOWN_EQ): Likewise.
3447 (ASSERT_KNOWN_EQ_AT): Likewise.
3448 (ASSERT_NE): Likewise.
3449 (ASSERT_MAYBE_NE): Likewise.
3450 (ASSERT_MAYBE_NE_AT): Likewise.
3451 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3452 the assertion to pass.
3453 (ASSERT_STREQ_AT): Likewise.
3455 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3457 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3458 interaction with -pie.
3460 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3462 * selftest.h: Fix alphabetization of per-source-file selftest
3465 2018-04-30 Jason Merrill <jason@redhat.com>
3467 PR c++/61982 - dead stores to destroyed objects.
3468 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3471 2018-04-30 Jason Merrill <jason@redhat.com>
3473 * tree.c (build_clobber): New.
3474 * tree.h: Declare it.
3475 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3477 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3479 * diagnostic-show-locus.c (layout::layout): Update for
3480 location_get_source_line returning a char_span.
3481 (struct char_span): Move to input.h.
3482 (struct correction): Update for fields in char_span becoming
3484 (struct source_line): Update for location_get_source_line
3485 returning a char_span.
3486 (layout::print_line): Likewise.
3487 * edit-context.c (edited_file::print_content): Likewise.
3488 (edited_file::print_diff_hunk): Likewise.
3489 (edited_file::print_run_of_changed_lines): Likewise.
3490 (edited_file::get_num_lines): Likewise.
3491 (edited_line::edited_line): Likewise.
3492 * final.c (asm_show_source): Likewise.
3493 * input.c (location_get_source_line): Convert return type
3494 from const char * to char_span, losing the final "line_len"
3496 (dump_location_info): Update for the above.
3497 (get_substring_ranges_for_loc): Likewise. Use a char_span
3498 when handling the literal within the line.
3499 (test_reading_source_line): Update for location_get_source_line
3500 returning a char_span.
3501 * input.h (class char_span): Move here from
3502 diagnostic-show-locus.c, converting from a struct to a class.
3503 Make data members private.
3504 (char_span::operator bool): New.
3505 (char_span::length): New.
3506 (char_span::get_buffer): New.
3507 (char_span::operator[]): New.
3508 (char_span::subspan): Make const.
3509 (char_span::xstrdup): New.
3510 (location_get_source_line): Convert return type from const char *
3511 to char_span, losing the final "line_size" param.
3513 2018-04-30 Jan Hubicka <jh@suse.cz>
3515 * lto-wrapper.c (ltrans_priorities): New static var.
3516 (cmp_priority): New.
3517 (run_gcc): Read priorities and if doing parallel build order
3518 the Makefile by them.
3520 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3522 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3524 2018-04-30 Richard Biener <rguenther@suse.de>
3526 * tree-cfg.c (verify_address): Remove base argument, add
3527 flag whether to check TREE_ADDRESSABLE and do that.
3528 (verify_expr): Remove.
3529 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3530 (verify_gimple_assign_single): Likewise.
3531 (verify_gimple_switch): Likewise.
3532 (verify_expr_location_1): Dereference tp once. Add (disabled)
3533 piece from verify_expr.
3534 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3536 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3538 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3540 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3542 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3543 (small_data_pattern): Likewise.
3544 (arc_rewrite_small_data): Likewise.
3545 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3546 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3547 (get_symbol_alignment): New function.
3548 (legitimate_small_data_address_p): Likewise.
3549 (legitimate_scaled_address): Update, call
3550 legitimate_small_data_address_p.
3551 (output_sdata): New static variable.
3552 (arc_print_operand): Update how we handle small data operands.
3553 (arc_print_operand_address): Likewise.
3554 (arc_legitimate_address_p): Update, use
3555 legitimate_small_data_address_p.
3556 (arc_rewrite_small_data_p): Remove.
3557 (arc_rewrite_small_data_1): Likewise.
3558 (arc_rewrite_small_data): Likewise.
3559 (small_data_pattern): Likewise.
3560 (compact_sda_memory_operand): Update to use
3561 legitimate_small_data_address_p and get_symbol_alignment.
3562 (prepare_move_operands): Don't rewite sdata pattern.
3563 (prepare_extend_operands): Remove.
3564 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3566 (zero_extendqisi2): Likewise.
3567 (zero_extendhisi2): Likewise.
3568 (extendqihi2): Likewise.
3569 (extendqisi2): Likewise.
3570 (extendhisi2): Likewise.
3574 * config/arc/constraints.md (Usd): Change it to memory constraint.
3576 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3578 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3579 as source of std instructions.
3580 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3581 allow 6-bit constants as source for store instructions.
3582 (movdi_insn): Update instruction pattern to allow 6-bit constants
3583 as source for store instructions.
3585 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3587 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3589 2018-04-30 Nathan Sidwell <nathan@acm.org>
3590 Sandra Loosemore <sandra@codesourcery.com>
3592 * dumpfile.c (dump_open): Allow '-' for stdout.
3593 * doc/invoke.texi (Developer Options): Document dump filename
3594 determination early. Document stdin/stdout selection.
3596 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3598 Microblaze Target: PIC data text relative
3600 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3601 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3603 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3604 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3605 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3606 New addressing mode for data-text relative position indepenedent code.
3607 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3608 'ADDRESS_SYMBOLIC_TXT_REL'.
3609 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3610 (microblaze_legitimate_pic_operand): Exclude function calls from
3611 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3612 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3614 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3615 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3616 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3617 for 'address + offset'.
3618 (microblaze_expand_prologue): Add new function prologue call for
3620 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3621 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3622 table in case of TARGET_PIC_DATA_TEXT_REL.
3623 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3624 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3625 Add new macros 'UNSPEC_TEXT',
3626 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3627 + exclude function calls from 'UNSPEC_PLT' in case of data text
3629 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3630 new target hook for generating address diff vector tables in case of
3632 * doc/tm.texi : Regenerate.
3633 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3634 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3635 of addr diff vector generation.
3636 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3637 target hook definition.
3638 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3639 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3640 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3641 option for data text relative.
3643 2018-04-30 Richard Biener <rguenther@suse.de>
3645 * tree-chrec.h (evolution_function_is_constant_p): Remove
3647 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3649 2018-04-30 Richard Biener <rguenther@suse.de>
3652 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3654 2018-04-30 Richard Biener <rguenther@suse.de>
3656 PR tree-optimization/28364
3657 PR tree-optimization/85275
3658 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3659 copying first exit test.
3661 2018-04-28 Mark Wielaard <mark@klomp.org>
3663 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3665 (dwarf_AT): Handle DW_AT_addr_base.
3666 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3668 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3671 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3672 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3673 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3674 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3676 2018-04-28 Richard Biener <rguenther@suse.de>
3678 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3679 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3680 to reflect use. Only add interesting stmts.
3682 2018-04-27 Martin Jambor <mjambor@suse.cz>
3685 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3686 the jump function allows for passing through aggregate values.
3688 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3690 * input.h (in_system_header_at): Convert from macro to inline
3692 (from_macro_expansion_at): Likewise.
3693 (from_macro_definition_at): Likewise.
3695 2018-04-27 Jeff Law <law@redhat.com>
3697 * config.gcc: Mark tile* targets as deprecated/obsolete.
3699 2018-04-27 Richard Biener <rguenther@suse.de>
3701 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3704 2018-04-27 Richard Biener <rguenther@suse.de>
3706 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3708 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3710 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3711 with Yd constraint. Set "preferred_for_speed" attribute from
3712 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3714 (*movdi_internal): Ditto.
3715 (movti_interunit splitters): Remove
3716 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3717 (movdi_interunit splitters): Ditto.
3718 * config/i386/constraints.md (Ye): Remove.
3719 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3721 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3724 * config/aarch64/constraints.md (Usg): Limit to 31.
3727 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3729 PR tree-optimization/85529
3730 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3731 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3732 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3733 zero extension or masking of the MSB bit.
3734 (optimize_range_tests): Add FIRST_BB argument, pass it through
3735 to optimize_range_tests_var_bound.
3736 (maybe_optimize_range_tests, reassociate_bb): Adjust
3737 optimize_range_tests callers.
3739 2018-04-26 Richard Biener <rguenther@suse.de>
3740 Jakub Jelinek <jakub@redhat.com>
3742 * cgraph.h (symbol_table): Just declare debug method here.
3743 * symtab.c (symbol_table::debug): Define.
3745 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3747 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3749 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3751 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3752 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3753 (*movdi_internal): Substitute Yi and Yj constraint with x
3754 and Ym and Yn constraint with y constraint. Update "isa"
3755 attribute and set "preferred_for_speed" attribute from
3756 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3757 (*movsi_internal): Ditto.
3758 (*movdf_internal): Ditto.
3759 (*movsf_internal): Ditto.
3760 (*zero_extendsidi2): Ditto.
3761 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3762 (sse2_loadld): Ditto.
3763 (*vec_extract<ssevecmodelower>_0): Ditto.
3764 (*vec_extractv4si_0_zext_sse4): Ditto.
3765 (vec_concatv2di): Ditto.
3766 (*vec_dup<mode>): Ditto.
3767 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3768 * config/i386/constraints.md (Yi): Remove.
3773 2018-04-26 Nathan Sidwell <nathan@acm.org>
3775 * dumpfile.c (dump_open): New.
3776 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3777 (dump_finish): Detect stdio/stderr by value not name.
3779 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3781 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3783 2018-04-26 Tom de Vries <tom@codesourcery.com>
3786 * config/nvptx/nvptx.c (verify_neutering_jumps)
3787 (verify_neutering_labels): New function
3788 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3790 2018-04-26 Tom de Vries <tom@codesourcery.com>
3793 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3794 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3797 2018-04-26 Richard Biener <rguenther@suse.de>
3798 Tom de Vries <tom@codesourcery.com>
3801 * lto-streamer-out.c (output_function): Fixup loops if required to match
3802 discovery done in the reader.
3804 2018-04-26 Richard Biener <rguenther@suse.de>
3806 PR tree-optimization/85116
3807 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3808 have a loop exit from the single latch predecessor. Remove
3809 case of header with just condition.
3810 (ch_base::copy_headers): Exclude infinite loops from any
3812 (pass_ch::execute): Record exits.
3814 2018-04-26 Richard Biener <rguenther@suse.de>
3816 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3817 prologue cost vector and pass it to vect_get_load_cost.
3818 (vect_get_peeling_costs_all_drs): Likewise.
3819 (vect_peeling_hash_get_lowest_cost): Likewise.
3820 (vect_enhance_data_refs_alignment): Likewise.
3822 2018-04-26 Richard Biener <rguenther@suse.de>
3825 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3826 checking of integer<->pointer conversions.
3827 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3828 sign-/zero-extending pointer types.
3829 (expand_omp_for_static_chunk): Likewise.
3831 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3832 Jean Lee <xiaoyur347@gmail.com>
3834 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3835 (TARGET_ASAN_SHADOW_OFFSET): Define.
3836 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3837 true for -fsanitize=address.
3839 2018-04-25 Mark Wielaard <mark@klomp.org>
3841 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3844 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3846 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3847 than "alu", remove explicit "memory" and "imm_disp" attributes.
3848 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3851 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3852 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3853 gen_lowpart_no_emit.
3855 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3858 * config/i386/i386.c (ix86_expand_builtin): Change memory
3859 operand to XI, extend p0 to Pmode.
3860 * config/i386/i386.md: Change unspec volatile and operand
3861 1 mode to XI, change operand 0 mode to P.
3863 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3865 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3866 GET_MODE_MASK before any checking.
3867 (nds32_can_use_bset_p): Likewise.
3868 (nds32_can_use_btgl_p): Likewise.
3870 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3872 * config/nds32/nds32-doubleword.md: New define_split pattern for
3873 illegal register number.
3875 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3877 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3879 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3881 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3883 2018-04-25 Richard Biener <rguenther@suse.de>
3885 * lto-streamer.h (LTO_major_version): Bump to 8.
3887 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3889 * BASE-VER: Set to 9.0.0.
3891 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3893 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3894 in __abskf2 and __powikf2.
3896 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3899 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3900 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3901 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3902 Use the above on operand 2. Reindent.
3903 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3905 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3908 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3910 * config/i386/i386.opt (mcet): Removed.
3911 * doc/install.texi: Remove -mcet documentation.
3912 * doc/invoke.texi: Likewise.
3914 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3917 * doc/install.texi: Remove -mcet from bootstrap-cet.
3919 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3922 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3923 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3927 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3928 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3929 containing a CONST_VECTOR.
3931 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3933 * doc/install.texi: Update newlib dependency for nvptx.
3935 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3938 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3939 instead of INTVAL when shifting x left.
3941 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3943 PR tree-optimization/85478
3944 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3945 vect_grouped_store_supported for single element vectors.
3947 2018-04-24 Richard Biener <rguenther@suse.de>
3950 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3951 load cost increase to the case of non-constant step.
3953 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3956 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3957 destination if any_malformed_asm.
3959 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3962 * expr.c (store_field): In the bitfield case, if the value comes from
3963 a function call and is returned in registers by means of a PARALLEL,
3964 do not change the mode of the temporary unless BLKmode and VOIDmode.
3966 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3968 PR rtl-optimization/85423
3969 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3970 dependencies to debug insns when the previous insn is non-debug.
3972 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3974 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3975 enums into a single definition.
3976 (fls): Fix predicates and printing.
3979 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3981 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3982 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3983 and short u6 immediate.
3984 (check_if_valid_sleep_operand): Remove.
3985 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3987 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3989 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3990 flag_always_save_lp condition.
3991 * config/nds32/nds32.opt (malways-save-lp): New option.
3993 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3995 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3996 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3997 * config/nds32/nds32.h
3998 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3999 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
4001 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4003 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
4004 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
4006 2018-04-22 Shiva Chen <shiva0217@gmail.com>
4007 Chung-Ju Wu <jasonwucj@gmail.com>
4009 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
4011 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
4012 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
4014 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4016 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
4018 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4020 * config/nds32/nds32-protos.h (nds32_data_alignment,
4021 nds32_local_alignment): Declare.
4022 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
4023 nds32_local_alignment): New functions.
4024 (TARGET_CONSTANT_ALIGNMENT): Define.
4025 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
4027 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4029 * config/nds32/nds32.c
4030 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
4031 (TARGET_MODES_TIEABLE_P): Likewise.
4033 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
4035 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
4038 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
4039 Chung-Ju Wu <jasonwucj@gmail.com>
4041 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4042 for unaligned access.
4043 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4045 * config/nds32/nds32-intrinsic.md: Likewise.
4046 * config/nds32/nds32_intrinsic.h: Likewise.
4047 * config/nds32/nds32.h (nds32_builtins): Likewise.
4048 * config/nds32/nds32.opt (munaligned-access): New option.
4049 * config/nds32/nds32.c (nds32_asm_file_start): Display
4050 flag_unaligned_access status.
4052 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4054 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4055 -mno-relax is present.
4056 * config/riscv/linux.h (LINK_SPEC): Ditto.
4058 2018-04-20 Martin Sebor <msebor@redhat.com>
4061 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4063 (gimple_fold_builtin_stxcpy_chk): Same.
4064 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4066 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4069 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4070 __powikf2 when long double is IEEE 128-bit.
4072 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4074 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4075 step to make sure stack always aligned.
4077 2018-04-20 Carl Love <cel@us.ibm.com>
4080 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4081 size check for arg0.
4083 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4084 Tom de Vries <tom@codesourcery.com>
4087 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4088 Emit insns for calls too.
4089 (nvptx_find_par): Always look for worker-level predecessor insn.
4090 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4092 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4093 (nvptx_process_pars): Propagate frames for calls.
4095 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4098 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4100 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4101 (ix86_handle_option): Don't handle OPT_mibt.
4102 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4104 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4106 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4107 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4108 (ix86_target_macros): Define __CET__ with flag_cf_protection
4109 for -fcf-protection.
4110 * config/i386/i386.c (isa2_opts): Remove -mibt.
4111 * config/i386/i386.h (TARGET_IBT): Removed.
4112 (TARGET_IBT_P): Likewise.
4113 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4114 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4115 * config/i386/i386.opt (mcet): Update help message.
4118 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4119 -mcet as an alias for -mshstk.
4121 2018-04-20 Richard Biener <rguenther@suse.de>
4124 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4125 complexity by forcing a single use of the multiply operand.
4127 2018-04-20 Martin Jambor <mjambor@suse.cz>
4130 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4131 recursion dependency to only apply to non-clones.
4133 2018-04-20 Martin Jambor <mjambor@suse.cz>
4136 * ipa-cp.c (create_specialized_node): Check that clones of
4137 self-recursive edges exist during IPA-CP.
4139 2018-04-19 Toon Moene <toon@moene.org>
4141 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4144 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4146 PR tree-optimization/85467
4147 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4148 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4149 VECTOR_CST element to type.
4151 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4154 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4155 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4156 (builtin_longjmp): Likewise.
4157 (save_stack_nonlocal): New pattern.
4158 (restore_stack_nonlocal): Likewise.
4160 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4163 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4164 Replace ASM_OUTPUT_LABEL with fprintf.
4166 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4169 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4170 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4171 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4172 define __IBT__ and __SHSTK__ for -fcf-protection.
4173 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4175 (ix86_trampoline_init): Likewise.
4176 (x86_output_mi_thunk): Likewise.
4177 (ix86_notrack_prefixed_insn_p): Likewise.
4178 (ix86_option_override_internal): Don't disallow -fcf-protection.
4179 * config/i386/i386.md (rdssp<mode>): Also enable for
4181 (incssp<mode>): Likewise.
4182 (nop_endbr): Likewise.
4183 * config/i386/i386.opt (mcet): Change help message to built-in
4187 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4188 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4189 enable CET built-in functions.
4191 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4193 * common/config/i386/i386-common.c
4194 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4195 OPTION_MASK_ISA_MOVDIRI_UNSET,
4196 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4197 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4198 * config.gcc (movdirintrin.h): New header.
4199 * config/i386/cpuid.h (bit_MOVDIRI,
4200 bit_MOVDIR64B): New bits.
4201 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4203 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4204 (VOID, PVOID, PCVOID)): New function types.
4205 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4206 __builtin_ia32_directstoreu_u64,
4207 __builtin_ia32_movdir64b): New builtins.
4208 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4209 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4211 (ix86_valid_target_attribute_inner_p): Ditto.
4212 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4213 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4214 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4215 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4216 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4217 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4218 (movdiri<mode>, movdir64b_<mode>): New.
4219 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4220 * config/i386/immintrin.h: Include movdirintrin.h.
4221 * config/i386/movdirintrin.h: New file.
4222 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4224 2018-04-19 Richard Biener <rguenther@suse.de>
4227 * cfg.c (clear_bb_flags): When loop state says we have
4228 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4230 2018-04-19 Richard Biener <rguenther@suse.de>
4232 PR tree-optimization/84737
4233 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4234 copying restrict info.
4235 (vect_setup_realignment): Use it.
4236 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4237 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4238 the first DR to all generated stores.
4239 (vectorizable_load): Likewise for loads.
4241 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4243 PR tree-optimization/85446
4244 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4245 the integral and pointer types to have the same precision.
4247 * doc/install.texi: Document --disable-cet being the default and
4250 2018-04-18 Martin Liska <mliska@suse.cz>
4252 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4255 2018-04-18 Martin Liska <mliska@suse.cz>
4258 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4261 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4262 arguments if they are comparable.
4264 2018-04-18 Martin Liska <mliska@suse.cz>
4267 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4270 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4273 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4276 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4277 ENDBR after calling __morestack.
4279 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4282 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4283 by using gcc_base_ver to generate a gcc_driver_version, and use
4284 it when generating GCC_DRIVER_NAME.
4285 * configure: Regenerate.
4287 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4290 * config.gcc: Obsolete powerpc*-*-*spe*.
4292 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4295 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4296 (stabstr_D): Change type of unum from unsigned int to
4297 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4300 2018-04-17 Jim Wilson <jimw@sifive.com>
4303 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4304 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4305 Set arg_pointer_offset after using pretend_args_size.
4307 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4309 PR rtl-optimization/85431
4310 * dse.c (record_store): Ignore zero width stores.
4313 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4314 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4315 __builtin_stack_restore rather than after it.
4316 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4317 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4318 argument instead of virtual_dynamic_stack_rtx.
4320 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4322 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4324 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4325 Add note to error message to explain internal mapping of overloaded
4326 built-in function name to non-overloaded built-in function name.
4327 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4330 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4333 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4334 where the inputs overlap with the output.
4336 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4338 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4339 (=v, v) alternative and explicit "memory" attribute.
4340 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4341 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4343 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4344 "sselog1" type instead of "sselog".
4345 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4346 "sselog". Remove explicit "memory" attribute.
4347 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4348 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4350 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4351 "sselog1" type instead of "sselog". Remove explicit "memory"
4353 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4354 use "sselog1" type instead of "sselog". Remove explicit "memory"
4356 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4357 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4359 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4360 "sselog1" type instead of "sselog". Remove explicit "memory"
4362 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4363 use "sselog1" type instead of "sselog". Remove explicit "memory"
4367 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4370 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4373 2018-04-17 Martin Jambor <mjambor@suse.cz>
4376 * ipa-cp.c (create_specialized_node): Call
4377 expand_all_artificial_thunks if necessary.
4379 2018-04-17 Martin Liska <mliska@suse.cz>
4382 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4383 in message, remote space in between '_G' and '('.
4385 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4388 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4389 avx512f_vmcmp<mode>3<round_saeonly_name>,
4390 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4391 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4392 avx512f_rndscale<mode><round_saeonly_name>,
4393 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4394 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4395 Use %<iptr>2 instead of %2 for -masm=intel.
4396 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4397 avx512f_vcvttss2usi<round_saeonly_name>,
4398 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4400 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4401 avx512f_vcvttsd2usi<round_saeonly_name>,
4402 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4403 Use %q1 instead of %1 for -masm=intel.
4404 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4405 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4406 of %3 for -masm=intel.
4407 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4409 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4411 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4413 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4414 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4416 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4418 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4419 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4420 %g1 and one with %0 and %1.
4421 (avx512er_vmrcp28<mode><round_saeonly_name>,
4422 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4424 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4425 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4426 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4427 of %0 and %{%4%} for -masm=intel.
4428 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4429 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4430 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4431 order of %0 and %{%5%}%{z%} for -masm=intel.
4433 2018-04-17 Jan Hubicka <jh@suse.cz>
4436 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4438 2018-04-17 Martin Liska <mliska@suse.cz>
4441 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4442 for target_clone error message. Make default implementation
4443 clone to be a local declaration.
4444 (separate_attrs): Add new argument and check for an empty
4446 (expand_target_clones): Handle it.
4447 (ipa_target_clone): Make redirection just for target_clones
4450 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4451 Tom de Vries <tom@codesourcery.com>
4454 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4455 tiled basic blocks without omp continue statements.
4457 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4460 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4461 vec_extract expression as having side effects to make sure it gets
4464 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4467 * config/i386/i386.c (get_builtin_code_for_version): Check
4470 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4473 * gcc/config.gcc: Support "skylake".
4474 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4476 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4477 (processor_target_table): Add "skylake".
4478 (ix86_option_override_internal): Add "skylake".
4479 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4480 PROCESSOR_CANNONLAKE.
4481 (get_builtin_code_for_version): Fix priority for
4482 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4483 PROCESSOR_SKYLAKE-AVX512.
4484 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4485 (processor_type): Add PROCESSOR_SKYLAKE.
4487 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4488 Jason Merrill <jason@redhat.com>
4491 * convert.c (convert_to_integer_1): Use direct recursion for
4492 enumeral types and types with a precision less than the number
4493 of bits in their mode.
4495 2018-04-16 Julia Koval <julia.koval@intel.com>
4498 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4499 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4501 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4504 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4505 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4506 and -mno-direct-move.
4508 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4511 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4512 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4513 Ensure negative shifts result in {0}.
4515 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4517 PR rtl-optimization/79916
4518 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4519 regs (if any) to define how to gnerate SD moves when LRA is in
4522 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4524 PR rtl-optimization/85393
4525 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4526 * except.c (expand_dw2_landing_pad_for_region): Make static.
4527 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4528 a label and unconditional jump to old_bb, rather than
4529 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4532 PR rtl-optimization/85376
4533 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4534 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4535 instead of a specific value.
4537 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4538 Bin Cheng <bin.cheng@arm.com>
4540 PR tree-optimization/82965
4541 PR tree-optimization/83991
4542 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4543 by_profile_only parameter.
4544 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4545 information if the loop was predicted to iterate too many times.
4546 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4548 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4551 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4554 2018-04-13 Martin Liska <mliska@suse.cz>
4555 Jakub Jelinek <jakub@redhat.com>
4558 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4559 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4560 * builtins.c (expand_builtin_memory_copy_args): Use
4561 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4562 handle dest_addr == pc_rtx.
4564 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4567 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4568 asked to not generate direct moves.
4569 (fix_trunc<mode>si2_stfiwx): Similar.
4570 (fix_trunc<mode>si2_internal): Similar.
4572 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4575 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4576 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4577 lookup if dest in some wider mode is known to be const0_rtx and
4578 if so, record permanent equivalence for it to be ZERO_EXTEND of
4579 the narrower mode destination.
4581 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4583 * lto-streamer-out.c (output_function): Revert 259346.
4584 * omp-expand.c (expand_oacc_for): Likewise.
4586 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4588 PR rtl-optimization/85354
4589 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4590 * sel-sched.c (sel_global_init): ... here.
4592 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4595 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4596 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4597 mode for PE-COFF targets.
4598 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4599 (i386_pe_asm_lto_end): Likewise.
4600 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4601 (TARGET_ASM_LTO_END): Likewise.
4602 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4603 (i386_pe_asm_lto_start): New function.
4604 (i386_pe_asm_lto_end): Likewise.
4606 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4607 Richard Biener <rguenther@suse.de>
4610 * lto-streamer-out.c (output_function): Fix CFG loop state before
4612 * omp-expand.c (expand_oacc_for): Handle calls to internal
4613 functions like regular functions.
4615 2018-04-12 Richard Biener <rguenther@suse.de>
4618 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4619 for the early LTO debug to properly generate references to it
4620 during DIE emission. Do not re-use that for the skeleton for
4622 (dwarf2out_early_finish): Likewise.
4624 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4627 * config/i386/sse.md
4628 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4629 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4630 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4631 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4632 and output is a reg, avoid creating invalid lowpart subreg, but
4633 instead split into a 512-bit move. Don't split if not AVX512VL,
4634 input is xmm16+ reg and output is a mem.
4635 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4636 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4637 xmm16+ reg and output is a mem.
4639 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4641 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4642 also for flag_dwarf2_cfi_asm.
4644 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4646 PR rtl-optimization/85342
4647 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4648 a bool scalar var inside of the loop instead. Don't try to update
4649 recog_data.operand after failed apply_change_group.
4651 2018-04-12 Tom de Vries <tom@codesourcery.com>
4654 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4655 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4656 array with flexible array member as array without given dimension.
4657 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4658 argument for undefined param to true.
4660 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4663 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4664 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4665 from PowerPC section.
4666 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4667 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4668 help text that is too long.
4669 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4670 help text that is too long.
4671 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4672 help text that is too long.
4674 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4676 * config/alpha/alpha.md (stack_probe_internal): Rename
4677 from "probe_stack". Update all callers.
4679 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4681 PR rtl-optimization/84566
4682 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4683 sched_macro_fuse_insns.
4685 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4688 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4689 (compute_block_dependences): ... from here.
4691 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4693 PR tree-optimization/85331
4694 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4695 from int to HOST_WIDE_INT.
4697 2018-04-11 Martin Jambor <mjambor@suse.cz>
4700 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4701 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4702 not the same as the source val.
4703 (cgraph_edge_brings_value_p): New parameter.
4704 (gather_edges_for_value): Pass destination value to
4705 cgraph_edge_brings_value_p.
4706 (perhaps_add_new_callers): Likewise.
4707 (get_info_about_necessary_edges): Likewise and exclude values brought
4708 only by self-recursive edges.
4709 (create_specialized_node): Redirect only clones of self-calling edges.
4710 (+self_recursive_pass_through_p): New function.
4711 (find_more_scalar_values_for_callers_subset): Use it.
4712 (find_aggregate_values_for_callers_subset): Likewise.
4713 (known_aggs_to_agg_replacement_list): Removed.
4714 (decide_whether_version_node): Re-calculate known constants for all
4715 remaining context clones.
4717 2018-04-11 Richard Biener <rguenther@suse.de>
4720 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4721 from early DWARF output.
4722 (dwarf2out_early_finish): Output line info unconditionally into
4723 early DWARF and add reference to it.
4725 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4728 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4729 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4730 other than V2DFmode using iptr mode attribute.
4731 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4733 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4735 PR rtl-optimization/84659
4736 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4738 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4741 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4743 (output_loc_list): Pass address of a dummy size variable even in the
4744 locview handling loop.
4745 (index_location_lists): Add comment on why skip_loc_list_entry can't
4748 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4751 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4754 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4757 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4758 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4759 and -mstring-compare-inline-limit.
4761 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4764 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4765 for stack clash protection in a register whenever we need it to be in
4768 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4770 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4771 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4773 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4776 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4778 (mlong-double-): Ditto.
4779 * config/rs6000/sysv4.opt (msdata=): Ditto.
4780 (mtls-size=): Ditto.
4782 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4784 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4785 erroneous entries for
4786 "vector int vec_ldl (int, long int *)", and
4787 "vector unsigned int vec_ldl (int, unsigned long int *)".
4788 Add comments and entries for
4789 "vector bool char vec_ldl (int, bool char *)",
4790 "vector bool short vec_ldl (int, bool short *)",
4791 "vector bool int vec_ldl (int, bool int *)",
4792 "vector bool long long vec_ldl (int, bool long long *)",
4793 "vector pixel vec_ldl (int, pixel *)",
4794 "vector long long vec_ldl (int, long long *)",
4795 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4796 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4797 type tree bool_long_long_type_node and correct definition of
4798 bool_V2DI_type_node to make reference to this new type tree.
4799 (rs6000_mangle_type): Replace erroneous reference to
4800 bool_long_type_node with bool_long_long_type_node.
4801 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4802 comments to emphasize sign distinctions for char and int types and
4803 replace RS6000_BTI_bool_long constant with
4804 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4805 use of RS6000_BTI_pixel.
4806 (bool_long_type_node): Remove this macro definition.
4807 (bool_long_long_type_node): New macro definition
4809 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4811 PR rtl-optimization/85300
4812 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4813 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4814 simplify_unary_operation fails.
4816 2018-04-10 Martin Liska <mliska@suse.cz>
4818 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4819 cgraph_edge and ipa_ref.
4821 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4825 * config/i386/sse.md
4826 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4827 computation of the VEC_MERGE selector from mask.
4828 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4829 Fix decoding of the VEC_MERGE selector into mask.
4831 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4833 PR tree-optimization/85286
4834 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4836 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4838 * final.c (final_1): Set insn_last_address as well as
4839 insn_current_address.
4841 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4844 * explow.c (emit_stack_probe): Call validize_mem on memory location
4845 before passing it to gen_probe_stack. Create address operand and
4846 legitimize it for the probe_stack_address case.
4848 2018-04-09 Jan Hubicka <jh@suse.cz>
4851 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4852 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4853 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4855 (free_lang_data): Rebuild type inheritance graph.
4857 2018-04-09 Martin Sebor <msebor@redhat.com>
4859 * invoke.texi (-finline-small-functions): Mention other optimization
4861 (-findirect-inlining, -fpartial-inlining): Same.
4862 (-finline-functions-called-once): Same.
4863 (-freorder-blocks-and-partition): Same.
4865 2018-04-09 Jan Hubicka <jh@suse.cz>
4868 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4869 jumps; choose last target that matches the criteria (i.e.
4870 no partition changes for non-crossing jumps).
4871 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4872 support for redirecting crossing jumps to non-crossing.
4874 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4876 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4877 also for naked functions.
4879 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4881 * config/arc/arc.md (add_shift): New pattern.
4882 (add_shift2): Likewise.
4883 (sub_shift): Likewise.
4884 (sub_shift_cmp0_noout): Likewise.
4885 (compare_si_ashiftsi): Likewise.
4886 (xbfu_cmp0_noout): New combine pattern.
4887 (xbfu_cmp0"): Likewise.
4888 (movsi_set_cc_insn): Place the predicable variant first.
4889 (commutative_binary_cmp0_noout): Remove clobber.
4890 (commutative_binary_cmp0): New pattern.
4891 (noncommutative_binary_cmp0): Likewise.
4892 (noncommutative_binary_cmp0_noout): Likewise.
4893 (noncommutative_binary_comparison_result_used): Removed.
4894 (rsub_cmp0): New pattern.
4895 (rsub_cmp0_noout): Likewise.
4896 (extzvsi): Changed, keep only meaningful variants.
4897 (SQH, SEZ): New iterators.
4898 (SQH_postfix): New mode attribute.
4899 (SEZ_prefix): New code attribute.
4900 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4901 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4902 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4904 (noncommutative_operator): Check the availability of barrel
4907 2018-04-09 Richard Biener <rguenther@suse.de>
4909 PR tree-optimization/85284
4910 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4911 Only use the niter constraining form of simple_iv when the exit
4914 2018-04-09 Tom de Vries <tom@codesourcery.com>
4917 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4918 (define_expand "*memory_barrier"): New define_expand.
4919 (define_insn "memory_barrier"): New insn.
4921 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4923 PR rtl-optimization/80463
4924 PR rtl-optimization/83972
4925 PR rtl-optimization/83480
4927 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4928 correct producer for the insn.
4929 (tidy_control_flow): Fixup seqnos in case of debug insns.
4931 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4933 PR rtl-optimization/83913
4935 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4936 different sched-times when merging exprs.
4938 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4940 PR rtl-optimization/83962
4942 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4943 tidy_fallthru_edge and tidy_control_flow.
4945 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4947 PR rtl-optimization/83530
4949 * sel-sched.c (force_next_insn): New global variable.
4950 (remove_insn_for_debug): When force_next_insn is true, also leave only
4951 next insn in the ready list.
4952 (sel_sched_region): When the region wasn't scheduled, make another pass
4953 over it with force_next_insn set to 1.
4955 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4957 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4959 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4960 for interrupt control.
4961 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4962 functions for interrupt control.
4963 * config/nds32/nds32-intrinsic.md: Likewise.
4964 * config/nds32/nds32_intrinsic.h: Likewise.
4965 * config/nds32/nds32.h (nds32_builtins): Likewise.
4967 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config/nds32/nds32.c (nds32_init_machine_status,
4970 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4971 strict_aligned_p field.
4972 (nds32_expand_to_rtl_hook): New function.
4973 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4974 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4976 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4977 Chung-Ju Wu <jasonwucj@gmail.com>
4979 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4980 * config/nds32/nds32-n7.md: New file.
4981 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4982 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4984 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4985 * config/nds32/nds32.md (pipeline_model): Add n7.
4986 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4987 * config/nds32/pipelines.md: Include n7 settings.
4989 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4990 Chung-Ju Wu <jasonwucj@gmail.com>
4992 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4993 * config/nds32/nds32-e8.md: New file.
4994 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4995 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4997 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4998 * config/nds32/nds32.md (pipeline_model): Add e8.
4999 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
5000 * config/nds32/pipelines.md: Include e8 settings.
5002 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5003 Chung-Ju Wu <jasonwucj@gmail.com>
5005 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
5006 * config/nds32/nds32-n8.md: New file.
5007 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
5008 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
5010 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
5011 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
5012 * config/nds32/nds32.md (pipeline_model): Add n8.
5013 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
5014 * config/nds32/pipelines.md: Include n8 settings.
5016 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
5017 Chung-Ju Wu <jasonwucj@gmail.com>
5019 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
5020 * config/nds32/nds32-n9-2r1w.md: New file.
5021 * config/nds32/nds32-n9-3r2w.md: New file.
5022 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
5023 nds32_register_ports): New or modify for cpu n9.
5024 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
5026 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
5027 * config/nds32/nds32-utils.c: New file.
5028 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
5029 TARGET_MUL_SLOW): Define.
5030 * config/nds32/nds32.md (pipeline_model): New attribute.
5031 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
5032 New options that support cpu n9.
5033 * config/nds32/pipelines.md: Include n9 settings.
5034 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
5036 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
5038 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
5039 information if necessary.
5040 (output_cond_branch_compare_zero): Likewise.
5041 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
5042 (nds32_target_alignment): Refine for alignment.
5043 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
5044 (FUNCTION_BOUNDARY): Modify.
5045 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
5047 * config/nds32/nds32.opt (malways-align, malign-functions): New.
5049 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5051 * config/nds32/constants.md (unspec_volatile_element): Add values for
5052 TLB operation and data prefetch.
5053 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5054 functions for TLB operation and data prefetch.
5055 * config/nds32/nds32-intrinsic.md: Likewise.
5056 * config/nds32/nds32_intrinsic.h: Likewise.
5057 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5058 (nds32_print_operand): Likewise.
5059 * config/nds32/nds32.h (nds32_builtins): Likewise.
5061 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5062 Andrew Pinski <pinsika@gcc.gnu.org>
5065 * match.pd: Use constant_boolean_node of correct type instead of
5066 boolean_true_node or boolean_false_node for simplifying
5067 pointer comparisons to zero.
5069 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5071 PR tree-optimization/80021
5072 * tree.c (verify_type_variant): Make error call in verify_variant_match
5073 translatable and remove final full stop.
5075 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5077 * config/nds32/constants.md (unspec_volatile_element): Add
5078 UNSPEC_VOLATILE_EH_RETURN.
5079 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5080 nds32_output_stack_pop): Support dwarf exception handling process.
5081 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5082 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5083 exception handling process.
5084 (nds32_compute_stack_frame): Likewise.
5085 (nds32_return_addr_rtx): Likewise.
5086 (nds32_initial_elimination_offset): Likewise.
5087 (nds32_expand_prologue): Likewise.
5088 (nds32_expand_epilogue): Likewise.
5089 (nds32_dynamic_chain_address): New function.
5090 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5092 (DYNAMIC_CHAIN_ADDRESS): Define.
5093 (EH_RETURN_DATA_REGNO): Define.
5094 (EH_RETURN_STACKADJ_RTX): Define.
5095 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5096 patterns for dwarf exception handling.
5098 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5100 * config/nds32/nds32.h: Clean up obsolete macros.
5102 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5104 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5105 Add enum values for particular instructions.
5106 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5107 particular intrinsic functions.
5108 * config/nds32/nds32-intrinsic.md: Likewise.
5109 * config/nds32/nds32_intrinsic.h: Likewise.
5110 * config/nds32/nds32.h (nds32_builtins): Likewise.
5111 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5112 (btst, ave): New patterns for particular instructions.
5114 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5116 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5117 Add enum values for atomic load/store and memory sync.
5118 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5120 * config/nds32/nds32-intrinsic.md: Likewise.
5121 * config/nds32/nds32_intrinsic.h: Likewise.
5122 * config/nds32/nds32.h (nds32_builtins): Likewise.
5124 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5126 PR tree-optimization/85257
5127 * fold-const.c (native_encode_vector): If not all elts could fit
5128 and off is -1, return 0 rather than offset.
5129 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5130 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5131 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5132 adjust buffer in native_interpret_expr call.
5134 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5136 * config/nds32/constants.md (unspec_volatile_element): Add cache
5137 control enum values.
5138 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5139 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5140 * config/nds32/nds32.c (nds32_cctl_names): New.
5141 (nds32_print_operand): Handle cache control register names.
5142 * config/nds32/nds32.h (nds32_builtins): New enum values.
5143 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5145 * config/nds32/nds32.md (type): Add mmu.
5146 * config/nds32/pipelines.md (simple_insn): Add mmu.
5148 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5150 * config/nds32/nds32.md (type): Remove call.
5151 * config/nds32/pipelines.md (simple_insn): Likewise.
5153 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5155 * config/nds32/constants.md (unspec_volatile_element): Add
5156 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5157 UNSPEC_VOLATILE_FMFCFG.
5158 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5159 description for fmfcfg and fmfcsr.
5160 (bdesc_1arg): Add fmtcsr.
5161 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5162 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5163 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5164 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5165 unspec_fmfcfg): New patterns.
5166 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5167 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5168 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5169 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5170 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5171 __nds32__fmfcfg): Define.
5173 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5175 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5176 intrinsic register names.
5177 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5178 intrinsic register enum values and macros.
5180 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5182 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5183 for load/store addressing form.
5184 (nds32_print_operand_address): Likewise.
5186 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5189 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5190 based on LABEL_REF. Remove useless assertion.
5191 (pic_address_needs_scratch): Fix formatting.
5192 (sparc_legitimize_pic_address): Minor tweaks.
5193 (sparc_delegitimize_address): Adjust assertion accordingly.
5194 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5195 into symbolic_operand.
5196 (movsi_high_pic_label_ref): Likewise.
5197 (movsi_lo_sum_pic_label_ref): Likewise.
5198 (movdi_pic_label_ref): Likewise.
5199 (movdi_high_pic_label_ref): Likewise.
5200 (movdi_lo_sum_pic_label_ref): Likewise.
5202 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5204 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5205 custom LIB_SPEC setup.
5207 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5208 Kito Cheng <kito.cheng@gmail.com>
5210 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5211 * config/riscv/freebsd.h: New.
5213 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5215 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5216 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5219 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5220 Kito Cheng <kito.cheng@gmail.com>
5222 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5223 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5224 * config/nds32/nds32-protos.h (nds32_output_call,
5225 nds32_output_return): Declare.
5226 * config/nds32/nds32.md: Refine all the call and return patterns.
5228 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5231 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5232 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5234 PR rtl-optimization/84872
5235 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5236 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5239 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5241 * expr.c (copy_blkmode_to_reg): Revert 254862.
5242 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5244 2018-04-06 Richard Biener <rguenther@suse.de>
5247 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5248 after seeing a component reference with an adjacent field. Treat
5249 refs to arrays at struct end of external decls similar to
5250 refs to unconstrained commons.
5252 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5255 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5256 look through SAVE_EXPRs with non-side-effects argument. Adjust
5258 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5261 2018-04-06 Richard Biener <rguenther@suse.de>
5264 * alias.c (find_base_term): New wrapper around find_base_term
5265 unwinding CSELIB_VAL_PTR changes.
5266 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5269 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5271 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5273 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5274 constant definitions.
5275 ("nop"): lr 0,0 -> nopr r0
5276 ("nop_lr0", "nop_lr1"): New insn definitions.
5278 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5280 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5281 NDS32_V3PUSH_AVAILABLE_P macro.
5283 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5284 Chung-Ju Wu <jasonwucj@gmail.com>
5286 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5287 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5288 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5289 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5290 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5291 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5292 * config/nds32/constraints.md: New constraints and checking for hard
5293 float configuration.
5294 * config/nds32/iterators.md: New mode iterator and attribute for hard
5295 float configuration.
5296 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5298 * config/nds32/nds32-fpu.md: New file.
5299 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5300 deal with hard float code generation.
5301 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5303 (abi_type, float_reg_number): New enum type.
5304 * config/nds32/nds32-predicates.c: New predicates for hard float.
5305 * config/nds32/nds32-protos.h: Declare functions for hard float.
5306 * config/nds32/nds32.c: Implementation for hard float configuration.
5307 * config/nds32/nds32.h: Definitions for hard float configuration.
5308 * config/nds32/nds32.md: Include hard float machine description and
5309 modify patterns for hard float configuration.
5310 * config/nds32/nds32.opt: New options for hard float configuration.
5311 * config/nds32/predicates.md: New predicates for hard float
5314 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5316 * common/config/nds32/nds32-common.c
5317 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5319 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5322 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5323 CONSTRUCTOR_ELT (ctor, ...)->value.
5325 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5328 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5330 2018-04-05 Tom de Vries <tom@codesourcery.com>
5333 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5336 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5337 Kito Cheng <kito.cheng@gmail.com>
5339 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5340 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5341 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5342 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5344 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5345 Kito Cheng <kito.cheng@gmail.com>
5347 * config/nds32/constraints.md (Ufe): New memory constraint.
5348 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5349 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5350 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5352 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5353 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5355 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5357 * config/nds32/nds32.md: Use optimize_size in the condition for
5358 alu-shift instructions.
5360 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5362 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5364 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5366 * config/nds32/nds32.md (negsi2): Refine pattern.
5368 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5369 Chung-Ju Wu <jasonwucj@gmail.com>
5371 * config/nds32/iterators.md (shift_rotate): New code iterator.
5372 (shift): New code attribute.
5373 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5374 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5375 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5376 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5377 bit-wise operations.
5378 (andsi3, *andsi3): Ditto.
5379 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5380 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5381 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5382 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5383 nds32_ior_operand, nds32_xor_operand): New predicates.
5385 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5387 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5388 (addsi3, subsi3): ... this.
5390 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5392 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5394 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5396 * config/nds32/nds32.md: Adjust indention.
5398 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5400 * config/nds32/nds32.md (feature): New attribute.
5402 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5404 * config/nds32/nds32.md (subtype): New attribute.
5406 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5409 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5410 expansion to perform a bitwise AND of the argument followed by a
5411 boolean negation of the result.
5413 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5415 PR rtl-optimization/84878
5416 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5417 the basic block. Assert the use reference is not artificial and that
5418 it has an associated insn.
5420 2018-04-04 Michael Matz <matz@suse.de>
5422 * builtins.c (compute_objsize): Pass correct operand
5423 to array_at_struct_end_p.
5425 2018-04-04 Richard Biener <rguenther@suse.de>
5428 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5429 from contexts for DINFO_LEVEL_TERSE and below.
5431 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5433 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5434 resiter_operand condition.
5435 * config/nds32/nds32.md (*move<mode>): Ditto.
5437 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5438 Monk Chiang <sh.chiang04@gmail.com>
5440 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5442 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5444 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5446 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5447 Kito Cheng <kito.cheng@gmail.com>
5449 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5450 nds32_cond_code_str, output_cond_branch,
5451 output_cond_branch_compare_zero, nds32_expand_cbranch,
5452 nds32_expand_cstore, nds32_expand_movcc,
5453 nds32_output_cbranchsi4_equality_zero,
5454 nds32_output_cbranchsi4_equality_reg,
5455 nds32_output_cbranchsi4_equality_reg_or_const_int,
5456 nds32_output_cbranchsi4_greater_less_zero: New functions.
5457 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5458 nds32_expand_cstore, nds32_expand_movcc,
5459 nds32_output_cbranchsi4_equality_zero,
5460 nds32_output_cbranchsi4_equality_reg,
5461 nds32_output_cbranchsi4_equality_reg_or_const_int,
5462 nds32_output_cbranchsi4_greater_less_zero): Declare.
5463 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5464 nds32_rimm11s_operand): New predicates.
5465 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5466 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5469 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5471 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5472 * config/nds32/nds32.md: Ditto.
5473 * config/nds32/pipelines.md: Ditto.
5475 2018-04-04 Richard Biener <rguenther@suse.de>
5477 PR tree-optimization/85168
5478 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5479 propagating abnormals.
5481 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5483 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5485 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5486 Kito Cheng <kito.cheng@gmail.com>
5488 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5489 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5490 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5491 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5492 * config/nds32/nds32.md (sibcall_internal): New.
5493 (sibcall_register): Remove.
5494 (sibcall_immediate): Remove.
5495 (sibcall_value_internal): New.
5496 (sibcall_value_register): Remove.
5497 (sibcall_value_immediate): Remove.
5498 * config/nds32/predicates.md (nds32_general_register_operand): New.
5499 (nds32_call_address_operand): New.
5501 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5503 PR rtl-optimization/85167
5504 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5505 bb_defs if *split_p, instead preinitialize it to NULL.
5507 PR tree-optimization/85156
5508 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5509 evaluating the argument multiple times.
5511 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5513 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5515 (_mm_cvtpd_ps): Likewise.
5516 (_mm_cvttpd_epi32): Likewise.
5517 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5518 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5519 vector, pixel, and bool following altivec.h include.
5521 2018-04-03 Martin Sebor <msebor@redhat.com>
5523 * doc/extend.texi (Common Function Attributes): Clarify.
5524 (const attribute): Likewise.
5525 (pure attribute): Likewise.
5527 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5530 * config/i386/i386.c (ix86_expand_vector_set): Use
5531 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5533 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5535 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5536 instructions when changing rounding bits to preserve precision bits
5537 in the x87 control word.
5539 2018-04-03 Martin Liska <mliska@suse.cz>
5541 PR tree-optimization/82491
5542 * rtl.h (strip_offset_and_add): Replace += suboffset with
5543 poly_uint64 () + suboffset.
5545 2018-03-29 Martin Liska <mliska@suse.cz>
5546 Martin Jambor <mjambor@suse.cz>
5549 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5550 param_type is not an integral or pointer type.
5552 2018-04-03 Richard Biener <rguenther@suse.de>
5554 * sese.h (recompute_all_dominators): Remove.
5556 2018-04-02 Martin Sebor <msebor@redhat.com>
5558 * doc/invoke.texi (-Wrestrict): Fix typos.
5560 2018-04-02 Jim Wilson <jimw@sifive.com>
5562 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5563 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5564 (<optab>di3, <optab>si3_extend): Likewise.
5565 (<optab>si3_mask, <optab>si3_mask_1): New.
5566 (<optab>di3_mask, <optab>di3_mask_1): New.
5567 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5568 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5569 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5571 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5573 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5576 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5578 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5579 (nds32_canonicalize_comparison): New function.
5581 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5582 Kito Cheng <kito.cheng@gmail.com>
5583 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5585 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5586 * config/nds32/constants.md (unspec_volatile_element): Add
5587 UNSPEC_VOLATILE_RELAX_GROUP.
5588 * config/nds32/nds32-relax-opt.c: New file.
5589 * config/nds32/nds32-predicates.c
5590 (nds32_symbol_load_store_p): New function.
5591 * config/nds32/nds32-protos.h
5592 (nds32_symbol_load_store_p): Declare function.
5593 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5594 * config/nds32/nds32.c
5595 (nds32_register_pass): New function to register pass.
5596 (nds32_register_passes): New function to register passes.
5597 * config/nds32/nds32.md (relax_group): New pattern.
5598 * config/nds32/nds32.opt (mrelax-hint): New option.
5599 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5601 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5603 * config/nds32/t-nds32: Modify files dependency.
5605 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5607 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5608 (PROFILE_HOOK): Define its implementation.
5610 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5612 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5613 type and 32-bit size.
5615 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5618 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5619 (V_128_256): New mode iterator.
5620 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5621 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5622 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5624 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5625 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5627 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5630 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5631 NaN inputs correctly.
5633 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5636 * config/rs6000/vsx.md (??r): New mode attribute.
5637 (*vsx_mov<mode>_64bit): Use it.
5638 (*vsx_mov<mode>_32bit): Likewise.
5640 2018-03-30 Martin Sebor <msebor@redhat.com>
5642 PR tree-optimization/84818
5643 * builtins.c (check_access): Use warning_n.
5645 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5648 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5650 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5653 2018-03-30 Julia Koval <julia.koval@intel.com>
5656 * x86-tune.def (movx, partial_reg_dependency): Enable for
5659 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5662 * lra-constraints.c (process_alt_operands): Move setting
5663 this_alternative_matches below.
5665 2018-03-29 Martin Liska <mliska@suse.cz>
5668 * doc/invoke.texi: Document how LTO works with debug info.
5669 Describe auto-load support of binutils. Mention 'x86-64'
5670 as valid option value of -march option.
5672 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5674 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5677 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5678 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5679 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5682 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5685 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5686 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5687 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5688 for RS6000_BTM_POWERPC64.
5689 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5690 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5691 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5696 2018-03-28 Carl Love <cel@us.ibm.com>
5699 2017-09-27 Carl Love <cel@us.ibm.com>
5701 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5702 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5703 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5706 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5708 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5709 instead of __vector bool.
5710 (_mm_max_pu8): Likewise.
5711 (_mm_min_pi16): Likewise.
5713 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5716 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5717 (DIVWEUO): Likewise.
5719 (DIVDEUO): Likewise.
5720 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5721 DIVWEUO and DIVDEUO.
5722 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5723 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5724 (div_extend): Likewise.
5725 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5727 (__builtin_divweuo): Likewise.
5728 (__builtin_divdeo): Likewise.
5729 (__builtin_divdeuo): Likewise.
5731 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5734 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5735 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5737 PR tree-optimization/82004
5738 * gimple-match-head.c (optimize_pow_to_exp): New function.
5739 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5740 Don't fold to exp if optimize_pow_to_exp is false.
5742 2018-03-28 Martin Liska <mliska@suse.cz>
5745 * calls.c (initialize_argument_information): Fix trailing space.
5746 * common.opt: Fix typo and provide better explanation for
5747 -fsanitize-coverage option.
5748 * config/i386/i386.opt: Fix typo.
5750 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5751 Martin Liska <mliska@suse.cz>
5754 * gimplify.c (asan_poison_variable): Don't do the check for
5755 gimplify_omp_ctxp here.
5756 (gimplify_decl_expr): Do it here.
5757 (gimplify_target_expr): Likewise.
5759 2018-03-28 Martin Liska <mliska@suse.cz>
5762 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5763 chkp_type_bounds_count if MPX is not enabled.
5765 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5767 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5769 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5772 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5773 function to create the function decl for complex long double
5774 multiply and divide for -mabi=ieeelongdouble.
5775 (init_float128_ieee): Call it.
5777 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5780 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5781 -fcf-protection=branch -mibt.
5782 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5784 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5787 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5789 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5792 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5793 extern array declarations.
5795 2018-03-27 Richard Biener <rguenther@suse.de>
5798 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5799 explicit single_use checks.
5801 2018-03-27 Richard Biener <rguenther@suse.de>
5803 PR tree-optimization/85082
5804 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5807 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5809 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5810 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5811 Turn on fasynchronous-unwind-tables and funwind-tables.
5813 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5816 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5817 (*bmi_blsr_<mode>_ccz): Ditto.
5819 2018-03-26 Tom de Vries <tom@codesourcery.com>
5821 PR tree-optimization/85063
5822 * omp-general.c (offloading_function_p): New function. Factor out
5824 * omp-offload.c (pass_omp_target_link::gate): ... here.
5825 * omp-general.h (offloading_function_p): Declare.
5826 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5827 with attribute omp declare target for offloading functions.
5829 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5831 PR tree-optimization/84005
5832 * tree-data-ref.h (get_base_for_alignment): Declare.
5833 * tree-data-ref.c (get_base_for_alignment_1): New function.
5834 (get_base_for_alignment): Likewise.
5835 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5836 get_base_for_alignment to find a suitable base object, instead
5837 of always using drb->base_address.
5839 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5842 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5843 known size by default.
5845 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5848 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5849 and non BLKmode operands.
5851 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5854 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5855 Clean up attributes.
5857 2018-03-23 Richard Biener <rguenther@suse.de>
5860 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5861 we are going to emit early debug for LTO.
5863 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5866 * function.c (match_asm_constraints_1): Don't optimize if input
5867 doesn't satisfy general_operand predicate for output's mode.
5870 * alias.c (write_dependence_p): Don't require for x_canonicalized
5871 non-VOIDmode if x has VOIDmode.
5874 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5875 just don't try to optimize it rather than assert it never happens.
5877 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5879 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5880 macro expansions for definition of ST_INTERNAL_<mode> and
5881 LD_INTERNAL_<mode> builtins.
5882 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5884 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5886 (altivec_expand_st_builtin): Likewise.
5887 (altivec_expand_builtin): Remove calls to deleted functions.
5888 (rs6000_address_for_altivec): Delete this function.
5889 * config/rs6000/vector.md: Remove expands for
5890 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5892 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5895 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5896 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5897 re-computing once computed.
5898 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5899 (arm_init_machine_status): Initialize
5900 machine->static_chain_stack_bytes.
5902 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5905 * doc/extend.texi: Add four new prototypes for vec_ld.
5906 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5907 definitions for more logical presentation.
5908 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5909 entries for V1TI variants of __builtin_altivec_ld builtin.
5910 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5911 handling of V1TI variant of LVX icode pattern.
5912 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5913 (rs6000_gimple_fold_builtin): Likewise.
5914 (altivec_init_builtins): Add code to define
5915 __builtin_altivec_lvx_v1ti function.
5917 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5920 * function.c (match_asm_constraints_1): Don't do the optimization
5921 if input isn't a REG, SUBREG, MEM or constant.
5923 2018-03-22 Tom de Vries <tom@codesourcery.com>
5925 PR tree-optimization/84956
5926 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5927 bb_has_abnormal_pred.
5929 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5932 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5933 DECL_INITIAL (decl) to decl at the end.
5934 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5937 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5939 * doc/extend.texi (__builtin_tgmath): Document when complex
5940 integer types are treated as _Complex _Float64.
5942 2018-03-21 Tom de Vries <tom@codesourcery.com>
5944 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5946 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5948 PR tree-optimization/84960
5949 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5950 if it is ENTRY block, move them into single succ of ENTRY in that case.
5952 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5954 PR tree-optimization/84811
5955 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5956 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5957 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5958 (known_subrange_p): Update accordingly. Cast each value involved
5959 in the size comparison, rather than casting the result of the
5962 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5964 PR tree-optimization/84982
5965 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5966 by flipping the least significant bit rather than all bits from
5967 bitpos to bitpos + bitsize - 1.
5969 2018-03-21 Nathan Sidwell <nathan@acm.org>
5971 * doc/extend.texi (Deprecated Features): Remove mention of
5972 long-deleted deprecations.
5974 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5977 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5978 * configure: Regenerate.
5980 2018-03-21 Tom de Vries <tom@codesourcery.com>
5982 PR tree-optimization/83126
5983 * tree-parloops.c (num_phis): New function.
5984 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5986 2018-03-21 Nathan Sidwell <nathan@acm.org>
5988 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5989 mention anon-struct/union members and trailing attributes.
5991 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5993 PR tree-optimization/84969
5994 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5995 builtin memset partitions if they set different rhs values.
5997 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5999 PR rtl-optimization/84989
6000 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
6001 VEC_DUPLICATE with scalar result mode.
6003 2018-03-21 Martin Liska <mliska@suse.cz>
6006 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
6007 not intended return statement.
6009 2018-03-21 Martin Liska <mliska@suse.cz>
6012 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
6013 (chkp_find_bound_slots_1): Limit number of iterations.
6015 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
6018 * Minor grammar fixes for x86 options.
6020 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6023 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
6024 holding REG_CFA_RESTORE notes, instead turn them into a USE.
6026 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
6029 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
6030 (altivec_lvx_<mode>_1op): Likewise.
6031 (altivec_stvx_<mode>_2op): Likewise.
6032 (altivec_stvx_<mode>_1op): Likewise.
6033 (altivec_lvx_<VM2:mode>): New define_expand.
6034 (altivec_stvx_<VM2:mode>): Likewise.
6035 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
6036 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6037 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
6038 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
6039 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
6040 (rs6000_gen_lvx): Likewise.
6041 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
6042 (altivec_expand_stv_builtin): Likewise.
6043 (altivec_expand_builtin): Likewise.
6044 * config/rs6000/vector.md: Likewise.
6046 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6049 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6052 2018-03-20 Richard Biener <rguenther@suse.de>
6055 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6056 sign-conversions as zero, fall back to standard scalar_stmt
6059 2018-03-20 Martin Liska <mliska@suse.cz>
6062 * predict.c (rebuild_frequencies): Handle case when we have
6063 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6065 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6068 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6069 flag_section_anchors.
6070 * varasm.c (use_blocks_for_decl_p): Remove hack for
6071 dw2_force_const_mem.
6074 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6076 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6077 be created, use lowpart_subreg of operands[0] rather than operands[0]
6079 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6080 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6081 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6082 and n constraint instead of aarch64_shift_imm_di and Usd.
6083 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6084 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6086 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6089 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6090 to favor GPR over NEON registers.
6091 (<shift>di3_neon): Likewise.
6093 2018-03-20 Tom de Vries <tom@codesourcery.com>
6096 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6097 (nvptx_process_pars): Emit bar.sync asap and alap.
6099 2018-03-20 Tom de Vries <tom@codesourcery.com>
6102 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6103 seen_label if seen_label is already set.
6105 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6108 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6109 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6110 Use 1U instead of 1. Formatting fixes.
6113 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6114 instead of TREE_TYPE (s1) for the return value.
6116 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6118 PR tree-optimization/84946
6119 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6120 bitsize + bitsize in poly_uint64 rather than poly_int64.
6123 * dwarf2asm.c: Include fold-const.c.
6124 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6125 of decl rather than decl itself.
6127 PR rtl-optimization/84643
6128 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6130 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6133 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6134 calling assemble_variable.
6136 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6139 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6140 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6142 2018-03-19 Jim Wilson <jimw@sifive.com>
6145 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6146 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6147 (riscv_first_stack_step): Likewise.
6148 (riscv_option_override): Use STACK_BOUNDARY instead of
6150 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6152 (BIGGEST_ALIGNMENT): Set to 128.
6153 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6154 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6157 2018-03-19 Richard Biener <rguenther@suse.de>
6159 PR tree-optimization/84933
6160 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6161 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6163 2018-03-19 Richard Biener <rguenther@suse.de>
6165 PR tree-optimization/84859
6166 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6167 (cond_if_else_store_replacement): Perform sinking operation on
6168 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6169 Generalize what a BB with a single eligible store is.
6171 2018-03-19 Richard Biener <rguenther@suse.de>
6173 PR tree-optimization/84929
6174 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6175 chrec_is_positive against non-chrec arg.
6177 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6180 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6182 2018-03-18 Martin Liska <mliska@suse.cz>
6184 PR rtl-optimization/84635
6185 * regrename.c (build_def_use): Use matches_mode only when
6188 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6190 PR tree-optimization/84913
6191 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6192 vectorize chains of COND_EXPRs.
6194 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6196 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6198 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6200 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6202 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6204 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6206 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6207 Kito Cheng <kito.cheng@gmail.com>
6209 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6210 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6211 (nds32_adjust_reg_alloc_order): New function.
6212 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6214 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6216 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6217 nds32_print_operand, nds32_print_operand_address): Use
6218 HOST_WIDE_INT_PRINT_DEC instead.
6220 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6222 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6224 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6227 * config/i386/i386.c (initial_ix86_tune_features,
6228 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6230 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6231 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6232 rather than 1u << ix86_tune. Formatting fix.
6233 (ix86_option_override_internal): Change ix86_arch_mask from
6234 unsigned int to unsigned HOST_WIDE_INT, initialize to
6235 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6236 (ix86_function_specific_restore): Likewise.
6238 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6241 * postreload.c (reload_combine_recognize_pattern): Perform
6242 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6243 truncate_int_for_mode the result for the destination's mode.
6246 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6247 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6249 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6252 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6255 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6256 regno_allocno_class_array and sorted_pseudos.
6257 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6258 insns where regno is used.
6260 2018-03-16 Martin Liska <mliska@suse.cz>
6263 * multiple_target.c (create_dispatcher_calls): Redirect
6264 reference in the symbol table.
6266 2018-03-16 Martin Liska <mliska@suse.cz>
6269 * multiple_target.c (create_dispatcher_calls): Redirect also
6272 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6276 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6277 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6278 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6280 2018-03-16 Julia Koval <julia.koval@intel.com>
6282 * doc/invoke.texi (Skylake Server): Add CLWB.
6283 Cannonlake): Remove CLWB.
6285 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6287 PR tree-optimization/84841
6288 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6290 (FLOAT_ONE_CONST_TYPE): Define.
6291 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6292 (sort_by_operand_rank): Put entries with higher constant_type last
6293 rather than first to match comments.
6295 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6297 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6300 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6303 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6304 check and use address of target always.
6306 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6309 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6310 (indirect_thunk_bnd_needed): Likewise.
6311 (indirect_thunks_used): Likewise.
6312 (indirect_thunks_bnd_used): Likewise.
6313 (indirect_return_needed): New.
6314 (indirect_return_bnd_needed): Likewise.
6315 (output_indirect_thunk_function): Add a bool argument for
6317 (output_indirect_thunk_function): Don't generate alias for
6318 function return thunk.
6319 (ix86_code_end): Call output_indirect_thunk_function to generate
6320 function return thunks.
6321 (ix86_output_function_return): Set indirect_return_bnd_needed
6322 and indirect_return_needed instead of indirect_thunk_bnd_needed
6323 and indirect_thunk_needed.
6325 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6327 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6328 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6329 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6331 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6332 Paul Hua <paul.hua.gm@gmail.com>
6335 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6337 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6339 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6340 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6343 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6346 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6347 instead of GET_MODE_SIZE when comparing Units.
6349 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6352 * varasm.c (hash_section): Return an unchangeble hash value
6353 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6354 Return !aarch64_can_use_per_function_literal_pools_p ().
6356 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6359 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6360 rather than address of cmode as last argument to prepare_cmp_insn.
6362 2018-03-15 Julia Koval <julia.koval@intel.com>
6364 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6365 F_AVX512VNNI, F_AVX512BITALG): New.
6367 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6370 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6371 insn for floating-point loads and stores.
6373 2018-03-14 Carl Love <cel@us.ibm.com>
6375 * config/rs6000/rs6000-c.c: Add macro definitions for
6376 ALTIVEC_BUILTIN_VEC_PERMXOR.
6377 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6378 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6379 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6380 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6382 * config/doc/extend.texi: Add prototypes for vec_permxor.
6384 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6387 * diagnostic-show-locus.c (class layout_point): Convert m_line
6388 from int to linenum_type.
6389 (line_span::comparator): Use linenum "compare" function when
6390 comparing line numbers.
6391 (test_line_span): New function.
6392 (layout_range::contains_point): Convert param "row" from int to
6394 (layout_range::intersects_line_p): Likewise.
6395 (layout::will_show_line_p): Likewise.
6396 (layout::print_source_line): Likewise.
6397 (layout::should_print_annotation_line_p): Likewise.
6398 (layout::print_annotation_line): Likewise.
6399 (layout::print_leading_fixits): Likewise.
6400 (layout::annotation_line_showed_range_p): Likewise.
6401 (struct line_corrections): Likewise for field m_row.
6402 (line_corrections::line_corrections): Likewise for param "row".
6403 (layout::print_trailing_fixits): Likewise.
6404 (layout::get_state_at_point): Likewise.
6405 (layout::get_x_bound_for_row): Likewise.
6406 (layout::print_line): Likewise.
6407 (diagnostic_show_locus): Likewise for locals "last_line" and
6409 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6410 * input.c (selftest::test_linenum_comparisons): New function.
6411 (selftest::input_c_tests): Call it.
6412 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6413 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6414 * selftest.h (ASSERT_GT): New macro.
6415 (ASSERT_GT_AT): New macro.
6416 (ASSERT_LT): New macro.
6417 (ASSERT_LT_AT): New macro.
6419 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6421 PR rtl-optimization/84780
6422 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6424 2018-03-14 Martin Liska <mliska@suse.cz>
6426 * tree.c (record_node_allocation_statistics): Use
6427 get_stats_node_kind.
6428 (get_stats_node_kind): New function extracted from
6429 record_node_allocation_statistics.
6430 (free_node): Use get_stats_node_kind.
6432 2018-03-14 Richard Biener <rguenther@suse.de>
6434 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6435 that the value-set of ANTIC_IN doesn't grow.
6438 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6440 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6441 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6443 2018-03-14 Julia Koval <julia.koval@intel.com>
6445 * config.gcc (icelake-client, icelake-server): New.
6447 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6448 (initial_ix86_arch_features): Ditto.
6449 (PTA_SKYLAKE): Add SGX.
6450 (PTA_ICELAKE): Remove.
6451 (PTA_ICELAKE_CLIENT): New.
6452 (PTA_ICELAKE_SERVER): New.
6453 (ix86_option_override_internal): Split up icelake on icelake client and
6455 (get_builtin_code_for_version): Ditto.
6456 (fold_builtin_cpu): Ditto.
6457 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6458 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6459 * config/i386/i386.h (processor_type): Ditto.
6460 * doc/invoke.texi: Ditto.
6462 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6465 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6466 INTEGER_CST offset, add it together with bitpos / 8 and
6467 sign extend based on POINTER_SIZE.
6471 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6474 * config/i386/constraints.md (Yc): New register constraint.
6475 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6476 Use Yc constraint for alternative 2 of operand 0. Remove
6477 preferred_for_speed attribute.
6479 2018-03-14 Richard Biener <rguenther@suse.de>
6481 PR tree-optimization/84830
6482 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6483 with the old one to avoid oscillations.
6485 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6488 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6490 (assign_by_spills): Return a flag of reload assignment failure.
6491 Do not process the reload assignment failures. Do not spill other
6492 reload pseudos if they has the same reg class. Update n if
6494 (lra_assign): Add a return arg. Set up from the result of
6495 assign_by_spills call.
6496 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6497 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6498 usage_insns if it is not NULL.
6499 (spill_hard_reg_in_range): New function.
6500 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6501 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6502 function prototypes.
6503 (lra_assign): Change prototype.
6504 * lra.c (lra): Add code to deal with fails by splitting hard reg
6507 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6509 * config/riscv/riscv.opt (mrelax): New option.
6510 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6511 "norelax" when riscv_mrelax is disabled.
6512 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6514 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6517 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6518 reassociation for int modes.
6520 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6522 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6523 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6525 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6526 * config/aarch64/aarch64-sve.md
6527 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6528 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6529 (*extend<mode><Vwide>2): Rename to...
6530 (aarch64_sve_extend<mode><Vwide>2): ...this.
6531 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6532 renaming the old pattern to...
6533 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6535 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6536 define_expand, renaming the old pattern to...
6537 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6538 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6539 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6540 account when deciding which SVE instruction the optab should use.
6541 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6543 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6545 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6546 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6547 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6548 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6549 (tlsdesc_small_<mode>): Turn a define_expand and use
6550 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6551 (tlsdesc_small_advsimd_<mode>): ...this.
6552 (tlsdesc_small_sve_<mode>): New pattern.
6554 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6556 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6557 (UNSPEC_UMUL_HIGHPART): New constants.
6558 (MUL_HIGHPART): New int iteraor.
6559 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6560 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6562 (*<su>mul<mode>3_highpart): New define_insn.
6564 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6567 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6570 2018-03-13 Martin Liska <mliska@suse.cz>
6573 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6575 (sem_item_optimizer::~sem_item_optimizer): Release it.
6576 (sem_item_optimizer::merge_classes): Register variable aliases.
6577 (sem_item_optimizer::fixup_pt_set): New function.
6578 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6579 * ipa-icf.h: Declare new variables and functions.
6581 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6584 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6585 integer_pow2p@2 and test integer_pow2p in condition.
6586 (A < 0 ? C : 0): Similarly for @1.
6589 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6590 characters starting at p contain '\0' character, don't look beyond
6594 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6595 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6598 * reg-stack.c (change_stack): Change update_end var from int to
6599 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6600 also call set_block_for_insn on the newly added insns and rescan.
6603 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6604 on the last operand.
6607 * tree.c (stabilize_reference_1): Return save_expr (e) for
6608 STATEMENT_LIST even if it doesn't have side-effects.
6610 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6612 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6614 2018-03-12 Renlin Li <renlin.li@arm.com>
6616 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6617 aarch64_output_scalar_simd_mov_immediate.
6619 2018-03-12 Martin Sebor <msebor@redhat.com>
6621 PR tree-optimization/83456
6622 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6623 for perfectly overlapping calls to memcpy.
6624 (gimple_fold_builtin_memory_chk): Same.
6625 (gimple_fold_builtin_strcpy): Handle no-warning.
6626 (gimple_fold_builtin_stxcpy_chk): Same.
6627 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6629 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6631 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6632 parameter. Use it for SFmode.
6633 (rs6000_function_arg_advance_1): Adjust.
6634 (rs6000_function_arg): Adjust.
6635 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6637 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6639 PR rtl-optimization/84169
6640 PR rtl-optimization/84780
6641 * combine.c (can_combine_p): Check for a 2-insn combination whether
6642 the destination register is used between the two insns, too.
6644 2018-03-12 Richard Biener <rguenther@suse.de>
6646 PR tree-optimization/84803
6647 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6648 for refs DR analysis didn't process.
6650 2018-03-12 Richard Biener <rguenther@suse.de>
6652 PR tree-optimization/84777
6653 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6654 force-vectorize loops ignore whether we are optimizing for size.
6656 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6658 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6659 (TARGET_MD_ASM_ADJUST): Define.
6661 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6662 Kito Cheng <kito.cheng@gmail.com>
6663 Chung-Ju Wu <jasonwucj@gmail.com>
6665 * config/nds32/nds32.c (nds32_compute_stack_frame,
6666 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6667 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6668 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6669 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6670 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6671 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6672 * config/nds32/nds32.md (prologue, epilogue): Use macro
6673 NDS32_V3PUSH_AVAILABLE_P to do checking.
6675 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6678 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6679 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6680 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6681 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6682 addition of most attributes on !orig_type_die or the attribute not
6683 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6685 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6686 Chung-Ju Wu <jasonwucj@gmail.com>
6688 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6690 * config/nds32/nds32.opt (mvh): New option.
6692 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6693 Chung-Ju Wu <jasonwucj@gmail.com>
6695 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6697 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6698 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6701 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6702 Chung-Ju Wu <jasonwucj@gmail.com>
6704 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6706 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6707 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6709 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6710 Kito Cheng <kito.cheng@gmail.com>
6711 Chung-Ju Wu <jasonwucj@gmail.com>
6713 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6714 UNSPEC_FFMISM and UNSPEC_FLMISM.
6715 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6716 for ffb, ffmism and flmism.
6717 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6718 (unspec_ffmism): Ditto.
6719 (unspec_flmism): Ditto.
6720 (nds32_expand_builtin_impl): Check if string extension is available.
6721 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6722 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6724 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6727 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6730 * lra-assigns.c (assign_by_spills): Return a flag of reload
6731 assignment failure. Do not process the reload assignment
6732 failures. Do not spill other reload pseudos if they has the same
6734 (lra_assign): Add a return arg. Set up from the result of
6735 assign_by_spills call.
6736 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6737 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6738 usage_insns if it is not NULL.
6739 (spill_hard_reg_in_range): New function.
6740 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6741 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6742 function prototypes.
6743 (lra_assign): Change prototype.
6744 * lra.c (lra): Add code to deal with fails by splitting hard reg
6747 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6750 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6752 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6755 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6756 (dw_val_node): Add val_symbolic_view.
6757 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6758 (symview_upper_bound): New.
6759 (new_line_info_table): Initialize symviews_since_reset.
6760 (dwarf2out_source_line): Count symviews_since_reset and set
6761 symview_upper_bound.
6762 (dw_val_equal_p): Handle symview.
6763 (add_AT_symview): New.
6764 (print_dw_val): Handle symview.
6765 (attr_checksum, attr_checksum_ordered): Likewise.
6766 (same_dw_val_p, size_of_die): Likewise.
6767 (value_format, output_die): Likewise.
6768 (add_high_low_attributes): Use add_AT_symview for entry_view.
6769 (dwarf2out_finish): Reset symview_upper_bound, clear
6772 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6775 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6776 Add strict argument and use it.
6777 (rs6000_split_multireg_move): Update for new strict argument.
6778 (mem_operand_gpr): Disallow all non-offsettable addresses.
6779 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6781 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6784 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6785 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6786 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6789 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6790 decl, use remap_type if we want to use the type.
6792 2018-03-09 Martin Sebor <msebor@redhat.com>
6794 PR tree-optimization/84526
6795 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6797 (builtin_access::generic_overlap): Be prepared to handle non-array
6800 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6802 PR rtl-optimization/84682
6803 * lra-constraints.c (process_address_1): Check is_address flag
6804 for address constraints.
6805 (process_alt_operands): Likewise.
6806 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6807 preprocess_constraints.
6808 * recog.h (preprocess_constraints): Add oploc parameter.
6810 * recog.c (preprocess_constraints): Test address_operand for
6811 CT_ADDRESS constraints.
6813 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6816 * lra-assigns.c (assign_by_spills): Return a flag of reload
6817 assignment failure. Do not process the reload assignment
6818 failures. Do not spill other reload pseudos if they has the same
6820 (lra_assign): Add a return arg. Set up from the result of
6821 assign_by_spills call.
6822 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6823 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6824 usage_insns if it is not NULL.
6825 (spill_hard_reg_in_range): New function.
6826 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6827 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6828 function prototypes.
6829 (lra_assign): Change prototype.
6830 * lra.c (lra): Add code to deal with fails by splitting hard reg
6833 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6836 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6837 Accept complain bool parameter. Only emit errors if it is true.
6838 (arm_parse_cpu_option_name): Likewise.
6839 (arm_target_thumb_only): Adjust callers of the above.
6840 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6841 prototype to take a default true bool parameter.
6842 (arm_parse_arch_option_name): Likewise.
6844 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6845 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6849 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6850 * configure: Regenerate.
6851 * configure.ac ("linker --version-script option"): New.
6852 ("linker soname option"): New.
6854 2018-03-09 Richard Biener <rguenther@suse.de>
6856 PR tree-optimization/84775
6857 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6858 immediate uses of predicate stmts and mark them modified.
6861 PR tree-optimization/84178
6862 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6864 (version_loop_for_if_conversion): Delay update_ssa call.
6865 (tree_if_conversion): Delay update_ssa until after predicate
6868 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6871 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6872 when the function accesses prior frames.
6874 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6877 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6878 gen_llsym, otherwise call maybe_gen_llsym.
6881 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6882 has ',' character inside of it.
6884 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6887 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6888 as clobbering CC_REGNUM.
6890 2018-03-08 Richard Biener <rguenther@suse.de>
6893 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6894 (follow_copies_to_constant): Do not follow SSA names registered
6897 2018-03-08 Richard Biener <rguenther@suse.de>
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 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6908 PR tree-optimization/84178
6909 * tree-if-conv.c (release_bb_predicate): Remove the
6910 the assertion that the stmts have NULL use_ops.
6911 Discard the statements, asserting that they haven't
6912 yet been added to a BB.
6914 2018-03-08 Richard Biener <rguenther@suse.de>
6916 PR tree-optimization/84746
6917 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6918 (phi_translate): Pass in destination ANTIC_OUT set.
6919 (phi_translate_1): Likewise. For a simplified result lookup
6920 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6921 (phi_translate_set): Adjust.
6922 (do_pre_regular_insertion): Likewise.
6923 (do_pre_partial_partial_insertion): Likewise.
6925 2018-03-08 Martin Liska <mliska@suse.cz>
6927 PR gcov-profile/84735
6928 * doc/gcov.texi: Document usage of profile files.
6929 * gcov-io.h: Document changes in the format.
6931 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6935 * dwarf2out.c (struct dw_line_info_table): Update comments for
6937 (FORCE_RESET_NEXT_VIEW): New.
6938 (FORCE_RESETTING_VIEW_P): New.
6939 (RESETTING_VIEW_P): Check for -1 too.
6940 (ZERO_VIEW_P): Likewise.
6941 (new_line_info_table): Force-reset next view.
6942 (dwarf2out_begin_function): Likewise.
6943 (dwarf2out_source_line): Simplify zero_view_p initialization.
6944 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6945 view directly. Omit view when omitting .loc at line 0.
6947 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6949 PR tree-optimization/84740
6950 * tree-switch-conversion.c (process_switch): Call build_constructors
6951 only if info.phi_count is non-zero.
6953 PR tree-optimization/84739
6954 * tree-tailcall.c (find_tail_calls): Check call arguments against
6955 DECL_ARGUMENTS (current_function_decl) rather than
6956 DECL_ARGUMENTS (func) when checking for tail recursion.
6958 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6960 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6961 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6962 Volker Reichelt's entry and add entries for people that perform
6963 GCC fuzzy testing and report numerous bugs.
6965 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6968 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6969 readonly data in sdata, if that is disabled.
6970 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6971 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6972 -mreadonly-in-sdata option.
6974 2018-03-07 Martin Sebor <msebor@redhat.com>
6976 PR tree-optimization/84468
6977 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6978 basic block when looking for nul assignment.
6980 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6983 * except.h (output_function_exception_table): Adjust prototype.
6984 * except.c (output_function_exception_table): Remove FNNAME parameter
6985 and add SECTION parameter. Ouput one part of the table at a time.
6986 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6987 the first part of the exception table and emit unwind directives.
6988 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6989 (i386_pe_seh_cold_init): Likewise.
6990 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6991 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6992 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6993 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6994 * config/i386/winnt.c: Include except.h.
6995 (struct seh_frame_state): Add reg_offset, after_prologue and
6996 in_cold_section fields.
6997 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6998 (i386_pe_seh_cold_init): New function.
6999 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
7000 to seh->in_cold_section.
7001 (seh_emit_push): Record the offset of the push.
7002 (seh_emit_save): Record the offet of the save.
7003 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
7004 Test seh->after_prologue to disregard the epilogue.
7005 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
7006 (i386_pe_end_cold_function): New function.
7008 2018-03-07 Jakub Jelinek <jakub@redhat.com>
7011 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
7012 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
7015 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
7017 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7018 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
7021 * multiple_target.c: Include tree-inline.h and intl.h.
7022 (expand_target_clones): Diagnose and fail if node->definition and
7023 !tree_versionable_function_p (node->decl).
7025 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
7027 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
7029 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
7030 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
7031 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
7033 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7036 * combine.c (try_combine): Use reg_or_subregno instead of handling
7037 just paradoxical SUBREGs and REGs.
7039 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
7041 * config/arc/arc.c (arc_finalize_pic): Remove function.
7042 (arc_must_save_register): We use single base PIC register, remove
7043 checks to save/restore the PIC register.
7044 (arc_expand_prologue): Likewise.
7045 * config/arc/arc-protos.h (arc_set_default_type_attributes):
7047 (arc_verify_short): Likewise.
7048 (arc_attr_type): Likewise.
7049 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7050 (walk_stores): Likewise.
7051 (arc_address_cost): Make it static.
7052 (arc_verify_short): Likewise.
7053 (branch_dest): Likewise.
7054 (arc_attr_type): Likewise.
7055 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7056 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7057 (arc_final_prescan_insn): Remove inserting the nops due to
7058 hardware hazards. It is done in reorg step.
7059 (insn_length_variant_t): Remove.
7060 (insn_length_parameters_t): Likewise.
7061 (arc_insn_length_parameters): Likewise.
7062 (arc_get_insn_variants): Likewise.
7063 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7065 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7068 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7071 PR tree-optimization/84687
7072 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7074 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7076 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7078 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7079 Rename to ppc_speculation_barrier.
7080 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7081 __builtin_ppc_speculation_barrier.
7083 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7086 * combine.c (combine_simplify_rtx): Don't try to simplify if
7087 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7090 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7092 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7093 to 32 bytes when compiling for POWER9.
7095 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7098 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7099 regparm >= 3 with no arg reg available also for calls with
7100 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7103 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7105 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7107 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7110 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7112 2018-03-05 Richard Biener <rguenther@suse.de>
7114 PR tree-optimization/84486
7115 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7116 When inserting a __builtin_assume_aligned call set the LHS
7117 SSA name alignment info accordingly.
7119 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7121 PR tree-optimization/84114
7122 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7123 Avoid reassociation of FLOAT_MODE addition.
7125 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7127 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7128 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7129 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7130 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7131 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7132 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7133 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7135 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7136 __builtin_ia32_wbinvd): New builtins.
7137 (SPECIAL_ARGS2): New.
7138 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7139 (SPECIAL_ARGS2): New.
7140 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7141 (ix86_valid_target_attribute_inner_p): Ditto.
7142 (ix86_init_mmx_sse_builtins): Add special_args2.
7143 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7144 TARGET_WBNOINVD_P): New.
7145 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7146 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7147 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7148 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7149 * config/i386/pconfigintrin.h: New file.
7150 * config/i386/wbnoinvdintrin.h: Ditto.
7151 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7152 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7154 2018-03-05 Richard Biener <rguenther@suse.de>
7156 PR tree-optimization/84670
7157 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7159 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7160 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7161 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7162 doesn't grow if all successors (recursively) were visited at least
7165 2018-03-05 Richard Biener <rguenther@suse.de>
7167 PR tree-optimization/84650
7168 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7169 if executed in the loop pipeline.
7171 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7173 * doc/configfiles.texi (Configuration Files): Move info about
7174 conditionalizing $target-protos.h to...
7175 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7176 differs from $target-protos.h.
7178 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7179 Chung-Ju Wu <jasonwucj@gmail.com>
7181 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7182 * config/nds32/nds32-multiple.md (setmemsi): Define.
7183 * config/nds32/nds32-memory-manipulation.c
7184 (nds32_gen_dup_4_byte_to_word_value): New.
7185 (emit_setmem_word_loop): New.
7186 (emit_setmem_byte_loop): New.
7187 (nds32_expand_setmem_loop): New.
7188 (nds32_expand_setmem_loop_v3m): New.
7189 (nds32_expand_setmem_unroll): New.
7190 (nds32_expand_setmem): New.
7192 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7193 Chung-Ju Wu <jasonwucj@gmail.com>
7195 * config/nds32/nds32-memory-manipulation.c
7196 (nds32_emit_load_store): New.
7197 (nds32_emit_post_inc_load_store): New.
7198 (nds32_emit_mem_move): New.
7199 (nds32_emit_mem_move_block): New.
7200 (nds32_expand_movmemsi_loop_unknown_size): New.
7201 (nds32_expand_movmemsi_loop_known_size): New.
7202 (nds32_expand_movmemsi_loop): New.
7203 (nds32_expand_movmemsi_unroll): New.
7204 (nds32_expand_movmemqi): Rename ...
7205 (nds32_expand_movmemsi): ... to this.
7206 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7207 (movmemsi): ... to this.
7208 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7209 (nds32_expand_movmemsi): ... to this.
7211 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7212 Monk Chiang <sh.chiang04@gmail.com>
7213 Chung-Ju Wu <jasonwucj@gmail.com>
7215 * config/nds32/nds32-protos.h
7216 (nds32_expand_load_multiple): New arguments.
7217 (nds32_expand_store_multiple): Ditto.
7218 (nds32_valid_multiple_load_store): Rename ...
7219 (nds32_valid_multiple_load_store_p): ... to this.
7220 * config/nds32/nds32-memory-manipulation.c
7221 (nds32_expand_load_multiple): Refine implementation.
7222 (nds32_expand_store_multiple): Ditto.
7223 * config/nds32/nds32-multiple.md
7224 (load_multiple): Update nds32_expand_load_multiple interface.
7225 (store_multiple): Update nds32_expand_store_multiple interface.
7226 * config/nds32/nds32-predicates.c
7227 (nds32_valid_multiple_load_store): Rename ...
7228 (nds32_valid_multiple_load_store_p): ... to this and refine
7230 * config/nds32/predicates.md
7231 (nds32_load_multiple_and_update_address_operation): New predicate.
7232 (nds32_store_multiple_and_update_address_operation): New predicate.
7234 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7235 Chung-Ju Wu <jasonwucj@gmail.com>
7237 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7238 (combo): New attribute.
7239 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7241 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7243 * config/nds32/nds32.opt: Change -mcmodel= default value.
7245 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7246 Monk Chiang <sh.chiang04@gmail.com>
7247 Chung-Ju Wu <jasonwucj@gmail.com>
7249 * config/nds32/constants.md (unspec_element): New enum.
7250 * config/nds32/constraints.md (Umw): New constraint.
7251 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7252 * config/nds32/nds32-intrinsic.md: Likewise.
7253 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7254 (nds32_valid_smw_lwm_base_p): New.
7255 (nds32_output_smw_single_word): New.
7256 (nds32_output_lmw_single_word): New.
7257 (nds32_expand_unaligned_load): New.
7258 (nds32_expand_unaligned_store): New.
7259 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7260 (nds32_output_smw_single_word): Declare.
7261 (nds32_output_lmw_single_word): Declare.
7262 (nds32_expand_unaligned_load): Declare.
7263 (nds32_expand_unaligned_store): Declare.
7264 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7265 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7266 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7267 NDS32_BUILTIN_UASTORE_DW.
7268 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7271 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7272 Kito Cheng <kito.cheng@gmail.com>
7273 Chung-Ju Wu <jasonwucj@gmail.com>
7275 * config/nds32/nds32-intrinsic.c
7276 (nds32_expand_builtin_null_ftype_reg): Delete.
7277 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7278 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7279 (nds32_read_argument): New.
7280 (nds32_legitimize_target): Ditto.
7281 (nds32_legitimize_argument): Ditto.
7282 (nds32_check_constant_argument): Ditto.
7283 (nds32_expand_unop_builtin): Ditto.
7284 (nds32_expand_unopimm_builtin): Ditto.
7285 (nds32_expand_binop_builtin): Ditto.
7286 (nds32_builtin_decl_impl): Ditto.
7287 (builtin_description): Ditto.
7288 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7289 (nds32_init_builtins_impl): Ditto.
7290 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7291 (nds32_builtin_decl): New.
7292 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7293 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7295 2018-03-02 Jeff Law <law@redhat.com>
7297 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7298 (redundant_insn, fill_simple_delay_slots): Likewise.
7299 (fill_slots_from_thread): Likewise.
7300 * resource.c (mark_referenced_resources): Likewise.
7301 (mark_set_resources, find_dead_or_set_registers): Likewise.
7303 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7305 * substring-locations.h (format_warning_va): Formatting fix for
7307 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7309 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7310 * substring-locations.c: Include intl.h.
7311 (format_warning_va): Turned into small wrapper around
7312 format_warning_n_va, renamed to ...
7313 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7314 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7316 (format_warning_at_substring_n): New function.
7317 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7318 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7319 format_warning_at_substring with just a shorter name instead of
7320 const function pointer.
7321 (fmtwarn_n): New function.
7322 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7323 appropriate, get rid of all the fmtstr temporaries, move conditionals
7324 with G_() wrapped string literals directly into fmtwarn arguments,
7325 cast dir.len to (int), formatting fixes.
7327 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7329 * doc/invoke.texi: Remove "Cilk Plus" references.
7331 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7332 Richard Biener <rguenther@suse.de>
7335 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7336 for error or warning attributes if CALL_FROM_THUNK_P is set.
7339 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7342 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7343 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7346 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7347 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7349 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7350 instead of -1U in last predictors element's probability member.
7352 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7355 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7356 arguments if they are comparable.
7358 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7360 PR tree-optimization/84634
7361 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7362 masks and masked_loop_p with a single loop_masks, making sure it's
7363 null for bb vectorization.
7365 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7367 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7368 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7369 loop->force_vectorize to check whether there is no alias.
7371 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7374 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7376 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7378 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7379 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7380 instead of a loop around prev_real_insn.
7381 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7385 * config/i386/i386.c (ix86_print_operand): Use conditional
7386 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7389 2018-03-02 Richard Biener <rguenther@suse.de>
7391 PR tree-optimization/84427
7392 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7393 (bitmap_set_subtract_values): Rewrite to handle multiple
7396 (prune_clobbered_mems): Likewise.
7397 (phi_translate): Take edge instead of pred/phiblock.
7398 (phi_translate_1): Likewise.
7399 (phi_translate_set): Likewise. Insert all translated
7400 exprs for a value into the set, keeping possibly multiple
7401 expressions per value.
7402 (compute_antic_aux): Adjust for phi_translate changes.
7403 When intersecting union the expressions and prune those
7404 not in the final value set, keeping possibly multiple
7405 expressions per value. Do not use value-insertion
7406 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7407 all expressions. Add verification that the value-sets
7408 only shrink during iteration.
7409 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7410 (do_pre_regular_insertion): Likewise.
7411 (do_pre_partial_partial_insertion): Likewise.
7413 2018-03-02 Richard Biener <rguenther@suse.de>
7416 * config/darwin.c (saved_debug_info_level): New static global.
7417 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7418 (darwin_asm_lto_end): Restore debug info generation settings.
7420 2018-03-01 Martin Liska <mliska@suse.cz>
7423 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7426 2018-03-01 Richard Biener <rguenther@suse.de>
7429 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7430 type annotation with LTO when debug was disabled at compile-time.
7432 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7434 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7436 (mips_final_postscan_insn): Likewise.
7438 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7440 PR rtl-optimization/84528
7441 * alias.c (init_alias_target): Add commentary.
7442 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7443 a unique base value if the frame pointer is not eliminated
7444 to the stack pointer.
7446 2018-03-01 Tom de Vries <tom@codesourcery.com>
7448 PR rtl-optimization/83327
7449 * lra-int.h (hard_regs_spilled_into): Declare.
7450 * lra.c (hard_regs_spilled_into): Define.
7451 (init_reg_info): Init hard_regs_spilled_into.
7452 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7453 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7454 (process_bb_lives): Handle hard_regs_spilled_into.
7455 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7456 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7458 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7460 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7461 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7462 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7463 * config/rs6000/aix72.h: New file.
7465 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7467 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7468 instead of warning_at with conditional singular and plural messages
7472 * stor-layout.c (update_alignment_for_field): For
7473 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7474 && !DECL_PACKED (field), do the alignment update, just use
7475 only desired_align instead of MAX (type_align, desired_align)
7477 (place_field): Don't do known_align < desired_align handling
7478 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7479 is non-NULL, instead do it after rli->prev_field handling and
7480 only if not within a bitfield word. For DECL_PACKED (field)
7481 use type_align of BITS_PER_UNIT.
7483 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7485 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7486 superfluous parentheses and trailing spaces.
7488 2018-02-28 Richard Biener <rguenther@suse.de>
7490 PR tree-optimization/84584
7491 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7492 SCoPs with fake exit edge.
7494 2018-02-28 Martin Liska <mliska@suse.cz>
7497 * timevar.c (timer::print): Fix format to properly print 100%
7500 2018-02-28 Richard Biener <rguenther@suse.de>
7503 * genmatch.c (capture_info::walk_match): Do not mark
7504 captured expressions without operands as expr_p given
7505 they act more like predicates and should be subject to
7506 "lost tail" side-effect preserving.
7508 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7510 PR rtl-optimization/81611
7511 * auto-inc-dec.c (attempt_change): Move dead note from
7512 mem_insn if it's the next use of regno
7513 (find_address): Take address use of reg holding
7514 non-incremented value. Add parm to limit search to the named
7516 (merge_in_block): Attempt to use a mem insn that is the next
7517 use of the original regno.
7519 2018-02-27 Martin Sebor <msebor@redhat.com>
7522 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7523 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7525 2018-02-27 Martin Sebor <msebor@redhat.com>
7527 PR translation/84207
7528 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7529 n argument to unsigned HOST_WIDE_INT.
7530 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7531 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7532 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7533 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7535 2018-02-27 Richard Biener <rguenther@suse.de>
7537 PR tree-optimization/84512
7538 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7539 Do not use the estimate returned from record_stmt_cost for
7540 the scalar iteration cost but sum properly using add_stmt_cost.
7542 2018-02-27 Richard Biener <rguenther@suse.de>
7544 PR tree-optimization/84466
7545 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7546 Adjust last change to less strictly validate use operands.
7548 2018-02-27 Martin Liska <mliska@suse.cz>
7550 PR gcov-profile/84548
7551 * gcov.c (process_file): Allow partial overlap and consider it
7552 also as group functions.
7553 (output_lines): Properly calculate range of lines for a group.
7555 2018-02-27 Martin Liska <mliska@suse.cz>
7557 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7558 'ggc' suffixes. Change first column width.
7559 (timer::print): Fix formatting of the column.
7561 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7563 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7564 preserve inline entry blocks for the sake of debug inline
7565 entry point markers alone.
7566 (remove_unused_locals): Suggest in comments a better place to
7567 force the preservation of inline entry blocks that are
7568 otherwise unused, but do not preserve them.
7570 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7572 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7574 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7577 * config/i386/constraints.md (Bs): Replace
7578 ix86_indirect_branch_register with
7579 TARGET_INDIRECT_BRANCH_REGISTER.
7581 * config/i386/i386.md (indirect_jump): Likewise.
7582 (tablejump): Likewise.
7583 (*sibcall_memory): Likewise.
7584 (*sibcall_value_memory): Likewise.
7585 Peepholes of indirect call and jump via memory: Likewise.
7586 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7587 (*sibcall_value_GOT_32): Likewise.
7588 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7589 (GOT_memory_operand): Likewise.
7590 (call_insn_operand): Likewise.
7591 (sibcall_insn_operand): Likewise.
7592 (GOT32_symbol_operand): Likewise.
7593 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7595 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7597 PR rtl-optimization/83496
7598 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7599 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7600 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7601 redundant insn, if any.
7602 (relax_delay_slots): Likewise.
7603 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7605 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7607 PR tree-optimization/83965
7608 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7609 that grouped statements are part of a reduction chain. Return
7610 true if the statement is not marked as a reduction itself but
7612 (vect_recog_dot_prod_pattern): Don't check whether the statement
7613 is part of a group here.
7614 (vect_recog_sad_pattern): Likewise.
7615 (vect_recog_widen_sum_pattern): Likewise.
7617 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7620 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7622 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7625 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7627 (ix86_output_indirect_function_return): New prototype.
7628 (ix86_split_simple_return_pop_internal): Likewise.
7629 * config/i386/i386.c (indirect_return_via_cx): New.
7630 (indirect_return_via_cx_bnd): Likewise.
7631 (indirect_thunk_name): Handle return va CX_REG.
7632 (output_indirect_thunk_function): Create alias for
7633 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7634 (ix86_output_indirect_jmp): Remove the bool argument.
7635 (ix86_output_indirect_function_return): New function.
7636 (ix86_split_simple_return_pop_internal): Likewise.
7637 * config/i386/i386.md (*indirect_jump): Don't pass false
7638 to ix86_output_indirect_jmp.
7639 (*tablejump_1): Likewise.
7640 (simple_return_pop_internal): Change it to define_insn_and_split.
7641 Call ix86_split_simple_return_pop_internal to split it for
7643 (simple_return_indirect_internal): Call
7644 ix86_output_indirect_function_return instead of
7645 ix86_output_indirect_jmp.
7647 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7650 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7651 memset and value initialization afterwards.
7653 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7655 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7657 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7660 * common/config/aarch64/aarch64-common.c
7661 (aarch_option_optimization_table[]): Switch
7662 off fomit-frame-pointer
7664 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7665 Chung-Ju Wu <jasonwucj@gmail.com>
7667 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7669 (store_multiple): Ditto.
7671 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7673 * config.gcc: Add --with-cpu support for nds32 target.
7674 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7675 * config/nds32/nds32.opt: Add -mcpu= option.
7677 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7679 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7680 isel=yes): Warn for these deprecated options.
7682 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7684 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7685 ISA_2_5_MASKS_EMBEDDED.
7687 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7689 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7690 p->max as pointers rather than using iterative_hash_expr.
7692 2018-02-23 Carl Love <cel@us.ibm.com>
7694 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7695 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7697 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7698 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7699 P8V_BUILTIN_VEC_VUNSIGNED2.
7701 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7704 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7705 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7707 * lra-constraints.c (curr_insn_transform): Set up
7708 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7710 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7711 * lra-eliminations.c (spill_pseudos): Ditto.
7712 (process_insn_for_elimination): Ditto.
7713 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7714 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7715 LRA_NON_CLOBBERED_ALT.
7717 2018-02-22 Martin Sebor <msebor@redhat.com>
7719 PR tree-optimization/84480
7720 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7721 to maybe_diag_stxncpy_trunc. Call it.
7722 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7723 from gimple_fold_builtin_strcpy. Print inlining stack.
7724 (handle_builtin_stxncpy): Print inlining stack.
7725 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7727 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7730 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7731 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7732 and -fcheck-pointer-bounds are used together.
7733 (indirect_thunk_prefix): New enum.
7734 (indirect_thunk_need_prefix): New function.
7735 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7736 "_nt" instead of "_bnd" for NOTRACK prefix.
7737 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7738 (output_indirect_thunk_function): Likewise.
7740 (ix86_code_end): Update output_indirect_thunk_function calls.
7741 (ix86_output_indirect_branch_via_reg): Replace
7742 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7743 (ix86_output_indirect_branch_via_push): Likewise.
7744 (ix86_output_function_return): Likewise.
7745 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7746 incompatible with -fcf-protection=branch and
7747 -fcheck-pointer-bounds.
7749 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7752 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7753 Change gcc_assert call to output_operand_lossage.
7755 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7757 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7759 2018-02-22 DJ Delorie <dj@redhat.com>
7760 Sebastian Perta <sebastian.perta@renesas.com>
7761 Oleg Endo <olegendo@gcc.gnu.org>
7763 * config/rx/rx.c (rx_rtx_costs): New function.
7764 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7766 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7768 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7770 2018-02-22 Martin Liska <mliska@suse.cz>
7773 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7774 Add "native" as a possible value.
7776 2018-02-22 Martin Liska <mliska@suse.cz>
7779 * config/i386/i386.c (ix86_option_override_internal):
7780 Add "native" as a possible value for -march and -mtune.
7782 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7785 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7786 to all type variants.
7788 PR tree-optimization/84503
7789 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7790 width as info->bitpos + info->bitsize - start.
7791 (merged_store_group::merge_overlapping): Simplify width computation.
7792 (check_no_overlap): New function.
7793 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7794 start + width and last_order of the group, fail if check_no_overlap
7796 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7797 to group if check_no_overlap fails.
7799 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7801 * config/rs6000/altivec.md: Delete contraint arguments to
7802 define_expand, define_split, and define_peephole2, and in
7803 define_insn_and_split if always unused.
7804 * config/rs6000/darwin.md: Ditto.
7805 * config/rs6000/dfp.md: Ditto.
7806 * config/rs6000/rs6000.md: Ditto.
7807 * config/rs6000/sync.md: Ditto.
7808 * config/rs6000/vector.md: Ditto.
7809 * config/rs6000/vsx.md: Ditto.
7811 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7813 * config/rs6000/altivec.md: Write output control strings as braced
7814 blocks instead of double-quoted strings.
7815 * config/rs6000/darwin.md: Ditto.
7816 * config/rs6000/rs6000.md: Ditto.
7817 * config/rs6000/vector.md: Ditto.
7818 * config/rs6000/vsx.md: Ditto.
7820 2018-02-21 Jason Merrill <jason@redhat.com>
7822 PR c++/84314 - ICE with templates and fastcall attribute.
7823 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7825 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7827 * ipa-cp.c (determine_versionability): Fix comment typos.
7829 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7832 * ipa-cp.c (determine_versionability): Do not version functions caling
7835 2018-02-21 Martin Liska <mliska@suse.cz>
7838 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7839 Add "native" as a possible value.
7840 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7841 the macro when native cpu detection is available.
7843 2018-02-21 Martin Liska <mliska@suse.cz>
7846 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7847 Add "native" as a possible value.
7848 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7849 when native cpu detection is available.
7851 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7852 Martin Sebor <msebor@redhat.com>
7854 PR tree-optimization/84478
7855 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7857 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7858 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7859 support which is conservatively correct, for 2 only stay conservative
7860 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7861 argument to the 2 argument get_range_strlen, adjust 6 arg
7862 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7864 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7865 (gimple_fold_builtin_strlen): Pass true as last argument to
7868 2018-02-20 Martin Sebor <msebor@redhat.com>
7871 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7872 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7873 (builtin_memref::builtin_memref): Factor out parts into
7874 set_base_and_offset and call it.
7876 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7879 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7880 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7881 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7882 search at the associated MODE_INT.
7884 2018-02-20 Jeff Law <law@redhat.com>
7887 PR tree-optimization/81592
7889 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7890 for range data rather than using global data.
7891 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7892 range data rather than using global data.
7893 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7894 pass it to children as needed.
7895 (struct directive::fmtresult): Similarly.
7896 (struct directive::set_width): Similarly.
7897 (struct directive::set_precision): Similarly.
7898 (format_integer, format_directive, parse_directive): Similarly.
7899 (format_none): Accept unnamed vr_values parameter.
7900 (format_percent, format_floating, format_character): Similarly.
7901 (format_string, format_plain): Similarly.
7902 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7903 the EVRP range analyzer for range data rather than using global data.
7904 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7905 gimple-ssa-evrp-analyze.h
7906 (class sprintf_dom_walker): Add after_dom_children member function.
7907 Add evrp_range_analyzer member.
7908 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7909 range analyzer as needed.
7910 (sprintf_dom_walker::after_dom_children): New member function.
7911 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7913 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7914 (evrp_range_analyzer::pop_to_marker): Likewise.
7916 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7918 PR tree-optimization/84419
7919 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7920 with the required type if its current type is compatible but
7923 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7926 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7927 after vectorization.
7929 2018-02-20 Martin Liska <mliska@suse.cz>
7932 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7933 possible values if we don't have a hint.
7935 2018-02-20 Martin Liska <mliska@suse.cz>
7939 * final.c (shorten_branches): Build align_tab array with one
7941 * opts.c (finish_options): Add alignment option limit check.
7942 (MAX_CODE_ALIGN): Likewise.
7943 (MAX_CODE_ALIGN_VALUE): Likewise.
7944 * doc/invoke.texi: Document maximum allowed option value for
7945 all -falign-* options.
7947 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7950 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7951 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7952 * var-tracking.c (emit_note_insn_var_location): Remove all references
7953 to NOTE_INSN_CALL_ARG_LOCATION.
7954 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7955 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7956 Use copy_rtx_if_shared.
7957 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7958 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7959 (dwarf2out_var_location): Remove handling of
7960 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7962 * final.c (final_scan_insn): Remove all references to
7963 NOTE_INSN_CALL_ARG_LOCATION.
7964 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7965 before dumping final insns.
7966 * except.c (emit_note_eh_region_end): Remove all references to
7967 NOTE_INSN_CALL_ARG_LOCATION.
7968 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7969 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7970 * config/arc/arc.c (hwloop_optimize): Likewise.
7971 * config/arm/arm.c (create_fix_barrier): Likewise.
7972 * config/s390/s390.c (s390_chunkify_start): Likewise.
7973 * config/sh/sh.c (find_barrier): Likewise.
7974 * config/i386/i386.c (rest_of_insert_endbranch,
7975 ix86_seh_fixup_eh_fallthru): Likewise.
7976 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7977 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7978 * config/frv/frv.c (frv_function_prologue): Likewise.
7979 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7981 (note_outside_basic_block_p): Remove all references to
7982 NOTE_INSN_CALL_ARG_LOCATION.
7983 * gengtype.c (adjust_field_rtx_def): Likewise.
7984 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7986 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7987 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7990 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7993 PR tree-optimization/84452
7994 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7995 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7998 2018-02-19 Martin Liska <mliska@suse.cz>
8001 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
8003 2018-02-19 Martin Liska <mliska@suse.cz>
8004 Richard Sandiford <richard.sandiford@linaro.org>
8006 PR tree-optimization/82491
8007 * gimple-fold.c (get_base_constructor): Make earlier bail out
8010 2018-02-19 Carl Love <cel@us.ibm.com>
8012 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
8013 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
8015 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
8016 P8V_BUILTIN_VEC_NEG.
8018 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
8020 * config/rl78/rl78.md (movdf): New define expand.
8022 2018-02-19 Martin Liska <mliska@suse.cz>
8025 * doc/invoke.texi: Fix typo.
8026 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
8028 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
8030 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8031 handle rs6000_single_float and rs6000_double_float specially for
8034 2018-02-16 Jeff Law <law@redhat.com>
8036 * config/rx/rx.c (add_pop_cfi_notes): New function.;
8039 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8042 * ipa-inline.c (inline_small_functions): Fix a typo.
8044 2018-02-16 Nathan Sidwell <nathan@acm.org>
8046 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
8048 2018-02-16 Carl Love <cel@us.ibm.com>
8050 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8051 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8052 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8053 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8054 expansion to P8V_BUILTIN_VEC_FLOAT2.
8056 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8058 PR rtl-optimization/70023
8059 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8060 src_regno into account.
8062 2018-02-16 Carl Love <cel@us.ibm.com>
8064 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8065 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8066 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8067 * config/rs6000/rs6000.c: Remove case statements for
8068 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8069 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8070 and P9V_BUILTIN_VEC_VINSERT4B.
8071 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8072 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8073 * config/rs6000/vsx.md:
8074 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8077 2018-02-16 Carl Love <cel@us.ibm.com>
8079 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8081 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8083 * config/rs6000/rs6000-c.c: Add the definitions for
8084 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8085 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8086 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8087 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8088 definition for insert4b and define insn *insert3b_internal.
8089 * doc/extend.texi: Add documentation for vec_extract4b.
8091 2018-02-16 Nathan Sidwell <nathan@acm.org>
8093 * doc/extend.texi (Backwards Compatibility): Mention friend
8094 injection. Note for-scope is deprecated.
8095 * doc/invoke.texi (-ffriend-injection): Deprecate.
8097 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8099 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8100 that moved to I2, also allow destinations that are a paradoxical
8101 subreg (instead of a normal reg).
8103 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8106 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8109 2018-02-16 Richard Biener <rguenther@suse.de>
8111 PR tree-optimization/84037
8112 PR tree-optimization/84016
8114 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8115 Adjust vec_construct for the fact we need additional higher latency
8116 128bit inserts for AVX256 and AVX512 vector builds.
8117 (ix86_add_stmt_cost): Scale vector construction cost for
8120 2018-02-16 Richard Biener <rguenther@suse.de>
8122 PR tree-optimization/84417
8123 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8124 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8125 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8127 2018-02-16 Martin Liska <mliska@suse.cz>
8130 * internal-fn.def (ASAN_CHECK): Set proper flags.
8131 (ASAN_MARK): Likewise.
8133 2018-02-16 Julia Koval <julia.koval@intel.com>
8135 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8136 from PTA_CANNONLAKE.
8138 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8141 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8142 Use ++iter rather than iter++ for std::list iterators.
8143 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8144 defer deleting them until all nodes in the forest are processed. Do
8145 free even leaf nodes. Change to_process into auto_vec.
8148 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8149 * vec.h (vec_default_construct): Use memset instead of placement new
8150 if BROKEN_VALUE_INITIALIZATION is defined.
8151 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8152 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8155 PR rtl-optimization/83723
8156 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8157 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8158 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8160 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8162 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8164 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8166 PR rtl-optimization/81443
8167 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8168 from inner REGs to paradoxical SUBREGs.
8170 2018-02-16 Richard Biener <rguenther@suse.de>
8172 PR tree-optimization/84399
8173 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8174 For operands we can analyze at their definition make sure we can
8175 analyze them at each use as well.
8177 2018-02-16 Richard Biener <rguenther@suse.de>
8179 PR tree-optimization/84190
8180 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8181 volatile accesses if the decl isn't volatile.
8183 2018-02-15 Jason Merrill <jason@redhat.com>
8185 PR c++/84314 - ICE with templates and fastcall attribute.
8186 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8187 TYPE_CANONICAL on an existing type.
8189 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8191 PR tree-optimization/84383
8192 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8193 dstoff nor call operand_equal_p if dstbase is NULL.
8195 PR tree-optimization/84334
8196 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8197 also a CONSTANT_CLASS_P, punt.
8199 2018-02-14 Jim Wilson <jimw@sifive.com>
8201 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8202 first SMALL_OPERAND check. New local min_second_step. Move assert
8203 to where locals are set. Add TARGET_RVC support.
8204 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8206 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8208 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8210 2018-02-14 Martin Sebor <msebor@redhat.com>
8212 PR tree-optimization/83698
8213 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8214 arrays constrain the offset range to their bounds.
8215 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8216 (builtin_access::overlap): Avoid setting the size of overlap if it's
8218 (maybe_diag_overlap): Also consider arrays when deciding what values
8219 of offsets to include in diagnostics.
8221 2018-02-14 Martin Sebor <msebor@redhat.com>
8224 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8225 that correspond to the kind of a declaration.
8227 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8230 * config/pa/pa.md: Load address of PIC label using the linkage table
8231 if the label is nonlocal.
8233 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8235 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8236 warning message if user requests -maltivec=be.
8237 * doc/invoke.texi: Document deprecation of -maltivec=be.
8239 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8242 * config/rs6000/rs6000-c.c: Update definitions for
8243 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8244 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8246 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8249 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8250 add _get_ssp intrinsics. Remove argument from
8251 __builtin_ia32_rdssp[d|q].
8252 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8253 * config/i386/i386-builtin.def: Remove argument from
8254 __builtin_ia32_rdssp[d|q].
8255 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8256 ix86_expand_special_args_builtin for _rdssp[d|q].
8257 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8258 Clear register before usage.
8259 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8260 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8262 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8264 PR tree-optimization/84357
8265 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8266 operand 1 of an ARRAY_REF too.
8268 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8271 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8272 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8274 (set_of_reg): New struct.
8275 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8276 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8277 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8279 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8280 Split into bitclr, bitset, bitinvert patterns if appropriate.
8281 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8282 use rx_fuse_in_memory_bitop.
8283 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8284 to named insn, correct maximum insn length.
8286 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8289 * machmode.def: Define a complex mode for PARTIAL_INT.
8290 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8292 * doc/rtl.texi: Document CSPImode.
8293 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8295 (msp430_hard_regno_nregs_with_padding): Likewise.
8297 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8300 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8302 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8304 PR rtl-optimization/84169
8305 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8306 we generated a parallel as new i3 and we split that to new i2 and i3
8307 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8308 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8309 those to i2, not i1. Partially rewrite this scan code.
8311 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8314 * stor-layout.c (place_field): For variable length fields, adjust
8315 offset_align afterwards not just based on the field's alignment,
8316 but also on the size.
8319 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8320 of exps and logs in the use_exp2 case.
8322 2018-02-13 Jeff Law <law@redhat.com>
8324 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8327 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8330 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8334 * common.opt (gas-loc-support, gas-locview-support): New.
8335 (ginline-points, ginternal-reset-location-views): New.
8336 * doc/invoke.texi: Document them. Use @itemx where intended.
8337 (gvariable-location-views): Adjust.
8338 * target.def (reset_location_view): New.
8339 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8340 (TARGET_RESET_LOCATION_VIEW): New.
8341 * doc/tm.texi: Rebuilt.
8342 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8343 (dwarf2out_default_as_locview_support): New.
8344 (output_asm_line_debug_info): Use option variables.
8345 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8346 (output_loc_list): Likewise.
8347 (add_high_low_attributes): Check option variables.
8348 Don't output entry view attribute in strict mode.
8349 (gen_inlined_subroutine_die): Check option variables.
8350 (dwarf2out_inline_entry): Likewise.
8351 (init_sections_and_labels): Likewise.
8352 (dwarf2out_early_finish): Likewise.
8353 (maybe_reset_location_view): New, from...
8354 (dwarf2out_var_location): ... here. Call it.
8355 * debug.h (dwarf2out_default_as_loc_support): Declare.
8356 (dwarf2out_default_as_locview_support): Declare.
8357 * hooks.c (hook_int_rtx_insn_0): New.
8358 * hooks.h (hook_int_rtx_insn_0): Declare.
8359 * toplev.c (process_options): Take -gas-loc-support and
8360 -gas-locview-support from dwarf2out. Enable
8361 -gvariable-location-views by default only with locview
8362 assembler support. Enable -ginternal-reset-location-views by
8363 default only if the target defines the corresponding hook.
8364 Enable -ginline-points by default if location views are
8365 enabled; force it disabled if statement frontiers are
8367 * tree-inline.c (expand_call_inline): Check option variables.
8368 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8370 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8372 PR tree-optimization/84321
8373 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8374 handling. Also check whether the anti-range contains any values
8375 that satisfy the mask; switch to a VR_RANGE if not.
8377 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8380 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8382 2018-02-13 Martin Jambor <mjambor@suse.cz>
8385 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8386 of call statements, also set location of a load to a temporary.
8388 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8390 * config/rl78/rl78.c (add_vector_labels): New function.
8391 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8392 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8393 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8394 which checks that no arguments are passed.
8395 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8396 * doc/extend.texi: Documentation for the new attribute.
8398 2018-02-13 Andreas Schwab <schwab@suse.de>
8400 * config/riscv/linux.h (CPP_SPEC): Define.
8402 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8405 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8406 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8407 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8408 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8409 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8410 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8411 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8412 temporarily for AES and PCLMUL builtins.
8414 PR tree-optimization/84339
8415 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8416 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8420 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8421 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8422 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8424 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8426 * omp-simd-clone.h: New file.
8427 * omp-simd-clone.c: Include omp-simd-clone.h.
8428 (expand_simd_clones): No longer static.
8429 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8430 cgraph.h and omp-simd-clone.h.
8431 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8432 (vect_recog_widen_shift_pattern): Formatting fix.
8433 (vect_pattern_recog_1): Don't check optab for calls.
8436 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8437 operands[2] into a REG before using gen_lowpart on it.
8439 2018-02-12 Jeff Law <law@redhat.com>
8442 * config/sh/sh.c (find_barrier): Consider a sibling call
8445 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8446 successfully back substituting a reg.
8448 2018-02-12 Richard Biener <rguenther@suse.de>
8450 PR tree-optimization/84037
8451 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8452 parameter, move visited init to caller.
8453 (vect_slp_analyze_operations): Separate cost from validity
8454 check, initialize visited once for all instances.
8455 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8457 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8459 (vect_model_promotion_demotion_cost): Likewise.
8460 (vectorizable_bswap): Guard cost modeling with !slp_node
8461 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8463 (vectorizable_call): Likewise.
8464 (vectorizable_conversion): Likewise.
8465 (vectorizable_assignment): Likewise.
8466 (vectorizable_shift): Likewise.
8467 (vectorizable_operation): Likewise.
8468 (vectorizable_store): Likewise.
8469 (vectorizable_load): Likewise.
8470 (vectorizable_condition): Likewise.
8471 (vectorizable_comparison): Likewise.
8473 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8476 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8477 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8478 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8480 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8483 * params.def (inline-min-speedup): Increase from 8 to 15.
8484 (max-inline-insns-auto): Decrease from 40 to 30.
8485 * ipa-split.c (consider_split): Add some buffer for function to
8486 be considered inlining candidate.
8487 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8490 2018-02-12 Richard Biener <rguenther@suse.de>
8492 PR tree-optimization/84037
8493 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8494 matched stmts if we cannot swap the non-matched ones.
8496 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8498 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8499 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8500 _mm_maskz_scalef_round_ss): New intrinsics.
8501 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8502 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8503 __builtin_ia32_scalefss_round): Remove.
8504 (__builtin_ia32_scalefsd_mask_round,
8505 __builtin_ia32_scalefss_mask_round): New intrinsics.
8506 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8507 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8508 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8509 "<round_constraint>")): Changed to ...
8510 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8511 "<round_scalar_constraint>")): ... this.
8512 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8513 %0, %1, %2<round_op3>}"): Changed to ...
8514 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8515 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8516 %2<round_scalar_mask_op3>}"): ... this.
8517 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8519 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8521 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8522 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8523 (_mm_maskz_sqrt_round_ss): New intrinsics.
8524 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8525 (__builtin_ia32_sqrtsd_mask_round)
8526 (__builtin_ia32_sqrtss_mask_round): New builtins.
8527 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8528 (__builtin_ia32_sqrtss_round): Remove.
8529 (__builtin_ia32_sqrtsd_mask_round)
8530 (__builtin_ia32_sqrtss_mask_round): New builtins.
8531 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8532 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8533 ((match_operand:VF_128 1 "vector_operand"
8534 "xBm,<round_constraint>")): Changed to ...
8535 ((match_operand:VF_128 1 "vector_operand"
8536 "xBm,<round_scalar_constraint>")): ... this.
8537 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8538 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8539 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8540 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8541 %<iptr>1<round_scalar_mask_op3>}): ... this.
8542 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8543 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8545 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8548 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8549 Cast vec_cmpeq result to correct type.
8550 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8551 Cast vec_cmpgt result to correct type.
8553 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8555 * final.c (final_scan_insn_1): Renamed from...
8556 (final_scan_insn): ... this. New wrapper, to recover
8557 seen from the outermost call in recursive ones.
8558 * config/sparc/sparc.c (output_return): Drop seen from call.
8559 (output_sibcall): Likewise.
8560 * config/visium/visium.c (output_branch): Likewise.
8562 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8564 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8567 2018-02-10 Alan Modra <amodra@gmail.com>
8570 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8571 Specify LR as an input.
8573 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8576 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8577 remove_member_access_dummy_vars): New functions.
8578 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8579 lower_omp_1, execute_lower_omp): Use them.
8581 PR rtl-optimization/84308
8582 * shrink-wrap.c (spread_components): Release todo vector.
8584 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8586 PR rtl-optimization/57193
8587 * ira-color.c (struct allocno_color_data): Add member
8588 conflict_allocno_hard_prefs.
8589 (update_conflict_allocno_hard_prefs): New.
8590 (bucket_allocno_compare_func): Add a preference based on
8591 conflict_allocno_hard_prefs.
8592 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8593 (color_allocnos): Remove a dead code. Initiate
8594 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8596 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8599 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8600 constraint from =wa to wa. Avoid a subreg on the output operand,
8601 instead use a pseudo and subreg it in a move.
8602 (p9_xxbrd_<mode>): Changed to ...
8603 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8604 (p9_xxbrd_v2df): New expander.
8605 (p9_xxbrw_<mode>): Changed to ...
8606 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8607 (p9_xxbrw_v4sf): New expander.
8609 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8611 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8613 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8616 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8617 multiply in 32-bit mode.
8618 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8619 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8622 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8624 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8625 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8626 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8627 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8629 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8632 * dwarf2out.c (is_trivial_indirect_ref): New function.
8633 (dwarf2out_late_global_decl): Do not generate a location
8634 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8635 and that are not defined in the current unit.
8637 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8639 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8640 instead of a libcall for UNORDERED.
8642 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8645 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8646 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8648 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8651 * config/s390/s390.c (s390_set_current_function): Invoke
8652 s390_indirect_branch_settings also if fndecl didn't change.
8654 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8656 * config/rs6000/rs6000.md (blockage): Set length to zero.
8658 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8660 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8662 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8665 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8666 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8670 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8671 PARALLEL incoming that failed vt_get_decl_and_offset check.
8674 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8675 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8677 (get_variable_section): For decls in named .bss* sections pass true as
8678 second argument to bss_initializer_p.
8680 2018-02-09 Marek Polacek <polacek@redhat.com>
8681 Jakub Jelinek <jakub@redhat.com>
8684 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8685 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8686 Sync some changes from cxx_fold_indirect_ref.
8688 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8690 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8692 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8693 (BLOCK_INLINE_ENTRY_LABEL): New.
8694 (dwarf2out_var_location): Disregard inline entry markers.
8695 (inline_entry_data): New struct.
8696 (inline_entry_data_hasher): New hashtable type.
8697 (inline_entry_data_hasher::hash): New.
8698 (inline_entry_data_hasher::equal): New.
8699 (inline_entry_data_table): New variable.
8700 (add_high_low_attributes): Add DW_AT_entry_pc and
8701 DW_AT_GNU_entry_view attributes if a pending entry is found
8702 in inline_entry_data_table. Add old entry_pc attribute only
8703 if debug nonbinding markers are disabled.
8704 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8705 markers are enabled.
8706 (block_within_block_p, dwarf2out_inline_entry): New.
8707 (dwarf2out_finish): Check that no entries remained in
8708 inline_entry_data_table.
8709 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8710 (final_scan_insn, notice_source_line): Likewise.
8711 (rest_of_clean_state): Skip inline entry markers.
8712 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8714 * gimple.c (gimple_build_debug_inline_entry): New.
8715 * gimple.h (enum gimple_debug_subcode): Add
8716 GIMPLE_DEBUG_INLINE_ENTRY.
8717 (gimple_build_debug_inline_entry): Declare.
8718 (gimple_debug_inline_entry_p): New.
8719 (gimple_debug_nonbind_marker_p): Adjust.
8720 * insn-notes.def (INLINE_ENTRY): New.
8721 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8722 inline entry marker notes.
8723 (print_insn): Likewise.
8724 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8725 (INSN_DEBUG_MARKER_KIND): Likewise.
8726 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8727 * tree-inline.c (expand_call_inline): Build and insert
8728 debug_inline_entry stmt.
8729 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8730 inline entry blocks early, if nonbind markers are enabled.
8731 (dump_scope_block): Dump fragment info.
8732 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8733 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8734 (gimple_build_debug_inline_entry): New.
8735 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8736 Enable/disable inline entry points too.
8737 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8738 (DEBUG_INSN): Describe inline entry markers.
8740 * common.opt (gvariable-location-views): New.
8741 (gvariable-location-views=incompat5): New.
8742 * config.in: Rebuilt.
8743 * configure: Rebuilt.
8744 * configure.ac: Test assembler for view support.
8745 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8746 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8747 * dwarf2out.c (var_loc_view): New typedef.
8748 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8749 (dwarf2out_locviews_in_attribute): New.
8750 (dwarf2out_locviews_in_loclist): New.
8751 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8752 (enum dw_line_info_opcode): Add LI_adv_address.
8753 (struct dw_line_info_table): Add view.
8754 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8755 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8756 (zero_view_p): New variable.
8757 (ZERO_VIEW_P): New macro.
8758 (output_asm_line_debug_info): New.
8759 (struct var_loc_node): Add view.
8760 (add_AT_view_list, AT_loc_list): New.
8761 (add_var_loc_to_decl): Add view param. Test it against last.
8762 (new_loc_list): Add view params. Record them.
8763 (AT_loc_list_ptr): Handle loc and view lists.
8764 (view_list_to_loc_list_val_node): New.
8765 (print_dw_val): Handle dw_val_class_view_list.
8766 (size_of_die): Likewise.
8767 (value_format): Likewise.
8768 (loc_list_has_views): New.
8769 (gen_llsym): Set vl_symbol too.
8770 (maybe_gen_llsym, skip_loc_list_entry): New.
8771 (dwarf2out_maybe_output_loclist_view_pair): New.
8772 (output_loc_list): Output view list or entries too.
8773 (output_view_list_offset): New.
8774 (output_die): Handle dw_val_class_view_list.
8775 (output_dwarf_version): New.
8776 (output_compilation_unit_header): Use it.
8777 (output_skeleton_debug_sections): Likewise.
8778 (output_rnglists, output_line_info): Likewise.
8779 (output_pubnames, output_aranges): Update version comments.
8780 (output_one_line_info_table): Output view numbers in asm comments.
8781 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8782 Call maybe_gen_llsym.
8783 (loc_list_from_tree_1): Adjust.
8784 (add_AT_location_description): Create view list attribute if
8785 needed, check it's absent otherwise.
8786 (convert_cfa_to_fb_loc_list): Adjust.
8787 (maybe_emit_file): Call output_asm_line_debug_info for test.
8788 (dwarf2out_var_location): Reset views as needed. Precompute
8789 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8790 attribute. Set view.
8791 (new_line_info_table): Reset next view.
8792 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8793 (dwarf2out_source_line): Likewise. Output view resets and labels to
8794 the assembler, or select appropriate line info opcodes.
8795 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8796 (optimize_string_length): Catch it. Adjust.
8797 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8798 dw_val_class_view_list, and remove it if no longer needed.
8799 (hash_loc_list): Hash view numbers.
8800 (loc_list_hasher::equal): Compare them.
8801 (optimize_location_lists): Check whether a view list symbol is
8802 needed, and whether the locview attribute is present, and
8803 whether they match. Remove the locview attribute if no longer
8805 (index_location_lists): Call skip_loc_list_entry for test.
8806 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8807 Use output_dwarf_version.
8808 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8809 (struct dw_val_node): Add val_view_list.
8810 * final.c (SEEN_NEXT_VIEW): New.
8811 (set_next_view_needed): New.
8812 (clear_next_view_needed): New.
8813 (maybe_output_next_view): New.
8814 (final_start_function): Rename to...
8815 (final_start_function_1): ... this. Take pointer to FIRST,
8816 add SEEN parameter. Emit param bindings in the initial view.
8817 (final_start_function): Reintroduce SEEN-less interface.
8818 (final): Rename to...
8819 (final_1): ... this. Take SEEN parameter. Output final pending
8820 next view at the end.
8821 (final): Reintroduce seen-less interface.
8822 (final_scan_insn): Output pending next view before switching
8823 sections or ending a block. Mark the next view as needed when
8824 outputting variable locations. Notify debug backend of section
8825 changes, and of location view changes.
8826 (rest_of_handle_final): Adjust.
8827 * toplev.c (process_options): Autodetect value for debug variable
8828 location views option. Warn on incompat5 without -gdwarf-5.
8829 * doc/invoke.texi (gvariable-location-views): New.
8830 (gvariable-location-views=incompat5): New.
8831 (gno-variable-location-views): New.
8833 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8835 PR tree-optimization/84136
8836 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8837 that the result of find_edge is non-NULL.
8839 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8842 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8843 storing integer register in SImode. Fix cost of 256 and 512
8844 byte aligned SSE register store.
8846 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8848 * config/i386/i386.c (ix86_multiplication_cost): Fix
8849 multiplication cost for TARGET_AVX512DQ.
8851 2018-02-08 Marek Polacek <polacek@redhat.com>
8853 PR tree-optimization/84238
8854 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8857 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8859 PR tree-optimization/84265
8860 * tree-vect-stmts.c (vectorizable_store): Don't treat
8861 VMAT_CONTIGUOUS accesses as grouped.
8862 (vectorizable_load): Likewise.
8864 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8866 PR tree-optimization/81635
8867 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8868 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8869 (test_round_for_mask): New functions.
8870 (wide_int_cc_tests): Call test_round_for_mask.
8871 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8872 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8873 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8874 range returned by get_range_info.
8876 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8879 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8880 * symtab.c: Include builtins.h
8881 (symtab_node::output_to_lto_symbol_table_p): Move here
8882 from lto-streamer-out.c:output_symbol_p.
8883 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8884 (output_symbol_p): Move all logic to symtab.c
8885 (produce_symtab): Update.
8887 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8889 * config/s390/s390-opts.h (enum indirect_branch): Define.
8890 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8891 (s390_indirect_branch_via_thunk)
8892 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8893 (enum s390_indirect_branch_type): Define.
8894 * config/s390/s390.c (struct s390_frame_layout, struct
8895 machine_function): Remove.
8896 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8897 (indirect_branch_table_label_no, indirect_branch_table_name):
8899 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8900 (enum s390_indirect_branch_option): Define.
8901 (s390_return_addr_from_memory): New function.
8902 (s390_handle_string_attribute): New function.
8903 (s390_attribute_table): Add new attribute handler.
8904 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8905 (s390_indirect_branch_via_thunk): New function.
8906 (s390_indirect_branch_via_inline_thunk): New function.
8907 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8908 sibling call optimization for non z10 compiles.
8909 (s390_emit_call): Force indirect branch target to be a single
8910 register. Add r1 clobber for non-z10 compiles.
8911 (s390_emit_epilogue): Emit return jump via return_use expander.
8912 (s390_reorg): Handle JUMP_INSNs as execute targets.
8913 (s390_option_override_internal): Perform validity checks for the
8914 new command line options.
8915 (s390_indirect_branch_attrvalue): New function.
8916 (s390_indirect_branch_settings): New function.
8917 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8918 (s390_output_indirect_thunk_function): New function.
8919 (s390_code_end): Implement target hook.
8920 (s390_case_values_threshold): Implement target hook.
8921 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8923 * config/s390/s390.h (struct s390_frame_layout)
8924 (struct machine_function): Move here from s390.c.
8925 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8926 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8927 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8928 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8929 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8930 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8931 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8932 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8933 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8934 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8935 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8936 (mnemonic attribute): Add values which aren't recognized
8938 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8939 pattern for branch conversion. Fix mnemonic attribute.
8940 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8941 indirect branch via thunk if requested.
8942 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8943 ("*indirect_jump"): Disable for branch conversion using out of
8945 ("indirect_jump_via_thunk<mode>_z10")
8946 ("indirect_jump_via_thunk<mode>")
8947 ("indirect_jump_via_inlinethunk<mode>_z10")
8948 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8949 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8950 ("casesi_jump_via_inlinethunk<mode>_z10")
8951 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8952 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8953 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8954 ("*indirect2_jump"): Disable for branch conversion.
8955 ("casesi_jump"): Turn into expander and expand patterns for branch
8957 ("return_use"): New expander.
8958 ("*return"): Emit return via thunk and rename it to ...
8959 ("*return<mode>"): ... this one.
8960 * config/s390/s390.opt: Add new options and and enum for the
8963 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8965 * lra-constraints.c (match_reload): Unconditionally use
8966 gen_lowpart_SUBREG, rather than selecting between that
8967 and equivalent gen_rtx_SUBREG code.
8969 2018-02-08 Richard Biener <rguenther@suse.de>
8971 PR tree-optimization/84233
8972 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8973 changed flag instead of boguously re-using phi_inserted.
8975 2018-02-08 Martin Jambor <mjambor@suse.cz>
8977 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8978 static local variables.
8980 2018-02-08 Richard Biener <rguenther@suse.de>
8982 PR tree-optimization/84278
8983 * tree-vect-stmts.c (vectorizable_store): When looking for
8984 smaller vector types to perform grouped strided loads/stores
8985 make sure the mode is supported by the target.
8986 (vectorizable_load): Likewise.
8988 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8990 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8991 Increase LDP/STP opportunities by adding adjacent callee-saves.
8993 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8995 PR rtl-optimization/84068
8996 PR rtl-optimization/83459
8997 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8999 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
9001 PR tree-optimization/84224
9002 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
9003 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
9006 2018-02-07 Iain Sandoe <iain@codesourcery.com>
9009 * config/rs6000/altivec.md (*restore_world): Remove LR use.
9010 * config/rs6000/predicates.md (restore_world_operation): Adjust op
9011 count, remove one USE.
9013 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
9015 * doc/install.texi (Configuration): Document the
9016 --with-long-double-format={ibm,ieee} PowerPC configuration
9020 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
9021 Convert from define_expand to be define_insn_and_split. Rework
9022 float/double/_Float128 conversions to QI/HI/SImode to work with
9023 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
9024 conversions to QI/HImode types did a store and then a load to
9025 truncate the value. For conversions to VSX registers, don't split
9026 the insn, instead emit the code directly. Use the code iterator
9027 any_fix to combine signed and unsigned conversions.
9028 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
9029 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
9030 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9031 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
9032 (fix_<mode>di2_hw): Likewise.
9033 (fixuns_<mode>di2_hw): Likewise.
9034 (fix_<mode>si2_hw): Likewise.
9035 (fixuns_<mode>si2_hw): Likewise.
9036 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
9037 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
9038 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
9039 fix<uns>_trunc<SFDF:mode>si2_p8.
9040 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
9042 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
9043 (fix<uns>_<mode>_mem): Likewise.
9044 (fctiw<u>z_<mode>_mem): Likewise.
9045 (fix<uns>_<mode>_mem): Likewise.
9046 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
9047 the register allocator from doing a direct move to the GPRs to do
9048 a store, and instead use the ISA 3.0 store byte/half-word from
9049 vector register instruction. For IEEE 128-bit floating point,
9050 also optimize stores of 32-bit ints.
9051 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9053 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9055 * genextract.c (push_pathstr_operand): New function to support
9057 (walk_rtx): Call push_pathstr_operand.
9058 (print_path): Support [a-zA-Z].
9060 2018-02-07 Richard Biener <rguenther@suse.de>
9062 PR tree-optimization/84037
9063 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9064 (cse_and_gimplify_to_preheader): Declare.
9065 (vect_get_place_in_interleaving_chain): Likewise.
9066 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9068 (_loop_vec_info::~_loop_vec_info): Delete it.
9069 (cse_and_gimplify_to_preheader): New function.
9070 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9071 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9072 (vectorizable_load): Likewise. For grouped stores always base
9073 the IV on the first element.
9074 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9075 condition before gimplifying.
9077 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9079 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9080 *DIV_EXPR and *MOD_EXPR.
9082 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9085 * config/i386/i386.c (ix86_option_override_internal): Mask out
9086 the CF_SET bit when checking -fcf-protection.
9088 2018-02-07 Tom de Vries <tom@codesourcery.com>
9091 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9094 2018-02-07 Richard Biener <rguenther@suse.de>
9096 PR tree-optimization/84204
9097 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9100 PR tree-optimization/84205
9101 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9102 special-case isl_ast_op_zdiv_r.
9104 PR tree-optimization/84223
9105 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9106 Only add conditions from within the region.
9107 (gather_bbs::after_dom_children): Adjust.
9109 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9112 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9113 * config/avr/avr.md: Only post-reload split REG-REG moves if
9114 either register is GENERAL_REG_P.
9116 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9118 PR tree-optimization/84235
9119 * tree-ssa-scopedtables.c
9120 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9121 if the subtraction is performed in floating point type where NaNs are
9122 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9123 build 1. Formatting fix.
9125 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9128 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9129 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9130 and skip it regardless of bb boundaries. Use CALL_P macro,
9131 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9132 unnecessarily, formatting fix.
9134 2018-02-06 Michael Collison <michael.collison@arm.com>
9136 * config/arm/thumb2.md:
9137 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9138 (*thumb_mov_notscc): Ditto.
9140 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9143 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9144 unsigned_fix, not "s".
9146 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9148 * configure.ac (gcc_fn_eh_frame_ro): New function.
9149 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9150 correct .eh_frame permissions.
9151 * configure: Regenerate.
9153 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9155 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9158 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9160 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9161 Display warning message for -mno-speculate-indirect-jumps.
9163 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9165 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9167 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9169 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9171 PR tree-optimization/84225
9172 * tree-eh.c (find_trapping_overflow): Only call
9173 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9175 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9178 * config/i386/i386.c: Reimplement the check of possible options
9179 -mibt/-mshstk conbination. Change error messages.
9180 * doc/invoke.texi: Fix a typo: remove extra '='.
9182 2018-02-06 Marek Polacek <polacek@redhat.com>
9184 PR tree-optimization/84228
9185 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9187 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9190 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9191 emitted arch directives.
9192 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9193 __ARM_FEATURE_COPROC before changing architectures.
9195 2018-02-06 Richard Biener <rguenther@suse.de>
9197 * config/i386/i386.c (print_reg): Fix typo.
9198 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9200 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9202 * configure: Regenerate.
9204 2018-02-05 Martin Sebor <msebor@redhat.com>
9206 PR tree-optimization/83369
9207 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9210 2018-02-05 Martin Liska <mliska@suse.cz>
9212 * doc/invoke.texi: Cherry-pick upstream r323995.
9214 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9216 * ira.c (ira_init_register_move_cost): Adjust comment.
9218 2018-02-05 Martin Liska <mliska@suse.cz>
9220 PR gcov-profile/84137
9221 * doc/gcov.texi: Fix typo in documentation.
9223 2018-02-05 Martin Liska <mliska@suse.cz>
9225 PR gcov-profile/83879
9226 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9227 using dlopen functionality.
9229 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9231 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9232 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9233 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9234 _mm_maskz_range_round_ss): New intrinsics.
9235 (__builtin_ia32_rangesd128_round)
9236 (__builtin_ia32_rangess128_round): Remove.
9237 (__builtin_ia32_rangesd128_mask_round,
9238 __builtin_ia32_rangess128_mask_round): New builtins.
9239 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9240 __builtin_ia32_rangess128_round): Remove.
9241 (__builtin_ia32_rangesd128_mask_round,
9242 __builtin_ia32_rangess128_mask_round): New builtins.
9243 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9244 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9245 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9246 "<round_saeonly_constraint>")): Changed to ...
9247 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9248 "<round_saeonly_scalar_constraint>")): ... this.
9249 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9250 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9251 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9252 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9253 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9255 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9257 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9259 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9260 Remove all values except native, 8540 and 8548.
9262 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9264 * config/i386/i386.c (ix86_output_function_return): Pass
9265 INVALID_REGNUM, instead of -1, as invalid register number to
9266 indirect_thunk_name and output_indirect_thunk.
9268 2018-02-02 Julia Koval <julia.koval@intel.com>
9270 * config.gcc: Add -march=icelake.
9271 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9272 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9273 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9274 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9275 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9276 (processor_target_table): Add icelake.
9277 (ix86_option_override_internal): Handle new PTAs.
9278 (get_builtin_code_for_version): Handle icelake.
9279 (M_INTEL_COREI7_ICELAKE): New.
9280 (fold_builtin_cpu): Handle icelake.
9281 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9282 * doc/invoke.texi: Add -march=icelake.
9284 2018-02-02 Julia Koval <julia.koval@intel.com>
9286 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9287 to wide_int_bitmask.
9288 * wide-int-bitmask.h: New.
9290 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9293 * config/i386/i386.md: Replace Pmode with word_mode in
9294 builtin_setjmp_setup and builtin_longjmp to support x32.
9296 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9300 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9302 (rs6000_supported_cpu_names): New static variable.
9303 (linux_cpu_translation_table): Likewise.
9304 (elf_platform) <cpu>: Define new static variable and use it.
9305 Translate kernel AT_PLATFORM name to canonical name if needed.
9306 Error if platform name is unknown.
9308 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9311 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9313 2018-02-01 Jeff Law <law@redhat.com>
9316 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9317 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9318 the scratch if RELEASE_VIA_POP is false.
9319 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9320 If we have to save a temporary register, decrement SIZE appropriately.
9321 Pass new arguments to release_scratch_register_on_entry.
9322 (ix86_adjust_stack_and_probe): Likewise.
9323 (ix86_emit_probe_stack_range): Pass new arguments to
9324 release_scratch_register_on_entry.
9326 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9328 PR rtl-optimization/84157
9329 * combine.c (change_zero_ext): Use REG_P predicate in
9330 front of HARD_REGISTER_P predicate.
9332 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9334 * config/avr/avr.c (avr_option_override): Move disabling of
9335 -fdelete-null-pointer-checks to...
9336 * common/config/avr/avr-common.c (avr_option_optimization_table):
9339 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9341 PR tree-optimization/81635
9342 * tree-data-ref.c (split_constant_offset_1): For types that
9343 wrap on overflow, try to use range info to prove that wrapping
9346 2018-02-01 Renlin Li <renlin.li@arm.com>
9349 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9351 (aarch64_register_move_cost): Likewise.
9352 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9354 (REG_CLASS_NAMES): Likewise.
9355 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9356 TAILCALL_ADDR_REGS. Remove IP registers.
9357 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9359 2018-02-01 Richard Biener <rguenther@suse.de>
9361 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9362 for specifying RPO order and allow NULL for that.
9363 * domwalk.c (dom_walker::dom_walker): Likewise.
9364 (dom_walker::walk): Handle NULL RPO order.
9365 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9367 (rewrite_update_dom_walker): Likewise.
9368 (mark_def_dom_walker): Likewise.
9370 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9372 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9373 (aarch64_maybe_expand_sve_subreg_move): Declare.
9374 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9375 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9377 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9378 that are semantically a reverse operation.
9379 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9380 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9381 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9383 (aarch64_can_change_mode_class): For big-endian, forbid changes
9384 between two SVE modes if they have different element sizes.
9386 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9388 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9389 the TImode handling for big-endian targets.
9391 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9393 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9394 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9396 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9397 Remove BSWAP handing for big-endian targets and use the form of
9398 LD1RQ appropariate for the mode.
9400 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9402 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9403 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9406 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9409 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9410 check for operands that need to go through aarch64_sve_reload_be.
9412 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9414 PR tree-optimization/81661
9415 PR tree-optimization/84117
9416 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9417 * tree-eh.c: Include gimplify.h.
9418 (find_trapping_overflow, replace_trapping_overflow,
9419 rewrite_to_non_trapping_overflow): New functions.
9420 * tree-vect-loop.c: Include tree-eh.h.
9421 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9422 * tree-data-ref.c: Include tree-eh.h.
9423 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9425 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9427 PR rtl-optimization/84123
9428 * combine.c (change_zero_ext): Check if hard register satisfies
9429 can_change_dest_mode before calling gen_lowpart_SUBREG.
9431 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9434 * ira.c (ira_init_register_move_cost): Remove assert.
9436 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9438 PR rtl-optimization/84071
9439 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9440 * doc/tm.texi: Regenerate.
9442 2018-01-31 Richard Biener <rguenther@suse.de>
9444 PR tree-optimization/84132
9445 * tree-data-ref.c (analyze_miv_subscript): Properly
9446 check whether evolution_function_is_affine_multivariate_p
9447 before calling gcd_of_steps_may_divide_p.
9449 2018-01-31 Julia Koval <julia.koval@intel.com>
9452 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9453 * config/i386/i386.md (rdpid_rex64) New.
9454 (rdpid): Make 32bit only.
9456 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9459 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9460 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9462 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9465 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9467 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9469 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9471 PR rtl-optimization/84071
9472 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9473 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9475 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9477 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9478 (arc_attribute_table): Add 'aux' attribute.
9479 (arc_in_small_data_p): Consider aux like variables.
9480 (arc_is_aux_reg_p): New function.
9481 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9482 (arc_get_aux_arg): New function.
9483 (prepare_move_operands): Handle aux-register access.
9484 (arc_handle_aux_attribute): New function.
9485 * doc/extend.texi (ARC Variable attributes): Add subsection.
9487 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9489 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9490 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9491 (arc_attribute_table): Add 'uncached' attribute.
9492 (arc_print_operand): Print '.di' flag for uncached memory
9494 (arc_in_small_data_p): Do not consider for small data the uncached
9496 (arc_is_uncached_mem_p): New function.
9497 * config/arc/predicates.md (compact_store_memory_operand): Check
9498 for uncached memory accesses.
9499 (nonvol_nonimm_operand): Likewise.
9500 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9502 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9505 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9506 falign-loops=): Add Optimization flag.
9508 2018-01-30 Jeff Law <law@redhat.com>
9511 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9512 INT_REGISTERS_SAVED. Check it prior to calling
9513 get_scratch_register_on_entry.
9514 (ix86_adjust_stack_and_probe): Similarly.
9515 (ix86_emit_probe_stack_range): Similarly.
9516 (ix86_expand_prologue): Corresponding changes.
9518 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9521 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9522 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9524 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9527 * lra-constraints.c (curr_insn_transform): Process AND in the
9530 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9532 PR rtl-optimization/83986
9533 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9534 dependence against last_pending_memory_flush in addition to
9537 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9539 PR tree-optimization/81611
9540 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9543 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9546 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9549 2018-01-30 Richard Biener <rguenther@suse.de>
9550 Jakub Jelinek <jakub@redhat.com>
9552 PR tree-optimization/84111
9553 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9554 inner loops added during recursion, as they don't have up-to-date
9557 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9560 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9561 (can_inline_edge_by_limits_p): ... here.
9562 (can_early_inline_edge_p, check_callers,
9563 update_caller_keys, update_callee_keys, recursive_inlining,
9564 add_new_edges_to_heap, speculation_useful_p,
9565 inline_small_functions,
9566 inline_small_functions, flatten_function,
9567 inline_to_all_callers_1): Update.
9569 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9571 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9574 2018-01-30 Richard Biener <rguenther@suse.de>
9576 PR tree-optimization/83008
9577 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9578 invariant and constant vector uses in stmts when they need
9581 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9584 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9585 * configure: Regenerate.
9587 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9589 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9591 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9592 Use gen_rtx_REG rather than gen_lowpart.
9594 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9596 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9597 rather than 0 when creating partial subregs.
9599 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9601 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9604 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9607 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9608 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9609 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9610 flags. This restores the settings used before the 2017-07-24.
9611 Turning off pre increment/decrement/modify allows IVOPTS to
9612 optimize DF/SF loops where the index is an int.
9614 2018-01-29 Richard Biener <rguenther@suse.de>
9615 Kelvin Nilsen <kelvin@gcc.gnu.org>
9618 * tree-vect-stmts.c (vectorizable_call): Don't call
9619 targetm.vectorize_builtin_md_vectorized_function if callee is
9622 2018-01-22 Carl Love <cel@us.ibm.com>
9624 * doc/extend.tex: Fix typo in second arg in
9625 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9627 2018-01-29 Richard Biener <rguenther@suse.de>
9629 PR tree-optimization/84086
9630 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9631 (flush_ssaname_freelist): When SSA names were released reset
9632 the SCEV hash table.
9634 2018-01-29 Richard Biener <rguenther@suse.de>
9636 PR tree-optimization/84057
9637 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9638 removed paths when removing edges.
9640 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9642 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9643 -mfunction-return=@var{choice}.
9645 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9648 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9650 (layout::print_source_line): Likewise.
9651 (test_get_line_width_without_trailing_whitespace): Add test cases.
9653 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9656 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9659 2018-01-26 Jim Wilson <jimw@sifive.com>
9661 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9663 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9666 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9668 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9669 and CMP + SUB-immediate -> SUBS.
9671 2018-01-26 Martin Sebor <msebor@redhat.com>
9673 PR tree-optimization/83896
9674 * tree-ssa-strlen.c (get_string_len): Rename...
9675 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9676 Avoid assuming length is constant.
9677 (handle_char_store): Use HOST_WIDE_INT for string length.
9679 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9682 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9683 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9685 2018-01-26 Richard Biener <rguenther@suse.de>
9687 PR rtl-optimization/84003
9688 * dse.c (record_store): Only record redundant stores when
9689 the earlier store aliases at least all accesses the later one does.
9691 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9693 PR rtl-optimization/83985
9694 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9695 REG_CFA_RESTORE insns.
9696 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9697 REG_CFA_RESTORE insns here.
9700 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9701 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9703 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9705 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9706 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9707 (arc_init): Likewise.
9708 (arc_override_options): Likewise.
9709 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9711 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9713 * config/arc/arc.h (TARGET_DBNZ): Define.
9714 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9715 properly set the tune attribute.
9716 (dbnz): Use TARGET_DBNZ guard.
9717 * config/arc/arc.opt (mtune): Add core3 option.
9719 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9721 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9722 recognize new pic like addresses.
9723 (arc_delegitimize_address): Clean up.
9725 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9727 * config/arc/arc-arches.def: Option mrf16 valid for all
9729 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9730 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9731 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9732 * config/arc/arc-tables.opt: Regenerate.
9733 * config/arc/arc.c (arc_conditional_register_usage): Handle
9734 reduced register file case.
9735 (arc_file_start): Set must have build attributes.
9736 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9738 * config/arc/arc.opt (mrf16): Add new option.
9739 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9740 * config/arc/genmultilib.awk: Handle new mrf16 option.
9741 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9742 * config/arc/t-multilib: Regenerate.
9743 * doc/invoke.texi (ARC Options): Document mrf16 option.
9745 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9747 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9748 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9749 (arc_attribute_table): Add 'secure_call' attribute.
9750 (arc_print_operand): Print secure call operand.
9751 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9753 (arc_is_secure_call_p): New function. * config/arc/arc.md
9754 (call_i): Add support for sjli instruction.
9755 (call_value_i): Likewise.
9756 * config/arc/constraints.md (Csc): New constraint.
9758 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9759 John Eric Martin <John.Martin@emmicro-us.com>
9761 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9762 * config/arc/arc.c (_arc_jli_section): New struct.
9763 (arc_jli_section): New type.
9764 (rc_jli_sections): New static variable.
9765 (arc_handle_jli_attribute): New function.
9766 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9767 (arc_file_end): New function.
9768 (TARGET_ASM_FILE_END): Define.
9769 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9770 (arc_add_jli_section): New function.
9771 (jli_call_scan): Likewise.
9772 (arc_reorg): Call jli_call_scan.
9773 (arc_output_addsi): Remove 'S' from printing asm operand.
9774 (arc_is_jli_call_p): New function.
9775 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9777 (movhi_insn): Likewise.
9778 (movsi_insn): Likewise.
9779 (movsi_set_cc_insn): Likewise.
9780 (loadqi_update): Likewise.
9781 (load_zeroextendqisi_update): Likewise.
9782 (load_signextendqisi_update): Likewise.
9783 (loadhi_update): Likewise.
9784 (load_zeroextendhisi_update): Likewise.
9785 (load_signextendhisi_update): Likewise.
9786 (loadsi_update): Likewise.
9787 (loadsf_update): Likewise.
9788 (movsicc_insn): Likewise.
9789 (bset_insn): Likewise.
9790 (bxor_insn): Likewise.
9791 (bclr_insn): Likewise.
9792 (bmsk_insn): Likewise.
9793 (bicsi3_insn): Likewise.
9794 (cmpsi_cc_c_insn): Likewise.
9795 (movsi_ne): Likewise.
9796 (movsi_cond_exec): Likewise.
9797 (clrsbsi2): Likewise.
9808 (call_i): Remove 'S' asm letter, add jli instruction.
9809 (call_value_i): Likewise.
9810 * config/arc/arc.op (mjli-always): New option.
9811 * config/arc/constraints.md (Cji): New constraint.
9812 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9814 (subsf3_fpx): Likewise.
9815 (mulsf3_fpx): Likewise.
9816 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9818 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9820 * doc/invoke.texi (ARC): Document mjli-always option.
9822 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9824 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9825 avoid addition with 0 and use incw and decw where possible.
9827 2018-01-26 Richard Biener <rguenther@suse.de>
9829 PR tree-optimization/81082
9830 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9831 association if it requires casting to unsigned.
9832 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9833 from fold_plusminus_mult_expr to catch important cases late when
9834 range info is available.
9836 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9838 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9839 * configure.ac (hidden_linkonce): New test.
9840 * configure: Regenerate.
9841 * config.in: Regenerate.
9843 2018-01-26 Julia Koval <julia.koval@intel.com>
9845 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9846 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9847 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9848 _mm_mask_bitshuffle_epi64_mask): Fix type.
9849 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9850 USI_FTYPE_V4DI_V4DI_USI): Remove.
9851 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9852 __builtin_ia32_vpshufbitqmb256_mask,
9853 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9854 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9855 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9857 2018-01-26 Alan Modra <amodra@gmail.com>
9860 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9861 UNSPEC_VBPERMQ. Sort other unspecs.
9863 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9865 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9867 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9870 * predict.c (drop_profile): Do not push/pop cfun; update also
9872 (handle_missing_profiles): Fix logic looking for zero profiles.
9874 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9877 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9878 on functions with #pragma omp declare simd or functions with simd
9880 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9881 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9882 Remove trailing \n from warning_at calls.
9884 2018-01-25 Tom de Vries <tom@codesourcery.com>
9887 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9888 for neutered workers.
9890 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9893 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9894 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9896 2018-01-24 Jeff Law <law@redhat.com>
9899 * i386.c (get_probe_interval): Move to earlier point.
9900 (ix86_compute_frame_layout): If -fstack-clash-protection and
9901 the frame is larger than the probe interval, then use pushes
9902 to save registers rather than reg->mem moves.
9903 (ix86_expand_prologue): Remove conditional for int_registers_saved
9906 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9909 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9910 min/max for never referenced object.
9912 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9915 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9917 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9918 attributes from DECL_ATTRIBUTES (decl) without affecting
9919 DECL_ATTRIBUTES (current_function_decl).
9920 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9921 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9923 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9925 PR tree-optimization/83979
9926 * fold-const.c (fold_comparison): Use constant_boolean_node
9927 instead of boolean_{true,false}_node.
9929 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9931 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9934 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9936 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9937 Simplify the clause that sets the length attribute.
9938 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9939 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9940 clause that sets the length attribute.
9941 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9943 2018-01-24 Tom de Vries <tom@codesourcery.com>
9946 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9947 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9948 Add strict parameter.
9949 (prevent_branch_around_nothing): Insert dummy insn between branch to
9950 label and label with no ptx insn inbetween.
9951 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9953 2018-01-24 Tom de Vries <tom@codesourcery.com>
9956 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9957 for neutered threads in warp.
9958 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9960 2018-01-24 Richard Biener <rguenther@suse.de>
9962 PR tree-optimization/83176
9963 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9966 2018-01-24 Richard Biener <rguenther@suse.de>
9968 PR tree-optimization/82819
9969 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9970 code generating pluses that are no-ops in the target precision.
9972 2018-01-24 Richard Biener <rguenther@suse.de>
9975 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9977 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9979 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9980 to merge probabilities.
9981 * predict.c (probably_never_executed): Also mark as cold functions
9982 with global 0 profile and guessed local profile.
9983 * profile-count.c (profile_probability::combine_with_count): New
9985 * profile-count.h (profile_probability::operator*,
9986 profile_probability::operator*=, profile_probability::operator/,
9987 profile_probability::operator/=): Reduce precision to adjusted
9988 and set value to guessed on contradictory divisions.
9989 (profile_probability::combine_with_freq): Remove.
9990 (profile_probability::combine_wiht_count): Declare.
9991 (profile_count::force_nonzero):: Set to adjusted.
9992 (profile_count::probability_in):: Set quality to adjusted.
9993 * tree-ssa-tail-merge.c (replace_block_by): Use
9996 2018-01-23 Andrew Waterman <andrew@sifive.com>
9997 Jim Wilson <jimw@sifive.com>
9999 * config/riscv/riscv.c (riscv_stack_boundary): New.
10000 (riscv_option_override): Set riscv_stack_boundary. Handle
10001 riscv_preferred_stack_boundary_arg.
10002 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
10003 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
10004 (STACK_BOUNDARY): Set to riscv_stack_boundary.
10005 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
10006 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
10007 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
10009 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
10012 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
10013 of struct ix86_frame.
10014 (ix86_expand_epilogue): Likewise. Add a local variable for
10015 the reg_save_offset field in struct ix86_frame.
10017 2018-01-23 Bin Cheng <bin.cheng@arm.com>
10019 PR tree-optimization/82604
10020 * tree-loop-distribution.c (enum partition_kind): New enum item
10021 PKIND_PARTIAL_MEMSET.
10022 (partition_builtin_p): Support above new enum item.
10023 (generate_code_for_partition): Ditto.
10024 (compute_access_range): Differentiate cases that equality can be
10025 proven at all loops, the innermost loops or no loops.
10026 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
10027 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
10028 (finalize_partitions, distribute_loop): Don't fuse partition of
10029 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
10030 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
10031 parloop is enabled.
10033 2018-01-23 Martin Liska <mliska@suse.cz>
10035 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
10036 order to ignore the predictor.
10037 (PRED_POLYMORPHIC_CALL): Likewise.
10038 (PRED_RECURSIVE_CALL): Likewise.
10040 2018-01-23 Martin Liska <mliska@suse.cz>
10042 * tree-profile.c (tree_profiling): Print function header to
10043 aware reader which function we are working on.
10044 * value-prof.c (gimple_find_values_to_profile): Do not print
10045 not interesting value histograms.
10047 2018-01-23 Martin Liska <mliska@suse.cz>
10049 * profile-count.h (enum profile_quality): Add
10050 profile_uninitialized as the first value. Do not number values
10051 as they are zero based.
10052 (profile_count::verify): Update sanity check.
10053 (profile_probability::verify): Likewise.
10055 2018-01-23 Nathan Sidwell <nathan@acm.org>
10057 * doc/invoke.texi (ffor-scope): Deprecate.
10059 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10061 PR tree-optimization/83510
10062 * domwalk.c (set_all_edges_as_executable): New function.
10063 (dom_walker::dom_walker): Convert bool param
10064 "skip_unreachable_blocks" to enum reachability. Move setup of
10065 edge flags to set_all_edges_as_executable and only do it when
10066 reachability is REACHABLE_BLOCKS.
10067 * domwalk.h (enum dom_walker::reachability): New enum.
10068 (dom_walker::dom_walker): Convert bool param
10069 "skip_unreachable_blocks" to enum reachability.
10070 (set_all_edges_as_executable): New decl.
10071 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10072 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10074 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10075 but converting true to REACHABLE_BLOCKS.
10076 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10078 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10079 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10080 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10082 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10083 if check_all_array_refs will be called.
10085 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10087 * tree.c (selftest::test_location_wrappers): Add more test
10090 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10092 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10093 (selftest::test_bit_in_range): Likewise.
10095 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10098 * doc/sourcebuild.texi (vect_float): Say that the selector
10099 only describes the situation when -funsafe-math-optimizations is on.
10100 (vect_float_strict): Document.
10102 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10104 PR tree-optimization/83965
10105 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10106 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10107 instead of checking only for a reduction.
10108 (vect_recog_widen_sum_pattern): Likewise.
10110 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10112 * predict.c (probably_never_executed): Only use precise profile info.
10113 (compute_function_frequency): Skip after inlining hack since we now
10114 have quality checking.
10116 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10118 * profile-count.h (profile_probability::very_unlikely,
10119 profile_probability::unlikely, profile_probability::even): Set
10120 precision to guessed.
10122 2018-01-23 Richard Biener <rguenther@suse.de>
10124 PR tree-optimization/83963
10125 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10126 Properly terminate dominator walk when crossing the exit edge not
10127 when visiting its source block.
10129 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10132 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10133 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10135 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10137 PR tree-optimization/83957
10138 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10139 semicolon after for body surrounded by braces.
10141 PR tree-optimization/83081
10142 * profile-count.h (profile_probability::split): New method.
10143 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10144 Use profile_probability::split.
10145 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10146 when splitting a single conditional jump into 2.
10148 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10150 PR tree-optimization/69452
10151 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10154 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10156 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10157 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10158 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10160 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10162 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10163 * config/rl78/rl78.md (movdi): New define_expand.
10164 * config/rl78/rl78.c (rl78_split_movdi): New function.
10166 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10169 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10171 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10172 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10173 128-bit to produce an UNSPEC move to get the double word with the
10174 signbit and then a shift directly to do signbit.
10175 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10176 implementation with a new version that just does either a direct
10177 move or a regular move. Move memory interface to separate insns.
10178 Move insns so they are next to the expander.
10179 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10180 with signbit move. Split big and little endian case.
10181 (signbit<mode>2_dm_mem_le): Likewise.
10182 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10183 (signbit<mode>2_dm2): Likewise.
10185 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10187 * config/rl78/rl78.md (anddi3): New define_expand.
10189 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10191 * config/rl78/rl78.md (umindi3): New define_expand.
10193 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10195 * config/rl78/rl78.md (smindi3): New define_expand.
10197 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10199 * config/rl78/rl78.md (smaxdi3): New define_expand.
10201 2018-01-22 Carl Love <cel@us.ibm.com>
10203 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10204 LVX_V1TI): Add macro expansion.
10205 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10206 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10207 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10208 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10209 Change check to determine if the instruction is a byte reversing
10210 entry. Fix typo in comment.
10211 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10212 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10213 Add def_builtin calls for new builtins.
10214 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10215 Add define_insn expansion.
10217 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10219 * config/rl78/rl78.md (umaxdi3): New define_expand.
10221 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10223 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10224 for non-QImode registers.
10226 2018-01-22 Richard Biener <rguenther@suse.de>
10228 PR tree-optimization/83963
10229 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10230 including the loop exit block.
10231 (scop_detection::merge_sese): Likewise.
10232 (scop_detection::add_scop): Do it here instead.
10234 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10236 * doc/sourcebuild.texi (arm_softfloat): Document.
10238 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10241 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10242 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10243 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10245 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10246 David Edelsohn <dje.gcc@gmail.com>
10249 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10250 Change "crset eq" to "crset 2".
10251 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10252 (*call_indirect_aix<mode>_nospec): Likewise.
10253 (*call_value_indirect_aix<mode>_nospec): Likewise.
10254 (*call_indirect_elfv2<mode>_nospec): Likewise.
10255 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10256 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10257 change assembly output from . to $.
10258 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10259 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10260 (*tablejump<mode>_internal1_nospec): Likewise.
10262 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10265 * config/sh/sh_optimize_sett_clrt.cc:
10266 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10268 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10270 PR tree-optimization/83940
10271 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10272 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10273 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10274 use it to pass back the definition type.
10275 (vect_check_store_rhs): Likewise rhs_dt_out.
10276 (vect_build_gather_load_calls): Add a mask_dt argument and use
10277 it instead of a call to vect_is_simple_use.
10278 (vectorizable_store): Update calls to vect_check_load_store_mask
10279 and vect_check_store_rhs. Use the dt returned by the latter instead
10280 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10281 instead of calls to vect_is_simple_use. Pass the scalar rather
10282 than the vector operand to vect_is_simple_use when handling
10283 second and subsequent copies of an rhs value.
10284 (vectorizable_load): Update calls to vect_check_load_store_mask
10285 and vect_build_gather_load_calls. Use the cached mask_dt and
10286 gs_info.offset_dt instead of calls to vect_is_simple_use.
10288 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10290 PR middle-end/83945
10291 * tree-emutls.c: Include gimplify.h.
10292 (lower_emutls_2): New function.
10293 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10294 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10295 it before further processing.
10298 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10299 UINTVAL (trueop1) instead of INTVAL (op1).
10301 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10305 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10306 INCOMING_FRAME_SP_OFFSET if not defined.
10307 (scan_trace): Add ENTRY argument. If true and
10308 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10309 emit a note to adjust the CFA offset.
10310 (create_cfi_notes): Adjust scan_trace callers.
10311 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10312 INCOMING_FRAME_SP_OFFSET in the CIE.
10313 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10314 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10316 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10317 * doc/tm.texi: Regenerated.
10319 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10321 PR rtl-optimization/83147
10322 * lra-constraints.c (remove_inheritance_pseudos): Use
10323 lra_substitute_pseudo_within_insn.
10325 2018-01-19 Tom de Vries <tom@codesourcery.com>
10326 Cesar Philippidis <cesar@codesourcery.com>
10329 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10331 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10334 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10335 spaces for function labels.
10337 2018-01-19 Martin Liska <mliska@suse.cz>
10339 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10340 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10341 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10342 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10343 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10344 (PRED_CONST_RETURN): Change from 69 to 65.
10345 (PRED_NULL_RETURN): Change from 91 to 71.
10346 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10347 (PRED_LOOP_GUARD): Change from 66 to 73.
10349 2018-01-19 Martin Liska <mliska@suse.cz>
10351 * predict.c (predict_insn_def): Add new assert.
10352 (struct branch_predictor): Change type to signed integer.
10353 (test_prediction_value_range): Amend test to cover
10354 PROB_UNINITIALIZED.
10355 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10356 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10357 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10358 (PRED_LOOP_IV_COMPARE): Likewise.
10359 * predict.h (PROB_UNINITIALIZED): Define new constant.
10361 2018-01-19 Martin Liska <mliska@suse.cz>
10363 * predict.c (dump_prediction): Add new format for
10364 analyze_brprob.py script which is enabled with -details
10366 * profile-count.h (precise_p): New function.
10368 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10370 PR tree-optimization/83922
10371 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10372 there are no statements that need masking.
10373 (vect_active_double_reduction_p): New function.
10374 (vect_analyze_loop_operations): Use it when handling phis that
10375 are not in the loop header.
10377 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10379 PR tree-optimization/83914
10380 * tree-vect-loop.c (vectorizable_induction): Don't convert
10381 init_expr or apply the peeling adjustment for inductions
10382 that are nested within the vectorized loop.
10384 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10386 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10389 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10393 * function.h (gimplify_parameters): Add gimple_seq * argument.
10394 * function.c: Include gimple.h and options.h.
10395 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10396 for the added local temporaries if needed.
10397 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10398 if there are any parameter cleanups, wrap whole body into a
10399 try/finally with the cleanups.
10401 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10404 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10405 Use GET_MODE_CLASS for scalar floating point.
10407 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10411 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10412 Fix call of call_cgraph_insertion_hooks.
10414 2018-01-18 Martin Sebor <msebor@redhat.com>
10416 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10418 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10421 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10424 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10427 * common.opt: (-ffile-prefix-map): New option.
10428 * opts.c (common_handle_option): Defer it.
10429 * opts-global.c (handle_common_deferred_options): Handle it.
10430 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10431 * file-prefix-map.h: New file.
10432 (remap_debug_filename, add_debug_prefix_map): ...here.
10433 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10434 * final.c (debug_prefix_map, add_debug_prefix_map
10435 remap_debug_filename): Move to...
10436 * file-prefix-map.c: New file.
10437 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10438 generalize, get rid of alloca(), use strrchr() instead of strchr().
10439 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10440 Implement in terms of add_prefix_map().
10441 (remap_macro_filename, remap_debug_filename): Implement in term of
10443 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10444 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10445 * dbxout.c: Include file-prefix-map.h.
10446 * varasm.c: Likewise.
10447 * vmsdbgout.c: Likewise.
10448 * xcoffout.c: Likewise.
10449 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10450 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10451 * doc/invoke.texi (-ffile-prefix-map): Document.
10452 (-fdebug-prefix-map): Update description.
10454 2018-01-18 Martin Liska <mliska@suse.cz>
10456 * config/i386/i386.c (indirect_thunk_name): Document that also
10458 (output_indirect_thunk): Document why both instructions
10459 (pause and lfence) are generated.
10461 2018-01-18 Richard Biener <rguenther@suse.de>
10463 PR tree-optimization/83887
10464 * graphite-scop-detection.c
10465 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10466 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10467 (scop_detection::merge_sese): Re-implement with a flood-fill
10468 algorithm that properly finds a SESE region if it exists.
10470 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10473 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10474 pointer_diff optimizations use view_convert instead of convert.
10476 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10478 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10479 Generate different code for -mno-speculate-indirect-jumps.
10480 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10481 (*call_indirect_aix<mode>): Disable for
10482 -mno-speculate-indirect-jumps.
10483 (*call_indirect_aix<mode>_nospec): New define_insn.
10484 (*call_value_indirect_aix<mode>): Disable for
10485 -mno-speculate-indirect-jumps.
10486 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10487 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10488 -mno-speculate-indirect-jumps.
10489 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10491 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10493 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10494 long double type, set the flags for noting the default long double
10495 type, even if we don't pass or return a long double type.
10497 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10500 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10503 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10505 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10506 support for merge[hl].
10507 (fold_mergehl_helper): New helper function.
10508 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10509 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10510 (altivec_vmrglw_direct): Add xxmrglw insn.
10512 2018-01-17 Andrew Waterman <andrew@sifive.com>
10514 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10515 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10517 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10520 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10521 call the lto_location_cache before reading the
10522 DECL_SOURCE_LOCATION of the types.
10524 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10525 Richard Sandiford <richard.sandiford@linaro.org>
10527 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10528 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10529 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10530 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10531 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10533 * config/aarch64/constraints.md (aarch64_movti_operand):
10535 * config/aarch64/predicates.md (Uti): Add new constraint.
10537 2018-01-17 Carl Love <cel@us.ibm.com>
10539 * config/rs6000/vsx.md (define_expand xl_len_r,
10540 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10541 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10543 (define_expand, define_insn): Move the shift left from the
10544 define_insn to the define_expand for lxvl and stxvl instructions.
10545 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10546 and XL_LEN_R definitions to PURE.
10548 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10550 * config/i386/i386.c (indirect_thunk_name): Declare regno
10551 as unsigned int. Compare regno with INVALID_REGNUM.
10552 (output_indirect_thunk): Ditto.
10553 (output_indirect_thunk_function): Ditto.
10554 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10555 in the call to output_indirect_thunk_function.
10557 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10559 PR middle-end/83884
10560 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10561 rather than the size of inner_type to determine the stack slot size
10562 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10564 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10567 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10570 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10572 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10573 endian Linux systems to optionally enable multilibs for selecting
10574 the long double type if the user configured an explicit type.
10575 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10576 have no long double multilibs if not defined.
10577 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10578 warn if the user used -mabi={ieee,ibm}longdouble and we built
10579 multilibs for long double.
10580 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10581 appropriate multilib option.
10582 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10584 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10585 for building long double multilibs.
10586 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10588 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10590 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10593 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10595 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10598 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10601 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10604 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10606 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10607 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10609 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10611 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10612 different rtl trees depending on TARGET_64BIT.
10613 (rs6000_gen_lvx): Likewise.
10615 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10617 * config/visium/visium.md (nop): Tweak comment.
10618 (hazard_nop): Likewise.
10620 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10622 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10623 -mspeculate-indirect-jumps.
10624 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10625 for -mno-speculate-indirect-jumps.
10626 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10627 (*call_value_indirect_elfv2<mode>): Disable for
10628 -mno-speculate-indirect-jumps.
10629 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10630 (indirect_jump): Emit different RTL for
10631 -mno-speculate-indirect-jumps.
10632 (*indirect_jump<mode>): Disable for
10633 -mno-speculate-indirect-jumps.
10634 (*indirect_jump<mode>_nospec): New define_insn.
10635 (tablejump): Emit different RTL for
10636 -mno-speculate-indirect-jumps.
10637 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10638 (tablejumpsi_nospec): New define_expand.
10639 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10640 (tablejumpdi_nospec): New define_expand.
10641 (*tablejump<mode>_internal1): Disable for
10642 -mno-speculate-indirect-jumps.
10643 (*tablejump<mode>_internal1_nospec): New define_insn.
10644 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10647 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10649 * caller-save.c (insert_save): Drop unnecessary parameter. All
10652 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10653 Richard Biener <rguenth@suse.de>
10656 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10657 return early, inline manually is_gimple_sizepos. Make sure if we
10658 call gimplify_expr we don't end up with a gimple constant.
10659 * tree.c (variably_modified_type_p): Don't return true for
10660 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10661 * gimplify.h (is_gimple_sizepos): Remove.
10663 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10665 PR tree-optimization/83857
10666 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10667 vectorizable_live_operation for pure SLP statements.
10668 (vectorizable_live_operation): Handle PHIs.
10670 2018-01-16 Richard Biener <rguenther@suse.de>
10672 PR tree-optimization/83867
10673 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10674 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10676 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10679 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10680 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10681 If off is not INTEGER_CST, issue a may not be aligned warning
10682 rather than isn't aligned. Use isn%'t rather than isn't.
10683 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10685 <case MULT_EXPR>: Improve the case when bottom and one of the
10686 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10687 operand, in that case check if the other operand is multiple of
10688 bottom divided by the INTEGER_CST operand.
10690 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10693 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10694 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10695 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10696 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10697 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10698 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10699 (pa_function_arg_size): New function.
10701 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10703 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10704 in a separate statement.
10706 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10708 PR tree-optimization/83847
10709 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10710 group gathers and scatters.
10712 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10714 PR rtl-optimization/86620
10715 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10717 PR rtl-optimization/83213
10718 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10719 to last if both are JUMP_INSNs.
10721 PR tree-optimization/83843
10722 * gimple-ssa-store-merging.c
10723 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10724 store_immediate_info for bswap/nop orig_stores.
10726 2018-01-15 Andrew Waterman <andrew@sifive.com>
10728 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10730 <UDIV>: Increase cost if !TARGET_DIV.
10732 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10734 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10735 (define_attr "cr_logical_3op"): New.
10736 (cceq_ior_compare): Adjust.
10737 (cceq_ior_compare_complement): Adjust.
10738 (*cceq_rev_compare): Adjust.
10739 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10740 (is_cracked_insn): Adjust.
10741 (insn_must_be_first_in_group): Adjust.
10742 * config/rs6000/40x.md: Adjust.
10743 * config/rs6000/440.md: Adjust.
10744 * config/rs6000/476.md: Adjust.
10745 * config/rs6000/601.md: Adjust.
10746 * config/rs6000/603.md: Adjust.
10747 * config/rs6000/6xx.md: Adjust.
10748 * config/rs6000/7450.md: Adjust.
10749 * config/rs6000/7xx.md: Adjust.
10750 * config/rs6000/8540.md: Adjust.
10751 * config/rs6000/cell.md: Adjust.
10752 * config/rs6000/e300c2c3.md: Adjust.
10753 * config/rs6000/e500mc.md: Adjust.
10754 * config/rs6000/e500mc64.md: Adjust.
10755 * config/rs6000/e5500.md: Adjust.
10756 * config/rs6000/e6500.md: Adjust.
10757 * config/rs6000/mpc.md: Adjust.
10758 * config/rs6000/power4.md: Adjust.
10759 * config/rs6000/power5.md: Adjust.
10760 * config/rs6000/power6.md: Adjust.
10761 * config/rs6000/power7.md: Adjust.
10762 * config/rs6000/power8.md: Adjust.
10763 * config/rs6000/power9.md: Adjust.
10764 * config/rs6000/rs64.md: Adjust.
10765 * config/rs6000/titan.md: Adjust.
10767 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10769 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10770 ix86_indirect_branch_register logic.
10772 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10774 * config/i386/constraints.md (Bs): Update
10775 ix86_indirect_branch_register check. Don't check
10776 ix86_indirect_branch_register with GOT_memory_operand.
10778 * config/i386/predicates.md (GOT_memory_operand): Don't check
10779 ix86_indirect_branch_register here.
10780 (GOT32_symbol_operand): Likewise.
10782 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10784 * config/i386/predicates.md (constant_call_address_operand):
10785 Rewrite ix86_indirect_branch_register logic.
10786 (sibcall_insn_operand): Likewise.
10788 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10790 * config/i386/constraints.md (Bs): Replace
10791 ix86_indirect_branch_thunk_register with
10792 ix86_indirect_branch_register.
10794 * config/i386/i386.md (indirect_jump): Likewise.
10795 (tablejump): Likewise.
10796 (*sibcall_memory): Likewise.
10797 (*sibcall_value_memory): Likewise.
10798 Peepholes of indirect call and jump via memory: Likewise.
10799 * config/i386/i386.opt: Likewise.
10800 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10801 (GOT_memory_operand): Likewise.
10802 (call_insn_operand): Likewise.
10803 (sibcall_insn_operand): Likewise.
10804 (GOT32_symbol_operand): Likewise.
10806 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10808 PR middle-end/83837
10809 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10810 type rather than type addr's type points to.
10811 (expand_omp_atomic_mutex): Likewise.
10812 (expand_omp_atomic): Likewise.
10814 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10817 * config/i386/i386.c (output_indirect_thunk_function): Use
10818 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10819 for __x86_return_thunk.
10821 2018-01-15 Richard Biener <rguenther@suse.de>
10823 PR middle-end/83850
10824 * expmed.c (extract_bit_field_1): Fix typo.
10826 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10829 * config/arm/iterators.md (VF): New mode iterator.
10830 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10831 Remove integer-related logic from pattern.
10832 (neon_vabd<mode>_3): Likewise.
10834 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10836 PR middle-end/82694
10837 * common.opt (fstrict-overflow): No longer an alias.
10838 (fwrapv-pointer): New option.
10839 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10840 also for pointer types based on flag_wrapv_pointer.
10841 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10842 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10843 opts->x_flag_wrapv got set.
10844 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10845 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10846 POINTER_TYPE_OVERFLOW_UNDEFINED.
10847 * match.pd: Likewise in address comparison pattern.
10848 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10850 2018-01-15 Richard Biener <rguenther@suse.de>
10853 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10854 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10855 Reset type names to their identifier if their TYPE_DECL doesn't
10856 have linkage (and thus is used for ODR and devirt).
10857 (save_debug_info_for_decl): Remove.
10858 (save_debug_info_for_type): Likewise.
10859 (add_tree_to_fld_list): Adjust.
10860 * tree-pretty-print.c (dump_generic_node): Make dumping of
10861 type names more robust.
10863 2018-01-15 Richard Biener <rguenther@suse.de>
10865 * BASE-VER: Bump to 8.0.1.
10867 2018-01-14 Martin Sebor <msebor@redhat.com>
10870 * builtins.c (check_access): Avoid warning when the no-warning bit
10873 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10875 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10876 * ira-color (allocno_hard_regs_compare): Likewise.
10878 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10881 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10882 Use .pushsection/.popsection.
10884 2018-01-14 Martin Sebor <msebor@redhat.com>
10887 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10889 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10891 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10892 entry from extra_headers.
10893 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10894 extra_headers, make the list bitwise identical to the i?86-*-* one.
10896 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10898 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10899 -mcmodel=large with -mindirect-branch=thunk,
10900 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10901 -mfunction-return=thunk-extern.
10902 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10903 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10904 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10906 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10908 * config/i386/i386.c (print_reg): Print the name of the full
10909 integer register without '%'.
10910 (ix86_print_operand): Handle 'V'.
10911 * doc/extend.texi: Document 'V' modifier.
10913 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10915 * config/i386/constraints.md (Bs): Disallow memory operand for
10916 -mindirect-branch-register.
10918 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10919 (GOT_memory_operand): Likewise.
10920 (call_insn_operand): Likewise.
10921 (sibcall_insn_operand): Likewise.
10922 (GOT32_symbol_operand): Likewise.
10923 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10924 for -mindirect-branch-register.
10925 (tablejump): Likewise.
10926 (*sibcall_memory): Likewise.
10927 (*sibcall_value_memory): Likewise.
10928 Disallow peepholes of indirect call and jump via memory for
10929 -mindirect-branch-register.
10930 (*call_pop): Replace m with Bw.
10931 (*call_value_pop): Likewise.
10932 (*sibcall_pop_memory): Replace m with Bs.
10933 * config/i386/i386.opt (mindirect-branch-register): New option.
10934 * doc/invoke.texi: Document -mindirect-branch-register option.
10936 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10938 * config/i386/i386-protos.h (ix86_output_function_return): New.
10939 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10940 set function_return_type.
10941 (indirect_thunk_name): Add ret_p to indicate thunk for function
10943 (output_indirect_thunk_function): Pass false to
10944 indirect_thunk_name.
10945 (ix86_output_indirect_branch_via_reg): Likewise.
10946 (ix86_output_indirect_branch_via_push): Likewise.
10947 (output_indirect_thunk_function): Create alias for function
10948 return thunk if regno < 0.
10949 (ix86_output_function_return): New function.
10950 (ix86_handle_fndecl_attribute): Handle function_return.
10951 (ix86_attribute_table): Add function_return.
10952 * config/i386/i386.h (machine_function): Add
10953 function_return_type.
10954 * config/i386/i386.md (simple_return_internal): Use
10955 ix86_output_function_return.
10956 (simple_return_internal_long): Likewise.
10957 * config/i386/i386.opt (mfunction-return=): New option.
10958 (indirect_branch): Mention -mfunction-return=.
10959 * doc/extend.texi: Document function_return function attribute.
10960 * doc/invoke.texi: Document -mfunction-return= option.
10962 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10964 * config/i386/i386-opts.h (indirect_branch): New.
10965 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10966 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10967 with local indirect jump when converting indirect call and jump.
10968 (ix86_set_indirect_branch_type): New.
10969 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10970 (indirectlabelno): New.
10971 (indirect_thunk_needed): Likewise.
10972 (indirect_thunk_bnd_needed): Likewise.
10973 (indirect_thunks_used): Likewise.
10974 (indirect_thunks_bnd_used): Likewise.
10975 (INDIRECT_LABEL): Likewise.
10976 (indirect_thunk_name): Likewise.
10977 (output_indirect_thunk): Likewise.
10978 (output_indirect_thunk_function): Likewise.
10979 (ix86_output_indirect_branch_via_reg): Likewise.
10980 (ix86_output_indirect_branch_via_push): Likewise.
10981 (ix86_output_indirect_branch): Likewise.
10982 (ix86_output_indirect_jmp): Likewise.
10983 (ix86_code_end): Call output_indirect_thunk_function if needed.
10984 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10986 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10987 (ix86_attribute_table): Add indirect_branch.
10988 * config/i386/i386.h (machine_function): Add indirect_branch_type
10989 and has_local_indirect_jump.
10990 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10992 (tablejump): Likewise.
10993 (*indirect_jump): Use ix86_output_indirect_jmp.
10994 (*tablejump_1): Likewise.
10995 (simple_return_indirect_internal): Likewise.
10996 * config/i386/i386.opt (mindirect-branch=): New option.
10997 (indirect_branch): New.
11000 (thunk-inline): Likewise.
11001 (thunk-extern): Likewise.
11002 * doc/extend.texi: Document indirect_branch function attribute.
11003 * doc/invoke.texi: Document -mindirect-branch= option.
11005 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
11008 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
11010 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
11012 * ipa-inline.c (want_inline_small_function_p): Return false if
11013 inlining has already failed with CIF_FINAL_ERROR.
11014 (update_caller_keys): Call want_inline_small_function_p before
11016 (update_callee_keys): Likewise.
11018 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
11020 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
11022 (rs6000_quadword_masked_address_p): Likewise.
11023 (quad_aligned_load_p): Likewise.
11024 (quad_aligned_store_p): Likewise.
11025 (const_load_sequence_p): Add comment to describe the outer-most loop.
11026 (mimic_memory_attributes_and_flags): New function.
11027 (rs6000_gen_stvx): Likewise.
11028 (replace_swapped_aligned_store): Likewise.
11029 (rs6000_gen_lvx): Likewise.
11030 (replace_swapped_aligned_load): Likewise.
11031 (replace_swapped_load_constant): Capitalize argument name in
11032 comment describing this function.
11033 (rs6000_analyze_swaps): Add a third pass to search for vector loads
11034 and stores that access quad-word aligned addresses and replace
11035 with stvx or lvx instructions when appropriate.
11036 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
11037 New function prototype.
11038 (rs6000_quadword_masked_address_p): Likewise.
11039 (rs6000_gen_lvx): Likewise.
11040 (rs6000_gen_stvx): Likewise.
11041 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
11042 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
11043 when memory address is aligned.
11044 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
11045 this split to select lvx instruction when memory address is aligned.
11046 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
11047 instruction when memory address is aligned.
11048 (*vsx_le_perm_load_v16qi): Likewise.
11049 (four unnamed splitters): Modify to select the stvx instruction
11050 when memory is aligned.
11052 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11054 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11055 which appears in the queue multiple times.
11057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11058 Alan Hayward <alan.hayward@arm.com>
11059 David Sherwood <david.sherwood@arm.com>
11061 * tree-vectorizer.h (vec_lower_bound): New structure.
11062 (_loop_vec_info): Add check_nonzero and lower_bounds.
11063 (LOOP_VINFO_CHECK_NONZERO): New macro.
11064 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11065 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11066 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11067 fields. Make seg_len the distance travelled, not including the
11069 (dr_direction_indicator): Declare.
11070 (dr_zero_step_indicator): Likewise.
11071 (dr_known_forward_stride_p): Likewise.
11072 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11074 (runtime_alias_check_p): Allow runtime alias checks with
11076 (operator ==): Compare access_size and align.
11077 (prune_runtime_alias_test_list): Rework for new distinction between
11078 the access_size and seg_len.
11079 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11081 (get_segment_min_max): New function.
11082 (create_intersect_range_checks): Use it.
11083 (dr_step_indicator): New function.
11084 (dr_direction_indicator): Likewise.
11085 (dr_zero_step_indicator): Likewise.
11086 (dr_known_forward_stride_p): Likewise.
11087 * tree-loop-distribution.c (data_ref_segment_size): Return
11088 DR_STEP * (niters - 1).
11089 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11091 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11092 (vect_preserves_scalar_order_p): New function, split out from...
11093 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11094 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11095 (vect_vfa_access_size): New function.
11096 (vect_vfa_align): Likewise.
11097 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11098 (dump_lower_bound): New function.
11099 (vect_check_lower_bound): Likewise.
11100 (vect_small_gap_p): Likewise.
11101 (vectorizable_with_step_bound_p): Likewise.
11102 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11103 depencies if the vectorization factor is 1. Convert the checks
11104 for nonzero steps into checks on the bounds of DR_STEP. Try using
11105 a bunds check for variable steps if the minimum required step is
11106 relatively small. Update calls to the dr_with_seg_len
11107 constructor and to vect_compile_time_alias.
11108 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11110 (vect_loop_versioning): Call it.
11111 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11113 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11116 Alan Hayward <alan.hayward@arm.com>
11117 David Sherwood <david.sherwood@arm.com>
11119 * doc/sourcebuild.texi (vect_scatter_store): Document.
11120 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11122 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11124 * genopinit.c (main): Add supports_vec_scatter_store and
11125 supports_vec_scatter_store_cached to target_optabs.
11126 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11127 IFN_MASK_SCATTER_STORE.
11128 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11130 * internal-fn.h (internal_store_fn_p): Declare.
11131 (internal_fn_stored_value_index): Likewise.
11132 * internal-fn.c (scatter_store_direct): New macro.
11133 (expand_scatter_store_optab_fn): New function.
11134 (direct_scatter_store_optab_supported_p): New macro.
11135 (internal_store_fn_p): New function.
11136 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11137 IFN_MASK_SCATTER_STORE.
11138 (internal_fn_mask_index): Likewise.
11139 (internal_fn_stored_value_index): New function.
11140 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11141 for scatter stores.
11142 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11143 * optabs-query.c (supports_vec_scatter_store_p): New function.
11144 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11145 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11146 true for scatter stores.
11147 (vect_gather_scatter_fn_p): Handle scatter stores too.
11148 (vect_check_gather_scatter): Consider using scatter stores if
11149 supports_vec_scatter_store_p.
11150 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11151 scatter stores too.
11152 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11153 internal_fn_stored_value_index.
11154 (check_load_store_masking): Handle scatter stores too.
11155 (vect_get_store_rhs): Make public.
11156 (vectorizable_call): Use internal_store_fn_p.
11157 (vectorizable_store): Handle scatter store internal functions.
11158 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11159 when deciding whether the end of the group has been reached.
11160 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11161 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11162 (mask_scatter_store<mode>): New insns.
11164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11165 Alan Hayward <alan.hayward@arm.com>
11166 David Sherwood <david.sherwood@arm.com>
11168 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11169 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11170 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11172 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11173 Use vect_truncate_gather_scatter_offset if we can't treat the
11174 operation as a normal gather load or scatter store.
11175 (get_group_load_store_type): Take the gather_scatter_info
11176 as argument. Try using a gather load or scatter store for
11177 single-element groups.
11178 (get_load_store_type): Update calls to get_group_load_store_type
11179 and vect_use_strided_gather_scatters_p.
11181 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11182 Alan Hayward <alan.hayward@arm.com>
11183 David Sherwood <david.sherwood@arm.com>
11185 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11186 optional tree argument.
11187 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11189 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11190 but continue to use the current value as a fallback.
11191 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11192 to compare the updates.
11193 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11194 (get_load_store_type): Use it when handling a strided access.
11195 (vect_get_strided_load_store_ops): New function.
11196 (vect_get_data_ptr_increment): Likewise.
11197 (vectorizable_load): Handle strided gather loads. Always pass
11198 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11200 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11201 Alan Hayward <alan.hayward@arm.com>
11202 David Sherwood <david.sherwood@arm.com>
11204 * doc/md.texi (gather_load@var{m}): Document.
11205 (mask_gather_load@var{m}): Likewise.
11206 * genopinit.c (main): Add supports_vec_gather_load and
11207 supports_vec_gather_load_cached to target_optabs.
11208 * optabs-tree.c (init_tree_optimization_optabs): Use
11209 ggc_cleared_alloc to allocate target_optabs.
11210 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11211 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11213 * internal-fn.h (internal_load_fn_p): Declare.
11214 (internal_gather_scatter_fn_p): Likewise.
11215 (internal_fn_mask_index): Likewise.
11216 (internal_gather_scatter_fn_supported_p): Likewise.
11217 * internal-fn.c (gather_load_direct): New macro.
11218 (expand_gather_load_optab_fn): New function.
11219 (direct_gather_load_optab_supported_p): New macro.
11220 (direct_internal_fn_optab): New function.
11221 (internal_load_fn_p): Likewise.
11222 (internal_gather_scatter_fn_p): Likewise.
11223 (internal_fn_mask_index): Likewise.
11224 (internal_gather_scatter_fn_supported_p): Likewise.
11225 * optabs-query.c (supports_at_least_one_mode_p): New function.
11226 (supports_vec_gather_load_p): Likewise.
11227 * optabs-query.h (supports_vec_gather_load_p): Declare.
11228 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11229 and memory_type field.
11230 (NUM_PATTERNS): Bump to 15.
11231 * tree-vect-data-refs.c: Include internal-fn.h.
11232 (vect_gather_scatter_fn_p): New function.
11233 (vect_describe_gather_scatter_call): Likewise.
11234 (vect_check_gather_scatter): Try using internal functions for
11235 gather loads. Recognize existing calls to a gather load function.
11236 (vect_analyze_data_refs): Consider using gather loads if
11237 supports_vec_gather_load_p.
11238 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11239 (vect_get_gather_scatter_offset_type): Likewise.
11240 (vect_convert_mask_for_vectype): Likewise.
11241 (vect_add_conversion_to_patterm): Likewise.
11242 (vect_try_gather_scatter_pattern): Likewise.
11243 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11244 (vect_vect_recog_func_ptrs): Add it.
11245 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11246 internal_fn_mask_index and internal_gather_scatter_fn_p.
11247 (check_load_store_masking): Take the gather_scatter_info as an
11248 argument and handle gather loads.
11249 (vect_get_gather_scatter_ops): New function.
11250 (vectorizable_call): Check internal_load_fn_p.
11251 (vectorizable_load): Likewise. Handle gather load internal
11253 (vectorizable_store): Update call to check_load_store_masking.
11254 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11255 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11256 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11257 (aarch64_gather_scale_operand_d): New predicates.
11258 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11259 (mask_gather_load<mode>): New insns.
11261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11262 Alan Hayward <alan.hayward@arm.com>
11263 David Sherwood <david.sherwood@arm.com>
11265 * optabs.def (fold_left_plus_optab): New optab.
11266 * doc/md.texi (fold_left_plus_@var{m}): Document.
11267 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11268 * internal-fn.c (fold_left_direct): Define.
11269 (expand_fold_left_optab_fn): Likewise.
11270 (direct_fold_left_optab_supported_p): Likewise.
11271 * fold-const-call.c (fold_const_fold_left): New function.
11272 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11273 * tree-parloops.c (valid_reduction_p): New function.
11274 (gather_scalar_reductions): Use it.
11275 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11276 (vect_finish_replace_stmt): Declare.
11277 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11278 (needs_fold_left_reduction_p): New function, split out from...
11279 (vect_is_simple_reduction): ...here. Accept reductions that
11280 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11281 (vect_force_simple_reduction): Also store the reduction type in
11282 the assignment's STMT_VINFO_REDUC_TYPE.
11283 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11284 (merge_with_identity): New function.
11285 (vect_expand_fold_left): Likewise.
11286 (vectorize_fold_left_reduction): Likewise.
11287 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11288 scalar phi in place for it. Check for target support and reject
11289 cases that would reassociate the operation. Defer the transform
11290 phase to vectorize_fold_left_reduction.
11291 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11292 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11293 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11297 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11298 call to ifc_temp_var.
11300 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11301 Alan Hayward <alan.hayward@arm.com>
11302 David Sherwood <david.sherwood@arm.com>
11304 * target.def (legitimize_address_displacement): Take the original
11305 offset as a poly_int.
11306 * targhooks.h (default_legitimize_address_displacement): Update
11308 * targhooks.c (default_legitimize_address_displacement): Likewise.
11309 * doc/tm.texi: Regenerate.
11310 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11311 as an argument, moving assert of ad->disp == ad->disp_term to...
11312 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11313 Try calling targetm.legitimize_address_displacement before expanding
11314 the address rather than afterwards, and adjust for the new interface.
11315 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11316 Match the new hook interface. Handle SVE addresses.
11317 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11318 new hook interface.
11320 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11322 * Makefile.in (OBJS): Add early-remat.o.
11323 * target.def (select_early_remat_modes): New hook.
11324 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11325 * doc/tm.texi: Regenerate.
11326 * targhooks.h (default_select_early_remat_modes): Declare.
11327 * targhooks.c (default_select_early_remat_modes): New function.
11328 * timevar.def (TV_EARLY_REMAT): New timevar.
11329 * passes.def (pass_early_remat): New pass.
11330 * tree-pass.h (make_pass_early_remat): Declare.
11331 * early-remat.c: New file.
11332 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11334 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11337 Alan Hayward <alan.hayward@arm.com>
11338 David Sherwood <david.sherwood@arm.com>
11340 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11341 vfm1 with a bound_epilog parameter.
11342 (vect_do_peeling): Update calls accordingly, and move the prologue
11343 call earlier in the function. Treat the base bound_epilog as 0 for
11344 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11345 this base when peeling for gaps.
11346 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11347 with fully-masked loops.
11348 (vect_estimate_min_profitable_iters): Handle the single peeled
11349 iteration in that case.
11351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11352 Alan Hayward <alan.hayward@arm.com>
11353 David Sherwood <david.sherwood@arm.com>
11355 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11356 single-element interleaving even if the size is not a power of 2.
11357 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11358 accesses for single-element interleaving if the group size is
11361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11362 Alan Hayward <alan.hayward@arm.com>
11363 David Sherwood <david.sherwood@arm.com>
11365 * doc/md.texi (fold_extract_last_@var{m}): Document.
11366 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11367 * optabs.def (fold_extract_last_optab): New optab.
11368 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11369 * internal-fn.c (fold_extract_direct): New macro.
11370 (expand_fold_extract_optab_fn): Likewise.
11371 (direct_fold_extract_optab_supported_p): Likewise.
11372 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11373 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11374 EXTRACT_LAST_REDUCTION.
11375 (get_initial_def_for_reduction): Do not create an initial vector
11376 for EXTRACT_LAST_REDUCTION reductions.
11377 (vectorizable_reduction): Leave the scalar phi in place for
11378 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11379 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11380 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11381 transform phase to vectorizable_condition.
11382 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11384 (vect_finish_stmt_generation): ...here.
11385 (vect_finish_replace_stmt): New function.
11386 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11387 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11389 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11392 Alan Hayward <alan.hayward@arm.com>
11393 David Sherwood <david.sherwood@arm.com>
11395 * doc/md.texi (extract_last_@var{m}): Document.
11396 * optabs.def (extract_last_optab): New optab.
11397 * internal-fn.def (EXTRACT_LAST): New internal function.
11398 * internal-fn.c (cond_unary_direct): New macro.
11399 (expand_cond_unary_optab_fn): Likewise.
11400 (direct_cond_unary_optab_supported_p): Likewise.
11401 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11402 loops using EXTRACT_LAST.
11403 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11404 (extract_last_<mode>): ...this optab.
11405 (vec_extract<mode><Vel>): Update accordingly.
11407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11408 Alan Hayward <alan.hayward@arm.com>
11409 David Sherwood <david.sherwood@arm.com>
11411 * target.def (empty_mask_is_expensive): New hook.
11412 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11413 * doc/tm.texi: Regenerate.
11414 * targhooks.h (default_empty_mask_is_expensive): Declare.
11415 * targhooks.c (default_empty_mask_is_expensive): New function.
11416 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11417 if the target says that empty masks are expensive.
11418 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11420 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11423 Alan Hayward <alan.hayward@arm.com>
11424 David Sherwood <david.sherwood@arm.com>
11426 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11427 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11428 (vect_use_loop_mask_for_alignment_p): New function.
11429 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11430 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11431 niters_skip argument. Make sure that the first niters_skip elements
11432 of the first iteration are inactive.
11433 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11434 Update call to vect_set_loop_masks_directly.
11435 (get_misalign_in_elems): New function, split out from...
11436 (vect_gen_prolog_loop_niters): ...here.
11437 (vect_update_init_of_dr): Take a code argument that specifies whether
11438 the adjustment should be added or subtracted.
11439 (vect_update_init_of_drs): Likewise.
11440 (vect_prepare_for_masked_peels): New function.
11441 (vect_do_peeling): Skip prologue peeling if we're using a mask
11442 instead. Update call to vect_update_inits_of_drs.
11443 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11445 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11446 alignment. Do not include the number of peeled iterations in
11447 the minimum threshold in that case.
11448 (vectorizable_induction): Adjust the start value down by
11449 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11450 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11451 Take the number of skipped iterations into account when calculating
11453 * tree-vect-stmts.c (vect_gen_while_not): New function.
11455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11456 Alan Hayward <alan.hayward@arm.com>
11457 David Sherwood <david.sherwood@arm.com>
11459 * doc/sourcebuild.texi (vect_fully_masked): Document.
11460 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11461 default value to 0.
11462 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11464 (vect_analyze_loop_2): ...here. Don't check the vectorization
11465 factor against the number of loop iterations if the loop is
11468 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11469 Alan Hayward <alan.hayward@arm.com>
11470 David Sherwood <david.sherwood@arm.com>
11472 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11473 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11474 (dump_groups): Update accordingly.
11475 (iv_use::mem_type): New member variable.
11476 (address_p): New function.
11477 (record_use): Add a mem_type argument and initialize the new
11479 (record_group_use): Add a mem_type argument. Use address_p.
11480 Remove obsolete null checks of base_object. Update call to record_use.
11481 (find_interesting_uses_op): Update call to record_group_use.
11482 (find_interesting_uses_cond): Likewise.
11483 (find_interesting_uses_address): Likewise.
11484 (get_mem_type_for_internal_fn): New function.
11485 (find_address_like_use): Likewise.
11486 (find_interesting_uses_stmt): Try find_address_like_use before
11487 calling find_interesting_uses_op.
11488 (addr_offset_valid_p): Use the iv mem_type field as the type
11489 of the addressed memory.
11490 (add_autoinc_candidates): Likewise.
11491 (get_address_cost): Likewise.
11492 (split_small_address_groups_p): Use address_p.
11493 (split_address_groups): Likewise.
11494 (add_iv_candidate_for_use): Likewise.
11495 (autoinc_possible_for_pair): Likewise.
11496 (rewrite_groups): Likewise.
11497 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11498 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11499 (get_alias_ptr_type_for_ptr_address): New function.
11500 (rewrite_use_address): Rewrite address uses in calls that were
11501 identified by find_address_like_use.
11503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11504 Alan Hayward <alan.hayward@arm.com>
11505 David Sherwood <david.sherwood@arm.com>
11507 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11509 * gimple-expr.h (is_gimple_addressable: Likewise.
11510 * gimple-expr.c (is_gimple_address): Likewise.
11511 * internal-fn.c (expand_call_mem_ref): New function.
11512 (expand_mask_load_optab_fn): Use it.
11513 (expand_mask_store_optab_fn): Likewise.
11515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11516 Alan Hayward <alan.hayward@arm.com>
11517 David Sherwood <david.sherwood@arm.com>
11519 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11520 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11521 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11522 (cond_umax@var{mode}): Document.
11523 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11524 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11525 (cond_umin_optab, cond_umax_optab): New optabs.
11526 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11527 (COND_IOR, COND_XOR): New internal functions.
11528 * internal-fn.h (get_conditional_internal_fn): Declare.
11529 * internal-fn.c (cond_binary_direct): New macro.
11530 (expand_cond_binary_optab_fn): Likewise.
11531 (direct_cond_binary_optab_supported_p): Likewise.
11532 (get_conditional_internal_fn): New function.
11533 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11534 Cope with reduction statements that are vectorized as calls rather
11536 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11537 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11538 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11539 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11540 (UNSPEC_COND_EOR): New unspecs.
11541 (optab): Add mappings for them.
11542 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11543 (sve_int_op, sve_fp_op): New int attributes.
11545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11546 Alan Hayward <alan.hayward@arm.com>
11547 David Sherwood <david.sherwood@arm.com>
11549 * optabs.def (while_ult_optab): New optab.
11550 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11551 * internal-fn.def (WHILE_ULT): New internal function.
11552 * internal-fn.h (direct_internal_fn_supported_p): New override
11553 that takes two types as argument.
11554 * internal-fn.c (while_direct): New macro.
11555 (expand_while_optab_fn): New function.
11556 (convert_optab_supported_p): Likewise.
11557 (direct_while_optab_supported_p): New macro.
11558 * wide-int.h (wi::udiv_ceil): New function.
11559 * tree-vectorizer.h (rgroup_masks): New structure.
11560 (vec_loop_masks): New typedef.
11561 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11562 and fully_masked_p.
11563 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11564 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11565 (vect_max_vf): New function.
11566 (slpeel_make_loop_iterate_ntimes): Delete.
11567 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11568 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11569 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11570 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11571 internal-fn.h, stor-layout.h and optabs-query.h.
11572 (vect_set_loop_mask): New function.
11573 (add_preheader_seq): Likewise.
11574 (add_header_seq): Likewise.
11575 (interleave_supported_p): Likewise.
11576 (vect_maybe_permute_loop_masks): Likewise.
11577 (vect_set_loop_masks_directly): Likewise.
11578 (vect_set_loop_condition_masked): Likewise.
11579 (vect_set_loop_condition_unmasked): New function, split out from
11580 slpeel_make_loop_iterate_ntimes.
11581 (slpeel_make_loop_iterate_ntimes): Rename to..
11582 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11583 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11584 (vect_do_peeling): Update call accordingly.
11585 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11587 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11588 mask_compare_type, can_fully_mask_p and fully_masked_p.
11589 (release_vec_loop_masks): New function.
11590 (_loop_vec_info): Use it to free the loop masks.
11591 (can_produce_all_loop_masks_p): New function.
11592 (vect_get_max_nscalars_per_iter): Likewise.
11593 (vect_verify_full_masking): Likewise.
11594 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11595 retries, and free the mask rgroups before retrying. Check loop-wide
11596 reasons for disallowing fully-masked loops. Make the final decision
11597 about whether use a fully-masked loop or not.
11598 (vect_estimate_min_profitable_iters): Do not assume that peeling
11599 for the number of iterations will be needed for fully-masked loops.
11600 (vectorizable_reduction): Disable fully-masked loops.
11601 (vectorizable_live_operation): Likewise.
11602 (vect_halve_mask_nunits): New function.
11603 (vect_double_mask_nunits): Likewise.
11604 (vect_record_loop_mask): Likewise.
11605 (vect_get_loop_mask): Likewise.
11606 (vect_transform_loop): Handle the case in which the final loop
11607 iteration might handle a partial vector. Call vect_set_loop_condition
11608 instead of slpeel_make_loop_iterate_ntimes.
11609 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11610 (check_load_store_masking): New function.
11611 (prepare_load_store_mask): Likewise.
11612 (vectorizable_store): Handle fully-masked loops.
11613 (vectorizable_load): Likewise.
11614 (supportable_widening_operation): Use vect_halve_mask_nunits for
11616 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11617 (vect_gen_while): New function.
11618 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11619 (aarch64_uqdec<mode>): New insn.
11621 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11622 Alan Hayward <alan.hayward@arm.com>
11623 David Sherwood <david.sherwood@arm.com>
11625 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11626 (reduc_xor_scal_optab): New optabs.
11627 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11628 (reduc_xor_scal_@var{m}): Document.
11629 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11630 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11631 internal functions.
11632 * fold-const-call.c (fold_const_call): Handle them.
11633 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11634 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11635 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11636 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11637 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11638 (UNSPEC_XORV): New unspecs.
11639 (optab): Add entries for them.
11640 (BITWISEV): New int iterator.
11641 (bit_reduc_op): New int attributes.
11643 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11644 Alan Hayward <alan.hayward@arm.com>
11645 David Sherwood <david.sherwood@arm.com>
11647 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11648 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11649 * optabs.def (vec_shl_insert_optab): New optab.
11650 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11651 (duplicate_and_interleave): Likewise.
11652 * tree-vect-loop.c: Include internal-fn.h.
11653 (neutral_op_for_slp_reduction): New function, split out from
11654 get_initial_defs_for_reduction.
11655 (get_initial_def_for_reduction): Handle option 2 for variable-length
11656 vectors by loading the neutral value into a vector and then shifting
11657 the initial value into element 0.
11658 (get_initial_defs_for_reduction): Replace the code argument with
11659 the neutral value calculated by neutral_op_for_slp_reduction.
11660 Use gimple_build_vector for constant-length vectors.
11661 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11662 but the first group_size elements have a neutral value.
11663 Use duplicate_and_interleave otherwise.
11664 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11665 Update call to get_initial_defs_for_reduction. Handle SLP
11666 reductions for variable-length vectors by creating one vector
11667 result for each scalar result, with the elements associated
11668 with other scalar results stubbed out with the neutral value.
11669 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11670 Require IFN_VEC_SHL_INSERT for double reductions on
11671 variable-length vectors, or SLP reductions that have
11672 a neutral value. Require can_duplicate_and_interleave_p
11673 support for variable-length unchained SLP reductions if there
11674 is no neutral value, such as for MIN/MAX reductions. Also require
11675 the number of vector elements to be a multiple of the number of
11676 SLP statements when doing variable-length unchained SLP reductions.
11677 Update call to vect_create_epilog_for_reduction.
11678 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11679 and remove initial values.
11680 (duplicate_and_interleave): Make public.
11681 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11682 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11684 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11685 Alan Hayward <alan.hayward@arm.com>
11686 David Sherwood <david.sherwood@arm.com>
11688 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11689 (can_duplicate_and_interleave_p): New function.
11690 (vect_get_and_check_slp_defs): Take the vector of statements
11691 rather than just the current one. Remove excess parentheses.
11692 Restriction rejectinon of vect_constant_def and vect_external_def
11693 for variable-length vectors to boolean types, or types for which
11694 can_duplicate_and_interleave_p is false.
11695 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11696 (duplicate_and_interleave): New function.
11697 (vect_get_constant_vectors): Use gimple_build_vector for
11698 constant-length vectors and suitable variable-length constant
11699 vectors. Use duplicate_and_interleave for other variable-length
11700 vectors. Don't defer the update when inserting new statements.
11702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11703 Alan Hayward <alan.hayward@arm.com>
11704 David Sherwood <david.sherwood@arm.com>
11706 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11707 min_profitable_iters doesn't go negative.
11709 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11710 Alan Hayward <alan.hayward@arm.com>
11711 David Sherwood <david.sherwood@arm.com>
11713 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11714 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11715 * optabs.def (vec_mask_load_lanes_optab): New optab.
11716 (vec_mask_store_lanes_optab): Likewise.
11717 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11718 (MASK_STORE_LANES): Likewise.
11719 * internal-fn.c (mask_load_lanes_direct): New macro.
11720 (mask_store_lanes_direct): Likewise.
11721 (expand_mask_load_optab_fn): Handle masked operations.
11722 (expand_mask_load_lanes_optab_fn): New macro.
11723 (expand_mask_store_optab_fn): Handle masked operations.
11724 (expand_mask_store_lanes_optab_fn): New macro.
11725 (direct_mask_load_lanes_optab_supported_p): Likewise.
11726 (direct_mask_store_lanes_optab_supported_p): Likewise.
11727 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11729 (vect_load_lanes_supported): Likewise.
11730 * tree-vect-data-refs.c (strip_conversion): New function.
11731 (can_group_stmts_p): Likewise.
11732 (vect_analyze_data_ref_accesses): Use it instead of checking
11733 for a pair of assignments.
11734 (vect_store_lanes_supported): Take a masked_p parameter.
11735 (vect_load_lanes_supported): Likewise.
11736 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11737 vect_store_lanes_supported and vect_load_lanes_supported.
11738 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11739 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11740 parameter. Don't allow gaps for masked accesses.
11741 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11742 and vect_load_lanes_supported.
11743 (get_load_store_type): Take a masked_p parameter and update
11744 call to get_group_load_store_type.
11745 (vectorizable_store): Update call to get_load_store_type.
11746 Handle IFN_MASK_STORE_LANES.
11747 (vectorizable_load): Update call to get_load_store_type.
11748 Handle IFN_MASK_LOAD_LANES.
11750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11751 Alan Hayward <alan.hayward@arm.com>
11752 David Sherwood <david.sherwood@arm.com>
11754 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11756 * config/aarch64/aarch64-protos.h
11757 (aarch64_sve_struct_memory_operand_p): Declare.
11758 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11759 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11760 (VPRED, vpred): Handle SVE structure modes.
11761 * config/aarch64/constraints.md (Utx): New constraint.
11762 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11763 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11764 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11765 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11766 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11767 structure modes. Split into pieces after RA.
11768 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11769 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11771 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11772 SVE structure modes.
11773 (aarch64_classify_address): Likewise.
11774 (sizetochar): Move earlier in file.
11775 (aarch64_print_operand): Handle SVE register lists.
11776 (aarch64_array_mode): New function.
11777 (aarch64_sve_struct_memory_operand_p): Likewise.
11778 (TARGET_ARRAY_MODE): Redefine.
11780 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11781 Alan Hayward <alan.hayward@arm.com>
11782 David Sherwood <david.sherwood@arm.com>
11784 * target.def (array_mode): New target hook.
11785 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11786 * doc/tm.texi: Regenerate.
11787 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11788 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11789 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11790 targetm.array_mode.
11791 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11795 Alan Hayward <alan.hayward@arm.com>
11796 David Sherwood <david.sherwood@arm.com>
11798 * fold-const.c (fold_binary_loc): Check the argument types
11799 rather than the result type when testing for a vector operation.
11801 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11803 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11804 * doc/tm.texi: Regenerate.
11806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11807 Alan Hayward <alan.hayward@arm.com>
11808 David Sherwood <david.sherwood@arm.com>
11810 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11811 (sve): Document new AArch64 extension.
11812 * doc/md.texi (w): Extend the description of the AArch64
11813 constraint to include SVE vectors.
11814 (Upl, Upa): Document new AArch64 predicate constraints.
11815 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11817 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11818 (msve-vector-bits=): New option.
11819 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11820 SVE when these are disabled.
11821 (sve): New extension.
11822 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11823 modes. Adjust their number of units based on aarch64_sve_vg.
11824 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11825 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11826 aarch64_addr_query_type.
11827 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11828 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11829 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11830 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11831 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11832 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11833 (aarch64_simd_imm_zero_p): Delete.
11834 (aarch64_check_zero_based_sve_index_immediate): Declare.
11835 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11836 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11837 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11838 (aarch64_sve_float_mul_immediate_p): Likewise.
11839 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11840 rather than an rtx.
11841 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11842 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11843 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11844 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11845 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11846 (aarch64_regmode_natural_size): Likewise.
11847 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11848 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11850 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11851 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11852 for VG and the SVE predicate registers.
11853 (V_ALIASES): Add a "z"-prefixed alias.
11854 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11855 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11856 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11857 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11858 (REG_CLASS_NAMES): Add entries for them.
11859 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11860 and the predicate registers.
11861 (aarch64_sve_vg): Declare.
11862 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11863 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11864 (REGMODE_NATURAL_SIZE): Define.
11865 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11867 * config/aarch64/aarch64.c: Include cfgrtl.h.
11868 (simd_immediate_info): Add a constructor for series vectors,
11869 and an associated step field.
11870 (aarch64_sve_vg): New variable.
11871 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11872 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11873 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11874 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11875 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11876 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11877 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11878 (aarch64_get_mask_mode): New functions.
11879 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11880 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11881 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11882 predicate modes and predicate registers. Explicitly restrict
11883 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11884 to store a vector mode if it is recognized by
11885 aarch64_classify_vector_mode.
11886 (aarch64_regmode_natural_size): New function.
11887 (aarch64_hard_regno_caller_save_mode): Return the original mode
11889 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11890 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11891 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11892 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11894 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11895 does not overlap dest if the function is frame-related. Handle
11897 (aarch64_split_add_offset): New function.
11898 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11899 them aarch64_add_offset.
11900 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11901 and update call to aarch64_sub_sp.
11902 (aarch64_add_cfa_expression): New function.
11903 (aarch64_expand_prologue): Pass extra temporary registers to the
11904 functions above. Handle the case in which we need to emit new
11905 DW_CFA_expressions for registers that were originally saved
11906 relative to the stack pointer, but now have to be expressed
11907 relative to the frame pointer.
11908 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11910 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11911 IP0 and IP1 values for SVE frames.
11912 (aarch64_expand_vec_series): New function.
11913 (aarch64_expand_sve_widened_duplicate): Likewise.
11914 (aarch64_expand_sve_const_vector): Likewise.
11915 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11916 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11917 into the register, rather than emitting a SET directly.
11918 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11919 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11920 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11921 (offset_9bit_signed_scaled_p): New functions.
11922 (aarch64_replicate_bitmask_imm): New function.
11923 (aarch64_bitmask_imm): Use it.
11924 (aarch64_cannot_force_const_mem): Reject expressions involving
11925 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11926 (aarch64_classify_index): Handle SVE indices, by requiring
11927 a plain register index with a scale that matches the element size.
11928 (aarch64_classify_address): Handle SVE addresses. Assert that
11929 the mode of the address is VOIDmode or an integer mode.
11930 Update call to aarch64_classify_symbol.
11931 (aarch64_classify_symbolic_expression): Update call to
11932 aarch64_classify_symbol.
11933 (aarch64_const_vec_all_in_range_p): New function.
11934 (aarch64_print_vector_float_operand): Likewise.
11935 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11936 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11937 and the FP immediates 1.0 and 0.5.
11938 (aarch64_print_address_internal): Handle SVE addresses.
11939 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11940 (aarch64_regno_regclass): Handle predicate registers.
11941 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11943 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11944 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11945 (aarch64_convert_sve_vector_bits): New function.
11946 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11947 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11948 rather than an rtx.
11949 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11950 Handle SVE vector and predicate modes. Accept VL-based constants
11951 that need only one temporary register, and VL offsets that require
11952 no temporary registers.
11953 (aarch64_conditional_register_usage): Mark the predicate registers
11954 as fixed if SVE isn't available.
11955 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11956 Return true for SVE vector and predicate modes.
11957 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11958 rather than an unsigned int. Handle SVE modes.
11959 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11961 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11963 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11964 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11965 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11966 (aarch64_sve_float_mul_immediate_p): New functions.
11967 (aarch64_sve_valid_immediate): New function.
11968 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11969 Explicitly reject structure modes. Check for INDEX constants.
11970 Handle PTRUE and PFALSE constants.
11971 (aarch64_check_zero_based_sve_index_immediate): New function.
11972 (aarch64_simd_imm_zero_p): Delete.
11973 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11974 vector modes. Accept constants in the range of CNT[BHWD].
11975 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11976 ask for an Advanced SIMD mode.
11977 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11978 (aarch64_simd_vector_alignment): Handle SVE predicates.
11979 (aarch64_vectorize_preferred_vector_alignment): New function.
11980 (aarch64_simd_vector_alignment_reachable): Use it instead of
11982 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11983 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11985 (MAX_VECT_LEN): Delete.
11986 (expand_vec_perm_d): Add a vec_flags field.
11987 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11988 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11989 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11991 (aarch64_evpc_rev): Rename to...
11992 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11993 (aarch64_evpc_rev_global): New function.
11994 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11995 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11997 (aarch64_evpc_sve_tbl): New function.
11998 (aarch64_expand_vec_perm_const_1): Update after rename of
11999 aarch64_evpc_rev. Handle SVE permutes too, trying
12000 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
12001 than aarch64_evpc_tbl.
12002 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
12003 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
12004 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
12005 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
12006 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
12007 (aarch64_expand_sve_vcond): New functions.
12008 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
12009 of aarch64_vector_mode_p.
12010 (aarch64_dwarf_poly_indeterminate_value): New function.
12011 (aarch64_compute_pressure_classes): Likewise.
12012 (aarch64_can_change_mode_class): Likewise.
12013 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
12014 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
12015 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
12016 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
12017 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
12018 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
12019 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
12020 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
12022 (Dn, Dl, Dr): Accept const as well as const_vector.
12023 (Dz): Likewise. Compare against CONST0_RTX.
12024 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
12025 of "vector" where appropriate.
12026 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
12027 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
12028 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
12029 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
12030 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
12031 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
12032 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
12033 (v_int_equiv): Extend to SVE modes.
12034 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
12036 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
12037 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
12038 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
12039 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
12040 (SVE_COND_FP_CMP): New int iterators.
12041 (perm_hilo): Handle the new unpack unspecs.
12042 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
12044 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
12045 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
12046 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
12047 (aarch64_equality_operator, aarch64_constant_vector_operand)
12048 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
12049 (aarch64_sve_nonimmediate_operand): Likewise.
12050 (aarch64_sve_general_operand): Likewise.
12051 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12052 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12053 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12054 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12055 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12056 (aarch64_sve_float_arith_immediate): Likewise.
12057 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12058 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12059 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12060 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12061 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12062 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12063 (aarch64_sve_float_arith_operand): Likewise.
12064 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12065 (aarch64_sve_float_mul_operand): Likewise.
12066 (aarch64_sve_vec_perm_operand): Likewise.
12067 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12068 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12069 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12070 as well as const_vector.
12071 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12072 in file. Use CONST0_RTX and CONSTM1_RTX.
12073 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12074 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12075 Use aarch64_simd_imm_zero.
12076 * config/aarch64/aarch64-sve.md: New file.
12077 * config/aarch64/aarch64.md: Include it.
12078 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12079 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12080 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12081 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12082 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12083 (sve): New attribute.
12084 (enabled): Disable instructions with the sve attribute unless
12086 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12087 aarch64_expand_mov_immediate.
12088 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12089 CNT[BHSD] immediates.
12090 (movti): Split CONST_POLY_INT moves into two halves.
12091 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12092 Split additions that need a temporary here if the destination
12093 is the stack pointer.
12094 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12095 (*add<mode>3_poly_1): New instruction.
12096 (set_clobber_cc): New expander.
12098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12100 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12101 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12102 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12103 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12104 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12105 Change innermode from fixed_mode_size to machine_mode.
12106 (simplify_subreg): Update call accordingly. Handle a constant-sized
12107 subreg of a variable-length CONST_VECTOR.
12109 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12110 Alan Hayward <alan.hayward@arm.com>
12111 David Sherwood <david.sherwood@arm.com>
12113 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12114 (add_offset_to_base): New function, split out from...
12115 (create_mem_ref): ...here. When handling a scale other than 1,
12116 check first whether the address is valid without the offset.
12117 Add it into the base if so, leaving the index and scale as-is.
12119 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12122 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12123 fold_for_warn before checking if arg2 is INTEGER_CST.
12125 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12127 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12128 (store_multiple_operation): Delete.
12129 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12130 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12131 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12132 guarded by TARGET_STRING.
12133 (rs6000_output_load_multiple): Delete.
12134 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12135 OPTION_MASK_STRING / TARGET_STRING handling.
12136 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12137 (const rs6000_opt_masks) <"string">: Change mask to 0.
12138 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12139 (MASK_STRING): Delete.
12140 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12142 (load_multiple): Delete.
12149 (store_multiple): Delete.
12156 (movmemsi_8reg): Delete.
12157 (corresponding unnamed define_insn): Delete.
12158 (movmemsi_6reg): Delete.
12159 (corresponding unnamed define_insn): Delete.
12160 (movmemsi_4reg): Delete.
12161 (corresponding unnamed define_insn): Delete.
12162 (movmemsi_2reg): Delete.
12163 (corresponding unnamed define_insn): Delete.
12164 (movmemsi_1reg): Delete.
12165 (corresponding unnamed define_insn): Delete.
12166 * config/rs6000/rs6000.opt (mno-string): New.
12167 (mstring): Replace by deprecation warning stub.
12168 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12170 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12172 * regrename.c (regrename_do_replace): If replacing the same
12173 reg multiple times, try to reuse last created gen_raw_REG.
12176 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12177 main to workaround a bug in GDB.
12179 2018-01-12 Tom de Vries <tom@codesourcery.com>
12182 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12184 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12186 PR rtl-optimization/80481
12187 * ira-color.c (get_cap_member): New function.
12188 (allocnos_conflict_by_live_ranges_p): Use it.
12189 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12190 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12192 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12195 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12196 (*saddl_se_1): Ditto.
12197 (*ssubsi_1): Ditto.
12198 (*ssubl_se_1): Ditto.
12200 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12202 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12203 rather than wi::to_widest for DR_INITs.
12204 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12205 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12206 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12208 (vect_analyze_group_access_1): Note that here.
12210 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12212 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12213 polynomial type sizes.
12215 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12217 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12218 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12219 (gimple_add_tmp_var): Likewise.
12221 2018-01-12 Martin Liska <mliska@suse.cz>
12223 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12224 (gimple_alloc_sizes): Likewise.
12225 (dump_gimple_statistics): Use PRIu64 in printf format.
12226 * gimple.h: Change uint64_t to int.
12228 2018-01-12 Martin Liska <mliska@suse.cz>
12230 * tree-core.h: Use uint64_t instead of int.
12231 * tree.c (tree_node_counts): Likewise.
12232 (tree_node_sizes): Likewise.
12233 (dump_tree_statistics): Use PRIu64 in printf format.
12235 2018-01-12 Martin Liska <mliska@suse.cz>
12237 * Makefile.in: As qsort_chk is implemented in vec.c, add
12238 vec.o to linkage of gencfn-macros.
12239 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12240 passing the info to record_node_allocation_statistics.
12241 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12243 * ggc-common.c (struct ggc_usage): Add operator== and use
12244 it in operator< and compare function.
12245 * mem-stats.h (struct mem_usage): Likewise.
12246 * vec.c (struct vec_usage): Remove operator< and compare
12247 function. Can be simply inherited.
12249 2018-01-12 Martin Jambor <mjambor@suse.cz>
12252 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12253 * tree-ssa-math-opts.c: Include domwalk.h.
12254 (convert_mult_to_fma_1): New function.
12255 (fma_transformation_info): New type.
12256 (fma_deferring_state): Likewise.
12257 (cancel_fma_deferring): New function.
12258 (result_of_phi): Likewise.
12259 (last_fma_candidate_feeds_initial_phi): Likewise.
12260 (convert_mult_to_fma): Added deferring logic, split actual
12261 transformation to convert_mult_to_fma_1.
12262 (math_opts_dom_walker): New type.
12263 (math_opts_dom_walker::after_dom_children): New method, body moved
12264 here from pass_optimize_widening_mul::execute, added deferring logic
12266 (pass_optimize_widening_mul::execute): Moved most of code to
12267 math_opts_dom_walker::after_dom_children.
12268 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12269 * config/i386/i386.c (ix86_option_override_internal): Added
12270 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12272 2018-01-12 Richard Biener <rguenther@suse.de>
12275 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12276 inline instance vars.
12278 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12281 * config/rx/rx.c (rx_is_restricted_memory_address):
12282 Handle SUBREG case.
12284 2018-01-12 Richard Biener <rguenther@suse.de>
12286 PR tree-optimization/80846
12287 * target.def (split_reduction): New target hook.
12288 * targhooks.c (default_split_reduction): New function.
12289 * targhooks.h (default_split_reduction): Declare.
12290 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12291 target requests first reduce vectors by combining low and high
12293 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12294 (get_vectype_for_scalar_type_and_size): Export.
12295 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12296 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12297 * doc/tm.texi: Regenerate.
12298 * config/i386/i386.c (ix86_split_reduction): Implement
12299 TARGET_VECTORIZE_SPLIT_REDUCTION.
12301 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12304 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12305 in PIC mode except for TARGET_VXWORKS_RTP.
12306 * config/sparc/sparc.c: Include cfgrtl.h.
12307 (TARGET_INIT_PIC_REG): Define.
12308 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12309 (sparc_pic_register_p): New predicate.
12310 (sparc_legitimate_address_p): Use it.
12311 (sparc_legitimize_pic_address): Likewise.
12312 (sparc_delegitimize_address): Likewise.
12313 (sparc_mode_dependent_address_p): Likewise.
12314 (gen_load_pcrel_sym): Remove 4th parameter.
12315 (load_got_register): Adjust call to above. Remove obsolete stuff.
12316 (sparc_expand_prologue): Do not call load_got_register here.
12317 (sparc_flat_expand_prologue): Likewise.
12318 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12319 (sparc_use_pseudo_pic_reg): New function.
12320 (sparc_init_pic_reg): Likewise.
12321 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12322 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12324 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12326 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12327 Add item for branch_cost.
12329 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12331 PR rtl-optimization/83565
12332 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12333 not extend the result to a larger mode for rotate operations.
12334 (num_sign_bit_copies1): Likewise.
12336 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12339 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12341 Use values-Xc.o for -pedantic.
12342 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12344 2018-01-12 Martin Liska <mliska@suse.cz>
12347 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12349 (possible_polymorphic_call_targets): Use it.
12350 (ipa_devirt): Likewise.
12352 2018-01-12 Martin Liska <mliska@suse.cz>
12354 * profile-count.h (enum profile_quality): Use 0 as invalid
12355 enum value of profile_quality.
12357 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12359 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12360 -mext-string options.
12362 2018-01-12 Richard Biener <rguenther@suse.de>
12364 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12365 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12366 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12368 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12370 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12372 * configure.ac (--with-long-double-format): Add support for the
12373 configuration option to change the default long double format on
12375 * config.gcc (powerpc*-linux*-*): Likewise.
12376 * configure: Regenerate.
12377 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12378 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12379 used without modification.
12381 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12383 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12384 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12385 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12386 MISC_BUILTIN_SPEC_BARRIER.
12387 (rs6000_init_builtins): Likewise.
12388 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12390 (speculation_barrier): New define_insn.
12391 * doc/extend.texi: Document __builtin_speculation_barrier.
12393 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12396 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12397 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12398 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12400 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12401 integral modes instead of "ss" and "sd".
12402 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12403 vectors with 32-bit and 64-bit elements.
12404 (vecdupssescalarmodesuffix): New mode attribute.
12405 (vec_dup<mode>): Use it.
12407 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12410 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12411 frame if argument is passed on stack.
12413 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12416 * ree.c (combine_reaching_defs): Optimize also
12417 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12418 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12420 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12422 PR middle-end/83189
12423 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12425 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12427 PR middle-end/83718
12428 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12429 after they are computed.
12431 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12433 PR tree-optimization/83695
12434 * gimple-loop-linterchange.cc
12435 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12436 reset cached scev information after interchange.
12437 (pass_linterchange::execute): Remove call to scev_reset_htab.
12439 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12441 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12442 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12443 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12444 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12445 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12446 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12447 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12448 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12449 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12450 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12451 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12452 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12453 (V_lane_reg): Likewise.
12454 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12456 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12457 (vfmal_lane_low<mode>_intrinsic,
12458 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12459 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12460 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12461 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12462 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12463 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12465 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12467 * config/arm/arm-cpus.in (fp16fml): New feature.
12468 (ALL_SIMD): Add fp16fml.
12469 (armv8.2-a): Add fp16fml as an option.
12470 (armv8.3-a): Likewise.
12471 (armv8.4-a): Add fp16fml as part of fp16.
12472 * config/arm/arm.h (TARGET_FP16FML): Define.
12473 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12475 * config/arm/arm-modes.def (V2HF): Define.
12476 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12477 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12478 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12479 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12480 vfmsl_low, vfmsl_high): New set of builtins.
12481 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12482 (vfml_op): New code attribute.
12483 (VFMLHALVES): New int iterator.
12484 (VFML, VFMLSEL): New mode attributes.
12485 (V_reg): Define mapping for V2HF.
12486 (V_hi, V_lo): New mode attributes.
12487 (VF_constraint): Likewise.
12488 (vfml_half, vfml_half_selector): New int attributes.
12489 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12491 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12492 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12494 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12495 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12496 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12497 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12499 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12500 Document new effective target and option set.
12502 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12504 * config/arm/arm-cpus.in (armv8_4): New feature.
12505 (ARMv8_4a): New fgroup.
12506 (armv8.4-a): New arch.
12507 * config/arm/arm-tables.opt: Regenerate.
12508 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12509 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12510 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12511 Add matching rules for -march=armv8.4-a and extensions.
12512 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12514 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12517 * config/rx/rx.md (BW): New mode attribute.
12518 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12520 2018-01-11 Richard Biener <rguenther@suse.de>
12522 PR tree-optimization/83435
12523 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12524 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12525 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12527 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12528 Alan Hayward <alan.hayward@arm.com>
12529 David Sherwood <david.sherwood@arm.com>
12531 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12533 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12534 (aarch64_print_address_internal): Use it to check for a zero offset.
12536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12537 Alan Hayward <alan.hayward@arm.com>
12538 David Sherwood <david.sherwood@arm.com>
12540 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12541 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12542 Return a poly_int64 rather than a HOST_WIDE_INT.
12543 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12544 rather than a HOST_WIDE_INT.
12545 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12546 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12547 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12548 final_offset from HOST_WIDE_INT to poly_int64.
12549 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12550 to_constant when getting the number of units in an Advanced SIMD
12552 (aarch64_builtin_vectorized_function): Check for a constant number
12554 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12556 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12557 attribute instead of GET_MODE_NUNITS.
12558 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12559 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12560 GET_MODE_SIZE for fixed-size registers.
12561 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12562 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12563 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12564 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12565 (aarch64_print_operand, aarch64_print_address_internal)
12566 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12567 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12568 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12569 Handle polynomial GET_MODE_SIZE.
12570 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12571 wider than SImode without modification.
12572 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12573 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12574 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12575 passing and returning SVE modes.
12576 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12577 rather than GEN_INT.
12578 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12579 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12580 (aarch64_allocate_and_probe_stack_space): Likewise.
12581 (aarch64_layout_frame): Cope with polynomial offsets.
12582 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12583 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12584 polynomial offsets.
12585 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12586 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12587 poly_int64 rather than a HOST_WIDE_INT.
12588 (aarch64_get_separate_components, aarch64_process_components)
12589 (aarch64_expand_prologue, aarch64_expand_epilogue)
12590 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12591 (aarch64_anchor_offset): New function, split out from...
12592 (aarch64_legitimize_address): ...here.
12593 (aarch64_builtin_vectorization_cost): Handle polynomial
12594 TYPE_VECTOR_SUBPARTS.
12595 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12597 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12598 number of elements from the PARALLEL rather than the mode.
12599 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12600 rather than GET_MODE_BITSIZE.
12601 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12602 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12603 (aarch64_expand_vec_perm_const_1): Handle polynomial
12604 d->perm.length () and d->perm elements.
12605 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12606 Apply to_constant to d->perm elements.
12607 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12608 polynomial CONST_VECTOR_NUNITS.
12609 (aarch64_move_pointer): Take amount as a poly_int64 rather
12611 (aarch64_progress_pointer): Avoid temporary variable.
12612 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12613 the mode attribute instead of GET_MODE.
12615 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12616 Alan Hayward <alan.hayward@arm.com>
12617 David Sherwood <david.sherwood@arm.com>
12619 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12620 x exists before using it.
12621 (aarch64_add_constant_internal): Rename to...
12622 (aarch64_add_offset_1): ...this. Replace regnum with separate
12623 src and dest rtxes. Handle the case in which they're different,
12624 including when the offset is zero. Replace scratchreg with an rtx.
12625 Use 2 additions if there is no spare register into which we can
12626 move a 16-bit constant.
12627 (aarch64_add_constant): Delete.
12628 (aarch64_add_offset): Replace reg with separate src and dest
12629 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12630 Use aarch64_add_offset_1.
12631 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12632 an rtx rather than an int. Take the delta as a poly_int64
12633 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12634 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12635 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12636 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12637 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12638 and aarch64_add_sp.
12639 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12640 aarch64_add_constant.
12642 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12644 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12645 Use scalar_float_mode.
12647 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12649 * config/aarch64/aarch64-simd.md
12650 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12651 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12652 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12653 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12654 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12655 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12656 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12657 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12658 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12659 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12661 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12664 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12665 targ_options->x_arm_arch_string is non NULL.
12667 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12669 * config/aarch64/aarch64.h
12670 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12672 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12675 * expmed.c (emit_store_flag_force): Swap if const op0
12676 and change VOIDmode to mode of op0.
12678 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12680 PR rtl-optimization/83761
12681 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12682 than bytes to mode_for_size.
12684 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12686 PR middle-end/83189
12687 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12688 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12691 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12693 PR middle-end/83575
12694 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12695 when in layout mode.
12696 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12697 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12700 2018-01-10 Michael Collison <michael.collison@arm.com>
12702 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12703 * config/aarch64/aarch64-option-extension.def: Add
12704 AARCH64_OPT_EXTENSION of 'fp16fml'.
12705 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12706 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12707 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12708 * config/aarch64/constraints.md (Ui7): New constraint.
12709 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12710 (VFMLA_SEL_W): Ditto.
12713 (VFMLA16_LOW): New int iterator.
12714 (VFMLA16_HIGH): Ditto.
12715 (UNSPEC_FMLAL): New unspec.
12716 (UNSPEC_FMLSL): Ditto.
12717 (UNSPEC_FMLAL2): Ditto.
12718 (UNSPEC_FMLSL2): Ditto.
12719 (f16mac): New code attribute.
12720 * config/aarch64/aarch64-simd-builtins.def
12721 (aarch64_fmlal_lowv2sf): Ditto.
12722 (aarch64_fmlsl_lowv2sf): Ditto.
12723 (aarch64_fmlalq_lowv4sf): Ditto.
12724 (aarch64_fmlslq_lowv4sf): Ditto.
12725 (aarch64_fmlal_highv2sf): Ditto.
12726 (aarch64_fmlsl_highv2sf): Ditto.
12727 (aarch64_fmlalq_highv4sf): Ditto.
12728 (aarch64_fmlslq_highv4sf): Ditto.
12729 (aarch64_fmlal_lane_lowv2sf): Ditto.
12730 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12731 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12732 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12733 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12734 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12735 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12736 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12737 (aarch64_fmlal_lane_highv2sf): Ditto.
12738 (aarch64_fmlsl_lane_highv2sf): Ditto.
12739 (aarch64_fmlal_laneq_highv2sf): Ditto.
12740 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12741 (aarch64_fmlalq_lane_highv4sf): Ditto.
12742 (aarch64_fmlsl_lane_highv4sf): Ditto.
12743 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12744 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12745 * config/aarch64/aarch64-simd.md:
12746 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12747 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12748 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12749 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12750 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12751 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12752 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12753 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12754 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12755 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12756 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12757 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12758 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12759 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12760 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12761 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12762 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12763 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12764 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12765 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12766 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12767 (vfmlsl_low_u32): Ditto.
12768 (vfmlalq_low_u32): Ditto.
12769 (vfmlslq_low_u32): Ditto.
12770 (vfmlal_high_u32): Ditto.
12771 (vfmlsl_high_u32): Ditto.
12772 (vfmlalq_high_u32): Ditto.
12773 (vfmlslq_high_u32): Ditto.
12774 (vfmlal_lane_low_u32): Ditto.
12775 (vfmlsl_lane_low_u32): Ditto.
12776 (vfmlal_laneq_low_u32): Ditto.
12777 (vfmlsl_laneq_low_u32): Ditto.
12778 (vfmlalq_lane_low_u32): Ditto.
12779 (vfmlslq_lane_low_u32): Ditto.
12780 (vfmlalq_laneq_low_u32): Ditto.
12781 (vfmlslq_laneq_low_u32): Ditto.
12782 (vfmlal_lane_high_u32): Ditto.
12783 (vfmlsl_lane_high_u32): Ditto.
12784 (vfmlal_laneq_high_u32): Ditto.
12785 (vfmlsl_laneq_high_u32): Ditto.
12786 (vfmlalq_lane_high_u32): Ditto.
12787 (vfmlslq_lane_high_u32): Ditto.
12788 (vfmlalq_laneq_high_u32): Ditto.
12789 (vfmlslq_laneq_high_u32): Ditto.
12790 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12791 (AARCH64_FL_FOR_ARCH8_4): New.
12792 (AARCH64_ISA_F16FML): New ISA flag.
12793 (TARGET_F16FML): New feature flag for fp16fml.
12794 (doc/invoke.texi): Document new fp16fml option.
12796 2018-01-10 Michael Collison <michael.collison@arm.com>
12798 * config/aarch64/aarch64-builtins.c:
12799 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12800 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12801 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12802 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12803 (AARCH64_ISA_SHA3): New ISA flag.
12804 (TARGET_SHA3): New feature flag for sha3.
12805 * config/aarch64/iterators.md (sha512_op): New int attribute.
12806 (CRYPTO_SHA512): New int iterator.
12807 (UNSPEC_SHA512H): New unspec.
12808 (UNSPEC_SHA512H2): Ditto.
12809 (UNSPEC_SHA512SU0): Ditto.
12810 (UNSPEC_SHA512SU1): Ditto.
12811 * config/aarch64/aarch64-simd-builtins.def
12812 (aarch64_crypto_sha512hqv2di): New builtin.
12813 (aarch64_crypto_sha512h2qv2di): Ditto.
12814 (aarch64_crypto_sha512su0qv2di): Ditto.
12815 (aarch64_crypto_sha512su1qv2di): Ditto.
12816 (aarch64_eor3qv8hi): Ditto.
12817 (aarch64_rax1qv2di): Ditto.
12818 (aarch64_xarqv2di): Ditto.
12819 (aarch64_bcaxqv8hi): Ditto.
12820 * config/aarch64/aarch64-simd.md:
12821 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12822 (aarch64_crypto_sha512su0qv2di): Ditto.
12823 (aarch64_crypto_sha512su1qv2di): Ditto.
12824 (aarch64_eor3qv8hi): Ditto.
12825 (aarch64_rax1qv2di): Ditto.
12826 (aarch64_xarqv2di): Ditto.
12827 (aarch64_bcaxqv8hi): Ditto.
12828 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12829 (vsha512h2q_u64): Ditto.
12830 (vsha512su0q_u64): Ditto.
12831 (vsha512su1q_u64): Ditto.
12832 (veor3q_u16): Ditto.
12833 (vrax1q_u64): Ditto.
12834 (vxarq_u64): Ditto.
12835 (vbcaxq_u16): Ditto.
12836 * config/arm/types.md (crypto_sha512): New type attribute.
12837 (crypto_sha3): Ditto.
12838 (doc/invoke.texi): Document new sha3 option.
12840 2018-01-10 Michael Collison <michael.collison@arm.com>
12842 * config/aarch64/aarch64-builtins.c:
12843 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12844 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12845 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12846 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12847 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12848 (AARCH64_ISA_SM4): New ISA flag.
12849 (TARGET_SM4): New feature flag for sm4.
12850 * config/aarch64/aarch64-simd-builtins.def
12851 (aarch64_sm3ss1qv4si): Ditto.
12852 (aarch64_sm3tt1aq4si): Ditto.
12853 (aarch64_sm3tt1bq4si): Ditto.
12854 (aarch64_sm3tt2aq4si): Ditto.
12855 (aarch64_sm3tt2bq4si): Ditto.
12856 (aarch64_sm3partw1qv4si): Ditto.
12857 (aarch64_sm3partw2qv4si): Ditto.
12858 (aarch64_sm4eqv4si): Ditto.
12859 (aarch64_sm4ekeyqv4si): Ditto.
12860 * config/aarch64/aarch64-simd.md:
12861 (aarch64_sm3ss1qv4si): Ditto.
12862 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12863 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12864 (aarch64_sm4eqv4si): Ditto.
12865 (aarch64_sm4ekeyqv4si): Ditto.
12866 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12867 (sm3part_op): Ditto.
12868 (CRYPTO_SM3TT): Ditto.
12869 (CRYPTO_SM3PART): Ditto.
12870 (UNSPEC_SM3SS1): New unspec.
12871 (UNSPEC_SM3TT1A): Ditto.
12872 (UNSPEC_SM3TT1B): Ditto.
12873 (UNSPEC_SM3TT2A): Ditto.
12874 (UNSPEC_SM3TT2B): Ditto.
12875 (UNSPEC_SM3PARTW1): Ditto.
12876 (UNSPEC_SM3PARTW2): Ditto.
12877 (UNSPEC_SM4E): Ditto.
12878 (UNSPEC_SM4EKEY): Ditto.
12879 * config/aarch64/constraints.md (Ui2): New constraint.
12880 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12881 * config/arm/types.md (crypto_sm3): New type attribute.
12882 (crypto_sm4): Ditto.
12883 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12884 (vsm3tt1aq_u32): Ditto.
12885 (vsm3tt1bq_u32): Ditto.
12886 (vsm3tt2aq_u32): Ditto.
12887 (vsm3tt2bq_u32): Ditto.
12888 (vsm3partw1q_u32): Ditto.
12889 (vsm3partw2q_u32): Ditto.
12890 (vsm4eq_u32): Ditto.
12891 (vsm4ekeyq_u32): Ditto.
12892 (doc/invoke.texi): Document new sm4 option.
12894 2018-01-10 Michael Collison <michael.collison@arm.com>
12896 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12897 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12898 (AARCH64_FL_FOR_ARCH8_4): New.
12899 (AARCH64_FL_V8_4): New flag.
12900 (doc/invoke.texi): Document new armv8.4-a option.
12902 2018-01-10 Michael Collison <michael.collison@arm.com>
12904 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12905 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12906 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12907 * config/aarch64/aarch64-option-extension.def: Add
12908 AARCH64_OPT_EXTENSION of 'sha2'.
12909 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12910 (crypto): Disable sha2 and aes if crypto disabled.
12911 (crypto): Enable aes and sha2 if enabled.
12912 (simd): Disable sha2 and aes if simd disabled.
12913 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12915 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12916 (TARGET_SHA2): New feature flag for sha2.
12917 (TARGET_AES): New feature flag for aes.
12918 * config/aarch64/aarch64-simd.md:
12919 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12920 conditional on TARGET_AES.
12921 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12922 (aarch64_crypto_sha1hsi): Make pattern conditional
12924 (aarch64_crypto_sha1hv4si): Ditto.
12925 (aarch64_be_crypto_sha1hv4si): Ditto.
12926 (aarch64_crypto_sha1su1v4si): Ditto.
12927 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12928 (aarch64_crypto_sha1su0v4si): Ditto.
12929 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12930 (aarch64_crypto_sha256su0v4si): Ditto.
12931 (aarch64_crypto_sha256su1v4si): Ditto.
12932 (doc/invoke.texi): Document new aes and sha2 options.
12934 2018-01-10 Martin Sebor <msebor@redhat.com>
12936 PR tree-optimization/83781
12937 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12940 2018-01-11 Martin Sebor <msebor@gmail.com>
12941 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12943 PR tree-optimization/83501
12944 PR tree-optimization/81703
12946 * tree-ssa-strlen.c (get_string_cst): Rename...
12947 (get_string_len): ...to this. Handle global constants.
12948 (handle_char_store): Adjust.
12950 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12951 Jim Wilson <jimw@sifive.com>
12953 * config/riscv/riscv-protos.h (riscv_output_return): New.
12954 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12955 (riscv_attribute_table, riscv_output_return),
12956 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12957 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12958 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12960 (riscv_expand_prologue): Add early return for naked function.
12961 (riscv_expand_epilogue): Likewise.
12962 (riscv_function_ok_for_sibcall): Return false for naked function.
12963 (riscv_set_current_function): New.
12964 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12965 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12966 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12967 * doc/extend.texi (RISC-V Function Attributes): New.
12969 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12971 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12972 check for 128-bit long double before checking TCmode.
12973 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12974 128-bit long doubles before checking TFmode or TCmode.
12975 (FLOAT128_IBM_P): Likewise.
12977 2018-01-10 Martin Sebor <msebor@redhat.com>
12979 PR tree-optimization/83671
12980 * builtins.c (c_strlen): Unconditionally return zero for the empty
12982 Use -Warray-bounds for warnings.
12983 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12984 for non-constant array indices with COMPONENT_REF, arrays of
12985 arrays, and pointers to arrays.
12986 (gimple_fold_builtin_strlen): Determine and set length range for
12987 non-constant character arrays.
12989 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12991 PR middle-end/81897
12992 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12995 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12997 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12999 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13002 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
13003 VECTOR_MEM_ALTIVEC_OR_VSX_P.
13004 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
13005 indexed_or_indirect_operand predicate.
13006 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
13007 (*vsx_le_perm_load_v8hi): Likewise.
13008 (*vsx_le_perm_load_v16qi): Likewise.
13009 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
13010 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
13011 (*vsx_le_perm_store_v8hi): Likewise.
13012 (*vsx_le_perm_store_v16qi): Likewise.
13013 (eight unnamed splitters): Likewise.
13015 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
13017 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
13018 * config/rs6000/emmintrin.h: Likewise.
13019 * config/rs6000/mmintrin.h: Likewise.
13020 * config/rs6000/xmmintrin.h: Likewise.
13022 2018-01-10 David Malcolm <dmalcolm@redhat.com>
13025 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
13027 * tree.c (tree_nop_conversion): Return true for location wrapper
13029 (maybe_wrap_with_location): New function.
13030 (selftest::check_strip_nops): New function.
13031 (selftest::test_location_wrappers): New function.
13032 (selftest::tree_c_tests): Call it.
13033 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
13034 (maybe_wrap_with_location): New decl.
13035 (EXPR_LOCATION_WRAPPER_P): New macro.
13036 (location_wrapper_p): New inline function.
13037 (tree_strip_any_location_wrapper): New inline function.
13039 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
13042 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
13043 stack_realign_offset for the largest alignment of stack slot
13045 (ix86_find_max_used_stack_alignment): New function.
13046 (ix86_finalize_stack_frame_flags): Use it. Set
13047 max_used_stack_alignment if we don't realign stack.
13048 * config/i386/i386.h (machine_function): Add
13049 max_used_stack_alignment.
13051 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13053 * config/arm/arm.opt (-mbranch-cost): New option.
13054 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13057 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13060 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13061 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13063 2018-01-10 Richard Biener <rguenther@suse.de>
13066 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13067 early out so it also covers the case where we have a non-NULL
13070 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13072 PR tree-optimization/83753
13073 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13074 for non-strided grouped accesses if the number of elements is 1.
13076 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13079 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13080 * i386.h (TARGET_USE_GATHER): Define.
13081 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13083 2018-01-10 Martin Liska <mliska@suse.cz>
13086 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13087 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13089 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13090 CLEANUP_NO_PARTITIONING is not set.
13092 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13094 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13095 for vectors, as a partial revert of r254296.
13096 * rtl.h (const_vec_p): Delete.
13097 (const_vec_duplicate_p): Don't test for vector CONSTs.
13098 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13099 * expmed.c (make_tree): Likewise.
13102 * common.md (E, F): Use CONSTANT_P instead of checking for
13104 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13105 checking for CONST_VECTOR.
13107 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13109 PR middle-end/83575
13110 * predict.c (force_edge_cold): Handle in more sane way edges
13111 with no prediction.
13113 2018-01-09 Carl Love <cel@us.ibm.com>
13115 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13117 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13118 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13119 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13120 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13121 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13122 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13123 * config/rs6000/rs6000-protos.h: Add extern defition for
13124 rs6000_generate_float2_double_code.
13125 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13127 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13128 (float2_v2df): Add define_expand.
13130 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13133 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13134 op_mode in the force_to_mode call.
13136 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13138 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13139 instead of checking each element individually.
13140 (aarch64_evpc_uzp): Likewise.
13141 (aarch64_evpc_zip): Likewise.
13142 (aarch64_evpc_ext): Likewise.
13143 (aarch64_evpc_rev): Likewise.
13144 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13145 instead of checking each element individually. Return true without
13147 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13148 whether all selected elements come from the same input, instead of
13149 checking each element individually. Remove calls to gen_rtx_REG,
13150 start_sequence and end_sequence and instead assert that no rtl is
13153 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13155 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13156 order of HIGH and CONST checks.
13158 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13160 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13161 if the destination isn't an SSA_NAME.
13163 2018-01-09 Richard Biener <rguenther@suse.de>
13165 PR tree-optimization/83668
13166 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13168 (canonicalize_loop_form): ... here, renamed from ...
13169 (canonicalize_loop_closed_ssa_form): ... this and amended to
13170 swap successor edges for loop exit blocks to make us use
13171 the RPO order we need for initial schedule generation.
13173 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13175 PR tree-optimization/64811
13176 * match.pd: When optimizing comparisons with Inf, avoid
13177 introducing or losing exceptions from comparisons with NaN.
13179 2018-01-09 Martin Liska <mliska@suse.cz>
13182 * asan.c (shadow_mem_size): Add gcc_assert.
13184 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13186 Don't save registers in main().
13189 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13190 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13191 * config/avr/avr.c (avr_set_current_function): Don't error if
13192 naked, OS_task or OS_main are specified at the same time.
13193 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13195 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13197 * common/config/avr/avr-common.c (avr_option_optimization_table):
13198 Switch on -mmain-is-OS_task for optimizing compilations.
13200 2018-01-09 Richard Biener <rguenther@suse.de>
13202 PR tree-optimization/83572
13203 * graphite.c: Include cfganal.h.
13204 (graphite_transform_loops): Connect infinite loops to exit
13205 and remove fake edges at the end.
13207 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13209 * ipa-inline.c (edge_badness): Revert accidental checkin.
13211 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13214 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13215 symbols; not inline clones.
13217 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13220 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13221 hard registers. Formatting fixes.
13223 PR preprocessor/83722
13224 * gcc.c (try_generate_repro): Pass
13225 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13226 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13229 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13230 Kito Cheng <kito.cheng@gmail.com>
13232 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13233 (riscv_leaf_function_p): Delete.
13234 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13236 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13238 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13240 (do_ifelse): New function.
13241 (do_isel): New function.
13242 (do_sub3): New function.
13243 (do_add3): New function.
13244 (do_load_mask_compare): New function.
13245 (do_overlap_load_compare): New function.
13246 (expand_compare_loop): New function.
13247 (expand_block_compare): Call expand_compare_loop() when appropriate.
13248 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13249 option description.
13250 (-mblock-compare-inline-loop-limit): New option.
13252 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13255 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13256 Reverse order of second and third operands in first alternative.
13257 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13258 of first and second elements in UNSPEC_VPERMR vector.
13259 (altivec_expand_vec_perm_le): Likewise.
13261 2018-01-08 Jeff Law <law@redhat.com>
13263 PR rtl-optimizatin/81308
13264 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13265 (process_switch): If group_case_labels makes a change, then set
13267 (pass_convert_switch::execute): If a switch is converted, then
13268 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13270 PR rtl-optimization/81308
13271 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13274 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13276 PR target/83663 - Revert r255946
13277 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13278 generation for cases where splatting a value is not useful.
13279 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13280 across a vec_duplicate and a paradoxical subreg forming a vector
13281 mode to a vec_concat.
13283 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13285 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13286 -march=armv8.3-a variants.
13287 * config/arm/t-multilib: Likewise.
13288 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13290 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13292 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13294 (cceq_ior_compare_complement): Give it a name so I can use it, and
13295 change boolean_or_operator predicate to boolean_operator so it can
13296 be used to generate a crand.
13297 (eqne): New code iterator.
13298 (bd/bd_neg): New code_attrs.
13299 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13300 a single define_insn.
13301 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13302 decrement (bdnzt/bdnzf/bdzt/bdzf).
13303 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13304 with the new names of the branch decrement patterns, and added the
13305 names of the branch decrement conditional patterns.
13307 2018-01-08 Richard Biener <rguenther@suse.de>
13309 PR tree-optimization/83563
13310 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13313 2018-01-08 Richard Biener <rguenther@suse.de>
13315 PR middle-end/83713
13316 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13318 2018-01-08 Richard Biener <rguenther@suse.de>
13320 PR tree-optimization/83685
13321 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13322 references to abnormals.
13324 2018-01-08 Richard Biener <rguenther@suse.de>
13327 * dwarf2out.c (output_indirect_strings): Handle empty
13328 skeleton_debug_str_hash.
13329 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13331 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13333 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13334 (emit_store_direct): Likewise.
13335 (arc_trampoline_adjust_address): Likewise.
13336 (arc_asm_trampoline_template): New function.
13337 (arc_initialize_trampoline): Use asm_trampoline_template.
13338 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13339 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13340 * config/arc/arc.md (flush_icache): Delete pattern.
13342 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13344 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13345 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13348 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13351 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13352 by not USED_FOR_TARGET.
13353 (make_pass_resolve_sw_modes): Likewise.
13355 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13357 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13360 2018-01-08 Richard Biener <rguenther@suse.de>
13362 PR middle-end/83580
13363 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13365 2018-01-08 Richard Biener <rguenther@suse.de>
13367 PR middle-end/83517
13368 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13370 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13372 PR middle-end/81897
13373 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13374 basic blocks with a small number of successors.
13375 (convert_control_dep_chain_into_preds): Improve handling of
13377 (dump_predicates): Split apart into...
13378 (dump_pred_chain): ...here...
13379 (dump_pred_info): ...and here.
13380 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13381 (can_chain_union_be_invalidated_p): Improve check for invalidation
13383 (uninit_uses_cannot_happen): Avoid unnecessary if
13384 convert_control_dep_chain_into_preds yielded nothing.
13386 2018-01-06 Martin Sebor <msebor@redhat.com>
13388 PR tree-optimization/83640
13389 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13390 subtracting negative offset from size.
13391 (builtin_access::overlap): Adjust offset bounds of the access to fall
13392 within the size of the object if possible.
13394 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13396 PR rtl-optimization/83699
13397 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13398 extract_bit_field_as_subreg to cases in which the extracted
13399 value is also a vector.
13401 * lra-constraints.c (process_alt_operands): Test for the equivalence
13402 substitutions when detecting a possible reload cycle.
13404 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13407 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13408 by default if flag_selective_schedling{,2}. Formatting fixes.
13410 PR rtl-optimization/83682
13411 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13412 if it has non-VECTOR_MODE element mode.
13413 (vec_duplicate_p): Likewise.
13415 PR middle-end/83694
13416 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13417 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13419 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13422 * config/i386/i386-builtin.def
13423 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13424 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13425 Require also OPTION_MASK_ISA_AVX512F in addition to
13426 OPTION_MASK_ISA_GFNI.
13427 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13428 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13429 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13430 to OPTION_MASK_ISA_GFNI.
13431 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13432 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13433 OPTION_MASK_ISA_AVX512BW.
13434 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13435 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13436 addition to OPTION_MASK_ISA_GFNI.
13437 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13438 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13439 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13440 to OPTION_MASK_ISA_GFNI.
13441 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13442 a requirement for all ISAs rather than any of them with a few
13444 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13446 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13447 bitmasks to be enabled with 3 exceptions, instead of requiring any
13448 enabled ISA with lots of exceptions.
13449 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13450 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13451 Change avx512bw in isa attribute to avx512f.
13452 * config/i386/sgxintrin.h: Add license boilerplate.
13453 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13454 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13455 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13456 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13458 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13459 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13460 temporarily sse2 rather than sse if not enabled already.
13463 * config/i386/sse.md (VI248_VLBW): Rename to ...
13464 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13465 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13466 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13467 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13468 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13469 mode iterator instead of VI248_VLBW.
13471 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13473 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13474 (record_modified): Skip clobbers; add debug output.
13475 (param_change_prob): Use sreal frequencies.
13477 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13479 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13480 punt for user-aligned variables.
13482 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13484 * tree-chrec.c (chrec_contains_symbols): Return true for
13487 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13490 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13491 of (x|y) == x for BICS pattern.
13493 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13495 PR tree-optimization/83605
13496 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13497 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13500 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13502 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13503 * config/epiphany/rtems.h: New file.
13505 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13506 Uros Bizjak <ubizjak@gmail.com>
13509 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13510 QIreg_operand instead of register_operand predicate.
13511 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13512 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13513 comments instead of -fmitigate[-_]rop.
13515 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13518 * cgraphunit.c (symbol_table::compile): Switch to text_section
13519 before calling assembly_start debug hook.
13520 * run-rtl-passes.c (run_rtl_passes): Likewise.
13523 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13525 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13526 range_int_cst_p rather than !symbolic_range_p before calling
13527 extract_range_from_multiplicative_op_1.
13529 2018-01-04 Jeff Law <law@redhat.com>
13531 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13532 redundant test in assertion.
13534 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13536 * doc/rtl.texi: Document machine_mode wrapper classes.
13538 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13540 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13541 using tree_to_uhwi.
13543 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13545 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13546 the VEC_PERM_EXPR fold to fail.
13548 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13551 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13552 to switched_sections.
13554 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13557 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13558 test for d.testing.
13560 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13563 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13564 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13566 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13569 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13570 is BLKmode and bitpos not zero or mode change is needed.
13572 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13575 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13578 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13581 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13582 instead of MULT rtx. Update all corresponding splitters.
13583 (*saddl_se): Ditto.
13584 (*ssub<modesuffix>): Ditto.
13585 (*ssubl_se): Ditto.
13586 (*cmp_sadd_di): Update split patterns.
13587 (*cmp_sadd_si): Ditto.
13588 (*cmp_sadd_sidi): Ditto.
13589 (*cmp_ssub_di): Ditto.
13590 (*cmp_ssub_si): Ditto.
13591 (*cmp_ssub_sidi): Ditto.
13592 * config/alpha/predicates.md (const23_operand): New predicate.
13593 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13594 Look for ASHIFT, not MULT inner operand.
13595 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13597 2018-01-04 Martin Liska <mliska@suse.cz>
13599 PR gcov-profile/83669
13600 * gcov.c (output_intermediate_file): Add version to intermediate
13602 * doc/gcov.texi: Document new field 'version' in intermediate
13603 file format. Fix location of '-k' option of gcov command.
13605 2018-01-04 Martin Liska <mliska@suse.cz>
13608 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13610 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13612 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13614 2018-01-03 Martin Sebor <msebor@redhat.com>
13616 PR tree-optimization/83655
13617 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13618 checking calls with invalid arguments.
13620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13622 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13623 (vectorizable_mask_load_store): Delete.
13624 (vectorizable_call): Return false for masked loads and stores.
13625 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13626 instead of gimple_assign_rhs1.
13627 (vectorizable_load): Handle IFN_MASK_LOAD.
13628 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13632 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13634 (vectorizable_mask_load_store): ...here.
13635 (vectorizable_load): ...and here.
13637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13639 * tree-vect-stmts.c (vect_build_all_ones_mask)
13640 (vect_build_zero_merge_argument): New functions, split out from...
13641 (vectorizable_load): ...here.
13643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13645 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13647 (vectorizable_mask_load_store): ...here.
13648 (vectorizable_store): ...and here.
13650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13652 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13654 (vectorizable_mask_load_store): ...here.
13656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13658 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13659 (vect_model_store_cost): Take a vec_load_store_type instead of a
13661 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13662 (vect_model_store_cost): Take a vec_load_store_type instead of a
13664 (vectorizable_mask_load_store): Update accordingly.
13665 (vectorizable_store): Likewise.
13666 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13670 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13671 IFN_MASK_LOAD calls here rather than...
13672 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13675 Alan Hayward <alan.hayward@arm.com>
13676 David Sherwood <david.sherwood@arm.com>
13678 * expmed.c (extract_bit_field_1): For vector extracts,
13679 fall back to extract_bit_field_as_subreg if vec_extract
13682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13683 Alan Hayward <alan.hayward@arm.com>
13684 David Sherwood <david.sherwood@arm.com>
13686 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13687 they are variable or constant sized.
13688 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13689 slots for constant-sized data.
13691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13692 Alan Hayward <alan.hayward@arm.com>
13693 David Sherwood <david.sherwood@arm.com>
13695 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13696 handling COND_EXPRs with boolean comparisons, try to find a better
13697 basis for the mask type than the boolean itself.
13699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13701 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13702 is calculated and how it can be overridden.
13703 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13704 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13706 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13709 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13710 Alan Hayward <alan.hayward@arm.com>
13711 David Sherwood <david.sherwood@arm.com>
13713 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13714 Remove the mode argument.
13715 (aarch64_simd_valid_immediate): Remove the mode and inverse
13717 * config/aarch64/iterators.md (bitsize): New iterator.
13718 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13719 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13720 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13721 aarch64_simd_valid_immediate.
13722 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13723 (aarch64_reg_or_bic_imm): Likewise.
13724 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13725 with an insn_type enum and msl with a modifier_type enum.
13726 Replace element_width with a scalar_mode. Change the shift
13727 to unsigned int. Add constructors for scalar_float_mode and
13728 scalar_int_mode elements.
13729 (aarch64_vect_float_const_representable_p): Delete.
13730 (aarch64_can_const_movi_rtx_p)
13731 (aarch64_simd_scalar_immediate_valid_for_move)
13732 (aarch64_simd_make_constant): Update call to
13733 aarch64_simd_valid_immediate.
13734 (aarch64_advsimd_valid_immediate_hs): New function.
13735 (aarch64_advsimd_valid_immediate): Likewise.
13736 (aarch64_simd_valid_immediate): Remove mode and inverse
13737 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13738 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13739 and aarch64_float_const_representable_p on the result.
13740 (aarch64_output_simd_mov_immediate): Remove mode argument.
13741 Update call to aarch64_simd_valid_immediate and use of
13742 simd_immediate_info.
13743 (aarch64_output_scalar_simd_mov_immediate): Update call
13746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13747 Alan Hayward <alan.hayward@arm.com>
13748 David Sherwood <david.sherwood@arm.com>
13750 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13751 (mode_nunits): Likewise CONST_MODE_NUNITS.
13752 * machmode.def (ADJUST_NUNITS): Document.
13753 * genmodes.c (mode_data::need_nunits_adj): New field.
13754 (blank_mode): Update accordingly.
13755 (adj_nunits): New variable.
13756 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13758 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13759 listed in adj_nunits.
13760 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13761 listed in adj_nunits. Don't emit case statements for such modes.
13762 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13763 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13764 nothing if adj_nunits is nonnull.
13765 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13766 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13767 (emit_mode_fbit): Update use of print_maybe_const_decl.
13768 (emit_move_size): Likewise. Treat the array as non-const
13770 (emit_mode_adjustments): Handle adj_nunits.
13772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13774 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13775 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13776 (VECTOR_MODES): Use it.
13777 (make_vector_modes): Take the prefix as an argument.
13779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13780 Alan Hayward <alan.hayward@arm.com>
13781 David Sherwood <david.sherwood@arm.com>
13783 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13784 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13785 for MODE_VECTOR_BOOL.
13786 * machmode.def (VECTOR_BOOL_MODE): Document.
13787 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13788 (make_vector_bool_mode): New function.
13789 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13791 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13792 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13794 * stor-layout.c (int_mode_for_mode): Likewise.
13795 * tree.c (build_vector_type_for_mode): Likewise.
13796 * varasm.c (output_constant_pool_2): Likewise.
13797 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13798 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13799 for MODE_VECTOR_BOOL.
13800 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13801 of mode class checks.
13802 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13803 instead of a list of mode class checks.
13804 (expand_vector_scalar_condition): Likewise.
13805 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13808 Alan Hayward <alan.hayward@arm.com>
13809 David Sherwood <david.sherwood@arm.com>
13811 * machmode.h (mode_size): Change from unsigned short to
13813 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13814 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13815 or if measurement_type is not polynomial.
13816 (fixed_size_mode::includes_p): Check for constant-sized modes.
13817 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13818 return a poly_uint16 rather than an unsigned short.
13819 (emit_mode_size): Change the type of mode_size from unsigned short
13820 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13821 (emit_mode_adjustments): Cope with polynomial vector sizes.
13822 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13824 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13826 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13827 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13828 * caller-save.c (setup_save_areas): Likewise.
13829 (replace_reg_with_saved_mem): Likewise.
13830 * calls.c (emit_library_call_value_1): Likewise.
13831 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13832 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13833 (gen_lowpart_for_combine): Likewise.
13834 * convert.c (convert_to_integer_1): Likewise.
13835 * cse.c (equiv_constant, cse_insn): Likewise.
13836 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13837 (cselib_subst_to_values): Likewise.
13838 * dce.c (word_dce_process_block): Likewise.
13839 * df-problems.c (df_word_lr_mark_ref): Likewise.
13840 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13841 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13842 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13843 (rtl_for_decl_location): Likewise.
13844 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13845 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13846 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13847 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13848 (expand_expr_real_1): Likewise.
13849 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13850 (pad_below): Likewise.
13851 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13852 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13853 * ira.c (get_subreg_tracking_sizes): Likewise.
13854 * ira-build.c (ira_create_allocno_objects): Likewise.
13855 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13856 (ira_sort_regnos_for_alter_reg): Likewise.
13857 * ira-costs.c (record_operand_costs): Likewise.
13858 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13859 (resolve_simple_move): Likewise.
13860 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13861 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13862 (lra_constraints): Likewise.
13863 (CONST_POOL_OK_P): Reject variable-sized modes.
13864 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13865 (add_pseudo_to_slot, lra_spill): Likewise.
13866 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13867 * optabs-query.c (get_best_extraction_insn): Likewise.
13868 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13869 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13870 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13871 * recog.c (offsettable_address_addr_space_p): Likewise.
13872 * regcprop.c (maybe_mode_change): Likewise.
13873 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13874 * regrename.c (build_def_use): Likewise.
13875 * regstat.c (dump_reg_info): Likewise.
13876 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13877 (find_reloads, find_reloads_subreg_address): Likewise.
13878 * reload1.c (eliminate_regs_1): Likewise.
13879 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13880 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13881 (simplify_binary_operation_1, simplify_subreg): Likewise.
13882 * targhooks.c (default_function_arg_padding): Likewise.
13883 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13884 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13885 (verify_gimple_assign_ternary): Likewise.
13886 * tree-inline.c (estimate_move_cost): Likewise.
13887 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13888 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13889 (get_address_cost_ainc): Likewise.
13890 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13891 (vect_supportable_dr_alignment): Likewise.
13892 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13893 (vectorizable_reduction): Likewise.
13894 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13895 (vectorizable_operation, vectorizable_load): Likewise.
13896 * tree.c (build_same_sized_truth_vector_type): Likewise.
13897 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13898 * var-tracking.c (emit_note_insn_var_location): Likewise.
13899 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13900 (ADDR_VEC_ALIGN): Likewise.
13902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13903 Alan Hayward <alan.hayward@arm.com>
13904 David Sherwood <david.sherwood@arm.com>
13906 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13908 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13909 or if measurement_type is polynomial.
13910 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13911 * combine.c (make_extraction): Likewise.
13912 * dse.c (find_shift_sequence): Likewise.
13913 * dwarf2out.c (mem_loc_descriptor): Likewise.
13914 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13915 (extract_bit_field, extract_low_bits): Likewise.
13916 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13917 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13918 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13919 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13920 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13921 * reload.c (find_reloads): Likewise.
13922 * reload1.c (alter_reg): Likewise.
13923 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13924 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13925 * tree-if-conv.c (predicate_mem_writes): Likewise.
13926 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13927 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13928 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13929 * valtrack.c (dead_debug_insert_temp): Likewise.
13930 * varasm.c (mergeable_constant_section): Likewise.
13931 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13934 Alan Hayward <alan.hayward@arm.com>
13935 David Sherwood <david.sherwood@arm.com>
13937 * expr.c (expand_assignment): Cope with polynomial mode sizes
13938 when assigning to a CONCAT.
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_precision): Change from unsigned short to
13946 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13948 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13949 or if measurement_type is not polynomial.
13950 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13951 in which the mode is already known to be a scalar_int_mode.
13952 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13953 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13955 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13956 for GET_MODE_PRECISION.
13957 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13958 for GET_MODE_PRECISION.
13959 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13961 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13962 (expand_field_assignment, make_extraction): Likewise.
13963 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13964 (get_last_value): Likewise.
13965 * convert.c (convert_to_integer_1): Likewise.
13966 * cse.c (cse_insn): Likewise.
13967 * expr.c (expand_expr_real_1): Likewise.
13968 * lra-constraints.c (simplify_operand_subreg): Likewise.
13969 * optabs-query.c (can_atomic_load_p): Likewise.
13970 * optabs.c (expand_atomic_load): Likewise.
13971 (expand_atomic_store): Likewise.
13972 * ree.c (combine_reaching_defs): Likewise.
13973 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13974 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13975 * tree.h (type_has_mode_precision_p): Likewise.
13976 * ubsan.c (instrument_si_overflow): Likewise.
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 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13983 polynomial numbers of units.
13984 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13985 (valid_vector_subparts_p): New function.
13986 (build_vector_type): Remove temporary shim and take the number
13987 of units as a poly_uint64 rather than an int.
13988 (build_opaque_vector_type): Take the number of units as a
13989 poly_uint64 rather than an int.
13990 * tree.c (build_vector_from_ctor): Handle polynomial
13991 TYPE_VECTOR_SUBPARTS.
13992 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13993 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13994 (build_vector_from_val): If the number of units is variable,
13995 use build_vec_duplicate_cst for constant operands and
13996 VEC_DUPLICATE_EXPR otherwise.
13997 (make_vector_type): Remove temporary is_constant ().
13998 (build_vector_type, build_opaque_vector_type): Take the number of
13999 units as a poly_uint64 rather than an int.
14000 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
14002 * cfgexpand.c (expand_debug_expr): Likewise.
14003 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
14004 (store_constructor, expand_expr_real_1): Likewise.
14005 (const_scalar_mask_from_tree): Likewise.
14006 * fold-const-call.c (fold_const_reduction): Likewise.
14007 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
14008 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
14009 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
14010 (fold_relational_const): Likewise.
14011 (native_interpret_vector): Likewise. Change the size from an
14012 int to an unsigned int.
14013 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
14014 TYPE_VECTOR_SUBPARTS.
14015 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
14016 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
14017 duplicating a non-constant operand into a variable-length vector.
14018 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
14019 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
14020 * ipa-icf.c (sem_variable::equals): Likewise.
14021 * match.pd: Likewise.
14022 * omp-simd-clone.c (simd_clone_subparts): Likewise.
14023 * print-tree.c (print_node): Likewise.
14024 * stor-layout.c (layout_type): Likewise.
14025 * targhooks.c (default_builtin_vectorization_cost): Likewise.
14026 * tree-cfg.c (verify_gimple_comparison): Likewise.
14027 (verify_gimple_assign_binary): Likewise.
14028 (verify_gimple_assign_ternary): Likewise.
14029 (verify_gimple_assign_single): Likewise.
14030 * tree-pretty-print.c (dump_generic_node): Likewise.
14031 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14032 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
14033 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
14034 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
14035 (vect_shift_permute_load_chain): Likewise.
14036 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
14037 (expand_vector_condition, optimize_vector_constructor): Likewise.
14038 (lower_vec_perm, get_compute_type): Likewise.
14039 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14040 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
14041 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
14042 (vect_recog_mask_conversion_pattern): Likewise.
14043 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
14044 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
14045 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14046 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
14047 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
14048 (vectorizable_shift, vectorizable_operation, vectorizable_store)
14049 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14050 (supportable_widening_operation): Likewise.
14051 (supportable_narrowing_operation): Likewise.
14052 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14054 * varasm.c (output_constant): Likewise.
14056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14057 Alan Hayward <alan.hayward@arm.com>
14058 David Sherwood <david.sherwood@arm.com>
14060 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14061 so that both the length == 3 and length != 3 cases set up their
14062 own permute vectors. Add comments explaining why we know the
14063 number of elements is constant.
14064 (vect_permute_load_chain): Likewise.
14066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14067 Alan Hayward <alan.hayward@arm.com>
14068 David Sherwood <david.sherwood@arm.com>
14070 * machmode.h (mode_nunits): Change from unsigned char to
14072 (ONLY_FIXED_SIZE_MODES): New macro.
14073 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14074 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14075 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14077 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14078 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14079 or if measurement_type is not polynomial.
14080 * genmodes.c (ZERO_COEFFS): New macro.
14081 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14083 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14084 Use ZERO_COEFFS when emitting initializers.
14085 * data-streamer.h (bp_pack_poly_value): New function.
14086 (bp_unpack_poly_value): Likewise.
14087 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14088 for GET_MODE_NUNITS.
14089 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14090 for GET_MODE_NUNITS.
14091 * tree.c (make_vector_type): Remove temporary shim and make
14092 the real function take the number of units as a poly_uint64
14093 rather than an int.
14094 (build_vector_type_for_mode): Handle polynomial nunits.
14095 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14096 * emit-rtl.c (const_vec_series_p_1): Likewise.
14097 (gen_rtx_CONST_VECTOR): Likewise.
14098 * fold-const.c (test_vec_duplicate_folding): Likewise.
14099 * genrecog.c (validate_pattern): Likewise.
14100 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14101 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14102 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14103 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14104 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14105 * rtlanal.c (subreg_get_info): Likewise.
14106 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14107 (vect_grouped_load_supported): Likewise.
14108 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14109 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14110 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14111 (simplify_const_unary_operation, simplify_binary_operation_1)
14112 (simplify_const_binary_operation, simplify_ternary_operation)
14113 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14114 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14115 instead of CONST_VECTOR_NUNITS.
14116 * varasm.c (output_constant_pool_2): Likewise.
14117 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14118 explicit-encoded elements in the XVEC for variable-length vectors.
14120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14122 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14125 Alan Hayward <alan.hayward@arm.com>
14126 David Sherwood <david.sherwood@arm.com>
14128 * coretypes.h (fixed_size_mode): Declare.
14129 (fixed_size_mode_pod): New typedef.
14130 * builtins.h (target_builtins::x_apply_args_mode)
14131 (target_builtins::x_apply_result_mode): Change type to
14132 fixed_size_mode_pod.
14133 * builtins.c (apply_args_size, apply_result_size, result_vector)
14134 (expand_builtin_apply_args_1, expand_builtin_apply)
14135 (expand_builtin_return): Update accordingly.
14137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14139 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14140 * cselib.c (cselib_hash_rtx): Likewise.
14141 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14142 CONST_VECTOR encoding.
14144 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14145 Jeff Law <law@redhat.com>
14148 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14149 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14150 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14151 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14154 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14155 explicitly probe *sp in a noreturn function if there were any callee
14156 register saves or frame pointer is needed.
14158 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14161 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14162 BLKmode for ternary, binary or unary expressions.
14165 * var-tracking.c (delete_vta_debug_insn): New inline function.
14166 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14167 insns from get_insns () to NULL instead of each bb separately.
14168 Use delete_vta_debug_insn. No longer static.
14169 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14170 delete_vta_debug_insns callers.
14171 * rtl.h (delete_vta_debug_insns): Declare.
14172 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14173 instead of variable_tracking_main.
14175 2018-01-03 Martin Sebor <msebor@redhat.com>
14177 PR tree-optimization/83603
14178 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14179 arguments past the endof the argument list in functions declared
14180 without a prototype.
14181 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14182 Avoid checking when arguments are null.
14184 2018-01-03 Martin Sebor <msebor@redhat.com>
14187 * doc/extend.texi (attribute const): Fix a typo.
14188 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14189 issuing -Wsuggest-attribute for void functions.
14191 2018-01-03 Martin Sebor <msebor@redhat.com>
14193 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14194 offset_int::from instead of wide_int::to_shwi.
14195 (maybe_diag_overlap): Remove assertion.
14196 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14197 * gimple-ssa-sprintf.c (format_directive): Same.
14198 (parse_directive): Same.
14199 (sprintf_dom_walker::compute_format_length): Same.
14200 (try_substitute_return_value): Same.
14202 2018-01-03 Jeff Law <law@redhat.com>
14204 PR middle-end/83654
14205 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14206 non-constant residual for zero at runtime and avoid probing in
14207 that case. Reorganize code for trailing problem to mirror handling
14210 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14212 PR tree-optimization/83501
14213 * tree-ssa-strlen.c (get_string_cst): New.
14214 (handle_char_store): Call get_string_cst.
14216 2018-01-03 Martin Liska <mliska@suse.cz>
14218 PR tree-optimization/83593
14219 * tree-ssa-strlen.c: Include tree-cfg.h.
14220 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14221 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14222 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14224 (strlen_dom_walker::before_dom_children): Call
14225 gimple_purge_dead_eh_edges. Dump tranformation with details
14227 (strlen_dom_walker::before_dom_children): Update call by adding
14228 new argument cleanup_eh.
14229 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14231 2018-01-03 Martin Liska <mliska@suse.cz>
14234 * cif-code.def (VARIADIC_THUNK): New enum value.
14235 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14238 2018-01-03 Jan Beulich <jbeulich@suse.com>
14240 * sse.md (mov<mode>_internal): Tighten condition for when to use
14241 vmovdqu<ssescalarsize> for TI and OI modes.
14243 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14245 Update copyright years.
14247 2018-01-03 Martin Liska <mliska@suse.cz>
14250 * ipa-visibility.c (function_and_variable_visibility): Skip
14251 functions with noipa attribure.
14253 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14255 * gcc.c (process_command): Update copyright notice dates.
14256 * gcov-dump.c (print_version): Ditto.
14257 * gcov.c (print_version): Ditto.
14258 * gcov-tool.c (print_version): Ditto.
14259 * gengtype.c (create_file): Ditto.
14260 * doc/cpp.texi: Bump @copying's copyright year.
14261 * doc/cppinternals.texi: Ditto.
14262 * doc/gcc.texi: Ditto.
14263 * doc/gccint.texi: Ditto.
14264 * doc/gcov.texi: Ditto.
14265 * doc/install.texi: Ditto.
14266 * doc/invoke.texi: Ditto.
14268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14270 * vector-builder.h (vector_builder::m_full_nelts): Change from
14271 unsigned int to poly_uint64.
14272 (vector_builder::full_nelts): Update prototype accordingly.
14273 (vector_builder::new_vector): Likewise.
14274 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14275 (vector_builder::operator ==): Likewise.
14276 (vector_builder::finalize): Likewise.
14277 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14278 Take the number of elements as a poly_uint64 rather than an
14280 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14281 from unsigned int to poly_uint64.
14282 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14283 (vec_perm_indices::new_vector): Likewise.
14284 (vec_perm_indices::length): Likewise.
14285 (vec_perm_indices::nelts_per_input): Likewise.
14286 (vec_perm_indices::input_nelts): Likewise.
14287 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14288 number of elements per input as a poly_uint64 rather than an
14289 unsigned int. Use the original encoding for variable-length
14290 vectors, rather than clamping each individual element.
14291 For the second and subsequent elements in each pattern,
14292 clamp the step and base before clamping their sum.
14293 (vec_perm_indices::series_p): Handle polynomial element counts.
14294 (vec_perm_indices::all_in_range_p): Likewise.
14295 (vec_perm_indices_to_tree): Likewise.
14296 (vec_perm_indices_to_rtx): Likewise.
14297 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14298 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14299 (tree_vector_builder::new_binary_operation): Handle polynomial
14300 element counts. Return false if we need to know the number
14301 of elements at compile time.
14302 * fold-const.c (fold_vec_perm): Punt if the number of elements
14303 isn't known at compile time.
14305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14307 * vec-perm-indices.h (vec_perm_builder): Change element type
14308 from HOST_WIDE_INT to poly_int64.
14309 (vec_perm_indices::element_type): Update accordingly.
14310 (vec_perm_indices::clamp): Handle polynomial element_types.
14311 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14312 (vec_perm_indices::all_in_range_p): Likewise.
14313 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14315 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14316 polynomial vec_perm_indices element types.
14317 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14318 * fold-const.c (fold_vec_perm): Likewise.
14319 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14320 * tree-vect-generic.c (lower_vec_perm): Likewise.
14321 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14322 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14323 element type to HOST_WIDE_INT.
14325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14326 Alan Hayward <alan.hayward@arm.com>
14327 David Sherwood <david.sherwood@arm.com>
14329 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14330 rather than an int. Use plus_constant.
14331 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14332 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14335 Alan Hayward <alan.hayward@arm.com>
14336 David Sherwood <david.sherwood@arm.com>
14338 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14339 a HOST_WIDE_INT to a poly_int64.
14341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14342 Alan Hayward <alan.hayward@arm.com>
14343 David Sherwood <david.sherwood@arm.com>
14345 * calls.c (load_register_parameters): Cope with polynomial
14346 mode sizes. Require a constant size for BLKmode parameters
14347 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14348 forces a parameter to be padded at the lsb end in order to
14349 fill a complete number of words, require the parameter size
14350 to be ordered wrt UNITS_PER_WORD.
14352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14353 Alan Hayward <alan.hayward@arm.com>
14354 David Sherwood <david.sherwood@arm.com>
14356 * reload1.c (spill_stack_slot_width): Change element type
14357 from unsigned int to poly_uint64_pod.
14358 (alter_reg): Treat mode sizes as polynomial.
14360 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14361 Alan Hayward <alan.hayward@arm.com>
14362 David Sherwood <david.sherwood@arm.com>
14364 * reload.c (complex_word_subreg_p): New function.
14365 (reload_inner_reg_of_subreg, push_reload): Use it.
14367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14368 Alan Hayward <alan.hayward@arm.com>
14369 David Sherwood <david.sherwood@arm.com>
14371 * lra-constraints.c (process_alt_operands): Reject matched
14372 operands whose sizes aren't ordered.
14373 (match_reload): Refer to this check here.
14375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14376 Alan Hayward <alan.hayward@arm.com>
14377 David Sherwood <david.sherwood@arm.com>
14379 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14380 that the mode size is in the set {1, 2, 4, 8, 16}.
14382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14383 Alan Hayward <alan.hayward@arm.com>
14384 David Sherwood <david.sherwood@arm.com>
14386 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14387 Use plus_constant instead of gen_rtx_PLUS.
14389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14390 Alan Hayward <alan.hayward@arm.com>
14391 David Sherwood <david.sherwood@arm.com>
14393 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14394 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14395 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14396 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14397 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14398 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14399 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14400 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14401 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14402 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14404 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14405 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14406 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14407 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14408 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14409 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14410 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14411 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14412 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14413 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14415 * expr.c (emit_move_resolve_push): Treat the input and result
14416 of PUSH_ROUNDING as a poly_int64.
14417 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14418 (emit_push_insn): Likewise.
14419 * lra-eliminations.c (mark_not_eliminable): Likewise.
14420 * recog.c (push_operand): Likewise.
14421 * reload1.c (elimination_effects): Likewise.
14422 * rtlanal.c (nonzero_bits1): Likewise.
14423 * calls.c (store_one_arg): Likewise. Require the padding to be
14424 known at compile time.
14426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14427 Alan Hayward <alan.hayward@arm.com>
14428 David Sherwood <david.sherwood@arm.com>
14430 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14431 Use plus_constant instead of gen_rtx_PLUS.
14433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14434 Alan Hayward <alan.hayward@arm.com>
14435 David Sherwood <david.sherwood@arm.com>
14437 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14438 rather than an int.
14440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14441 Alan Hayward <alan.hayward@arm.com>
14442 David Sherwood <david.sherwood@arm.com>
14444 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14445 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14446 via stack temporaries. Treat the mode size as polynomial too.
14448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14449 Alan Hayward <alan.hayward@arm.com>
14450 David Sherwood <david.sherwood@arm.com>
14452 * expr.c (expand_expr_real_2): When handling conversions involving
14453 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14454 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14455 as a poly_uint64 too.
14457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14458 Alan Hayward <alan.hayward@arm.com>
14459 David Sherwood <david.sherwood@arm.com>
14461 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14464 Alan Hayward <alan.hayward@arm.com>
14465 David Sherwood <david.sherwood@arm.com>
14467 * combine.c (can_change_dest_mode): Handle polynomial
14468 REGMODE_NATURAL_SIZE.
14469 * expmed.c (store_bit_field_1): Likewise.
14470 * expr.c (store_constructor): Likewise.
14471 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14472 and polynomial REGMODE_NATURAL_SIZE.
14473 (gen_lowpart_common): Likewise.
14474 * reginfo.c (record_subregs_of_mode): Likewise.
14475 * rtlanal.c (read_modify_subreg_p): Likewise.
14477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14478 Alan Hayward <alan.hayward@arm.com>
14479 David Sherwood <david.sherwood@arm.com>
14481 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14482 numbers of elements.
14484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14485 Alan Hayward <alan.hayward@arm.com>
14486 David Sherwood <david.sherwood@arm.com>
14488 * match.pd: Cope with polynomial numbers of vector elements.
14490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14491 Alan Hayward <alan.hayward@arm.com>
14492 David Sherwood <david.sherwood@arm.com>
14494 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14495 in a POINTER_PLUS_EXPR.
14497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14498 Alan Hayward <alan.hayward@arm.com>
14499 David Sherwood <david.sherwood@arm.com>
14501 * omp-simd-clone.c (simd_clone_subparts): New function.
14502 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14503 (ipa_simd_modify_function_body): Likewise.
14505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14506 Alan Hayward <alan.hayward@arm.com>
14507 David Sherwood <david.sherwood@arm.com>
14509 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14510 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14511 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14512 (expand_vector_condition, vector_element): Likewise.
14513 (subparts_gt): New function.
14514 (get_compute_type): Use subparts_gt.
14515 (count_type_subparts): Delete.
14516 (expand_vector_operations_1): Use subparts_gt instead of
14517 count_type_subparts.
14519 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14520 Alan Hayward <alan.hayward@arm.com>
14521 David Sherwood <david.sherwood@arm.com>
14523 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14524 (vect_compile_time_alias): ...this new function. Do the calculation
14525 on poly_ints rather than trees.
14526 (vect_prune_runtime_alias_test_list): Update call accordingly.
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 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14534 (vect_schedule_slp_instance): Likewise.
14536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14537 Alan Hayward <alan.hayward@arm.com>
14538 David Sherwood <david.sherwood@arm.com>
14540 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14541 constant and extern definitions for variable-length vectors.
14542 (vect_get_constant_vectors): Note that the number of units
14543 is known to be constant.
14545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14546 Alan Hayward <alan.hayward@arm.com>
14547 David Sherwood <david.sherwood@arm.com>
14549 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14550 of units as polynomial. Choose between WIDE and NARROW based
14553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14554 Alan Hayward <alan.hayward@arm.com>
14555 David Sherwood <david.sherwood@arm.com>
14557 * tree-vect-stmts.c (simd_clone_subparts): New function.
14558 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14561 Alan Hayward <alan.hayward@arm.com>
14562 David Sherwood <david.sherwood@arm.com>
14564 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14565 vectors as polynomial. Use build_index_vector for
14566 IFN_GOMP_SIMD_LANE.
14568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14569 Alan Hayward <alan.hayward@arm.com>
14570 David Sherwood <david.sherwood@arm.com>
14572 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14573 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14574 for variable-length vectors.
14575 (vectorizable_mask_load_store): Treat the number of units as
14576 polynomial, asserting that it is constant if the condition has
14577 already been enforced.
14578 (vectorizable_store, vectorizable_load): Likewise.
14580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14581 Alan Hayward <alan.hayward@arm.com>
14582 David Sherwood <david.sherwood@arm.com>
14584 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14585 of units as polynomial. Punt if we can't tell at compile time
14586 which vector contains the final result.
14588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14589 Alan Hayward <alan.hayward@arm.com>
14590 David Sherwood <david.sherwood@arm.com>
14592 * tree-vect-loop.c (vectorizable_induction): Treat the number
14593 of units as polynomial. Punt on SLP inductions. Use an integer
14594 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14595 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
14603 * tree.c (build_index_vector): New function.
14604 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14605 of units as polynomial, forcibly converting it to a constant if
14606 vectorizable_reduction has already enforced the condition.
14607 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14608 to create a {1,2,3,...} vector.
14609 (vectorizable_reduction): Treat the number of units as polynomial.
14610 Choose vectype_in based on the largest scalar element size rather
14611 than the smallest number of units. Enforce the restrictions
14614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14615 Alan Hayward <alan.hayward@arm.com>
14616 David Sherwood <david.sherwood@arm.com>
14618 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14619 number of units as polynomial.
14621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14622 Alan Hayward <alan.hayward@arm.com>
14623 David Sherwood <david.sherwood@arm.com>
14625 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14626 * target.def (autovectorize_vector_sizes): Return the vector sizes
14627 by pointer, using vector_sizes rather than a bitmask.
14628 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14629 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14630 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14632 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14633 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14634 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14635 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14636 * omp-general.c (omp_max_vf): Likewise.
14637 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14638 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14639 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14640 * tree-vect-slp.c (vect_slp_bb): Likewise.
14641 * doc/tm.texi: Regenerate.
14642 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14644 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14645 the vector size as a poly_uint64 rather than an unsigned int.
14646 (current_vector_size): Change from an unsigned int to a poly_uint64.
14647 (get_vectype_for_scalar_type): Update accordingly.
14648 * tree.h (build_truth_vector_type): Take the size and number of
14649 units as a poly_uint64 rather than an unsigned int.
14650 (build_vector_type): Add a temporary overload that takes
14651 the number of units as a poly_uint64 rather than an unsigned int.
14652 * tree.c (make_vector_type): Likewise.
14653 (build_truth_vector_type): Take the number of units as a poly_uint64
14654 rather than an unsigned int.
14656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14657 Alan Hayward <alan.hayward@arm.com>
14658 David Sherwood <david.sherwood@arm.com>
14660 * target.def (get_mask_mode): Take the number of units and length
14661 as poly_uint64s rather than unsigned ints.
14662 * targhooks.h (default_get_mask_mode): Update accordingly.
14663 * targhooks.c (default_get_mask_mode): Likewise.
14664 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14665 * doc/tm.texi: Regenerate.
14667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14668 Alan Hayward <alan.hayward@arm.com>
14669 David Sherwood <david.sherwood@arm.com>
14671 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14672 * omp-general.c (omp_max_vf): Likewise.
14673 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14674 (expand_omp_simd): Handle polynomial safelen.
14675 * omp-low.c (omplow_simd_context): Add a default constructor.
14676 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14677 (lower_rec_simd_input_clauses): Update accordingly.
14678 (lower_rec_input_clauses): Likewise.
14680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14681 Alan Hayward <alan.hayward@arm.com>
14682 David Sherwood <david.sherwood@arm.com>
14684 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14685 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14686 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14687 (vect_analyze_slp_cost): Likewise.
14688 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14689 (vect_model_load_cost): Likewise.
14691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14692 Alan Hayward <alan.hayward@arm.com>
14693 David Sherwood <david.sherwood@arm.com>
14695 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14696 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14697 from an unsigned int * to a poly_uint64_pod *.
14698 (calculate_unrolling_factor): New function.
14699 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14702 Alan Hayward <alan.hayward@arm.com>
14703 David Sherwood <david.sherwood@arm.com>
14705 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14706 from an unsigned int to a poly_uint64.
14707 (_loop_vec_info::slp_unrolling_factor): Likewise.
14708 (_loop_vec_info::vectorization_factor): Change from an int
14710 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14711 (vect_get_num_vectors): New function.
14712 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14713 (vect_get_num_copies): Use vect_get_num_vectors.
14714 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14715 to an unsigned int *.
14716 (vect_analyze_data_refs): Change min_vf from an int * to a
14718 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14719 than an unsigned HOST_WIDE_INT.
14720 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14721 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14722 to an unsigned int *.
14723 (vect_analyze_data_ref_dependences): Likewise.
14724 (vect_compute_data_ref_alignment): Handle polynomial vf.
14725 (vect_enhance_data_refs_alignment): Likewise.
14726 (vect_prune_runtime_alias_test_list): Likewise.
14727 (vect_shift_permute_load_chain): Likewise.
14728 (vect_supportable_dr_alignment): Likewise.
14729 (dependence_distance_ge_vf): Take the vectorization factor as a
14730 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14731 (vect_analyze_data_refs): Change min_vf from an int * to a
14733 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14734 vfm1 as a poly_uint64 rather than an int. Make the same change
14735 for the returned bound_scalar.
14736 (vect_gen_vector_loop_niters): Handle polynomial vf.
14737 (vect_do_peeling): Likewise. Update call to
14738 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14739 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14741 * tree-vect-loop.c (vect_determine_vectorization_factor)
14742 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14743 (vect_get_known_peeling_cost): Likewise.
14744 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14745 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14746 (vect_transform_loop): Likewise. Use the lowest possible VF when
14747 updating the upper bounds of the loop.
14748 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14749 rather than an int.
14750 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14751 polynomial unroll factors.
14752 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14753 (vect_make_slp_decision): Likewise.
14754 (vect_supported_load_permutation_p): Likewise, and polynomial
14756 (vect_analyze_slp_cost): Handle polynomial vf.
14757 (vect_slp_analyze_node_operations): Likewise.
14758 (vect_slp_analyze_bb_1): Likewise.
14759 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14760 than an unsigned HOST_WIDE_INT.
14761 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14762 (vectorizable_load): Handle polynomial vf.
14763 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14765 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14768 Alan Hayward <alan.hayward@arm.com>
14769 David Sherwood <david.sherwood@arm.com>
14771 * match.pd: Handle bit operations involving three constants
14772 and try to fold one pair.
14774 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14776 * tree-vect-loop-manip.c: Include gimple-fold.h.
14777 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14778 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14779 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14780 Add a path that uses a step of VF instead of 1, but disable it
14782 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14783 and niters_no_overflow parameters. Update calls to
14784 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14785 Create a new SSA name if the latter choses to use a ste other
14786 than zero, and return it via niters_vector_mult_vf_var.
14787 * tree-vect-loop.c (vect_transform_loop): Update calls to
14788 vect_do_peeling, vect_gen_vector_loop_niters and
14789 slpeel_make_loop_iterate_ntimes.
14790 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14791 (vect_gen_vector_loop_niters): Update declarations after above changes.
14793 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14795 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14796 128-bit round to integer instructions.
14797 (ceil<mode>2): Likewise.
14798 (btrunc<mode>2): Likewise.
14799 (round<mode>2): Likewise.
14801 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14803 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14804 unaligned VSX load/store on P8/P9.
14805 (expand_block_clear): Allow the use of unaligned VSX
14806 load/store on P8/P9.
14808 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14810 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14812 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14813 swap associated with both a load and a store.
14815 2018-01-02 Andrew Waterman <andrew@sifive.com>
14817 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14818 * config/riscv/riscv.md (clear_cache): Use it.
14820 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14822 * web.c: Remove out-of-date comment.
14824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14826 * expr.c (fixup_args_size_notes): Check that any existing
14827 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14828 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14829 (emit_single_push_insn): ...here.
14831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14833 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14834 (const_vector_encoded_nelts): New function.
14835 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14836 (const_vector_int_elt, const_vector_elt): Declare.
14837 * emit-rtl.c (const_vector_int_elt_1): New function.
14838 (const_vector_elt): Likewise.
14839 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14840 of CONST_VECTOR_ELT.
14842 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14844 * expr.c: Include rtx-vector-builder.h.
14845 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14846 directly on the tree encoding.
14847 (const_vector_from_tree): Likewise.
14848 * optabs.c: Include rtx-vector-builder.h.
14849 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14850 sequence of "u" values.
14851 * vec-perm-indices.c: Include rtx-vector-builder.h.
14852 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14853 directly on the vec_perm_indices encoding.
14855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14857 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14858 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14859 * rtx-vector-builder.h: New file.
14860 * rtx-vector-builder.c: Likewise.
14861 * rtl.h (rtx_def::u2): Add a const_vector field.
14862 (CONST_VECTOR_NPATTERNS): New macro.
14863 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14864 (CONST_VECTOR_DUPLICATE_P): Likewise.
14865 (CONST_VECTOR_STEPPED_P): Likewise.
14866 (CONST_VECTOR_ENCODED_ELT): Likewise.
14867 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14868 (unwrap_const_vec_duplicate): Likewise.
14869 (const_vec_series_p): Check for a non-duplicated vector encoding.
14870 Say that the function only returns true for integer vectors.
14871 * emit-rtl.c: Include rtx-vector-builder.h.
14872 (gen_const_vec_duplicate_1): Delete.
14873 (gen_const_vector): Call gen_const_vec_duplicate instead of
14874 gen_const_vec_duplicate_1.
14875 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14876 (gen_const_vec_duplicate): Use rtx_vector_builder.
14877 (gen_const_vec_series): Likewise.
14878 (gen_rtx_CONST_VECTOR): Likewise.
14879 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14880 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14881 Build a new vector rather than modifying a CONST_VECTOR in-place.
14882 (handle_special_swappables): Update call accordingly.
14883 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14884 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14885 Build a new vector rather than modifying a CONST_VECTOR in-place.
14886 (handle_special_swappables): Update call accordingly.
14888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14890 * simplify-rtx.c (simplify_const_binary_operation): Use
14891 CONST_VECTOR_ELT instead of XVECEXP.
14893 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14895 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14896 the selector elements to be different from the data elements
14897 if the selector is a VECTOR_CST.
14898 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14899 ssizetype for the selector.
14901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14903 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14904 before testing each element individually.
14905 * tree-vect-generic.c (lower_vec_perm): Likewise.
14907 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14909 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14910 * selftest-run-tests.c (selftest::run_tests): Call it.
14911 * vector-builder.h (vector_builder::operator ==): New function.
14912 (vector_builder::operator !=): Likewise.
14913 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14914 (vec_perm_indices::all_from_input_p): New function.
14915 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14916 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14917 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14918 instead of reading the VECTOR_CST directly. Detect whether both
14919 vector inputs are the same before constructing the vec_perm_indices,
14920 and update the number of inputs argument accordingly. Use the
14921 utility functions added above. Only construct sel2 if we need to.
14923 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14925 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14926 the broadcast of the low byte.
14927 (expand_mult_highpart): Use an explicit encoding for the permutes.
14928 * optabs-query.c (can_mult_highpart_p): Likewise.
14929 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14930 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14931 (vectorizable_bswap): Likewise.
14932 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14933 explicit encoding for the power-of-2 permutes.
14934 (vect_permute_store_chain): Likewise.
14935 (vect_grouped_load_supported): Likewise.
14936 (vect_permute_load_chain): Likewise.
14938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14940 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14941 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14942 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14943 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14944 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14945 (vect_gen_perm_mask_any): Likewise.
14947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14949 * int-vector-builder.h: New file.
14950 * vec-perm-indices.h: Include int-vector-builder.h.
14951 (vec_perm_indices): Redefine as an int_vector_builder.
14952 (auto_vec_perm_indices): Delete.
14953 (vec_perm_builder): Redefine as a stand-alone class.
14954 (vec_perm_indices::vec_perm_indices): New function.
14955 (vec_perm_indices::clamp): Likewise.
14956 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14957 (vec_perm_indices::new_vector): New function.
14958 (vec_perm_indices::new_expanded_vector): Update for new
14959 vec_perm_indices class.
14960 (vec_perm_indices::rotate_inputs): New function.
14961 (vec_perm_indices::all_in_range_p): Operate directly on the
14962 encoded form, without computing elided elements.
14963 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14964 encoding. Update for new vec_perm_indices class.
14965 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14966 the given vec_perm_builder.
14967 (expand_vec_perm_var): Update vec_perm_builder constructor.
14968 (expand_mult_highpart): Use vec_perm_builder instead of
14969 auto_vec_perm_indices.
14970 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14971 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14972 or double series encoding as appropriate.
14973 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14974 vec_perm_indices instead of auto_vec_perm_indices.
14975 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14976 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14977 (vect_permute_store_chain): Likewise.
14978 (vect_grouped_load_supported): Likewise.
14979 (vect_permute_load_chain): Likewise.
14980 (vect_shift_permute_load_chain): Likewise.
14981 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14982 (vect_transform_slp_perm_load): Likewise.
14983 (vect_schedule_slp_instance): Likewise.
14984 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14985 (vectorizable_mask_load_store): Likewise.
14986 (vectorizable_bswap): Likewise.
14987 (vectorizable_store): Likewise.
14988 (vectorizable_load): Likewise.
14989 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14990 vec_perm_indices instead of auto_vec_perm_indices. Use
14991 tree_to_vec_perm_builder to read the vector from a tree.
14992 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14993 vec_perm_builder instead of a vec_perm_indices.
14994 (have_whole_vector_shift): Use vec_perm_builder and
14995 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14996 truncation to calc_vec_perm_mask_for_shift.
14997 (vect_create_epilog_for_reduction): Likewise.
14998 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14999 from auto_vec_perm_indices to vec_perm_indices.
15000 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15001 instead of changing individual elements.
15002 (aarch64_vectorize_vec_perm_const): Use new_vector to install
15003 the vector in d.perm.
15004 * config/arm/arm.c (expand_vec_perm_d::perm): Change
15005 from auto_vec_perm_indices to vec_perm_indices.
15006 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
15007 instead of changing individual elements.
15008 (arm_vectorize_vec_perm_const): Use new_vector to install
15009 the vector in d.perm.
15010 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
15011 Update vec_perm_builder constructor.
15012 (rs6000_expand_interleave): Likewise.
15013 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
15014 (rs6000_expand_interleave): Likewise.
15016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15018 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
15019 to qimode could truncate the indices.
15020 * optabs.c (expand_vec_perm_var): Likewise.
15022 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15024 * Makefile.in (OBJS): Add vec-perm-indices.o.
15025 * vec-perm-indices.h: New file.
15026 * vec-perm-indices.c: Likewise.
15027 * target.h (vec_perm_indices): Replace with a forward class
15029 (auto_vec_perm_indices): Move to vec-perm-indices.h.
15030 * optabs.h: Include vec-perm-indices.h.
15031 (expand_vec_perm): Delete.
15032 (selector_fits_mode_p, expand_vec_perm_var): Declare.
15033 (expand_vec_perm_const): Declare.
15034 * target.def (vec_perm_const_ok): Replace with...
15035 (vec_perm_const): ...this new hook.
15036 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
15037 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
15038 * doc/tm.texi: Regenerate.
15039 * optabs.def (vec_perm_const): Delete.
15040 * doc/md.texi (vec_perm_const): Likewise.
15041 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
15042 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
15043 expand_vec_perm for constant permutation vectors. Assert that
15044 the mode of variable permutation vectors is the integer equivalent
15045 of the mode that is being permuted.
15046 * optabs-query.h (selector_fits_mode_p): Declare.
15047 * optabs-query.c: Include vec-perm-indices.h.
15048 (selector_fits_mode_p): New function.
15049 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15050 is defined, instead of checking whether the vec_perm_const_optab
15051 exists. Use targetm.vectorize.vec_perm_const instead of
15052 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15053 fit in the vector mode before using a variable permute.
15054 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15055 vec_perm_indices instead of an rtx.
15056 (expand_vec_perm): Replace with...
15057 (expand_vec_perm_const): ...this new function. Take the selector
15058 as a vec_perm_indices rather than an rtx. Also take the mode of
15059 the selector. Update call to shift_amt_for_vec_perm_mask.
15060 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15061 Use vec_perm_indices::new_expanded_vector to expand the original
15062 selector into bytes. Check whether the indices fit in the vector
15063 mode before using a variable permute.
15064 (expand_vec_perm_var): Make global.
15065 (expand_mult_highpart): Use expand_vec_perm_const.
15066 * fold-const.c: Includes vec-perm-indices.h.
15067 * tree-ssa-forwprop.c: Likewise.
15068 * tree-vect-data-refs.c: Likewise.
15069 * tree-vect-generic.c: Likewise.
15070 * tree-vect-loop.c: Likewise.
15071 * tree-vect-slp.c: Likewise.
15072 * tree-vect-stmts.c: Likewise.
15073 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15075 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15076 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15077 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15078 (aarch64_vectorize_vec_perm_const): ...this new function.
15079 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15080 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15081 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15082 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15083 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15084 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15085 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15087 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15088 check for NEON modes.
15089 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15090 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15091 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15092 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15094 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15095 the old VEC_PERM_CONST conditions.
15096 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15097 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15098 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15099 (ia64_vectorize_vec_perm_const_ok): Merge into...
15100 (ia64_vectorize_vec_perm_const): ...this new function.
15101 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15102 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15103 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15104 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15105 * config/mips/mips.c (mips_expand_vec_perm_const)
15106 (mips_vectorize_vec_perm_const_ok): Merge into...
15107 (mips_vectorize_vec_perm_const): ...this new function.
15108 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15109 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15110 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15111 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15112 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15113 (rs6000_expand_vec_perm_const): Delete.
15114 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15116 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15117 (altivec_expand_vec_perm_const_le): Take each operand individually.
15118 Operate on constant selectors rather than rtxes.
15119 (altivec_expand_vec_perm_const): Likewise. Update call to
15120 altivec_expand_vec_perm_const_le.
15121 (rs6000_expand_vec_perm_const): Delete.
15122 (rs6000_vectorize_vec_perm_const_ok): Delete.
15123 (rs6000_vectorize_vec_perm_const): New function.
15124 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15125 an element count and rtx array.
15126 (rs6000_expand_extract_even): Update call accordingly.
15127 (rs6000_expand_interleave): Likewise.
15128 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15129 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15130 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15131 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15132 (rs6000_expand_vec_perm_const): Delete.
15133 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15134 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15135 (altivec_expand_vec_perm_const_le): Take each operand individually.
15136 Operate on constant selectors rather than rtxes.
15137 (altivec_expand_vec_perm_const): Likewise. Update call to
15138 altivec_expand_vec_perm_const_le.
15139 (rs6000_expand_vec_perm_const): Delete.
15140 (rs6000_vectorize_vec_perm_const_ok): Delete.
15141 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15142 reference to the SPE evmerge intructions.
15143 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15144 an element count and rtx array.
15145 (rs6000_expand_extract_even): Update call accordingly.
15146 (rs6000_expand_interleave): Likewise.
15147 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15148 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15150 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15154 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15155 vector mode and that that mode matches the mode of the data
15157 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15158 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15159 directly using expand_vec_perm_1 when forcing selectors into
15161 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15163 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15165 * optabs-query.h (can_vec_perm_p): Delete.
15166 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15167 * optabs-query.c (can_vec_perm_p): Split into...
15168 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15169 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15170 particular selector is valid.
15171 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15172 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15173 (vect_grouped_load_supported): Likewise.
15174 (vect_shift_permute_load_chain): Likewise.
15175 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15176 (vect_transform_slp_perm_load): Likewise.
15177 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15178 (vectorizable_bswap): Likewise.
15179 (vect_gen_perm_mask_checked): Likewise.
15180 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15181 implementations of variable permutation vectors into account
15182 when deciding which selector to use.
15183 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15184 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15185 with a false third argument.
15186 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15187 to test whether the constant selector is valid and can_vec_perm_var_p
15188 to test whether a variable selector is valid.
15190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15192 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15193 * optabs-query.c (can_vec_perm_p): Likewise.
15194 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15195 instead of vec_perm_indices.
15196 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15197 (vect_gen_perm_mask_checked): Likewise,
15198 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15199 (vect_gen_perm_mask_checked): Likewise,
15201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15203 * optabs-query.h (qimode_for_vec_perm): Declare.
15204 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15205 (qimode_for_vec_perm): ...this new function.
15206 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15208 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15210 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15211 does not have a conditional at the top.
15213 2018-01-02 Richard Biener <rguenther@suse.de>
15215 * ipa-inline.c (big_speedup_p): Fix expression.
15217 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15220 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15223 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15227 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15228 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15229 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15230 cond_taken_branch_cost 3->4.
15232 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15234 PR tree-optimization/83581
15235 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15236 TODO_cleanup_cfg if any changes have been made.
15238 PR middle-end/83608
15239 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15240 convert_modes if target mode has the right side, but different mode
15243 PR middle-end/83609
15244 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15245 last argument when extracting from CONCAT. If either from_real or
15246 from_imag is NULL, use expansion through memory. If result is not
15247 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15248 the parts directly to inner mode, if even that fails, use expansion
15251 PR middle-end/83623
15252 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15253 check for bswap in mode rather than HImode and use that in expand_unop
15256 Copyright (C) 2018 Free Software Foundation, Inc.
15258 Copying and distribution of this file, with or without modification,
15259 are permitted in any medium without royalty provided the copyright
15260 notice and this notice are preserved.