1 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
3 * config/s390/s390-builtin-types.def: Add void function type.
4 * config/s390/s390-builtins.def: Use the function type for the
7 2018-06-04 Jim Wilson <jimw@sifive.com>
9 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
11 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12 and maybe_eh_return. Change regno to unsigned int. Use new args to
13 handle EH_RETURN_DATA_REGNO registers properly.
14 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
15 (riscv_expand_epilogue): Update comment. Change argument name and
16 type. Update code to use new name and type. Pass new args to
17 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
19 * config/riscv/riscv.md (NORMAL_RETURN): New.
20 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
21 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
22 (eh_return): Call gen_eh_return_internal and emit barrier.
23 (eh_return_internal): Call riscv_expand_epilogue.
25 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
27 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
28 bit_insertion field and declare can_be_merged_into method.
29 (merged_store_group::can_be_merged_into): New method.
30 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
31 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
32 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
34 2018-06-04 Richard Biener <rguenther@suse.de>
36 PR tree-optimization/85955
37 * builtins.c (fold_builtin_sincos): Convert pointers to
38 destination to appropriate type before dereferencing.
40 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
42 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
44 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
46 * expr.c (expand_expr_real_1): Force the operand into memory if
47 its TYPE_MODE is BLKmode and if there is no integer mode for
48 the number of bits being extracted.
50 2018-06-04 Jakub Jelinek <jakub@redhat.com>
54 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
55 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
57 2018-06-04 Richard Biener <rguenther@suse.de>
59 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
60 (cleanup_tree_cfg_noloop): ... single caller. Do
61 start_recording_case_labels later.
63 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
65 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
66 to _IMMINTRIN_H_INCLUDED.
67 * config/i386/pconfigintrin.h: Ditto.
68 * config/i386/waitpkgintrin.h: Ditto.
69 * config/i386/immintrin.h: Add includes for sgxintrin.h,
70 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
71 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
72 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
73 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
74 waitpkgintrin.h and cldemoteintrin.h.
76 2018-06-04 Richard Biener <rguenther@suse.de>
78 PR tree-optimization/86038
79 * tracer.c (find_best_successor): Check probability for
80 being initialized, bail out if not.
82 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
85 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
86 of bits to ignore when comparing architectures.
88 2018-06-04 Jakub Jelinek <jakub@redhat.com>
90 PR tree-optimization/69615
91 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
92 maximum or minimum of the type, try to merge it also as if
93 range1 is + [-, x - 1] or + [x + 1, -].
96 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
98 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
100 PR tree-optimization/86034
101 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
102 the unsigned bitfield type in a bit insertion sequence if it does not
103 have a larger precision than the bitfield size.
104 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
106 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
108 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
110 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
112 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
113 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
114 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
115 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
117 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
119 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
120 Disable -fdelete-null-pointer-checks for ELF toolchain.
122 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
123 Kito Cheng <kito.cheng@gmail.com>
125 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
126 (nds32le-*-*, nds32be-*-*): Integrate checking process.
127 (nds32*-*-*): Add glibc and uclibc conditions.
128 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
129 (TARGET_EXCEPT_UNWIND_INFO): Define.
130 * config/nds32/elf.h: New file.
131 * config/nds32/linux.h: New file.
132 * config/nds32/nds32-elf.opt: New file.
133 * config/nds32/nds32-linux.opt: New file.
134 * config/nds32/nds32-fp-as-gp.c
135 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
136 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
138 (nds32_asm_file_end): Ditto.
139 (nds32_print_operand): Ditto.
140 (nds32_insert_attributes): Ditto.
141 (nds32_init_libfuncs): New function.
142 (TARGET_HAVE_TLS): Define.
143 (TARGET_INIT_LIBFUNCS): Define.
144 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
146 (TARGET_ELF): Apply different mcmodel setting.
147 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
148 been migrated into elf.h and linux.h files.
149 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
150 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
151 (mcmodel): The content has been migrated into nds32-elf.opt and
152 nds32-linux.opt files.
153 * config/nds32/t-elf: New file.
154 * config/nds32/t-linux: New file.
156 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
157 Shiva Chen <shiva0217@gmail.com>
159 * config/nds32/constants.md (unspec_volatile_element): Add
160 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
161 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
163 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
164 (make_pass_nds32_fp_as_gp): Declare.
165 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
167 (nds32_asm_function_end_prologue): Remove unused asm output.
168 (nds32_asm_function_begin_epilogue): Remove unused asm output.
169 (nds32_asm_file_start): Output necessary fp_as_gp information.
170 (nds32_option_override): Adjust register usage.
171 (nds32_expand_prologue): Consider fp_as_gp situation.
172 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
173 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
176 (simple_return): Ditto.
177 (omit_fp_begin): Output special directive for fp_as_gp.
178 (omit_fp_end): Output special directive for fp_as_gp.
179 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
180 mforbid-fp-as-gp): New options.
182 2018-06-01 Mark Wielaard <mark@klomp.org>
184 * dwarf2out.c (dwarf2out_finish): Remove generation of
187 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
189 * gimple-ssa-store-merging.c: Include gimple-fold.h.
190 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
191 (struct merged_store_group): Add bit_insertion field.
192 (dump_char_array): Use standard hexadecimal format.
193 (merged_store_group::merged_store_group): Set bit_insertion to false.
194 (merged_store_group::apply_stores): Use optimal buffer size. Deal
195 with BIT_INSERT_EXPR stores. Move up code updating the mask and
196 also print the mask in the dump file.
197 (pass_store_merging::gate): Minor tweak.
198 (imm_store_chain_info::coalesce_immediate): Fix wrong association
199 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
200 stores with INTEGER_CST stores.
201 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
202 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
203 and use it throughout. Generate bit insertion sequences if need be.
204 (pass_store_merging::process_store): Remove redundant condition.
205 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
207 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
209 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
210 the 128-bit floating point types. Fix function comment.
212 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
214 * config/aarch64/aarch64-simd.md
215 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
217 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
220 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
222 PR tree-optimization/85989
223 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
225 (backprop::intersect_uses): Check it when deciding whether this
226 is a backedge reference.
227 (backprop::process_block): Add each phi to m_visited_phis
228 after visiting it, then clear it at the end.
230 2018-06-01 Richard Biener <rguenther@suse.de>
232 * tree-vectorizer.h (vect_dr_stmt): New function.
233 (vect_get_load_cost): Adjust.
234 (vect_get_store_cost): Likewise.
235 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
236 Use vect_dr_stmt instead of DR_SMTT.
237 (vect_record_base_alignments): Likewise.
238 (vect_calculate_target_alignment): Likewise.
239 (vect_compute_data_ref_alignment): Likewise and make static.
240 (vect_update_misalignment_for_peel): Likewise.
241 (vect_verify_datarefs_alignment): Likewise.
242 (vector_alignment_reachable_p): Likewise.
243 (vect_get_data_access_cost): Likewise. Pass down
244 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
245 (vect_get_peeling_costs_all_drs): Likewise.
246 (vect_peeling_hash_get_lowest_cost): Likewise.
247 (vect_enhance_data_refs_alignment): Likewise.
248 (vect_find_same_alignment_drs): Likewise.
249 (vect_analyze_data_refs_alignment): Likewise.
250 (vect_analyze_group_access_1): Likewise.
251 (vect_analyze_group_access): Likewise.
252 (vect_analyze_data_ref_access): Likewise.
253 (vect_analyze_data_ref_accesses): Likewise.
254 (vect_vfa_segment_size): Likewise.
255 (vect_small_gap_p): Likewise.
256 (vectorizable_with_step_bound_p): Likewise.
257 (vect_prune_runtime_alias_test_list): Likewise.
258 (vect_analyze_data_refs): Likewise.
259 (vect_supportable_dr_alignment): Likewise.
260 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
261 (vect_gen_prolog_loop_niters): Likewise.
262 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
263 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
265 (vect_recog_mask_conversion_pattern): Likewise.
266 (vect_try_gather_scatter_pattern): Likewise.
267 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
268 to vect_get_store_cost.
269 (vect_get_store_cost): Get stmt_info instead of DR.
270 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
271 (vect_get_load_cost): Get stmt_info instead of DR.
273 2018-06-01 Richard Biener <rguenther@suse.de>
276 * gimple-fold.c (var_decl_component_p): Also allow offsetted
277 vars wrapped in MEM_REFs.
279 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
281 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
282 Fix subreg tests so that we only return a choice between
283 GENERAL_REGS and FP_REGS if the original classes included both.
285 2018-06-01 Richard Biener <rguenther@suse.de>
288 * tree-ssa-structalias.c (get_function_part_constraint):
289 Handle NULL fi->decl.
290 (find_func_aliases_for_call): Properly handle indirect
292 (find_func_clobbers): Likewise.
293 (ipa_pta_execute): Likewise.
294 (create_variable_info_for): For functions that are ifunc_resolver
295 resolve to a varinfo that contains the result of the resolver
297 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
300 2018-05-31 Michael Collison <michael.collison@arm.com>
302 * config/aarch64/aarch64.md:
303 (*fix_to_zero_extenddfdi2): New pattern.
304 * gcc.target/aarch64/fix_extend1.c: New testcase.
306 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
310 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
311 and BUILT_IN_STRNCMP_EQ.
312 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
314 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
315 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
316 (gimple_fold_builtin): Likewise.
317 * tree-ssa-strlen.c (compute_string_length): New function.
318 (determine_min_obsize): New function.
319 (handle_builtin_string_cmp): New function to handle calls to
320 string compare functions.
321 (strlen_optimize_stmt): Add handling to builtin string compare
323 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
324 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
325 * tree.c (build_common_builtin_nodes): Add new defines of
326 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
328 2018-05-31 Jakub Jelinek <jakub@redhat.com>
331 * bb-reorder.c (pass_partition_blocks::gate): Return false for
332 functions with naked attribute.
334 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
336 * config/i386/sse.md (avx_vec_concat<mode>):
337 Substitute concat_tg_mode mode attribute with xtg_mode.
338 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
339 (concat_tg_mode): Remove mode attribute.
341 2018-05-31 Martin Sebor <msebor@redhat.com>
344 * calls.c (alloc_max_size): Correct a logic error/typo.
345 Treat excessive arguments as infinite. Warn for invalid arguments.
346 * doc/invoke.texi (-Walloc-size-larger-than): Update.
348 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
351 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
352 and movx for Haswell.
354 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
355 Cesar Philippidis <cesar@codesourcery.com>
358 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
359 when emitting error on private/firstprivate reductions.
360 * omp-low.c (lower_omp_target): Avoid reference-type processing
361 on pointers for firstprivate clause.
363 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
365 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
368 * config/aarch64/aarch64-simd.md
369 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
370 (aarch64_ld1_x3_<mode>): Likewise
371 (aarch64_st1x2<VALLDIF:mode>): Likewise
372 (aarch64_st1_x2_<mode>): Likewise
373 (aarch64_st1x3<VALLDIF:mode>): Likewise
374 (aarch64_st1_x3_<mode>): Likewise
375 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
376 (vld1_s8_x3): Likewise.
377 (vld1_u16_x3): Likewise.
378 (vld1_s16_x3): Likewise.
379 (vld1_u32_x3): Likewise.
380 (vld1_s32_x3): Likewise.
381 (vld1_u64_x3): Likewise.
382 (vld1_s64_x3): Likewise.
383 (vld1_f16_x3): Likewise.
384 (vld1_f32_x3): Likewise.
385 (vld1_f64_x3): Likewise.
386 (vld1_p8_x3): Likewise.
387 (vld1_p16_x3): Likewise.
388 (vld1_p64_x3): Likewise.
389 (vld1q_u8_x3): Likewise.
390 (vld1q_s8_x3): Likewise.
391 (vld1q_u16_x3): Likewise.
392 (vld1q_s16_x3): Likewise.
393 (vld1q_u32_x3): Likewise.
394 (vld1q_s32_x3): Likewise.
395 (vld1q_u64_x3): Likewise.
396 (vld1q_s64_x3): Likewise.
397 (vld1q_f16_x3): Likewise.
398 (vld1q_f32_x3): Likewise.
399 (vld1q_f64_x3): Likewise.
400 (vld1q_p8_x3): Likewise.
401 (vld1q_p16_x3): Likewise.
402 (vld1q_p64_x3): Likewise.
403 (vst1_s64_x2): Likewise.
404 (vst1_u64_x2): Likewise.
405 (vst1_f64_x2): Likewise.
406 (vst1_s8_x2): Likewise.
407 (vst1_p8_x2): Likewise.
408 (vst1_s16_x2): Likewise.
409 (vst1_p16_x2): Likewise.
410 (vst1_s32_x2): Likewise.
411 (vst1_u8_x2): Likewise.
412 (vst1_u16_x2): Likewise.
413 (vst1_u32_x2): Likewise.
414 (vst1_f16_x2): Likewise.
415 (vst1_f32_x2): Likewise.
416 (vst1_p64_x2): Likewise.
417 (vst1q_s8_x2): Likewise.
418 (vst1q_p8_x2): Likewise.
419 (vst1q_s16_x2): Likewise.
420 (vst1q_p16_x2): Likewise.
421 (vst1q_s32_x2): Likewise.
422 (vst1q_s64_x2): Likewise.
423 (vst1q_u8_x2): Likewise.
424 (vst1q_u16_x2): Likewise.
425 (vst1q_u32_x2): Likewise.
426 (vst1q_u64_x2): Likewise.
427 (vst1q_f16_x2): Likewise.
428 (vst1q_f32_x2): Likewise.
429 (vst1q_f64_x2): Likewise.
430 (vst1q_p64_x2): Likewise.
431 (vst1_s64_x3): Likewise.
432 (vst1_u64_x3): Likewise.
433 (vst1_f64_x3): Likewise.
434 (vst1_s8_x3): Likewise.
435 (vst1_p8_x3): Likewise.
436 (vst1_s16_x3): Likewise.
437 (vst1_p16_x3): Likewise.
438 (vst1_s32_x3): Likewise.
439 (vst1_u8_x3): Likewise.
440 (vst1_u16_x3): Likewise.
441 (vst1_u32_x3): Likewise.
442 (vst1_f16_x3): Likewise.
443 (vst1_f32_x3): Likewise.
444 (vst1_p64_x3): Likewise.
445 (vst1q_s8_x3): Likewise.
446 (vst1q_p8_x3): Likewise.
447 (vst1q_s16_x3): Likewise.
448 (vst1q_p16_x3): Likewise.
449 (vst1q_s32_x3): Likewise.
450 (vst1q_s64_x3): Likewise.
451 (vst1q_u8_x3): Likewise.
452 (vst1q_u16_x3): Likewise.
453 (vst1q_u32_x3): Likewise.
454 (vst1q_u64_x3): Likewise.
455 (vst1q_f16_x3): Likewise.
456 (vst1q_f32_x3): Likewise.
457 (vst1q_f64_x3): Likewise.
458 (vst1q_p64_x3): Likewise.
460 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
462 * config/msp430/msp430.c (msp430_output_labelref): Prepend
463 user_label_prefix to name.
465 * tree-core.h: Update comment about the format of NAME string
466 passed to handler in attribute_spec.
468 * config/msp430/msp430.md: Remove erroneous subreg expression from
469 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
470 zero_extend{q,h}isi2.
472 2018-05-30 Borislav Petkov <bp@suse.de>
474 * doc/extend.texi: Document some architecture specific
475 constraints and sort entries.
477 2018-05-30 Martin Sebor <msebor@redhat.com>
480 * builtins.c (expand_builtin_stpcpy_1): New function.
481 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
482 only if the former succeeds.
484 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
486 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
489 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
491 * doc/invoke.texi (-flinker-output): Document
493 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
495 * passes.c (ipa_write_summaries): Only modify statements if body
497 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
498 incrementally linking.
499 (ipa_passes): Likewise.
500 * lto-cgraph.c (lto_output_node): When incrementally linking do not
501 pass down resolution info.
502 * common.opt (flag_incremental_link): Update info.
503 * gcc.c (plugin specs): Turn flinker-output=* to
504 -plugin-opt=-linker-output-known
505 * toplev.c (compile_file): Also cut compilation when doing incremental
507 * flag-types. (enum lto_partition_model): Add
508 LTO_LINKER_OUTPUT_NOLTOREL.
509 (invoke.texi): Add -flinker-output docs.
510 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
511 link same way as WPA; do not stream in dead initializers.
513 * dwarf2out.c (dwarf2out_die_ref_for_decl,
514 darf2out_register_external_decl): Support incremental link.
516 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
518 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
520 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
522 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
523 it down to simple_object_copy_lto_debug_sections.
524 (run_gcc): Determine incremental LTO link time and configure
525 lto1 into non-wpa mode, disable renaming of debug sections.
527 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
529 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
530 descriptions of various incorrectly documented functions.
532 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
535 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
536 address check not strict.
538 2018-05-30 Richard Biener <rguenther@suse.de>
540 PR tree-optimization/85964
541 * tracer.c (better_p): Drop initialized count check, we only
542 call the function with initialized counts now.
543 (find_best_successor): Do find a best edge if one
544 has uninitialized count.
545 (find_best_predecessor): Likewise. Do BB frequency check only
546 if count is initialized.
548 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
550 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
551 (aarch64_ldrstr_offset_compare): New.
552 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
553 load/store orderings.
554 (aarch64_gen_adjusted_ldpstp): Likewise.
556 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
558 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
559 Check for subset of GENERAL_REGS and FP_REGS.
560 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
563 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
565 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
566 and wi::to_poly_offset. Add the current offset and then check
567 whether the sum fits, rather than using an unchecked addition of
568 a checked term. Check for a shwi rather than a uhwi.
569 * expr.c (get_bit_range): Use tree_to_poly_uint64.
570 (store_constructor): Use poly_int_tree_p.
571 (expand_expr_real_1): Likewise.
572 * function.c (assign_temp): Likewise.
573 * fold-const.c (const_binop): Use poly_int_tree_p and
575 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
577 * ipa-icf-gimple.c (func_checker::compare_operand): Use
578 to_poly_offset for MEM offsets.
579 * ipa-icf.c (sem_variable::equals): Likewise.
580 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
581 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
582 wi::to_poly_offset for BIT_FIELD_REF offsets.
583 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
585 * var-tracking.c (emit_note_insn_var_location): Use
588 2018-05-29 Jim Wilson <jimw@sifive.com>
590 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
592 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
595 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
596 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
598 (sse4_1_round<mode>2): Use nonimmediate_operand
599 for operand 1 predicate.
601 2018-05-29 Martin Sebor <msebor@redhat.com>
602 Richard Biener <rguenther@suse.de>
605 * calls.c (get_size_range): Call determine_value_range instead
607 * tree-vrp.h (determine_value_range): Declared new function.
608 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
610 2018-05-29 Richard Biener <rguenther@suse.de>
612 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
613 sure to use non-pattern stmts for get_earlier_stmt arguments.
614 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
615 called on pattern stmts.
616 (get_later_stmt): Likewise.
618 2018-05-29 Martin Liska <mliska@suse.cz>
620 PR gcov-profile/85759
621 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
624 2018-05-29 Jakub Jelinek <jakub@redhat.com>
626 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
628 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
632 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
633 VEC_PACK_FLOAT_EXPR): New tree codes.
634 * tree-pretty-print.c (op_code_prio): Handle
635 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
636 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
637 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
638 * tree-inline.c (estimate_operator_cost): Likewise.
639 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
640 * fold-const.c (const_binop): Likewise.
641 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
642 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
643 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
644 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
645 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
646 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
647 * expr.c (expand_expr_real_2): Likewise.
648 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
649 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
650 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
652 * optabs.c (expand_widen_pattern_expr): For
653 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
654 sign from result type rather than operand's type.
655 (expand_binop_directly): For vec_packu_float_optab and
656 vec_packs_float_optab allow result type to be different from operand's
658 * optabs-tree.c (optab_for_tree_code): Handle
659 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
660 VEC_PACK_FLOAT_EXPR. Formatting fixes.
661 * tree-vect-generic.c (expand_vector_operations_1): Handle
662 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
664 * tree-vect-stmts.c (supportable_widening_operation): Handle
666 (supportable_narrowing_operation): Handle FLOAT_EXPR.
667 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
668 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
669 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
670 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
672 (vec_pack<floatprefix>_float_<mode>): New expander.
673 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
675 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
676 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
677 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
678 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
679 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
681 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
682 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
683 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
684 VEC_PACK_FLOAT_EXPR): Document.
686 2018-05-29 Richard Biener <rguenther@suse.de>
688 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
690 (stmt_vec_info_vec): Make pointer.
691 (init_stmt_vec_info_vec): Remove.
692 (free_stmt_vec_info_vec): Likewise.
693 (set_stmt_vec_info_vec): New function.
694 (free_stmt_vec_infos): Likewise.
695 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
696 (set_vinfo_for_stmt): Likewise.
697 (get_earlier_stmt): Likewise.
698 (get_later_stmt): Likewise.
699 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
700 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
701 (vec_info::~vec_info): Free stmt_vec_infos.
702 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
703 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
704 (pass_slp_vectorize::execute): Likewise.
705 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
706 (free_stmt_vec_info_vec): Likewise.
707 (set_stmt_vec_info_vec): New function.
708 (free_stmt_vec_infos): Likewise.
709 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
710 the global stmt_vec_info_vec.
711 * tree-parloops.c (gather_scalar_reductions): Use
712 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
715 2018-05-29 Richard Biener <rguenther@suse.de>
717 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
719 2018-05-29 Martin Liska <mliska@suse.cz>
720 David Malcolm <dmalcolm@redhat.com>
722 * vec.c (test_reverse): New.
723 (vec_c_tests): Add new test.
724 * vec.h (vl_ptr>::reverse): New function.
726 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
728 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
730 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
733 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
735 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
737 2018-05-28 Richard Biener <rguenther@suse.de>
739 PR tree-optimization/85933
740 * tree-vect-data-refs.c (vect_record_base_alignments): Only
741 look at stmts marked as vectorizable.
743 2018-05-28 Richard Biener <rguenther@suse.de>
745 PR tree-optimization/85934
746 * tree-vect-generic.c (expand_vector_operations_1): Hoist
747 vector boolean check before scalar optimization.
749 2018-05-28 Jakub Jelinek <jakub@redhat.com>
751 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
754 2018-05-28 Mark Wielaard <mark@klomp.org>
756 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
757 if it is an expression containing a minus sign.
759 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
761 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
763 2018-05-27 Paul Koning <ni1d@arrl.net>
765 * config/pdp11/pdp11.md (truncsihi2): Remove.
767 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
768 Chung-Ju Wu <jasonwucj@gmail.com>
770 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
772 (unaligned_store_dw): Ditto.
773 * config/nds32/nds32-memory-manipulation.c
774 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
775 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
776 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
777 (emit_setmem_word_loop): Rename to ...
778 (emit_setmem_doubleword_loop): ... this.
779 (nds32_gen_dup_4_byte_to_word_value): New function.
780 (nds32_gen_dup_8_byte_to_double_word_value): New function.
781 (nds32_expand_setmem_loop): Refine implementation.
782 (nds32_expand_setmem_loop_v3m): Ditto.
783 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
786 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
788 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
790 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
792 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
793 (nds32_init_machine_status): Initialize machine->attr_naked_p and
794 machine->attr_no_prologue_p.
795 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
796 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
797 (nds32_expand_epilogue): Consider attr_naked_p.
798 (nds32_expand_epilogue_v3pop): Likewise.
799 (nds32_can_use_return_insn): Likewise.
800 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
801 attr_no_prologue_p fields.
802 * config/nds32/nds32.opt (mret-in-naked-func): New option.
804 2018-05-27 Jakub Jelinek <jakub@redhat.com>
807 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
810 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
812 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
814 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
816 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
818 (*<floatsuffix>floatv2div2sf2): Rename to ...
819 (*float<floatunssuffix>v2div2sf2): ... this.
820 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
821 (float<floatunssuffix>v2div2sf2_mask): ... this.
822 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
823 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
824 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
826 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
828 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
830 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
832 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
834 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
836 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
837 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
838 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
839 gen_ufix_truncv8dfv8si2.
840 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
841 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
842 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
843 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
844 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
845 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
846 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
847 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
849 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
853 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
855 2018-05-25 Jim Wilson <jimw@sifive.com>
857 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
858 * config/riscv/riscv.c (struct machine_function): Add
859 interrupt_handler_p and attribute_checked_p fields.
860 (riscv_attribute_table): Add interrupt.
861 (riscv_interrupt_type_p): New.
862 (riscv_save_reg_p): Save extra regs for interrupt handler.
863 (riscv_use_save_libcall): Return false for interrupt handler.
864 (riscv_first_stack_step): Add forward declaration.
865 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
866 for interrupt handler with large frame. Use it for saved reg list.
867 (riscv_expand_prologue): Move flag_stack_usage_info support to
868 eliminate duplication.
869 (riscv_expand_epilogue): Generate mret for interrupt handler.
870 (riscv_epilogue_uses): New.
871 (riscv_can_use_return_insn): Return false for interrupt handler.
872 (riscv_function_ok_for_sibcall): Likewise.
873 (riscv_set_current_function): Add interrupt handler support.
874 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
875 * config/riscv/riscv.md (UNSPECV_MRET): New.
877 (riscv_frflags, riscv_fsflags): Use tab after opcode.
879 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
881 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
883 PR tree-optimization/85712
884 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
885 this candidate has already been replaced in-situ by a copy.
887 2018-05-25 Jason Merrill <jason@redhat.com>
889 PR c++/80485 - inline function non-zero address.
890 * symtab.c (nonzero_address): Check DECL_COMDAT.
892 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
895 * config/alpha/alpha.md (ashlsi3): New insn pattern.
896 (*ashlsi_se): Rename from *ashldi_se. Define as sign
897 extension of SImode operation. Use const123_operand predicate.
899 (*saddl_se_1): Ditto.
901 (*ssubl_se_1): Ditto.
902 * config/alpha/predicates.md (const123_operand): New predicate.
903 * config/alpha/constraints.md (P): Use IN_RANGE.
905 2018-05-25 Richard Biener <rguenther@suse.de>
907 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
909 (ref_maybe_used_by_stmt_p): Likewise.
910 (stmt_may_clobber_ref_p): Likewise.
911 (stmt_may_clobber_ref_p_1): Likewise.
912 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
914 (ref_maybe_used_by_stmt_p): Likewise.
915 (stmt_may_clobber_ref_p): Likewise.
916 (stmt_may_clobber_ref_p_1): Likewise.
917 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
918 the alias oracle to disambiguate DRs with stmts DR analysis
920 (vect_analyze_data_refs): Do not give up on not analyzable
921 DRs for BB vectorization. Remove code truncating the dataref
924 2018-05-25 Jakub Jelinek <jakub@redhat.com>
927 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
928 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
929 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
931 2018-05-25 Richard Biener <rguenther@suse.de>
933 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
934 function, combining stmt data ref gathering and fatal analysis
936 (vect_analyze_data_refs): Remove now redudnant code and simplify.
937 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
938 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
939 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
940 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
942 2018-05-25 Bin Cheng <bin.cheng@arm.com>
944 PR tree-optimization/85720
945 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
946 SCC if all partitions are builtins.
947 (version_loop_by_alias_check): New parameter. Generate cancelable
948 runtime alias check if all partitions are builtins.
949 (distribute_loop): Update call to above function.
951 2018-05-25 Bin Cheng <bin.cheng@arm.com>
953 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
954 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
955 (parm_default_def_partition_arg): Ditto.
956 (set_parm_default_def_partition): Ditto.
957 (get_parm_default_def_partitions): Ditto and make it static.
958 (get_undefined_value_partitions): Ditto and make it static.
959 (remove_ssa_form): Refactor call to init_var_map here.
960 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
961 computation for loop region.
962 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
963 (register_default_def): Delete.
964 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
965 (parm_default_def_partition_arg): Ditto.
966 (set_parm_default_def_partition): Ditto.
967 (get_parm_default_def_partitions): Ditto and make it static.
968 (get_undefined_value_partitions): Ditto and make it static.
969 (coalesce_with_default, coalesce_with_default): Update comment.
970 (create_coalesce_list_for_region): New func factored out from
971 create_outofssa_var_map.
972 (populate_coalesce_list_for_outofssa): New func factored out from
973 create_outofssa_var_map and coalesce_ssa_name.
974 (create_outofssa_var_map): Delete.
975 (coalesce_ssa_name): Refactor to support live range computation.
976 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
977 (get_parm_default_def_partitions): Delete.
978 (get_undefined_value_partitions): Ditto.
979 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
980 computation for loop region.
981 (new_tree_live_info, loe_visit_block): Ditto.
982 (live_worklist, set_var_live_on_entry): Ditto.
983 (calculate_live_on_exit, verify_live_on_entry): Ditto.
984 * tree-ssa-live.h (struct _var_map): New fields.
985 (init_var_map): Change decl.
986 (region_contains_p): New.
988 2018-05-25 Bin Cheng <bin.cheng@arm.com>
990 * tree-ssa-live.h (live_merge_and_clear): Delete.
992 2018-05-25 Richard Biener <rguenther@suse.de>
995 * tree-dump.c (dequeue_and_dump): Remove access to removed
996 operand 2 of a SWITCH_EXPR.
998 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1000 * doc/sourcebuild.texi (vect_double_cond_arith): Include
1001 multiplication and division.
1002 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
1003 (cond_udiv@var{m}, cond_umod@var{m}): Document.
1004 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
1005 (cond_udiv_optab, cond_umod_optab): New optabs.
1006 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1007 (IFN_COND_RDIV): New internal functions.
1008 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1009 TRUNC_MOD_EXPR and RDIV_EXPR.
1010 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1011 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1013 (SVE_INT_BINARY): Include mult.
1014 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1015 (optab, sve_int_op): Handle mult.
1016 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1018 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1019 for SVE_INT_BINARY_SD.
1021 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1023 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1024 (optab, sve_int_op): Handle div and udiv.
1025 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1026 for SVE_INT_BINARY_SD.
1027 (*<optab><mode>3): New insn for the same.
1029 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1031 * tree-vect-patterns.c: Include predict.h.
1032 (vect_recog_divmod_pattern): Restrict check for division support
1033 to when optimizing for size.
1035 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1037 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1038 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1039 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1040 (gimple_match_op::set_op): Likewise.
1041 (gimple_resimplify4): Declare.
1042 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1043 (expr::gen_transform): Likewise.
1044 (decision_tree::gen): Generate a simplification routine for 4 operands.
1045 * gimple-match-head.c (gimple_simplify): Add an overload for
1046 4 operands. In the top-level function, handle up to 4 call
1047 arguments and call gimple_resimplify4.
1048 (gimple_resimplify4): New function.
1049 (build_call_internal): Pass a fourth operand.
1050 (maybe_push_to_seq): Likewise.
1051 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1052 Fold VEC_COND_EXPRs of an operation and a default value into
1053 an IFN_COND_* function if possible.
1054 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1056 (SVE_COND_FP_BINARY): Include them.
1057 (optab, sve_fp_op): Handle them.
1058 (SVE_INT_BINARY_REV): New code iterator.
1059 (SVE_COND_FP_BINARY_REV): New int iterator.
1060 (commutative): New int attribute.
1061 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1063 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1065 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1066 (*cond_<optab><mode>): New patterns for reversed operands.
1068 2018-05-25 Richard Biener <rguenther@suse.de>
1070 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1071 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1072 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1073 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1074 * tree-vect-data-refs.c (everywhere): Adjust users.
1075 * tree-vect-loop.c (everywhere): Likewise.
1076 * tree-vect-slp.c (everywhere): Likewise.
1077 * tree-vect-stmts.c (everywhere): Likewise.
1078 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1080 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1082 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1084 (gcc_cv_as_section_exclude): ... this.
1085 Try Solaris as #exclude syntax.
1086 * configure: Regenerate.
1087 * config.in: Regenerate.
1088 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1090 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1091 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1093 * varasm.c (default_elf_asm_named_section): Don't check if
1094 HAVE_GAS_SECTION_EXCLUDE is defined.
1096 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1098 * doc/md.texi: Update the documentation of the cond_* optabs
1099 to mention the new final operand. Fix GET_MODE_NUNITS call.
1100 Describe the scalar case too.
1101 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1102 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1104 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1105 (get_conditional_internal_fn): Update comment.
1106 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1107 accumulator value as a final argument to conditional functions.
1108 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1109 a define_expand and add an "else" operand. Assert for now that
1110 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1111 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1112 (*cond_<optab><mode>): New patterns.
1113 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1114 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1115 (UNSPEC_COND_EOR): Delete.
1116 (optab): Remove associated mappings.
1117 (SVE_INT_BINARY): New code iterator.
1118 (sve_int_op): Remove int attribute and add "minus" to the code
1120 (SVE_COND_INT_OP): Delete.
1121 (SVE_COND_FP_OP): Rename to...
1122 (SVE_COND_FP_BINARY): ...this.
1124 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1126 * optabs.c (can_reuse_operands_p): New function.
1127 (maybe_legitimize_operands): Try to reuse the results for
1130 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1132 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1133 Add {q} suffix to insn mnemonic.
1135 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1137 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1138 (msp430_warn_func_return): New.
1140 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1142 * fold-const.c (tree_nonzero_bits): New function.
1143 * fold-const.h (tree_nonzero_bits): Likewise.
1144 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1145 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1147 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1151 * varasm.c (assemble_alias): Check ifunc_resolver only on
1154 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1157 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1158 when memory input operand is handled.
1160 2018-05-24 Luis Machado <luis.machado@linaro.org>
1162 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1164 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1166 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1168 * match.pd: Delay FMA folds until after vectorization.
1170 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1173 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1174 address check not strict.
1176 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1178 * gimple-match.h (gimple_match_op): New class.
1179 (mprts_hook): Replace parameters with a gimple_match_op *.
1180 (maybe_build_generic_op): Likewise.
1181 (gimple_simplified_result_is_gimple_val): Replace parameters with
1182 a const gimple_match_op *.
1183 (gimple_simplify): Replace code_helper * and tree * parameters with
1184 a gimple_match_op * parameter.
1185 (gimple_resimplify1): Replace code_helper *, tree and tree *
1186 parameters with a gimple_match_op * parameter.
1187 (gimple_resimplify2): Likewise.
1188 (gimple_resimplify3): Likewise.
1189 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1190 parameters with a gimple_match_op * parameter.
1191 * gimple-match-head.c (gimple_simplify): Change prototypes of
1192 auto-generated functions to take a gimple_match_op * instead of
1193 separate code_helper * and tree * parameters. Make the same
1194 change in the top-level overload and update calls to the
1195 gimple_resimplify routines. Update calls to the auto-generated
1196 functions and to maybe_push_res_to_seq in the publicly-facing
1197 operation-specific gimple_simplify overloads.
1198 (gimple_match_op::MAX_NUM_OPS): Define.
1199 (gimple_resimplify1): Replace rcode and ops with a single res_op
1200 parameter. Update call to gimple_simplify.
1201 (gimple_resimplify2): Likewise.
1202 (gimple_resimplify3): Likewise.
1203 (mprts_hook): Replace parameters with a gimple_match_op *.
1204 (maybe_build_generic_op): Likewise.
1205 (build_call_internal): Replace type, nargs and ops with
1206 a gimple_match_op *.
1207 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1208 with a single gimple_match_op *. Update calls to mprts_hook,
1209 build_call_internal and gimple_simplified_result_is_gimple_val.
1210 Factor out code that is common to the tree_code and combined_fn cases.
1211 * genmatch.c (expr::gen_transform): Replace tem_code and
1212 tem_ops with a gimple_match_op called tem_op. Update calls
1213 to the gimple_resimplify functions and maybe_push_res_to_seq.
1214 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1215 res_ops. Update call to the gimple_resimplify functions.
1216 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1217 (decision_tree::gen): Make the functions take a gimple_match_op *
1218 called res_op instead of separate res_code and res_ops parameters.
1219 Update call accordingly.
1220 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1221 and ops with a single res_op parameter. Update calls to
1222 maybe_build_generic_op and maybe_push_res_to_seq.
1223 (fold_stmt_1): Update calls to gimple_simplify and
1224 replace_stmt_with_simplification.
1225 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1226 and gimple_simplified_result_is_gimple_val.
1227 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1229 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1230 with a gimple_match_op *.
1231 (vn_nary_build_or_lookup): Likewise. Update call to
1232 vn_nary_build_or_lookup_1.
1233 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1234 gimple_match_op *. Update calls to the gimple_resimplify routines
1235 and to gimple_simplified_result_is_gimple_val.
1236 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1237 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1238 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1239 (visit_nary_op): Likewise.
1240 (visit_reference_op_load): Likewise.
1242 2018-05-23 Luis Machado <luis.machado@linaro.org>
1244 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1245 modifier for printing the step amount.
1247 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1250 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1253 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1255 * doc/sourcebuild.texi (Endianness): New subsubsection.
1257 2018-05-23 Luis Machado <luis.machado@linaro.org>
1259 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1260 <prefetch_dynamic_strides>: New const bool field.
1261 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1262 prefetch_dynamic_strides.
1263 (exynosm1_prefetch_tune): Likewise.
1264 (thunderxt88_prefetch_tune): Likewise.
1265 (thunderx_prefetch_tune): Likewise.
1266 (thunderx2t99_prefetch_tune): Likewise.
1267 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1269 (aarch64_override_options_internal): Update to set
1270 PARAM_PREFETCH_DYNAMIC_STRIDES.
1271 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1272 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1273 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1274 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1275 prefetch-dynamic-strides setting.
1277 2018-05-23 Luis Machado <luis.machado@linaro.org>
1279 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1280 <minimum_stride>: New const int field.
1281 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1282 minimum_stride field defaulting to -1.
1283 (exynosm1_prefetch_tune): Likewise.
1284 (thunderxt88_prefetch_tune): Likewise.
1285 (thunderx_prefetch_tune): Likewise.
1286 (thunderx2t99_prefetch_tune): Likewise.
1287 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1288 <default_opt_level>: Set to 3.
1289 (aarch64_override_options_internal): Update to set
1290 PARAM_PREFETCH_MINIMUM_STRIDE.
1291 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1292 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1293 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1294 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1295 stride is constant and is below the minimum stride threshold.
1297 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1299 * config/arm/arm-cpus.in (mode26): Delete.
1300 (armv4): Delete mode26 reference.
1301 * config/arm/arm.c (arm_configure_build_target): Delete use of
1304 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1306 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1308 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1309 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1311 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1312 (floatunsdidf2): Ditto.
1314 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1316 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1317 (fixuns_trunc<mode>si2_avx512f): Ditto.
1318 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1319 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1320 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1322 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1324 PR rtl-optimization/79985
1325 * df-scan.c (df_insn_refs_collect): Remove special case for
1326 global registers and asm statements.
1328 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1330 * extend.texi (Global Register Variables): Rewrite the bullet list.
1331 Note that the register is available for allocation. Note that access
1332 via inline asm must use constraints. Add note about async-signal
1333 handlers. Remove paragraph about automagic register selection.
1335 2018-05-23 Richard Biener <rguenther@suse.de>
1337 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1338 of fixed offset from memset VN.
1340 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1342 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1344 (alloc_cand_and_find_basis): Initialize first_interp field.
1345 (slsr_process_mul): Modify first_interp field.
1346 (slsr_process_add): Likewise.
1347 (slsr_process_cast): Modify first_interp field for each new
1349 (slsr_process_copy): Likewise.
1350 (dump_candidate): Dump first_interp field.
1351 (replace_mult_candidate): Process all interpretations, not just
1353 (replace_rhs_if_not_dup): Likewise.
1354 (replace_one_candidate): Likewise.
1356 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1358 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1360 (aarch64_needs_frame_chain): New function.
1361 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1363 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1366 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1367 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1368 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1369 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1370 as true for strict-align.
1371 (aarch64_can_inline_p): Perform checks even when callee has no
1372 attributes to check for strict alignment.
1373 * doc/extend.texi (AArch64 Function Attributes): Document
1375 * doc/invoke.texi: (AArch64 Options): Likewise.
1377 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1379 PR tree-optimization/85853
1380 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1381 the handling of the root of the node to...
1382 (vect_slp_analyze_node_operations_1): ...this new function,
1383 and run the whole thing with the child nodes' def types
1384 set according to their SLP node's def type.
1386 2018-05-23 Richard Biener <rguenther@suse.de>
1389 * tree-data-ref.c (create_runtime_alias_checks): Defer
1390 and ignore overflow warnings.
1392 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1394 PR tree-optimization/85822
1395 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1398 2018-05-23 Richard Biener <rguenther@suse.de>
1400 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1401 memset constants via native_interpret_expr.
1403 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1406 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1408 (cgraph_node::create_alias): Likewise.
1409 (cgraph_node::get_availability): Check ifunc_resolver instead
1410 of looking up ifunc attribute.
1411 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1412 * varasm.c (do_assemble_alias): Likewise.
1413 (assemble_alias): Likewise.
1414 (default_binds_local_p_3): Likewise.
1415 * cgraph.h (cgraph_node): Add ifunc_resolver.
1416 (cgraph_node::only_called_directly_or_aliased_p): Return false
1418 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1420 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1421 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1422 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1423 instead of looking up ifunc attribute.
1425 2018-05-22 Luis Machado <luis.machado@linaro.org>
1427 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1429 2018-05-22 Martin Sebor <msebor@redhat.com>
1432 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1433 only when expasion succeeds.
1434 (expand_builtin_strcmp): Same.
1435 (expand_builtin_strncmp): Same.
1437 2018-05-22 Martin Sebor <msebor@redhat.com>
1439 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1441 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1442 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1444 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1445 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1447 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1448 Add check for legitimate_address.
1449 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1450 (aarch64_swap_ldrstr_operands): New.
1451 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1454 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1455 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1457 * config/aarch64/aarch64.md: New patterns to generate stp
1459 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1460 single words and double words.
1461 (load_pair_sw, load_pair_dw): Likewise.
1462 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1464 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1466 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1467 for different mode ldpstp and add peephole for merged zero stores.
1469 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1471 (aarch64_gen_store_pair): Rename calls to match new patterns.
1472 (aarch64_gen_load_pair): Rename calls to match new patterns.
1473 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1474 (load_pair<DREG:mode><DREG2:mode>): ... This.
1475 (store_pair<mode>): Rename to...
1476 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1477 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1479 (V_INT_EQUIV): Handle SImode.
1480 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1483 2018-05-22 Martin Sebor <msebor@redhat.com>
1486 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1487 or ajust the presumed bound on an operation to avoid unnecessary
1490 2018-05-22 Martin Sebor <msebor@redhat.com>
1492 PR tree-optimization/85826
1493 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1494 assuming that a DECL necesarily has a constant size.
1496 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1499 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1501 2018-05-22 Richard Biener <rguenther@suse.de>
1503 PR tree-optimization/85834
1504 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1505 non-constant and non-zero memset arguments.
1507 2018-05-22 Martin Liska <mliska@suse.cz>
1510 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1512 2018-05-22 Richard Biener <rguenther@suse.de>
1514 PR tree-optimization/85863
1515 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1516 comparisons when vectype is specified.
1517 (vectorizable_condition): Do not specify vectype for
1518 vect_is_simple_cond when SLP vectorizing.
1520 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1523 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1524 define __ibm128 as long double.
1525 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1526 as a distinct type when IEEE 128-bit support is enabled.
1527 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1528 128-bit types to use the correct functions.
1529 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1530 convert between 128-bit floating point types that have different
1531 modes but the same representation, instead of using gen_lowpart to
1533 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1535 (IFKF_reg): New attributes to give the register constraints for
1537 (extend<mode>tf2_internal): New insns to mark an explicit
1538 conversion between 128-bit floating point types that have a
1539 different mode but share the same representation.
1541 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1543 PR tree-optimization/85814
1544 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1545 a null return from get_strinfo when unsharing the next
1546 strinfo in the chain.
1548 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1551 * varasm.c (weak_finish): Clean up weak_decls.
1553 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1555 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1556 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1557 UNSPEC_UADALP values.
1558 * config/aarch64/iterators.md (ABAL): New int iterator.
1561 (sur): Add mappings for the above.
1562 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1564 (aarch64_<sur>abal<mode>_4): Likewise.
1565 (aarch64_<sur>adalp<mode>_3): Likewise.
1566 (<sur>sadv16qi): New define_expand.
1568 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1570 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1571 (*movdf_internal): Ditto.
1572 (*rcpsf2_sse): Ditto.
1573 (*rsqrtsf2_sse): Ditto.
1574 (*sqrt<mode>2_sse): Ditto.
1576 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1578 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1580 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1581 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1582 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1583 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1585 * config/aarch64/arm_neon.h: Likewise.
1586 * config/aarch64/iterators.md (VQ_I): New.
1588 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1590 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1592 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1595 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1597 * config/nds32/constraints.md (S): New constraint.
1598 * config/nds32/nds32.md (call_internal): Use constraint S.
1599 (call_value_internal): Likewise.
1600 (sibcall_internal): Likewise.
1601 (sibcall_value_internal): Likewise.
1603 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1604 Chung-Ju Wu <jasonwucj@gmail.com>
1606 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1609 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1610 Chung-Ju Wu <jasonwucj@gmail.com>
1612 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1613 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1614 (nds32_rtx_costs_impl): Simplify.
1615 (nds32_address_cost_impl): Simplify.
1616 (nds32_init_rtx_costs): New function.
1617 (nds32_rtx_costs_speed_prefer): Likewise.
1618 (nds32_rtx_costs_size_prefer): Likewise.
1619 (nds32_address_cost_speed_prefer): Likewise.
1620 (nds32_address_cost_speed_fwprop): Likewise.
1621 (nds32_address_cost_size_prefer): Likewise.
1622 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1623 * config/nds32/nds32.c (nds32_option_override): Use
1624 nds32_init_rtx_costs function.
1626 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1628 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1629 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1630 (TARGET_PIPELINE_N8): Likewise.
1631 (TARGET_PIPELINE_N10): Likewise.
1632 (TARGET_PIPELINE_N13): Likewise.
1633 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1635 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1637 * config/nds32/nds32-fpu.md: Update copyright year.
1639 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1641 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1643 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1645 * config/nds32/nds32.c
1646 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1647 * config/nds32/nds32.opt (minline-asm-r15): New option.
1649 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1651 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1653 * config/nds32/nds32.md (abssi2): New pattern.
1655 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1657 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1658 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1659 Merge insn pattern from sse_cvtsi2ss<round_name> and
1660 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1661 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1662 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1663 using SWI48 mode iterator.
1664 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1665 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1666 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1667 pattern from sse_cvttss2si<round_saeonly_name>
1668 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1669 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1670 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1671 using SWI48 mode iterator.
1672 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1673 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1674 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1675 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1676 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1677 using SWI48 mode iterator.
1678 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1679 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1680 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1681 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1682 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1683 SWI48 mode iterator.
1684 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1685 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1686 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1687 pattern from sse_cvttsd2si<round_saeonly_name>
1688 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1690 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1692 * config/nds32/nds32-md-auxiliary.c
1693 (nds32_valid_smw_lwm_base_p): Refine.
1694 (nds32_output_smw_single_word): Refine.
1695 (nds32_output_smw_double_word): New.
1696 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1698 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1700 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1701 (nds32_output_stack_pop): Refine.
1702 (nds32_expand_unaligned_load): Refine.
1703 (nds32_expand_unaligned_store): Refine.
1705 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1706 Chung-Ju Wu <jasonwucj@gmail.com>
1708 * config/nds32/constants.md: Add TP_REGNUM constant.
1709 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1710 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1712 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1713 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1714 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1715 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1716 and PIC code generation.
1717 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1719 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1721 * config/nds32/nds32.md: Support TLS and PIC.
1722 * config/nds32/nds32.c: Support TLS and PIC.
1723 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1724 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1727 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1730 mode with E_ prefix.
1732 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1733 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1736 * config/nds32/nds32-md-auxiliary.c
1737 (symbolic_reference_mentioned_p): New.
1738 (nds32_legitimize_ict_address): New.
1739 (nds32_expand_ict_move): New.
1740 (nds32_indirect_call_referenced_p): New.
1741 (nds32_symbol_binds_local_p): Delete.
1742 (nds32_long_call_p): Modify.
1743 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1744 * config/nds32/nds32-protos.h
1745 (symbolic_reference_mentioned_p): Declare.
1746 (nds32_legitimize_ict_address): Declare.
1747 (nds32_expand_ict_move): Declare.
1748 (nds32_indirect_call_referenced_p): Declare.
1749 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1750 (nds32_relax_group): Use nds32_ict_const_p as condition.
1751 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1752 (nds32_asm_file_start): Output ict_model directive in asm code.
1753 (nds32_legitimate_address_p): Consider indirect call.
1754 (nds32_print_operand): Consider indirect call.
1755 (nds32_print_operand_address): Consider indirect call.
1756 (nds32_insert_attributes): Handle "indirect_call" attribute.
1757 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1758 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1759 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1760 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1761 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1762 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1763 (TARGET_ICT_MODEL_SMALL): Define.
1764 (TARGET_ICT_MODEL_LARGE): Define.
1765 * config/nds32/nds32.md (movsi): Consider ict model.
1766 (call, call_value): Consider ict model.
1767 (sibcall, sibcall_value): Consider ict model.
1768 * config/nds32/nds32.opt (mict-model): New option.
1769 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1772 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1773 Monk Chiang <sh.chiang04@gmail.com>
1774 Jim Wilson <jimw@sifive.com>
1776 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1777 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1778 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1779 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1780 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1781 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1782 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1783 compute save_libcall_adjustment properly.
1784 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1785 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1786 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1787 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1788 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1789 (ABI_SPEC): Handle mabi=ilp32e.
1790 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1791 (RVE): Add RVE mask.
1792 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1793 <-march>: Add rv32e as an example.
1795 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1798 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1799 (intra_create_variable_infos): Handle C++ constructors.
1801 2018-05-18 Martin Liska <mliska@suse.cz>
1803 * passes.def: Remove a redundant pass.
1805 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1808 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1810 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1812 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1814 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1816 (armv2, armv2a, armv3, armv3m): Delete architectures.
1817 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1818 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1819 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1821 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1822 (*mulsidi3adddi): Likewise.
1823 (mulsidi3): Likewise.
1824 (*mulsidi3_nov6): Likewise.
1825 (umulsidi3): Likewise.
1826 (umulsidi3_nov6): Likewise.
1827 (umaddsidi4): Likewise.
1828 (*umulsidi3adddi): Likewise.
1829 (smulsi3_highpart): Likewise.
1830 (*smulsi3_highpart_nov6): Likewise.
1831 (umulsi3_highpart): Likewise.
1832 (*umulsi3_highpart_nov6): Likewise.
1833 * config/arm/arm.h (arm_arch3m): Delete.
1834 * config/arm/arm.c (arm_arch3m): Delete.
1835 (arm_option_override_internal): Update armv3-related comment.
1836 (arm_configure_build_target): Delete use of isa_bit_mode32.
1837 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1838 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1839 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1841 (mulusa3): Likewise.
1842 * config/arm/arm-protos.h (arm_arch3m): Delete.
1843 * config/arm/arm-tables.opt: Regenerate.
1844 * config/arm/arm-tune.md: Likewise.
1845 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1846 deleted architectures.
1848 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1850 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1851 (armv5t, armv5te): New features.
1852 (ARMv5, ARMv5e): Delete fgroups.
1853 (ARMv5t, ARMv5te): Adjust for above changes.
1855 (armv5, armv5e): Delete arches.
1856 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1858 (*call_reg_arm): Likewise.
1859 (*call_value_reg_armv5): Likewise.
1860 (*call_value_reg_arm): Likewise.
1861 (*call_symbol): Likewise.
1862 (*call_value_symbol): Likewise.
1863 (*sibcall_insn): Likewise.
1864 (*sibcall_value_insn): Likewise.
1866 (prefetch): Likewise.
1867 (define_split and define_peephole2 dependent on arm_arch5):
1869 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1871 (TARGET_ARM_QBIT): Likewise.
1872 (TARGET_DSP_MULTIPLY): Likewise.
1873 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1874 (arm_arch5, arm_arch5e): Delete.
1875 (arm_arch5t, arm_arch5te): Declare.
1876 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1877 (arm_arch5t): Declare.
1878 (arm_option_reconfigure_globals): Update for the above.
1879 (arm_options_perform_arch_sanity_checks): Update comment, replace
1880 use of arm_arch5 with arm_arch5t.
1881 (use_return_insn): Likewise.
1882 (arm_emit_call_insn): Likewise.
1883 (output_return_instruction): Likewise.
1884 (arm_final_prescan_insn): Likewise.
1885 (arm_coproc_builtin_available): Likewise.
1886 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1887 arm_arch5e with arm_arch5t and arm_arch5te.
1888 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1889 (arm_arch5t, arm_arch5te): Declare.
1890 * config/arm/arm-tables.opt: Regenerate.
1891 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1892 * config/arm/t-multilib: Likewise.
1893 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1894 instead of arm_arch5.
1895 (*call_reg_thumb1): Likewise.
1896 (*call_value_reg_thumb1_v5): Likewise.
1897 (*call_value_reg_thumb1): Likewise.
1898 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1900 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1902 2018-05-18 Martin Liska <mliska@suse.cz>
1904 PR gcov-profile/84846
1905 * doc/gcov.texi: Document -t option of gcov tool.
1907 2018-05-18 Martin Liska <mliska@suse.cz>
1909 PR gcov-profile/84846
1910 * gcov.c (print_usage): Add new -t option.
1911 (process_args): Handle the option.
1912 (generate_results): Use stdout as output when requested by
1915 2018-05-18 Martin Liska <mliska@suse.cz>
1917 PR gcov-profile/84846
1918 * coverage.c (coverage_init): Write PWD to .gcno file.
1919 * doc/gcov.texi: Document how working directory is printed.
1920 * gcov-dump.c (dump_gcov_file): Print PWD.
1921 * gcov.c (output_intermediate_file): Likewise.
1922 (read_graph_file): Read PWD string.
1923 (output_lines): Print PWD.
1925 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1928 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1929 for retval and return false if all args to phi are zero.
1931 2018-05-18 Richard Biener <rguenther@suse.de>
1933 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1935 (evrp_dom_walker::before_dom_children): Call it.
1937 2018-05-18 Richard Biener <rguenther@suse.de>
1939 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1940 results when processing array refs with variable index.
1942 2018-05-18 Toon Moene <toon@moene.org>
1944 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1945 directly after that of -floop-interchange. Indicate that both
1946 options are enabled by default when specifying -O3.
1948 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1950 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1951 iterator. Delete separate integer-mode vec_set<mode> expander.
1952 (aarch64_simd_vec_setv2di): Delete.
1953 (vec_setv2di): Delete.
1954 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1955 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1956 the "w, r" alternative.
1958 2018-05-18 Martin Liska <mliska@suse.cz>
1960 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1961 * tree-pass.h (make_pass_lower_switch_O0): New function.
1962 * tree-switch-conversion.c (node_has_low_bound): Remove.
1963 (node_has_high_bound): Likewise.
1964 (node_is_bounded): Likewise.
1965 (class pass_lower_switch): Make it a template type and create
1967 (pass_lower_switch::execute): Add template argument.
1968 (make_pass_lower_switch): New function.
1969 (make_pass_lower_switch_O0): New function.
1970 (do_jump_if_equal): Remove.
1971 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1972 all the hard work to tree optimization passes.
1974 2018-05-18 Martin Liska <mliska@suse.cz>
1976 * dbgcnt.c (limit_low): Renamed from limit.
1977 (limit_high): New variable.
1978 (dbg_cnt_is_enabled): Check for upper limit.
1979 (dbg_cnt): Adjust dumping.
1980 (dbg_cnt_set_limit_by_index): Add new argument for high
1982 (dbg_cnt_set_limit_by_name): Likewise.
1983 (dbg_cnt_process_single_pair): Parse new format.
1984 (dbg_cnt_process_opt): Use strtok.
1985 (dbg_cnt_list_all_counters): Remove 'value' and add
1987 * doc/invoke.texi: Document changes.
1989 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1991 * doc/sourcebuild.texi (scalar_all_fma): Document.
1992 * tree.def (FMA_EXPR): Delete.
1993 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1994 * internal-fn.c (ternary_direct): New macro.
1995 (expand_ternary_optab_fn): Likewise.
1996 (direct_ternary_optab_supported_p): Likewise.
1997 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1998 * builtins.c (fold_builtin_fma): Delete.
1999 (fold_builtin_3): Don't call it.
2000 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
2001 * expr.c (expand_expr_real_2): Likewise.
2002 * fold-const.c (operand_equal_p): Likewise.
2003 (fold_ternary_loc): Likewise.
2004 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
2005 * gimple.c (DEFTREECODE): Likewise.
2006 * gimplify.c (gimplify_expr): Likewise.
2007 * optabs-tree.c (optab_for_tree_code): Likewise.
2008 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2009 * tree-eh.c (operation_could_trap_p): Likewise.
2010 (stmt_could_throw_1_p): Likewise.
2011 * tree-inline.c (estimate_operator_cost): Likewise.
2012 * tree-pretty-print.c (dump_generic_node): Likewise.
2013 (op_code_prio): Likewise.
2014 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2015 * tree-ssa-operands.c (get_expr_operands): Likewise.
2016 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2017 * fold-const-call.h (fold_fma): Delete.
2018 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2019 CFN_FNMA and CFN_FNMS.
2021 * genmatch.c (combined_fn): New enum.
2022 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2023 (commutative_op): New function.
2024 (commutate): Use it. Handle more than 2 operands.
2025 (dt_operand::gen_gimple_expr): Use commutative_op.
2026 (parser::parse_expr): Allow :c to be used with non-binary
2027 operators if the commutative operand is known.
2028 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2029 CFN_FMS, CFN_FNMA and CFN_FNMS.
2030 (backprop::process_assign_use): Remove FMA_EXPR handling.
2031 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2032 (gen_hsa_fma): New function.
2033 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2034 IFN_FNMA and IFN_FNMS.
2035 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2036 * gimple-fold.h (follow_all_ssa_edges): Declare.
2037 * gimple-fold.c (follow_all_ssa_edges): New function.
2038 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2039 gimple_build interface and use follow_all_ssa_edges to fold the result.
2040 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2041 instead of checking for optabs directly.
2042 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2043 rather than FMA_EXPRs.
2044 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2045 call to IFN_FMA instead of an FMA_EXPR.
2047 2018-05-17 Jim Wilson <jimw@sifive.com>
2049 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2050 sign extended promoted subreg, and the range does not have the sign bit
2051 set, then do a sign extend.
2053 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2054 test, check for sign extended subreg and/or constant operands, and
2055 do a sign extend in that case.
2057 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2059 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2060 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2062 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2063 Change logics_shift_reg to logics_shift_imm.
2064 (thunderx2t99_fp_loadpair_basic): Delete.
2065 (thunderx2t99_fp_storepair_basic): Delete.
2066 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2067 (thunderx2t99_asimd_polynomial): Delete.
2068 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2069 and neon_fp_mul_d_scalar_q.
2070 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2071 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2072 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2073 (thunderx2t99_asimd_lut): Add missing tbl types.
2074 (thunderx2t99_asimd_ext): Delete.
2075 (thunderx2t99_asimd_load1_1_mult): Delete.
2076 (thunderx2t99_asimd_load1_2_mult): Delete.
2077 (thunderx2t99_asimd_load1_ldp): New.
2078 (thunderx2t99_asimd_load1): New.
2079 (thunderx2t99_asimd_load2): Add missing *load2* types.
2080 (thunderx2t99_asimd_load3): New.
2081 (thunderx2t99_asimd_load4): New.
2082 (thunderx2t99_asimd_store1_1_mult): Delete.
2083 (thunderx2t99_asimd_store1_2_mult): Delete.
2084 (thunderx2t99_asimd_store2_mult): Delete.
2085 (thunderx2t99_asimd_store2_onelane): Delete.
2086 (thunderx2t99_asimd_store_stp): New.
2087 (thunderx2t99_asimd_store1): New.
2088 (thunderx2t99_asimd_store2): New.
2089 (thunderx2t99_asimd_store3): New.
2090 (thunderx2t99_asimd_store4): New.
2092 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2094 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2095 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2097 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2098 Segher Boessenkool <segher@kernel.crashing.org>
2101 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2104 2018-05-17 Richard Biener <rguenther@suse.de>
2106 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2107 for pruning loop and prune defs feeding only already visited PHIs.
2109 2018-05-17 Richard Biener <rguenther@suse.de>
2111 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2113 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2114 Richard Biener <rguenther@suse.de>
2116 PR tree-optimization/85793
2117 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2118 for VMAT_ELEMENTWISE.
2120 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2122 * internal-fn.h (lookup_internal_fn): Declare
2123 * internal-fn.c (lookup_internal_fn): New function.
2124 * gimple.c (gimple_build_call_from_tree): Handle calls to
2126 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2127 internal function names.
2128 * tree-pretty-print.c (dump_generic_node): Likewise.
2129 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2131 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2133 * gimple-fold.h (gimple_build): Make the function forms take
2134 combined_fn rather than built_in_function.
2135 (gimple_simplify): Likewise.
2136 * gimple-match-head.c (gimple_simplify): Likewise.
2137 * gimple-fold.c (gimple_build): Likewise.
2138 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2139 rather than gimple_build_call_internal.
2140 (get_initial_defs_for_reduction): Likewise.
2141 (vect_create_epilog_for_reduction): Likewise.
2142 (vectorizable_live_operation): Likewise.
2144 2018-05-17 Martin Liska <mliska@suse.cz>
2146 * gimple-ssa-sprintf.c (format_directive): Do not use
2147 space in between 'G_' and '('.
2149 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2152 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2153 even if the mask is not all ones.
2156 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2158 (ix86_gimple_fold_builtin): Likewise.
2161 * config/i386/i386.c: Include tree-vector-builder.h.
2162 (ix86_vector_shift_count): New function.
2163 (ix86_fold_builtin): Fold shift builtins by scalar count.
2164 (ix86_gimple_fold_builtin): Likewise.
2166 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2167 _mm512_setzero): New intrinsics.
2169 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2170 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2172 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2173 code generation for cases where splatting a value is not useful.
2174 * simplify-rtx.c (simplify_ternary_operation): Simplify
2175 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2176 mode to a vec_concat.
2178 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2180 * config.gcc: Support "goldmont-plus".
2181 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2183 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2184 PROCESSOR_GOLDMONT_PLUS.
2185 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2186 (processor_target_table): Add "goldmont-plus".
2187 (PTA_GOLDMONT_PLUS): Define.
2188 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2189 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2190 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2191 (fold_builtin_cpu): Add "goldmont-plus".
2192 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2193 (ix86_option_override_internal): Add "goldmont-plus".
2194 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2195 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2196 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2197 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2199 2018-05-17 Richard Biener <rguenther@suse.de>
2201 PR tree-optimization/85757
2202 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2203 remove defs that only feed that PHI from further processing.
2205 2018-05-16 Jim Wilson <jimw@sifive.com>
2207 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2209 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2211 2018-05-16 Mark Wielaard <mark@klomp.org>
2213 * dwarf2out.c (count_index_strings): New function.
2214 (output_indirect_strings): Call count_index_strings and generate
2215 header for dwarf_version >= 5.
2217 2018-05-16 Mark Wielaard <mark@klomp.org>
2219 * dwarf2out.c (dwarf_FORM): New function.
2220 (set_indirect_string): Use dwarf_FORM.
2221 (reset_indirect_string): Likewise.
2222 (size_of_die): Likewise.
2223 (value_format): Likewise.
2224 (output_die): Likewise.
2225 (add_skeleton_AT_string): Likewise.
2226 (output_macinfo_op): Likewise.
2227 (index_string): Likewise.
2228 (output_index_string_offset): Likewise.
2229 (output_index_string): Likewise.
2230 (count_index_strings): Likewise.
2232 2018-05-16 Carl Love <cel@us.ibm.com>
2234 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2235 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2237 2018-05-16 Martin Jambor <mjambor@suse.cz>
2239 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2240 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2242 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2244 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2245 (fnma<mode>4): Likewise.
2246 (fms<mode>4): Likewise.
2247 (fnms<mode>4): Likewise.
2248 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2249 (aarch64_fnma<mode>4): Likewise.
2250 (aarch64_fms<mode>4): Likewise.
2251 (aarch64_fnms<mode>4): Likewise.
2252 (aarch64_fnmadd<mode>4): Likewise.
2254 2018-05-16 Jason Merrill <jason@redhat.com>
2256 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2258 2018-05-16 Richard Biener <rguenther@suse.de>
2260 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2261 (dump_stmt_cost): Declare.
2262 (add_stmt_cost): Dump cost we add.
2263 (add_stmt_costs): New function.
2264 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2266 (vect_analyze_stmt): Adjust prototype.
2267 (vectorizable_condition): Likewise.
2268 (vectorizable_live_operation): Likewise.
2269 (vectorizable_reduction): Likewise.
2270 (vectorizable_induction): Likewise.
2271 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2272 cost vector to pass to vectorizable_ and record afterwards.
2273 (vect_model_reduction_cost): Take cost vector argument and adjust.
2274 (vect_model_induction_cost): Likewise.
2275 (vectorizable_reduction): Likewise.
2276 (vectorizable_induction): Likewise.
2277 (vectorizable_live_operation): Likewise.
2278 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2279 SLP_TREE_NUMBER_OF_VEC_STMTS.
2280 (vect_analyze_slp_cost_1): Remove.
2281 (vect_analyze_slp_cost): Likewise.
2282 (vect_slp_analyze_node_operations): Take visited args and
2283 a target cost vector. Avoid processing already visited stmt sets.
2284 (vect_slp_analyze_operations): Use a local cost vector to gather
2285 costs and register those of non-discarded instances.
2286 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2287 (vect_schedule_slp_instance): Remove copying of
2288 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2290 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2291 adding cost. Record cost entry location.
2292 (vect_prologue_cost_for_slp_op): Function to compute cost of
2293 a constant or invariant generated for SLP vect in the prologue,
2294 split out from vect_analyze_slp_cost_1.
2295 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2296 (vect_model_promotion_demotion_cost): Likewise.
2297 (vect_model_store_cost): Likewise, make static.
2298 (vect_model_load_cost): Likewise.
2299 (vectorizable_bswap): Add cost vector arg and adjust.
2300 (vectorizable_call): Likewise.
2301 (vectorizable_simd_clone_call): Likewise.
2302 (vectorizable_conversion): Likewise.
2303 (vectorizable_assignment): Likewise.
2304 (vectorizable_shift): Likewise.
2305 (vectorizable_operation): Likewise.
2306 (vectorizable_store): Likewise.
2307 (vectorizable_load): Likewise.
2308 (vectorizable_condition): Likewise.
2309 (vectorizable_comparison): Likewise.
2310 (can_vectorize_live_stmts): Likewise.
2311 (vect_analyze_stmt): Likewise.
2312 (vect_transform_stmt): Adjust calls to vectorizable_*.
2313 * tree-vectorizer.c: Include gimple-pretty-print.h.
2314 (dump_stmt_cost): New function.
2316 2018-05-16 Richard Biener <rguenther@suse.de>
2318 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2319 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2320 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2321 (check_name): New callback.
2322 (dse_classify_store): Track cycles via a visited bitmap of PHI
2323 defs and simplify handling of in-loop and across loop dead stores
2324 and properly fail for loop-variant refs. Handle byte-tracking with
2325 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2328 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2330 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2331 (vect_get_mask_type_for_stmt): Likewise.
2332 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2334 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2335 to determine the statement's vector type and the vector type that
2336 should be used for calculating nunits. Deal with cases in which
2337 the type has to be deferred.
2338 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2339 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2340 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2341 (vect_determine_vf_for_stmt): New functions, split out from...
2342 (vect_determine_vectorization_factor): ...here.
2343 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2344 (vect_get_mask_type_for_stmt): New functions, split out from
2345 vect_determine_vectorization_factor.
2347 2018-05-16 Richard Biener <rguenther@suse.de>
2349 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2350 verify the [VEC_]COND_EXPR embedded comparison.
2352 2018-05-15 Martin Sebor <msebor@redhat.com>
2354 PR tree-optimization/85753
2355 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2356 RECORD_TYPE in addition to ARRAY_TYPE.
2358 2018-05-15 Martin Sebor <msebor@redhat.com>
2361 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2363 2018-05-15 Richard Biener <rguenther@suse.de>
2365 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2366 add by_clobber_p one. Change algorithm to collect all defs
2367 representing uses we need to walk and try reducing them to
2368 a single one before failing.
2369 (dse_dom_walker::dse_optimize_stmt): Adjust.
2371 2018-05-13 Mark Wielaard <mark@klomp.org>
2373 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2374 (size_of_loc_descr): Likewise.
2375 (output_loc_operands): Likewise.
2376 (output_loc_operands_raw): Likewise.
2377 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2378 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2379 (hash_loc_operands): Likewise.
2380 (compare_loc_operands): Likewise.
2382 2018-05-14 Mark Wielaard <mark@klomp.org>
2384 * dwarf2out.c (count_index_addrs): New function.
2385 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2387 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2389 PR tree-optimization/83648
2390 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2391 return value as malloc candidate.
2393 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2396 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2397 as true in call to suggest_attribute.
2399 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2401 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2402 -mreadonly-in-sdata.
2404 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2406 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2408 (aarch64_crypto_aesd_fused): Likewise.
2410 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2412 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2413 (movsi_aarch64): Likewise.
2414 (load_pairsi): Likewise.
2415 (load_pairdi): Likewise.
2416 (store_pairsi): Likewise.
2417 (store_pairdi): Likewise.
2418 (load_pairsf): Likewise.
2419 (load_pairdf): Likewise.
2420 (store_pairsf): Likewise.
2421 (store_pairdf): Likewise.
2422 (zero_extend): Likewise.
2423 (trunc): Swap alternatives.
2424 (fcvt_target): Add '?' to prefer w over r.
2426 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2429 * config/i386/i386.md: Disallow non-commutative arithmetics in
2430 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2431 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2432 in the peephole2 before it.
2434 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2436 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2437 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2438 (ix86_handle_option): Handle -mcldemote.
2439 * config.gcc: New header.
2440 * config/i386/cldemoteintrin.h: New file.
2441 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2442 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2444 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2445 OPTION_MASK_ISA_CLDEMOTE.
2446 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2447 (ix86_valid_target_attribute_inner_p): Ditto.
2448 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2449 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2450 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2451 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2452 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2454 * config/i386/i386.opt: Add -mcldemote.
2455 * config/i386/x86intrin.h: New header.
2456 * doc/invoke.texi: Add -mcldemote.
2458 2018-05-14 Richard Biener <rguenther@suse.de>
2460 * doc/match-and-simplify.texi: Adjust :s documentation.
2462 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2464 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2465 intended memcpy size.
2466 (REORDER_45): Likewise.
2468 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2470 * sort.cc: New file.
2471 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2472 * vec.c (qsort_chk): Use gcc_qsort.
2473 * Makefile.in (OBJS-libcommon): Add sort.o.
2474 (build/sort.o): New target. Use it...
2475 (BUILD_RTL): ... here, and...
2476 (build/gencfn-macros): ... here, and...
2477 (build/genmatch): ... here.
2479 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2480 Chung-Ju Wu <jasonwucj@gmail.com>
2482 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2483 * config/nds32/nds32-graywolf.md: New file.
2484 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2485 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2487 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2488 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2489 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2490 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2491 * config/nds32/pipelines.md: Include n15 settings.
2493 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2494 Chung-Ju Wu <jasonwucj@gmail.com>
2496 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2497 * config/nds32/nds32-n13.md: New file.
2498 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2499 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2501 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2502 * config/nds32/nds32.md (pipeline_model): Add n13.
2503 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2504 * config/nds32/pipelines.md: Include n13 settings.
2506 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2507 Chung-Ju Wu <jasonwucj@gmail.com>
2509 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2510 * config/nds32/nds32-n10.md: New file.
2511 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2512 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2514 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2515 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2516 * config/nds32/nds32.md (pipeline_model): Add n10.
2517 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2518 * config/nds32/pipelines.md: Include n10 settings.
2520 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2521 Kito Cheng <kito.cheng@gmail.com>
2522 Chung-Ju Wu <jasonwucj@gmail.com>
2524 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2525 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2526 Add enum values for DSP extension instructions.
2527 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2529 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2530 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2532 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2533 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2534 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2535 * config/nds32/nds32-intrinsic.md: Likewise.
2536 * config/nds32/nds32_intrinsic.h: Likewise.
2537 * config/nds32/nds32-md-auxiliary.c: Likewise.
2538 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2539 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2540 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2541 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2542 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2543 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2544 TYPE_DMAC in switch statement.
2545 * config/nds32/nds32.c: New checking and implementation for DSP
2546 extension instructions.
2547 * config/nds32/nds32.h: Likewise.
2548 * config/nds32/nds32.md: Likewise.
2549 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2550 * config/nds32/predicates.md: Implement new predicates for DSP
2553 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2555 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2556 Reformat alternatives and attributes so it is easier to identify
2557 which constraints/attributes go with which instruction.
2558 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2559 (mov<mode>_softfloat32, FMOVE64): Likewise.
2560 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2561 (mov<mode>_softfloat64, FMOVE64): Likewise.
2563 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2565 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2567 (Basic PowerPC Built-in Functions): The new name of the
2568 subsection previously known as "PowerPC Built-in Functions".
2569 (Basic PowerPC Built-in Functions Available on all Configurations):
2571 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2572 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2573 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2574 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2576 2018-05-11 Martin Jambor <mjambor@suse.cz>
2579 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2582 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2585 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2587 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2589 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2590 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2591 (ix86_handle_option): Handle -mwaitpkg.
2592 * config.gcc: New header.
2593 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2594 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2595 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2597 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2598 OPTION_MASK_ISA_WAITPKG.
2599 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2600 (ix86_option_override_internal): Add PTA_WAITPKG.
2601 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2602 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2603 IX86_BUILTIN_TPAUSE.
2604 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2605 __builtin_ia32_umwait and __builtin_ia32_tpause.
2606 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2607 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2608 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2609 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2610 UNSPECV_TPAUSE): New.
2611 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2612 * config/i386/i386.opt: Add -mwaitpkg.
2613 * config/i386/waitpkgintrin.h: New file.
2614 * config/i386/x86intrin.h: New header.
2615 * doc/invoke.texi: Add -mwaitpkg.
2617 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2620 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2622 (cortex-m0): Use armv6s-m isa.
2623 (cortex-m0plus): Likewise.
2624 (cortex-m1): Likewise.
2625 (cortex-m0.small-multiply): Likewise.
2626 (cortex-m0plus.small-multiply): Likewise.
2627 (cortex-m1.small-multiply): Likewise.
2629 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2630 Jakub Jelinek <jakub@redhat.com>
2632 PR tree-optimization/85692
2633 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2634 source permute as well.
2636 2018-05-11 Martin Liska <mliska@suse.cz>
2639 * doc/extend.texi: Document LLVM style format for no_sanitize
2642 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2644 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2645 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2646 (mode_supports_vsx_dform_quad): Likewise.
2647 (mode_supports_vmx_dform): Move these functions to be next to the
2648 other mode_supports functions.
2649 (mode_supports_dq_form): Likewise.
2650 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2651 mode_supports_dq_form.
2652 (reg_offset_addressing_ok_p): Likewise.
2653 (offsettable_ok_by_alignment): Likewise.
2654 (rs6000_legitimate_offset_address_p): Likewise.
2655 (legitimate_lo_sum_address_p): Likewise.
2656 (rs6000_legitimize_address): Likewise.
2657 (rs6000_legitimize_reload_address): Likewise.
2658 (rs6000_secondary_reload_inner): Likewise.
2659 (rs6000_preferred_reload_class): Likewise.
2660 (rs6000_output_move_128bit): Likewise.
2662 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2664 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2665 Generate SImode target register for null target.
2666 <case IX86_BUILTIN_XGETBV>: Ditto.
2667 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2668 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2670 2018-05-10 Carl Love <cel@us.ibm.com>
2672 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2673 dcbtt and dcbtstt if operands[2] is 0.
2675 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2678 * config/i386/sse.md (usadv64qi): New expander.
2680 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2682 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2683 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2684 -maltivec=be support.
2685 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2686 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2687 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2688 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2689 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2690 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2691 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2692 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2693 altivec_vsumsws): Adjust.
2694 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2695 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2696 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2698 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2699 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2700 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2701 (altivec_lve<VI_char>x): Delete expand.
2702 (*altivec_lve<VI_char>x_internal): Rename to...
2703 (altivec_lve<VI_char>x): ... this.
2704 (altivec_lvxl_<mode>): Delete expand.
2705 (*altivec_lvxl_<mode>_internal): Rename to ...
2706 (altivec_lvxl_<mode>): ... this.
2707 (altivec_stvxl_<mode>): Delete expand.
2708 (*altivec_stvxl_<mode>_internal): Rename to ...
2709 (altivec_stvxl_<mode>): ... this.
2710 (altivec_stve<VI_char>x): Delete expand.
2711 (*altivec_stve<VI_char>x_internal): Rename to ...
2712 (altivec_stve<VI_char>x): ... this.
2713 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2714 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2715 reduc_plus_scal_<mode>): Adjust.
2716 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2718 (rs6000_cpu_cpp_builtins): Adjust.
2719 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2720 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2721 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2722 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2723 -maltivec=be support.
2724 (rs6000_split_vec_extract_var): Adjust.
2725 (rs6000_split_v4si_init): Adjust.
2726 (swap_selector_for_mode): Delete.
2727 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2728 altivec_expand_stvex_be): Delete.
2729 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2730 -maltivec=be support.
2731 (rs6000_gimple_fold_builtin): Ditto.
2732 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2734 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2735 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2736 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2737 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2738 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2739 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2740 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2741 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2742 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2743 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2744 anonymous split): Adjust.
2745 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2746 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2748 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2750 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2751 when --with-gxx-include-dir is also specified.
2752 * configure: Regenerate.
2754 2018-05-09 Jim Wilson <jimw@sifive.com>
2757 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2758 * config/riscv/t-withmultilib: New.
2759 * config/riscv/withmultilib.h: New.
2760 * doc/install.texi: Document RISC-V --with-multilib-list support.
2762 2018-05-09 Richard Biener <rguenther@suse.de>
2764 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2766 (vect_bb_vectorization_profitable_p): Adjust. Compute
2767 actual scalar cost using the cost vector and the add_stmt_cost
2770 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2772 PR rtl-optimization/85645
2773 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2774 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2776 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2778 PR rtl-optimization/85645
2779 * shrink-wrap.c (spread_components): Return a boolean saying if
2780 anything was changed.
2781 (try_shrink_wrapping_separate): Iterate spread_components until
2782 nothing changes anymore.
2784 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2786 PR rtl-optimization/85645
2787 * regrename.c (build_def_use): Also kill the chains that include the
2788 destination of a REG_CFA_REGISTER note.
2790 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2792 PR rtl-optimization/85645
2793 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2794 insn that has a REG_CFA_REGISTER note.
2796 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2798 * cfgexpand.c (expand_clobber): New function.
2799 (expand_gimple_stmt_1): Use it.
2800 * tree-vect-stmts.c (vect_clobber_variable): New function,
2802 (vectorizable_simd_clone_call): ...here.
2803 (vectorizable_store): Emit a clobber either side of an
2804 IFN_STORE_LANES sequence.
2805 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2807 2018-05-09 Tom de Vries <tom@codesourcery.com>
2810 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2811 (define_insn "trap_if_false"): Add exit after trap.
2813 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2815 PR rtl-optimization/85638
2816 * bb-reorder.c: Include common/common-target.h.
2817 (create_forwarder_block): New function extracted from...
2818 (fix_up_crossing_landing_pad): ...here. Rename into...
2819 (dw2_fix_up_crossing_landing_pad): ...this.
2820 (sjlj_fix_up_crossing_landing_pad): New function.
2821 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2822 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2823 from both partitions and exit the loop after one iteration.
2825 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2828 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2830 (Basic PowerPC Built-in Functions): The new name of the
2831 subsection previously known as "PowerPC Built-in Functions".
2832 (Basic PowerPC Built-in Functions Available on all Configurations):
2834 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2836 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2838 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2840 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2843 2018-05-08 Jim Wilson <jimw@sifive.com>
2845 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2846 (LD_EMUL_SUFFIX): New.
2847 (LINK_SPEC): Use it.
2849 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2851 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2853 (Basic PowerPC Built-in Functions): The new name of the
2854 subsection previously known as "PowerPC Built-in Functions".
2855 (Basic PowerPC Built-in Functions Available on all Configurations):
2857 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2859 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2861 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2863 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2866 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2869 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2870 after cmpelim optimization.
2872 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2874 * config.gcc: Support "goldmont".
2875 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2876 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2878 * config/i386/i386.c (m_GOLDMONT): Define.
2879 (processor_target_table): Add "goldmont".
2880 (PTA_GOLDMONT): Define.
2881 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2882 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2883 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2884 (fold_builtin_cpu): Add "goldmont".
2885 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2886 (ix86_option_override_internal): Add "goldmont".
2887 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2888 (processor_type): Add PROCESSOR_GOLDMONT.
2889 * config/i386/i386.md: Add CPU "glm".
2890 * config/i386/glm.md: New file.
2891 * config/i386/x86-tune.def: Add m_GOLDMONT.
2892 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2894 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2897 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2899 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2900 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2901 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2904 * config/i386/i386.c (ix86_fold_builtin): Handle
2905 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2908 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2909 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2911 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2914 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2915 (check_arch): Likewise.
2916 (check_fpu): Return the result rather than printing it.
2917 (end arch): Fix operator precedence.
2918 (end cpu): Likewise.
2919 (END): Print the result from check_fpu.
2921 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2922 Alan Hayward <alan.hayward@arm.com>
2923 David Sherwood <david.sherwood@arm.com>
2925 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2926 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2927 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2928 (*fcmuo<mode>_and): New patterns.
2930 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2932 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2933 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2934 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2935 (cmp_op, sve_imm_con): New code attributes.
2936 (SVE_COND_INT_CMP, imm_con): Delete.
2937 (cmp_op): Remove above unspecs from int attribute.
2938 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2940 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2941 comparison-specific unspecs.
2942 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2943 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2944 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2945 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2946 (*vec_fcm<cmp_op><mode>): Rename to...
2947 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2948 (*vec_fcmuo<mode>): Rename to...
2949 (*fcmuo<mode>): ...this and adjust likewise.
2950 (*pred_fcm<cmp_op><mode>): New pattern.
2951 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2952 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2954 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2956 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2957 (aarch64_emit_sve_predicated_cond): New function.
2958 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2959 (aarch64_emit_unspec_cond_or): Replace with...
2960 (aarch64_emit_sve_or_conds): ...this new function. Use
2961 aarch64_emit_sve_ptrue_op for the individual comparisons and
2962 aarch64_emit_binop to OR them together.
2963 (aarch64_emit_inverted_unspec_cond): Replace with...
2964 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2965 aarch64_emit_sve_ptrue_op for the comparison and
2966 aarch64_emit_unop to invert the result.
2967 (aarch64_expand_sve_vec_cmp_float): Update after the above
2968 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2970 2018-05-07 Nathan Sidwell <nathan@acm.org>
2972 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2973 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2974 (Backwards Compatibility): Likewise.
2976 2018-05-07 Luis Machado <luis.machado@linaro.org>
2980 2018-05-07 Luis Machado <luis.machado@linaro.org>
2982 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2983 <prefetch_dynamic_strides>: New const bool field.
2984 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2985 prefetch_dynamic_strides.
2986 (exynosm1_prefetch_tune): Likewise.
2987 (thunderxt88_prefetch_tune): Likewise.
2988 (thunderx_prefetch_tune): Likewise.
2989 (thunderx2t99_prefetch_tune): Likewise.
2990 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2991 (aarch64_override_options_internal): Update to set
2992 PARAM_PREFETCH_DYNAMIC_STRIDES.
2993 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2994 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2995 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2996 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2997 prefetch-dynamic-strides setting.
2999 2018-05-07 Luis Machado <luis.machado@linaro.org>
3001 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3002 <minimum_stride>: New const int field.
3003 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3004 minimum_stride field.
3005 (exynosm1_prefetch_tune): Likewise.
3006 (thunderxt88_prefetch_tune): Likewise.
3007 (thunderx_prefetch_tune): Likewise.
3008 (thunderx2t99_prefetch_tune): Likewise.
3009 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3010 (aarch64_override_options_internal): Update to set
3011 PARAM_PREFETCH_MINIMUM_STRIDE.
3012 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3013 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3014 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3015 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3016 stride is constant and is below the minimum stride threshold.
3018 2018-05-07 Luis Machado <luis.machado@linaro.org>
3020 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3023 2018-05-07 Luis Machado <luis.machado@linaro.org>
3025 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3026 <prefetch_dynamic_strides>: New const bool field.
3027 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3028 prefetch_dynamic_strides.
3029 (exynosm1_prefetch_tune): Likewise.
3030 (thunderxt88_prefetch_tune): Likewise.
3031 (thunderx_prefetch_tune): Likewise.
3032 (thunderx2t99_prefetch_tune): Likewise.
3033 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3034 (aarch64_override_options_internal): Update to set
3035 PARAM_PREFETCH_DYNAMIC_STRIDES.
3036 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3037 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3038 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3039 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3040 prefetch-dynamic-strides setting.
3042 2018-05-07 Luis Machado <luis.machado@linaro.org>
3044 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3045 <minimum_stride>: New const int field.
3046 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3047 minimum_stride field.
3048 (exynosm1_prefetch_tune): Likewise.
3049 (thunderxt88_prefetch_tune): Likewise.
3050 (thunderx_prefetch_tune): Likewise.
3051 (thunderx2t99_prefetch_tune): Likewise.
3052 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3053 (aarch64_override_options_internal): Update to set
3054 PARAM_PREFETCH_MINIMUM_STRIDE.
3055 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3056 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3057 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3058 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3059 stride is constant and is below the minimum stride threshold.
3061 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3064 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3065 the type is addressable. Don't force op into register if it has
3068 2018-05-05 Roland McGrath <mcgrathr@google.com>
3071 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3072 any section for which we don't know a specific type it should have,
3073 regardless of name. Previously this was done only for the exact
3074 names ".init_array", ".fini_array", and ".preinit_array".
3075 (default_elf_asm_named_section): Add comment about
3076 relationship with default_section_type_flags and SECTION_NOTYPE.
3077 (get_section): Don't consider it a type conflict if one side has
3078 SECTION_NOTYPE and the other doesn't, as long as neither has the
3079 SECTION_BSS et al used in the default_section_type_flags logic.
3081 2018-05-05 Tom de Vries <tom@codesourcery.com>
3084 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3085 (workaround_barsyncs): New function.
3086 (nvptx_reorg): Use workaround_barsyncs.
3087 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3088 (define_expand "nvptx_membar_cta"): New define_expand.
3089 (define_insn "*nvptx_membar_cta"): New insn.
3091 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3093 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3094 To improve optimization opportunities.
3095 * builtin-types.def: The new needed builtin types for the above.
3097 2018-05-04 Richard Biener <rguenther@suse.de>
3099 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3100 * gimple-ssa-store-merging.c
3101 (imm_store_chain_info::output_merged_store): Remove redundant create,
3102 release split_store vector contents on failure.
3103 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3104 scalar stmt vector on cache hit.
3106 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3108 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3110 * config.gcc (powerpc-xilinx-eabi*): Remove.
3111 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3113 (fusion_addis_mem_combo_load): Ditto.
3114 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3116 (rs6000_cpu_cpp_builtins): Ditto.
3117 * config/rs6000/rs6000-linux.c
3118 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3119 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3120 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3122 (rs6000_setup_reg_addr_masks): Ditto.
3123 (rs6000_init_hard_regno_mode_ok): Ditto.
3124 (rs6000_option_override_internal): Ditto.
3125 (legitimate_lo_sum_address_p): Ditto.
3126 (rs6000_legitimize_address): Ditto.
3127 (rs6000_legitimize_reload_address): Ditto.
3128 (rs6000_legitimate_address_p): Ditto.
3129 (abi_v4_pass_in_fpr): Ditto.
3130 (setup_incoming_varargs): Ditto.
3131 (rs6000_gimplify_va_arg): Ditto.
3132 (rs6000_split_multireg_move): Ditto.
3133 (rs6000_savres_strategy): Ditto.
3134 (rs6000_emit_prologue_components): Ditto.
3135 (rs6000_emit_epilogue_components): Ditto.
3136 (rs6000_emit_prologue): Ditto.
3137 (rs6000_emit_epilogue): Ditto.
3138 (rs6000_elf_file_end): Ditto.
3139 (rs6000_function_value): Ditto.
3140 (rs6000_libcall_value): Ditto.
3141 * config/rs6000/rs6000.h: Ditto.
3142 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3143 (TARGET_MINMAX): ... this. New.
3144 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3145 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3146 (*movsi_internal1_single): Delete.
3147 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3148 mfpu=, mxilinx-fpu): Delete.
3149 * config/rs6000/singlefp.h: Delete.
3150 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3151 * config/rs6000/t-rs6000: Ditto.
3152 * config/rs6000/t-xilinx: Delete.
3153 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3154 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3155 (VStype_simple): Delete.
3156 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3157 * config/rs6000/xfpu.h: Delete.
3158 * config/rs6000/xfpu.md: Delete.
3159 * config/rs6000/xilinx.h: Delete.
3160 * config/rs6000/xilinx.opt: Delete.
3161 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3162 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3164 2018-05-04 Tom de Vries <tom@codesourcery.com>
3167 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3170 2018-05-04 Richard Biener <rguenther@suse.de>
3173 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3174 (expand_complex_div_wide): Likewise.
3175 (expand_complex_operations_1): Likewise.
3176 (expand_complex_libcall): Preserve EH info of the original stmt.
3177 (tree_lower_complex): Handle removed blocks.
3178 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3179 on complex multiplication and division libcall builtins.
3181 2018-05-04 Richard Biener <rguenther@suse.de>
3184 * fold-const.c (negate_expr_p): Restrict negation of operand
3185 zero of a division to when we know that can happen without
3187 (fold_negate_expr_1): Likewise.
3189 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3192 * real.h (real_nextafter): Declare.
3193 * real.c (real_nextafter): New function.
3194 * fold-const-call.c (fold_const_nextafter): New function.
3195 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3196 CASE_CFN_NEXTTOWARD.
3197 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3198 even when arg1_mode is different from arg0_mode.
3200 2018-05-03 Nathan Sidwell <nathan@acm.org>
3202 * doc/extend.texi (Deprecated Features): Remove
3204 (Backwards Compatibility): Likewise.
3205 * doc/invoke.texi (C++ Language Options): Likewise.
3206 (C++ Dialect Options): Likewise.
3208 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3211 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3212 _mm512_mask_mullox_epi64): New intrinsics.
3214 2018-05-03 Tom de Vries <tom@codesourcery.com>
3217 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3218 dump files): Add offload-tree.
3220 2018-05-03 Richard Biener <rguenther@suse.de>
3222 PR tree-optimization/85615
3223 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3224 to loops not nested in BBs loop father to avoid creating multi-entry
3227 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3229 PR tree-optimization/70291
3230 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3231 arguments. Change return type to tree. Emit libcall as a new
3232 statement rather than replacing existing one when inplace_p is true.
3233 (expand_complex_multiplication_components): New function.
3234 (expand_complex_multiplication): Expand floating-point complex
3235 multiplication using the above.
3236 (expand_complex_division): Rename inner_type parameter to type.
3237 Update expand_complex_libcall call-site.
3238 (expand_complex_operations_1): Update expand_complex_multiplication
3239 and expand_complex_division call-sites.
3241 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3244 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3245 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3246 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3247 the highest significant bit of the shift count mask is clear. In
3248 check whether and[sq]i3 is needed verify that all significant bits
3249 of the shift count other than the highest are set.
3251 2018-05-02 Tom de Vries <tom@codesourcery.com>
3254 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3255 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3256 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3257 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3258 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3259 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3260 __builtin_goacc_parlevel_size.
3262 2018-05-02 Richard Biener <rguenther@suse.de>
3264 PR tree-optimization/85597
3265 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3266 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3269 2018-05-02 Tom de Vries <tom@codesourcery.com>
3272 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3273 dump files): Add ltrans-tree.
3275 2018-05-02 Tom de Vries <tom@codesourcery.com>
3278 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3279 dump files): Add wpa-ipa.
3281 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3283 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3284 powerpc*-*-linux*paired* target.
3285 * config/rs6000/750cl.h: Delete.
3286 * config/rs6000/paired.h: Delete.
3287 * config/rs6000/paired.md: Delete.
3288 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3290 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3291 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3292 comment. Remove paired float support.
3293 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3294 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3296 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3297 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3299 * config/rs6000/rs6000.c: Remove paired float support.
3300 (paired_expand_vector_init, paired_expand_vector_move,
3301 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3302 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3303 paired_expand_builtin, paired_expand_predicate_builtin,
3304 paired_init_builtins): Delete.
3305 * config/rs6000/rs6000.h: Remove paired float support.
3306 * config/rs6000/rs6000.md: Remove paired float support.
3307 (move_from_CR_ov_bit): Delete.
3308 * config/rs6000/rs6000.opt (mpaired): Delete.
3309 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3310 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3312 2018-05-02 Richard Biener <rguenther@suse.de>
3315 * gimplify.c (gimplify_save_expr): When in SSA form allow
3316 SAVE_EXPRs to compute to SSA vars.
3318 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3321 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3322 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3323 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3324 clobber operands[2], instead use a new pseudo. Formatting fixes.
3326 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3328 PR tree-optimization/85586
3329 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3330 exit early for statements in the same group if the accesses are
3333 2018-05-02 Tom de Vries <tom@codesourcery.com>
3336 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3339 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3341 PR tree-optimization/85143
3342 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3344 2018-05-01 Tom de Vries <tom@codesourcery.com>
3347 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3348 not found" error message.
3350 2018-05-01 Tom de Vries <tom@codesourcery.com>
3353 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3354 * vec.c (test_ordered_remove_if): New function.
3355 (vec_c_tests): Call test_ordered_remove_if.
3356 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3357 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3358 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3359 VEC_ORDERED_REMOVE_IF.
3361 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3363 PR tree-optimization/82665
3364 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3365 pointer subtraction where arguments come from a memchr call.
3367 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3369 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3370 --push-state --as-needed and --pop-state instead of --as-needed and
3371 --no-as-needed if ld supports it.
3372 * configure: Regenerated.
3375 * doc/install.texi2html: Replace _002d with - and _002a with * in
3376 generated html files using sed.
3378 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3381 * gcc-rich-location.c (blank_line_before_p): New function.
3382 (use_new_line): New function.
3383 (gcc_rich_location::add_fixit_insert_formatted): New function.
3384 * gcc-rich-location.h
3385 (gcc_rich_location::add_fixit_insert_formatted): New function.
3387 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3389 * selftest.c (assert_streq): Rename "expected" and "actual" to
3390 "val1" and "val2". Extend NULL-handling to cover both inputs
3391 symmetrically, while still requiring both to be non-NULL for a pass.
3392 * selftest.h (assert_streq): Rename "expected" and "actual" to
3394 (ASSERT_EQ): Likewise.
3395 (ASSERT_EQ_AT): Likewise.
3396 (ASSERT_KNOWN_EQ): Likewise.
3397 (ASSERT_KNOWN_EQ_AT): Likewise.
3398 (ASSERT_NE): Likewise.
3399 (ASSERT_MAYBE_NE): Likewise.
3400 (ASSERT_MAYBE_NE_AT): Likewise.
3401 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3402 the assertion to pass.
3403 (ASSERT_STREQ_AT): Likewise.
3405 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3407 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3408 interaction with -pie.
3410 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3412 * selftest.h: Fix alphabetization of per-source-file selftest
3415 2018-04-30 Jason Merrill <jason@redhat.com>
3417 PR c++/61982 - dead stores to destroyed objects.
3418 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3421 2018-04-30 Jason Merrill <jason@redhat.com>
3423 * tree.c (build_clobber): New.
3424 * tree.h: Declare it.
3425 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3427 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3429 * diagnostic-show-locus.c (layout::layout): Update for
3430 location_get_source_line returning a char_span.
3431 (struct char_span): Move to input.h.
3432 (struct correction): Update for fields in char_span becoming
3434 (struct source_line): Update for location_get_source_line
3435 returning a char_span.
3436 (layout::print_line): Likewise.
3437 * edit-context.c (edited_file::print_content): Likewise.
3438 (edited_file::print_diff_hunk): Likewise.
3439 (edited_file::print_run_of_changed_lines): Likewise.
3440 (edited_file::get_num_lines): Likewise.
3441 (edited_line::edited_line): Likewise.
3442 * final.c (asm_show_source): Likewise.
3443 * input.c (location_get_source_line): Convert return type
3444 from const char * to char_span, losing the final "line_len"
3446 (dump_location_info): Update for the above.
3447 (get_substring_ranges_for_loc): Likewise. Use a char_span
3448 when handling the literal within the line.
3449 (test_reading_source_line): Update for location_get_source_line
3450 returning a char_span.
3451 * input.h (class char_span): Move here from
3452 diagnostic-show-locus.c, converting from a struct to a class.
3453 Make data members private.
3454 (char_span::operator bool): New.
3455 (char_span::length): New.
3456 (char_span::get_buffer): New.
3457 (char_span::operator[]): New.
3458 (char_span::subspan): Make const.
3459 (char_span::xstrdup): New.
3460 (location_get_source_line): Convert return type from const char *
3461 to char_span, losing the final "line_size" param.
3463 2018-04-30 Jan Hubicka <jh@suse.cz>
3465 * lto-wrapper.c (ltrans_priorities): New static var.
3466 (cmp_priority): New.
3467 (run_gcc): Read priorities and if doing parallel build order
3468 the Makefile by them.
3470 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3472 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3474 2018-04-30 Richard Biener <rguenther@suse.de>
3476 * tree-cfg.c (verify_address): Remove base argument, add
3477 flag whether to check TREE_ADDRESSABLE and do that.
3478 (verify_expr): Remove.
3479 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3480 (verify_gimple_assign_single): Likewise.
3481 (verify_gimple_switch): Likewise.
3482 (verify_expr_location_1): Dereference tp once. Add (disabled)
3483 piece from verify_expr.
3484 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3486 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3488 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3490 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3492 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3493 (small_data_pattern): Likewise.
3494 (arc_rewrite_small_data): Likewise.
3495 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3496 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3497 (get_symbol_alignment): New function.
3498 (legitimate_small_data_address_p): Likewise.
3499 (legitimate_scaled_address): Update, call
3500 legitimate_small_data_address_p.
3501 (output_sdata): New static variable.
3502 (arc_print_operand): Update how we handle small data operands.
3503 (arc_print_operand_address): Likewise.
3504 (arc_legitimate_address_p): Update, use
3505 legitimate_small_data_address_p.
3506 (arc_rewrite_small_data_p): Remove.
3507 (arc_rewrite_small_data_1): Likewise.
3508 (arc_rewrite_small_data): Likewise.
3509 (small_data_pattern): Likewise.
3510 (compact_sda_memory_operand): Update to use
3511 legitimate_small_data_address_p and get_symbol_alignment.
3512 (prepare_move_operands): Don't rewite sdata pattern.
3513 (prepare_extend_operands): Remove.
3514 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3516 (zero_extendqisi2): Likewise.
3517 (zero_extendhisi2): Likewise.
3518 (extendqihi2): Likewise.
3519 (extendqisi2): Likewise.
3520 (extendhisi2): Likewise.
3524 * config/arc/constraints.md (Usd): Change it to memory constraint.
3526 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3528 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3529 as source of std instructions.
3530 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3531 allow 6-bit constants as source for store instructions.
3532 (movdi_insn): Update instruction pattern to allow 6-bit constants
3533 as source for store instructions.
3535 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3537 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3539 2018-04-30 Nathan Sidwell <nathan@acm.org>
3540 Sandra Loosemore <sandra@codesourcery.com>
3542 * dumpfile.c (dump_open): Allow '-' for stdout.
3543 * doc/invoke.texi (Developer Options): Document dump filename
3544 determination early. Document stdin/stdout selection.
3546 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3548 Microblaze Target: PIC data text relative
3550 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3551 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3553 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3554 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3555 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3556 New addressing mode for data-text relative position indepenedent code.
3557 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3558 'ADDRESS_SYMBOLIC_TXT_REL'.
3559 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3560 (microblaze_legitimate_pic_operand): Exclude function calls from
3561 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3562 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3564 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3565 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3566 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3567 for 'address + offset'.
3568 (microblaze_expand_prologue): Add new function prologue call for
3570 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3571 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3572 table in case of TARGET_PIC_DATA_TEXT_REL.
3573 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3574 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3575 Add new macros 'UNSPEC_TEXT',
3576 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3577 + exclude function calls from 'UNSPEC_PLT' in case of data text
3579 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3580 new target hook for generating address diff vector tables in case of
3582 * doc/tm.texi : Regenerate.
3583 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3584 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3585 of addr diff vector generation.
3586 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3587 target hook definition.
3588 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3589 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3590 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3591 option for data text relative.
3593 2018-04-30 Richard Biener <rguenther@suse.de>
3595 * tree-chrec.h (evolution_function_is_constant_p): Remove
3597 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3599 2018-04-30 Richard Biener <rguenther@suse.de>
3602 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3604 2018-04-30 Richard Biener <rguenther@suse.de>
3606 PR tree-optimization/28364
3607 PR tree-optimization/85275
3608 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3609 copying first exit test.
3611 2018-04-28 Mark Wielaard <mark@klomp.org>
3613 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3615 (dwarf_AT): Handle DW_AT_addr_base.
3616 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3618 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3621 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3622 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3623 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3624 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3626 2018-04-28 Richard Biener <rguenther@suse.de>
3628 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3629 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3630 to reflect use. Only add interesting stmts.
3632 2018-04-27 Martin Jambor <mjambor@suse.cz>
3635 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3636 the jump function allows for passing through aggregate values.
3638 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3640 * input.h (in_system_header_at): Convert from macro to inline
3642 (from_macro_expansion_at): Likewise.
3643 (from_macro_definition_at): Likewise.
3645 2018-04-27 Jeff Law <law@redhat.com>
3647 * config.gcc: Mark tile* targets as deprecated/obsolete.
3649 2018-04-27 Richard Biener <rguenther@suse.de>
3651 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3654 2018-04-27 Richard Biener <rguenther@suse.de>
3656 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3658 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3660 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3661 with Yd constraint. Set "preferred_for_speed" attribute from
3662 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3664 (*movdi_internal): Ditto.
3665 (movti_interunit splitters): Remove
3666 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3667 (movdi_interunit splitters): Ditto.
3668 * config/i386/constraints.md (Ye): Remove.
3669 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3671 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3674 * config/aarch64/constraints.md (Usg): Limit to 31.
3677 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3679 PR tree-optimization/85529
3680 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3681 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3682 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3683 zero extension or masking of the MSB bit.
3684 (optimize_range_tests): Add FIRST_BB argument, pass it through
3685 to optimize_range_tests_var_bound.
3686 (maybe_optimize_range_tests, reassociate_bb): Adjust
3687 optimize_range_tests callers.
3689 2018-04-26 Richard Biener <rguenther@suse.de>
3690 Jakub Jelinek <jakub@redhat.com>
3692 * cgraph.h (symbol_table): Just declare debug method here.
3693 * symtab.c (symbol_table::debug): Define.
3695 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3697 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3699 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3701 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3702 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3703 (*movdi_internal): Substitute Yi and Yj constraint with x
3704 and Ym and Yn constraint with y constraint. Update "isa"
3705 attribute and set "preferred_for_speed" attribute from
3706 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3707 (*movsi_internal): Ditto.
3708 (*movdf_internal): Ditto.
3709 (*movsf_internal): Ditto.
3710 (*zero_extendsidi2): Ditto.
3711 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3712 (sse2_loadld): Ditto.
3713 (*vec_extract<ssevecmodelower>_0): Ditto.
3714 (*vec_extractv4si_0_zext_sse4): Ditto.
3715 (vec_concatv2di): Ditto.
3716 (*vec_dup<mode>): Ditto.
3717 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3718 * config/i386/constraints.md (Yi): Remove.
3723 2018-04-26 Nathan Sidwell <nathan@acm.org>
3725 * dumpfile.c (dump_open): New.
3726 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3727 (dump_finish): Detect stdio/stderr by value not name.
3729 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3731 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3733 2018-04-26 Tom de Vries <tom@codesourcery.com>
3736 * config/nvptx/nvptx.c (verify_neutering_jumps)
3737 (verify_neutering_labels): New function
3738 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3740 2018-04-26 Tom de Vries <tom@codesourcery.com>
3743 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3744 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3747 2018-04-26 Richard Biener <rguenther@suse.de>
3748 Tom de Vries <tom@codesourcery.com>
3751 * lto-streamer-out.c (output_function): Fixup loops if required to match
3752 discovery done in the reader.
3754 2018-04-26 Richard Biener <rguenther@suse.de>
3756 PR tree-optimization/85116
3757 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3758 have a loop exit from the single latch predecessor. Remove
3759 case of header with just condition.
3760 (ch_base::copy_headers): Exclude infinite loops from any
3762 (pass_ch::execute): Record exits.
3764 2018-04-26 Richard Biener <rguenther@suse.de>
3766 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3767 prologue cost vector and pass it to vect_get_load_cost.
3768 (vect_get_peeling_costs_all_drs): Likewise.
3769 (vect_peeling_hash_get_lowest_cost): Likewise.
3770 (vect_enhance_data_refs_alignment): Likewise.
3772 2018-04-26 Richard Biener <rguenther@suse.de>
3775 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3776 checking of integer<->pointer conversions.
3777 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3778 sign-/zero-extending pointer types.
3779 (expand_omp_for_static_chunk): Likewise.
3781 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3782 Jean Lee <xiaoyur347@gmail.com>
3784 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3785 (TARGET_ASAN_SHADOW_OFFSET): Define.
3786 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3787 true for -fsanitize=address.
3789 2018-04-25 Mark Wielaard <mark@klomp.org>
3791 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3794 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3796 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3797 than "alu", remove explicit "memory" and "imm_disp" attributes.
3798 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3801 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3802 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3803 gen_lowpart_no_emit.
3805 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3808 * config/i386/i386.c (ix86_expand_builtin): Change memory
3809 operand to XI, extend p0 to Pmode.
3810 * config/i386/i386.md: Change unspec volatile and operand
3811 1 mode to XI, change operand 0 mode to P.
3813 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3815 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3816 GET_MODE_MASK before any checking.
3817 (nds32_can_use_bset_p): Likewise.
3818 (nds32_can_use_btgl_p): Likewise.
3820 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3822 * config/nds32/nds32-doubleword.md: New define_split pattern for
3823 illegal register number.
3825 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3827 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3829 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3831 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3833 2018-04-25 Richard Biener <rguenther@suse.de>
3835 * lto-streamer.h (LTO_major_version): Bump to 8.
3837 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3839 * BASE-VER: Set to 9.0.0.
3841 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3843 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3844 in __abskf2 and __powikf2.
3846 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3849 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3850 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3851 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3852 Use the above on operand 2. Reindent.
3853 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3855 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3858 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3860 * config/i386/i386.opt (mcet): Removed.
3861 * doc/install.texi: Remove -mcet documentation.
3862 * doc/invoke.texi: Likewise.
3864 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3867 * doc/install.texi: Remove -mcet from bootstrap-cet.
3869 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3872 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3873 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3877 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3878 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3879 containing a CONST_VECTOR.
3881 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3883 * doc/install.texi: Update newlib dependency for nvptx.
3885 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3888 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3889 instead of INTVAL when shifting x left.
3891 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3893 PR tree-optimization/85478
3894 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3895 vect_grouped_store_supported for single element vectors.
3897 2018-04-24 Richard Biener <rguenther@suse.de>
3900 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3901 load cost increase to the case of non-constant step.
3903 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3906 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3907 destination if any_malformed_asm.
3909 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3912 * expr.c (store_field): In the bitfield case, if the value comes from
3913 a function call and is returned in registers by means of a PARALLEL,
3914 do not change the mode of the temporary unless BLKmode and VOIDmode.
3916 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3918 PR rtl-optimization/85423
3919 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3920 dependencies to debug insns when the previous insn is non-debug.
3922 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3924 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3925 enums into a single definition.
3926 (fls): Fix predicates and printing.
3929 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3931 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3932 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3933 and short u6 immediate.
3934 (check_if_valid_sleep_operand): Remove.
3935 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3937 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3939 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3940 flag_always_save_lp condition.
3941 * config/nds32/nds32.opt (malways-save-lp): New option.
3943 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3945 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3946 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3947 * config/nds32/nds32.h
3948 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3949 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3951 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3953 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3954 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3956 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3957 Chung-Ju Wu <jasonwucj@gmail.com>
3959 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3961 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3962 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3964 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3966 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3968 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3970 * config/nds32/nds32-protos.h (nds32_data_alignment,
3971 nds32_local_alignment): Declare.
3972 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3973 nds32_local_alignment): New functions.
3974 (TARGET_CONSTANT_ALIGNMENT): Define.
3975 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3977 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3979 * config/nds32/nds32.c
3980 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3981 (TARGET_MODES_TIEABLE_P): Likewise.
3983 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3985 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3988 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3989 Chung-Ju Wu <jasonwucj@gmail.com>
3991 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3992 for unaligned access.
3993 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3995 * config/nds32/nds32-intrinsic.md: Likewise.
3996 * config/nds32/nds32_intrinsic.h: Likewise.
3997 * config/nds32/nds32.h (nds32_builtins): Likewise.
3998 * config/nds32/nds32.opt (munaligned-access): New option.
3999 * config/nds32/nds32.c (nds32_asm_file_start): Display
4000 flag_unaligned_access status.
4002 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4004 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
4005 -mno-relax is present.
4006 * config/riscv/linux.h (LINK_SPEC): Ditto.
4008 2018-04-20 Martin Sebor <msebor@redhat.com>
4011 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4013 (gimple_fold_builtin_stxcpy_chk): Same.
4014 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4016 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4019 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4020 __powikf2 when long double is IEEE 128-bit.
4022 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4024 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4025 step to make sure stack always aligned.
4027 2018-04-20 Carl Love <cel@us.ibm.com>
4030 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4031 size check for arg0.
4033 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4034 Tom de Vries <tom@codesourcery.com>
4037 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4038 Emit insns for calls too.
4039 (nvptx_find_par): Always look for worker-level predecessor insn.
4040 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4042 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4043 (nvptx_process_pars): Propagate frames for calls.
4045 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4048 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4050 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4051 (ix86_handle_option): Don't handle OPT_mibt.
4052 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4054 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4056 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4057 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4058 (ix86_target_macros): Define __CET__ with flag_cf_protection
4059 for -fcf-protection.
4060 * config/i386/i386.c (isa2_opts): Remove -mibt.
4061 * config/i386/i386.h (TARGET_IBT): Removed.
4062 (TARGET_IBT_P): Likewise.
4063 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4064 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4065 * config/i386/i386.opt (mcet): Update help message.
4068 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4069 -mcet as an alias for -mshstk.
4071 2018-04-20 Richard Biener <rguenther@suse.de>
4074 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4075 complexity by forcing a single use of the multiply operand.
4077 2018-04-20 Martin Jambor <mjambor@suse.cz>
4080 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4081 recursion dependency to only apply to non-clones.
4083 2018-04-20 Martin Jambor <mjambor@suse.cz>
4086 * ipa-cp.c (create_specialized_node): Check that clones of
4087 self-recursive edges exist during IPA-CP.
4089 2018-04-19 Toon Moene <toon@moene.org>
4091 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4094 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4096 PR tree-optimization/85467
4097 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4098 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4099 VECTOR_CST element to type.
4101 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4104 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4105 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4106 (builtin_longjmp): Likewise.
4107 (save_stack_nonlocal): New pattern.
4108 (restore_stack_nonlocal): Likewise.
4110 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4113 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4114 Replace ASM_OUTPUT_LABEL with fprintf.
4116 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4119 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4120 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4121 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4122 define __IBT__ and __SHSTK__ for -fcf-protection.
4123 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4125 (ix86_trampoline_init): Likewise.
4126 (x86_output_mi_thunk): Likewise.
4127 (ix86_notrack_prefixed_insn_p): Likewise.
4128 (ix86_option_override_internal): Don't disallow -fcf-protection.
4129 * config/i386/i386.md (rdssp<mode>): Also enable for
4131 (incssp<mode>): Likewise.
4132 (nop_endbr): Likewise.
4133 * config/i386/i386.opt (mcet): Change help message to built-in
4137 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4138 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4139 enable CET built-in functions.
4141 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4143 * common/config/i386/i386-common.c
4144 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4145 OPTION_MASK_ISA_MOVDIRI_UNSET,
4146 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4147 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4148 * config.gcc (movdirintrin.h): New header.
4149 * config/i386/cpuid.h (bit_MOVDIRI,
4150 bit_MOVDIR64B): New bits.
4151 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4153 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4154 (VOID, PVOID, PCVOID)): New function types.
4155 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4156 __builtin_ia32_directstoreu_u64,
4157 __builtin_ia32_movdir64b): New builtins.
4158 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4159 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4161 (ix86_valid_target_attribute_inner_p): Ditto.
4162 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4163 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4164 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4165 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4166 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4167 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4168 (movdiri<mode>, movdir64b_<mode>): New.
4169 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4170 * config/i386/immintrin.h: Include movdirintrin.h.
4171 * config/i386/movdirintrin.h: New file.
4172 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4174 2018-04-19 Richard Biener <rguenther@suse.de>
4177 * cfg.c (clear_bb_flags): When loop state says we have
4178 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4180 2018-04-19 Richard Biener <rguenther@suse.de>
4182 PR tree-optimization/84737
4183 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4184 copying restrict info.
4185 (vect_setup_realignment): Use it.
4186 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4187 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4188 the first DR to all generated stores.
4189 (vectorizable_load): Likewise for loads.
4191 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4193 PR tree-optimization/85446
4194 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4195 the integral and pointer types to have the same precision.
4197 * doc/install.texi: Document --disable-cet being the default and
4200 2018-04-18 Martin Liska <mliska@suse.cz>
4202 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4205 2018-04-18 Martin Liska <mliska@suse.cz>
4208 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4211 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4212 arguments if they are comparable.
4214 2018-04-18 Martin Liska <mliska@suse.cz>
4217 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4220 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4223 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4226 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4227 ENDBR after calling __morestack.
4229 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4232 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4233 by using gcc_base_ver to generate a gcc_driver_version, and use
4234 it when generating GCC_DRIVER_NAME.
4235 * configure: Regenerate.
4237 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4240 * config.gcc: Obsolete powerpc*-*-*spe*.
4242 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4245 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4246 (stabstr_D): Change type of unum from unsigned int to
4247 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4250 2018-04-17 Jim Wilson <jimw@sifive.com>
4253 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4254 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4255 Set arg_pointer_offset after using pretend_args_size.
4257 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4259 PR rtl-optimization/85431
4260 * dse.c (record_store): Ignore zero width stores.
4263 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4264 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4265 __builtin_stack_restore rather than after it.
4266 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4267 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4268 argument instead of virtual_dynamic_stack_rtx.
4270 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4272 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4274 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4275 Add note to error message to explain internal mapping of overloaded
4276 built-in function name to non-overloaded built-in function name.
4277 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4280 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4283 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4284 where the inputs overlap with the output.
4286 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4288 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4289 (=v, v) alternative and explicit "memory" attribute.
4290 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4291 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4293 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4294 "sselog1" type instead of "sselog".
4295 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4296 "sselog". Remove explicit "memory" attribute.
4297 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4298 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4300 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4301 "sselog1" type instead of "sselog". Remove explicit "memory"
4303 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4304 use "sselog1" type instead of "sselog". Remove explicit "memory"
4306 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4307 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4309 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4310 "sselog1" type instead of "sselog". Remove explicit "memory"
4312 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4313 use "sselog1" type instead of "sselog". Remove explicit "memory"
4317 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4320 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4323 2018-04-17 Martin Jambor <mjambor@suse.cz>
4326 * ipa-cp.c (create_specialized_node): Call
4327 expand_all_artificial_thunks if necessary.
4329 2018-04-17 Martin Liska <mliska@suse.cz>
4332 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4333 in message, remote space in between '_G' and '('.
4335 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4338 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4339 avx512f_vmcmp<mode>3<round_saeonly_name>,
4340 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4341 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4342 avx512f_rndscale<mode><round_saeonly_name>,
4343 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4344 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4345 Use %<iptr>2 instead of %2 for -masm=intel.
4346 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4347 avx512f_vcvttss2usi<round_saeonly_name>,
4348 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4350 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4351 avx512f_vcvttsd2usi<round_saeonly_name>,
4352 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4353 Use %q1 instead of %1 for -masm=intel.
4354 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4355 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4356 of %3 for -masm=intel.
4357 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4359 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4361 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4363 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4364 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4366 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4368 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4369 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4370 %g1 and one with %0 and %1.
4371 (avx512er_vmrcp28<mode><round_saeonly_name>,
4372 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4374 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4375 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4376 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4377 of %0 and %{%4%} for -masm=intel.
4378 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4379 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4380 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4381 order of %0 and %{%5%}%{z%} for -masm=intel.
4383 2018-04-17 Jan Hubicka <jh@suse.cz>
4386 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4388 2018-04-17 Martin Liska <mliska@suse.cz>
4391 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4392 for target_clone error message. Make default implementation
4393 clone to be a local declaration.
4394 (separate_attrs): Add new argument and check for an empty
4396 (expand_target_clones): Handle it.
4397 (ipa_target_clone): Make redirection just for target_clones
4400 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4401 Tom de Vries <tom@codesourcery.com>
4404 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4405 tiled basic blocks without omp continue statements.
4407 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4410 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4411 vec_extract expression as having side effects to make sure it gets
4414 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4417 * config/i386/i386.c (get_builtin_code_for_version): Check
4420 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4423 * gcc/config.gcc: Support "skylake".
4424 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4426 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4427 (processor_target_table): Add "skylake".
4428 (ix86_option_override_internal): Add "skylake".
4429 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4430 PROCESSOR_CANNONLAKE.
4431 (get_builtin_code_for_version): Fix priority for
4432 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4433 PROCESSOR_SKYLAKE-AVX512.
4434 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4435 (processor_type): Add PROCESSOR_SKYLAKE.
4437 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4438 Jason Merrill <jason@redhat.com>
4441 * convert.c (convert_to_integer_1): Use direct recursion for
4442 enumeral types and types with a precision less than the number
4443 of bits in their mode.
4445 2018-04-16 Julia Koval <julia.koval@intel.com>
4448 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4449 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4451 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4454 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4455 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4456 and -mno-direct-move.
4458 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4461 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4462 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4463 Ensure negative shifts result in {0}.
4465 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4467 PR rtl-optimization/79916
4468 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4469 regs (if any) to define how to gnerate SD moves when LRA is in
4472 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4474 PR rtl-optimization/85393
4475 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4476 * except.c (expand_dw2_landing_pad_for_region): Make static.
4477 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4478 a label and unconditional jump to old_bb, rather than
4479 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4482 PR rtl-optimization/85376
4483 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4484 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4485 instead of a specific value.
4487 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4488 Bin Cheng <bin.cheng@arm.com>
4490 PR tree-optimization/82965
4491 PR tree-optimization/83991
4492 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4493 by_profile_only parameter.
4494 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4495 information if the loop was predicted to iterate too many times.
4496 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4498 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4501 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4504 2018-04-13 Martin Liska <mliska@suse.cz>
4505 Jakub Jelinek <jakub@redhat.com>
4508 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4509 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4510 * builtins.c (expand_builtin_memory_copy_args): Use
4511 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4512 handle dest_addr == pc_rtx.
4514 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4517 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4518 asked to not generate direct moves.
4519 (fix_trunc<mode>si2_stfiwx): Similar.
4520 (fix_trunc<mode>si2_internal): Similar.
4522 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4525 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4526 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4527 lookup if dest in some wider mode is known to be const0_rtx and
4528 if so, record permanent equivalence for it to be ZERO_EXTEND of
4529 the narrower mode destination.
4531 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4533 * lto-streamer-out.c (output_function): Revert 259346.
4534 * omp-expand.c (expand_oacc_for): Likewise.
4536 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4538 PR rtl-optimization/85354
4539 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4540 * sel-sched.c (sel_global_init): ... here.
4542 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4545 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4546 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4547 mode for PE-COFF targets.
4548 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4549 (i386_pe_asm_lto_end): Likewise.
4550 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4551 (TARGET_ASM_LTO_END): Likewise.
4552 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4553 (i386_pe_asm_lto_start): New function.
4554 (i386_pe_asm_lto_end): Likewise.
4556 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4557 Richard Biener <rguenther@suse.de>
4560 * lto-streamer-out.c (output_function): Fix CFG loop state before
4562 * omp-expand.c (expand_oacc_for): Handle calls to internal
4563 functions like regular functions.
4565 2018-04-12 Richard Biener <rguenther@suse.de>
4568 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4569 for the early LTO debug to properly generate references to it
4570 during DIE emission. Do not re-use that for the skeleton for
4572 (dwarf2out_early_finish): Likewise.
4574 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4577 * config/i386/sse.md
4578 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4579 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4580 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4581 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4582 and output is a reg, avoid creating invalid lowpart subreg, but
4583 instead split into a 512-bit move. Don't split if not AVX512VL,
4584 input is xmm16+ reg and output is a mem.
4585 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4586 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4587 xmm16+ reg and output is a mem.
4589 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4591 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4592 also for flag_dwarf2_cfi_asm.
4594 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4596 PR rtl-optimization/85342
4597 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4598 a bool scalar var inside of the loop instead. Don't try to update
4599 recog_data.operand after failed apply_change_group.
4601 2018-04-12 Tom de Vries <tom@codesourcery.com>
4604 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4605 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4606 array with flexible array member as array without given dimension.
4607 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4608 argument for undefined param to true.
4610 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4613 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4614 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4615 from PowerPC section.
4616 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4617 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4618 help text that is too long.
4619 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4620 help text that is too long.
4621 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4622 help text that is too long.
4624 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4626 * config/alpha/alpha.md (stack_probe_internal): Rename
4627 from "probe_stack". Update all callers.
4629 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4631 PR rtl-optimization/84566
4632 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4633 sched_macro_fuse_insns.
4635 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4638 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4639 (compute_block_dependences): ... from here.
4641 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4643 PR tree-optimization/85331
4644 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4645 from int to HOST_WIDE_INT.
4647 2018-04-11 Martin Jambor <mjambor@suse.cz>
4650 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4651 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4652 not the same as the source val.
4653 (cgraph_edge_brings_value_p): New parameter.
4654 (gather_edges_for_value): Pass destination value to
4655 cgraph_edge_brings_value_p.
4656 (perhaps_add_new_callers): Likewise.
4657 (get_info_about_necessary_edges): Likewise and exclude values brought
4658 only by self-recursive edges.
4659 (create_specialized_node): Redirect only clones of self-calling edges.
4660 (+self_recursive_pass_through_p): New function.
4661 (find_more_scalar_values_for_callers_subset): Use it.
4662 (find_aggregate_values_for_callers_subset): Likewise.
4663 (known_aggs_to_agg_replacement_list): Removed.
4664 (decide_whether_version_node): Re-calculate known constants for all
4665 remaining context clones.
4667 2018-04-11 Richard Biener <rguenther@suse.de>
4670 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4671 from early DWARF output.
4672 (dwarf2out_early_finish): Output line info unconditionally into
4673 early DWARF and add reference to it.
4675 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4678 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4679 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4680 other than V2DFmode using iptr mode attribute.
4681 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4683 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4685 PR rtl-optimization/84659
4686 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4688 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4691 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4693 (output_loc_list): Pass address of a dummy size variable even in the
4694 locview handling loop.
4695 (index_location_lists): Add comment on why skip_loc_list_entry can't
4698 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4701 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4704 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4707 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4708 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4709 and -mstring-compare-inline-limit.
4711 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4714 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4715 for stack clash protection in a register whenever we need it to be in
4718 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4720 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4721 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4723 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4726 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4728 (mlong-double-): Ditto.
4729 * config/rs6000/sysv4.opt (msdata=): Ditto.
4730 (mtls-size=): Ditto.
4732 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4734 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4735 erroneous entries for
4736 "vector int vec_ldl (int, long int *)", and
4737 "vector unsigned int vec_ldl (int, unsigned long int *)".
4738 Add comments and entries for
4739 "vector bool char vec_ldl (int, bool char *)",
4740 "vector bool short vec_ldl (int, bool short *)",
4741 "vector bool int vec_ldl (int, bool int *)",
4742 "vector bool long long vec_ldl (int, bool long long *)",
4743 "vector pixel vec_ldl (int, pixel *)",
4744 "vector long long vec_ldl (int, long long *)",
4745 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4746 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4747 type tree bool_long_long_type_node and correct definition of
4748 bool_V2DI_type_node to make reference to this new type tree.
4749 (rs6000_mangle_type): Replace erroneous reference to
4750 bool_long_type_node with bool_long_long_type_node.
4751 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4752 comments to emphasize sign distinctions for char and int types and
4753 replace RS6000_BTI_bool_long constant with
4754 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4755 use of RS6000_BTI_pixel.
4756 (bool_long_type_node): Remove this macro definition.
4757 (bool_long_long_type_node): New macro definition
4759 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4761 PR rtl-optimization/85300
4762 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4763 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4764 simplify_unary_operation fails.
4766 2018-04-10 Martin Liska <mliska@suse.cz>
4768 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4769 cgraph_edge and ipa_ref.
4771 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4775 * config/i386/sse.md
4776 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4777 computation of the VEC_MERGE selector from mask.
4778 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4779 Fix decoding of the VEC_MERGE selector into mask.
4781 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4783 PR tree-optimization/85286
4784 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4786 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4788 * final.c (final_1): Set insn_last_address as well as
4789 insn_current_address.
4791 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4794 * explow.c (emit_stack_probe): Call validize_mem on memory location
4795 before passing it to gen_probe_stack. Create address operand and
4796 legitimize it for the probe_stack_address case.
4798 2018-04-09 Jan Hubicka <jh@suse.cz>
4801 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4802 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4803 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4805 (free_lang_data): Rebuild type inheritance graph.
4807 2018-04-09 Martin Sebor <msebor@redhat.com>
4809 * invoke.texi (-finline-small-functions): Mention other optimization
4811 (-findirect-inlining, -fpartial-inlining): Same.
4812 (-finline-functions-called-once): Same.
4813 (-freorder-blocks-and-partition): Same.
4815 2018-04-09 Jan Hubicka <jh@suse.cz>
4818 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4819 jumps; choose last target that matches the criteria (i.e.
4820 no partition changes for non-crossing jumps).
4821 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4822 support for redirecting crossing jumps to non-crossing.
4824 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4826 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4827 also for naked functions.
4829 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4831 * config/arc/arc.md (add_shift): New pattern.
4832 (add_shift2): Likewise.
4833 (sub_shift): Likewise.
4834 (sub_shift_cmp0_noout): Likewise.
4835 (compare_si_ashiftsi): Likewise.
4836 (xbfu_cmp0_noout): New combine pattern.
4837 (xbfu_cmp0"): Likewise.
4838 (movsi_set_cc_insn): Place the predicable variant first.
4839 (commutative_binary_cmp0_noout): Remove clobber.
4840 (commutative_binary_cmp0): New pattern.
4841 (noncommutative_binary_cmp0): Likewise.
4842 (noncommutative_binary_cmp0_noout): Likewise.
4843 (noncommutative_binary_comparison_result_used): Removed.
4844 (rsub_cmp0): New pattern.
4845 (rsub_cmp0_noout): Likewise.
4846 (extzvsi): Changed, keep only meaningful variants.
4847 (SQH, SEZ): New iterators.
4848 (SQH_postfix): New mode attribute.
4849 (SEZ_prefix): New code attribute.
4850 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4851 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4852 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4854 (noncommutative_operator): Check the availability of barrel
4857 2018-04-09 Richard Biener <rguenther@suse.de>
4859 PR tree-optimization/85284
4860 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4861 Only use the niter constraining form of simple_iv when the exit
4864 2018-04-09 Tom de Vries <tom@codesourcery.com>
4867 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4868 (define_expand "*memory_barrier"): New define_expand.
4869 (define_insn "memory_barrier"): New insn.
4871 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4873 PR rtl-optimization/80463
4874 PR rtl-optimization/83972
4875 PR rtl-optimization/83480
4877 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4878 correct producer for the insn.
4879 (tidy_control_flow): Fixup seqnos in case of debug insns.
4881 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4883 PR rtl-optimization/83913
4885 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4886 different sched-times when merging exprs.
4888 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4890 PR rtl-optimization/83962
4892 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4893 tidy_fallthru_edge and tidy_control_flow.
4895 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4897 PR rtl-optimization/83530
4899 * sel-sched.c (force_next_insn): New global variable.
4900 (remove_insn_for_debug): When force_next_insn is true, also leave only
4901 next insn in the ready list.
4902 (sel_sched_region): When the region wasn't scheduled, make another pass
4903 over it with force_next_insn set to 1.
4905 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4907 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4909 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4910 for interrupt control.
4911 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4912 functions for interrupt control.
4913 * config/nds32/nds32-intrinsic.md: Likewise.
4914 * config/nds32/nds32_intrinsic.h: Likewise.
4915 * config/nds32/nds32.h (nds32_builtins): Likewise.
4917 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4919 * config/nds32/nds32.c (nds32_init_machine_status,
4920 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4921 strict_aligned_p field.
4922 (nds32_expand_to_rtl_hook): New function.
4923 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4924 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4926 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4927 Chung-Ju Wu <jasonwucj@gmail.com>
4929 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4930 * config/nds32/nds32-n7.md: New file.
4931 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4932 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4934 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4935 * config/nds32/nds32.md (pipeline_model): Add n7.
4936 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4937 * config/nds32/pipelines.md: Include n7 settings.
4939 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4940 Chung-Ju Wu <jasonwucj@gmail.com>
4942 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4943 * config/nds32/nds32-e8.md: New file.
4944 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4945 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4947 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4948 * config/nds32/nds32.md (pipeline_model): Add e8.
4949 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4950 * config/nds32/pipelines.md: Include e8 settings.
4952 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4953 Chung-Ju Wu <jasonwucj@gmail.com>
4955 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4956 * config/nds32/nds32-n8.md: New file.
4957 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4958 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4960 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4961 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4962 * config/nds32/nds32.md (pipeline_model): Add n8.
4963 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4964 * config/nds32/pipelines.md: Include n8 settings.
4966 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4967 Chung-Ju Wu <jasonwucj@gmail.com>
4969 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4970 * config/nds32/nds32-n9-2r1w.md: New file.
4971 * config/nds32/nds32-n9-3r2w.md: New file.
4972 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4973 nds32_register_ports): New or modify for cpu n9.
4974 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4976 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4977 * config/nds32/nds32-utils.c: New file.
4978 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4979 TARGET_MUL_SLOW): Define.
4980 * config/nds32/nds32.md (pipeline_model): New attribute.
4981 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4982 New options that support cpu n9.
4983 * config/nds32/pipelines.md: Include n9 settings.
4984 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4986 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4988 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4989 information if necessary.
4990 (output_cond_branch_compare_zero): Likewise.
4991 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4992 (nds32_target_alignment): Refine for alignment.
4993 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4994 (FUNCTION_BOUNDARY): Modify.
4995 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4997 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4999 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
5001 * config/nds32/constants.md (unspec_volatile_element): Add values for
5002 TLB operation and data prefetch.
5003 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
5004 functions for TLB operation and data prefetch.
5005 * config/nds32/nds32-intrinsic.md: Likewise.
5006 * config/nds32/nds32_intrinsic.h: Likewise.
5007 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5008 (nds32_print_operand): Likewise.
5009 * config/nds32/nds32.h (nds32_builtins): Likewise.
5011 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5012 Andrew Pinski <pinsika@gcc.gnu.org>
5015 * match.pd: Use constant_boolean_node of correct type instead of
5016 boolean_true_node or boolean_false_node for simplifying
5017 pointer comparisons to zero.
5019 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5021 PR tree-optimization/80021
5022 * tree.c (verify_type_variant): Make error call in verify_variant_match
5023 translatable and remove final full stop.
5025 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5027 * config/nds32/constants.md (unspec_volatile_element): Add
5028 UNSPEC_VOLATILE_EH_RETURN.
5029 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5030 nds32_output_stack_pop): Support dwarf exception handling process.
5031 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5032 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5033 exception handling process.
5034 (nds32_compute_stack_frame): Likewise.
5035 (nds32_return_addr_rtx): Likewise.
5036 (nds32_initial_elimination_offset): Likewise.
5037 (nds32_expand_prologue): Likewise.
5038 (nds32_expand_epilogue): Likewise.
5039 (nds32_dynamic_chain_address): New function.
5040 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5042 (DYNAMIC_CHAIN_ADDRESS): Define.
5043 (EH_RETURN_DATA_REGNO): Define.
5044 (EH_RETURN_STACKADJ_RTX): Define.
5045 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5046 patterns for dwarf exception handling.
5048 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5050 * config/nds32/nds32.h: Clean up obsolete macros.
5052 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5054 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5055 Add enum values for particular instructions.
5056 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5057 particular intrinsic functions.
5058 * config/nds32/nds32-intrinsic.md: Likewise.
5059 * config/nds32/nds32_intrinsic.h: Likewise.
5060 * config/nds32/nds32.h (nds32_builtins): Likewise.
5061 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5062 (btst, ave): New patterns for particular instructions.
5064 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5066 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5067 Add enum values for atomic load/store and memory sync.
5068 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5070 * config/nds32/nds32-intrinsic.md: Likewise.
5071 * config/nds32/nds32_intrinsic.h: Likewise.
5072 * config/nds32/nds32.h (nds32_builtins): Likewise.
5074 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5076 PR tree-optimization/85257
5077 * fold-const.c (native_encode_vector): If not all elts could fit
5078 and off is -1, return 0 rather than offset.
5079 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5080 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5081 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5082 adjust buffer in native_interpret_expr call.
5084 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5086 * config/nds32/constants.md (unspec_volatile_element): Add cache
5087 control enum values.
5088 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5089 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5090 * config/nds32/nds32.c (nds32_cctl_names): New.
5091 (nds32_print_operand): Handle cache control register names.
5092 * config/nds32/nds32.h (nds32_builtins): New enum values.
5093 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5095 * config/nds32/nds32.md (type): Add mmu.
5096 * config/nds32/pipelines.md (simple_insn): Add mmu.
5098 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5100 * config/nds32/nds32.md (type): Remove call.
5101 * config/nds32/pipelines.md (simple_insn): Likewise.
5103 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5105 * config/nds32/constants.md (unspec_volatile_element): Add
5106 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5107 UNSPEC_VOLATILE_FMFCFG.
5108 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5109 description for fmfcfg and fmfcsr.
5110 (bdesc_1arg): Add fmtcsr.
5111 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5112 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5113 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5114 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5115 unspec_fmfcfg): New patterns.
5116 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5117 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5118 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5119 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5120 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5121 __nds32__fmfcfg): Define.
5123 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5125 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5126 intrinsic register names.
5127 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5128 intrinsic register enum values and macros.
5130 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5132 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5133 for load/store addressing form.
5134 (nds32_print_operand_address): Likewise.
5136 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5139 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5140 based on LABEL_REF. Remove useless assertion.
5141 (pic_address_needs_scratch): Fix formatting.
5142 (sparc_legitimize_pic_address): Minor tweaks.
5143 (sparc_delegitimize_address): Adjust assertion accordingly.
5144 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5145 into symbolic_operand.
5146 (movsi_high_pic_label_ref): Likewise.
5147 (movsi_lo_sum_pic_label_ref): Likewise.
5148 (movdi_pic_label_ref): Likewise.
5149 (movdi_high_pic_label_ref): Likewise.
5150 (movdi_lo_sum_pic_label_ref): Likewise.
5152 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5154 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5155 custom LIB_SPEC setup.
5157 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5158 Kito Cheng <kito.cheng@gmail.com>
5160 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5161 * config/riscv/freebsd.h: New.
5163 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5165 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5166 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5169 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5170 Kito Cheng <kito.cheng@gmail.com>
5172 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5173 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5174 * config/nds32/nds32-protos.h (nds32_output_call,
5175 nds32_output_return): Declare.
5176 * config/nds32/nds32.md: Refine all the call and return patterns.
5178 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5181 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5182 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5184 PR rtl-optimization/84872
5185 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5186 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5189 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5191 * expr.c (copy_blkmode_to_reg): Revert 254862.
5192 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5194 2018-04-06 Richard Biener <rguenther@suse.de>
5197 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5198 after seeing a component reference with an adjacent field. Treat
5199 refs to arrays at struct end of external decls similar to
5200 refs to unconstrained commons.
5202 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5205 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5206 look through SAVE_EXPRs with non-side-effects argument. Adjust
5208 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5211 2018-04-06 Richard Biener <rguenther@suse.de>
5214 * alias.c (find_base_term): New wrapper around find_base_term
5215 unwinding CSELIB_VAL_PTR changes.
5216 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5219 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5221 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5223 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5224 constant definitions.
5225 ("nop"): lr 0,0 -> nopr r0
5226 ("nop_lr0", "nop_lr1"): New insn definitions.
5228 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5230 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5231 NDS32_V3PUSH_AVAILABLE_P macro.
5233 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5234 Chung-Ju Wu <jasonwucj@gmail.com>
5236 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5237 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5238 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5239 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5240 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5241 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5242 * config/nds32/constraints.md: New constraints and checking for hard
5243 float configuration.
5244 * config/nds32/iterators.md: New mode iterator and attribute for hard
5245 float configuration.
5246 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5248 * config/nds32/nds32-fpu.md: New file.
5249 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5250 deal with hard float code generation.
5251 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5253 (abi_type, float_reg_number): New enum type.
5254 * config/nds32/nds32-predicates.c: New predicates for hard float.
5255 * config/nds32/nds32-protos.h: Declare functions for hard float.
5256 * config/nds32/nds32.c: Implementation for hard float configuration.
5257 * config/nds32/nds32.h: Definitions for hard float configuration.
5258 * config/nds32/nds32.md: Include hard float machine description and
5259 modify patterns for hard float configuration.
5260 * config/nds32/nds32.opt: New options for hard float configuration.
5261 * config/nds32/predicates.md: New predicates for hard float
5264 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5266 * common/config/nds32/nds32-common.c
5267 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5269 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5272 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5273 CONSTRUCTOR_ELT (ctor, ...)->value.
5275 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5278 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5280 2018-04-05 Tom de Vries <tom@codesourcery.com>
5283 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5286 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5287 Kito Cheng <kito.cheng@gmail.com>
5289 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5290 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5291 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5292 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5294 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5295 Kito Cheng <kito.cheng@gmail.com>
5297 * config/nds32/constraints.md (Ufe): New memory constraint.
5298 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5299 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5300 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5302 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5303 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5305 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5307 * config/nds32/nds32.md: Use optimize_size in the condition for
5308 alu-shift instructions.
5310 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5312 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5314 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5316 * config/nds32/nds32.md (negsi2): Refine pattern.
5318 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5319 Chung-Ju Wu <jasonwucj@gmail.com>
5321 * config/nds32/iterators.md (shift_rotate): New code iterator.
5322 (shift): New code attribute.
5323 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5324 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5325 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5326 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5327 bit-wise operations.
5328 (andsi3, *andsi3): Ditto.
5329 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5330 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5331 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5332 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5333 nds32_ior_operand, nds32_xor_operand): New predicates.
5335 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5337 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5338 (addsi3, subsi3): ... this.
5340 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5342 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5344 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5346 * config/nds32/nds32.md: Adjust indention.
5348 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5350 * config/nds32/nds32.md (feature): New attribute.
5352 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5354 * config/nds32/nds32.md (subtype): New attribute.
5356 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5359 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5360 expansion to perform a bitwise AND of the argument followed by a
5361 boolean negation of the result.
5363 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5365 PR rtl-optimization/84878
5366 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5367 the basic block. Assert the use reference is not artificial and that
5368 it has an associated insn.
5370 2018-04-04 Michael Matz <matz@suse.de>
5372 * builtins.c (compute_objsize): Pass correct operand
5373 to array_at_struct_end_p.
5375 2018-04-04 Richard Biener <rguenther@suse.de>
5378 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5379 from contexts for DINFO_LEVEL_TERSE and below.
5381 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5383 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5384 resiter_operand condition.
5385 * config/nds32/nds32.md (*move<mode>): Ditto.
5387 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5388 Monk Chiang <sh.chiang04@gmail.com>
5390 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5392 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5394 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5396 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5397 Kito Cheng <kito.cheng@gmail.com>
5399 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5400 nds32_cond_code_str, output_cond_branch,
5401 output_cond_branch_compare_zero, nds32_expand_cbranch,
5402 nds32_expand_cstore, nds32_expand_movcc,
5403 nds32_output_cbranchsi4_equality_zero,
5404 nds32_output_cbranchsi4_equality_reg,
5405 nds32_output_cbranchsi4_equality_reg_or_const_int,
5406 nds32_output_cbranchsi4_greater_less_zero: New functions.
5407 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5408 nds32_expand_cstore, nds32_expand_movcc,
5409 nds32_output_cbranchsi4_equality_zero,
5410 nds32_output_cbranchsi4_equality_reg,
5411 nds32_output_cbranchsi4_equality_reg_or_const_int,
5412 nds32_output_cbranchsi4_greater_less_zero): Declare.
5413 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5414 nds32_rimm11s_operand): New predicates.
5415 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5416 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5419 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5421 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5422 * config/nds32/nds32.md: Ditto.
5423 * config/nds32/pipelines.md: Ditto.
5425 2018-04-04 Richard Biener <rguenther@suse.de>
5427 PR tree-optimization/85168
5428 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5429 propagating abnormals.
5431 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5433 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5435 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5436 Kito Cheng <kito.cheng@gmail.com>
5438 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5439 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5440 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5441 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5442 * config/nds32/nds32.md (sibcall_internal): New.
5443 (sibcall_register): Remove.
5444 (sibcall_immediate): Remove.
5445 (sibcall_value_internal): New.
5446 (sibcall_value_register): Remove.
5447 (sibcall_value_immediate): Remove.
5448 * config/nds32/predicates.md (nds32_general_register_operand): New.
5449 (nds32_call_address_operand): New.
5451 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5453 PR rtl-optimization/85167
5454 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5455 bb_defs if *split_p, instead preinitialize it to NULL.
5457 PR tree-optimization/85156
5458 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5459 evaluating the argument multiple times.
5461 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5463 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5465 (_mm_cvtpd_ps): Likewise.
5466 (_mm_cvttpd_epi32): Likewise.
5467 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5468 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5469 vector, pixel, and bool following altivec.h include.
5471 2018-04-03 Martin Sebor <msebor@redhat.com>
5473 * doc/extend.texi (Common Function Attributes): Clarify.
5474 (const attribute): Likewise.
5475 (pure attribute): Likewise.
5477 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5480 * config/i386/i386.c (ix86_expand_vector_set): Use
5481 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5483 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5485 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5486 instructions when changing rounding bits to preserve precision bits
5487 in the x87 control word.
5489 2018-04-03 Martin Liska <mliska@suse.cz>
5491 PR tree-optimization/82491
5492 * rtl.h (strip_offset_and_add): Replace += suboffset with
5493 poly_uint64 () + suboffset.
5495 2018-03-29 Martin Liska <mliska@suse.cz>
5496 Martin Jambor <mjambor@suse.cz>
5499 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5500 param_type is not an integral or pointer type.
5502 2018-04-03 Richard Biener <rguenther@suse.de>
5504 * sese.h (recompute_all_dominators): Remove.
5506 2018-04-02 Martin Sebor <msebor@redhat.com>
5508 * doc/invoke.texi (-Wrestrict): Fix typos.
5510 2018-04-02 Jim Wilson <jimw@sifive.com>
5512 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5513 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5514 (<optab>di3, <optab>si3_extend): Likewise.
5515 (<optab>si3_mask, <optab>si3_mask_1): New.
5516 (<optab>di3_mask, <optab>di3_mask_1): New.
5517 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5518 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5519 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5521 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5523 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5526 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5528 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5529 (nds32_canonicalize_comparison): New function.
5531 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5532 Kito Cheng <kito.cheng@gmail.com>
5533 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5535 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5536 * config/nds32/constants.md (unspec_volatile_element): Add
5537 UNSPEC_VOLATILE_RELAX_GROUP.
5538 * config/nds32/nds32-relax-opt.c: New file.
5539 * config/nds32/nds32-predicates.c
5540 (nds32_symbol_load_store_p): New function.
5541 * config/nds32/nds32-protos.h
5542 (nds32_symbol_load_store_p): Declare function.
5543 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5544 * config/nds32/nds32.c
5545 (nds32_register_pass): New function to register pass.
5546 (nds32_register_passes): New function to register passes.
5547 * config/nds32/nds32.md (relax_group): New pattern.
5548 * config/nds32/nds32.opt (mrelax-hint): New option.
5549 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5551 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5553 * config/nds32/t-nds32: Modify files dependency.
5555 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5557 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5558 (PROFILE_HOOK): Define its implementation.
5560 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5562 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5563 type and 32-bit size.
5565 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5568 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5569 (V_128_256): New mode iterator.
5570 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5571 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5572 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5574 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5575 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5577 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5580 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5581 NaN inputs correctly.
5583 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5586 * config/rs6000/vsx.md (??r): New mode attribute.
5587 (*vsx_mov<mode>_64bit): Use it.
5588 (*vsx_mov<mode>_32bit): Likewise.
5590 2018-03-30 Martin Sebor <msebor@redhat.com>
5592 PR tree-optimization/84818
5593 * builtins.c (check_access): Use warning_n.
5595 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5598 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5600 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5603 2018-03-30 Julia Koval <julia.koval@intel.com>
5606 * x86-tune.def (movx, partial_reg_dependency): Enable for
5609 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5612 * lra-constraints.c (process_alt_operands): Move setting
5613 this_alternative_matches below.
5615 2018-03-29 Martin Liska <mliska@suse.cz>
5618 * doc/invoke.texi: Document how LTO works with debug info.
5619 Describe auto-load support of binutils. Mention 'x86-64'
5620 as valid option value of -march option.
5622 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5624 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5627 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5628 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5629 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5632 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5635 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5636 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5637 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5638 for RS6000_BTM_POWERPC64.
5639 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5640 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5641 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5646 2018-03-28 Carl Love <cel@us.ibm.com>
5649 2017-09-27 Carl Love <cel@us.ibm.com>
5651 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5652 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5653 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5656 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5658 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5659 instead of __vector bool.
5660 (_mm_max_pu8): Likewise.
5661 (_mm_min_pi16): Likewise.
5663 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5666 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5667 (DIVWEUO): Likewise.
5669 (DIVDEUO): Likewise.
5670 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5671 DIVWEUO and DIVDEUO.
5672 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5673 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5674 (div_extend): Likewise.
5675 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5677 (__builtin_divweuo): Likewise.
5678 (__builtin_divdeo): Likewise.
5679 (__builtin_divdeuo): Likewise.
5681 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5684 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5685 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5687 PR tree-optimization/82004
5688 * gimple-match-head.c (optimize_pow_to_exp): New function.
5689 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5690 Don't fold to exp if optimize_pow_to_exp is false.
5692 2018-03-28 Martin Liska <mliska@suse.cz>
5695 * calls.c (initialize_argument_information): Fix trailing space.
5696 * common.opt: Fix typo and provide better explanation for
5697 -fsanitize-coverage option.
5698 * config/i386/i386.opt: Fix typo.
5700 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5701 Martin Liska <mliska@suse.cz>
5704 * gimplify.c (asan_poison_variable): Don't do the check for
5705 gimplify_omp_ctxp here.
5706 (gimplify_decl_expr): Do it here.
5707 (gimplify_target_expr): Likewise.
5709 2018-03-28 Martin Liska <mliska@suse.cz>
5712 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5713 chkp_type_bounds_count if MPX is not enabled.
5715 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5717 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5719 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5722 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5723 function to create the function decl for complex long double
5724 multiply and divide for -mabi=ieeelongdouble.
5725 (init_float128_ieee): Call it.
5727 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5730 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5731 -fcf-protection=branch -mibt.
5732 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5734 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5737 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5739 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5742 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5743 extern array declarations.
5745 2018-03-27 Richard Biener <rguenther@suse.de>
5748 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5749 explicit single_use checks.
5751 2018-03-27 Richard Biener <rguenther@suse.de>
5753 PR tree-optimization/85082
5754 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5757 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5759 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5760 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5761 Turn on fasynchronous-unwind-tables and funwind-tables.
5763 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5766 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5767 (*bmi_blsr_<mode>_ccz): Ditto.
5769 2018-03-26 Tom de Vries <tom@codesourcery.com>
5771 PR tree-optimization/85063
5772 * omp-general.c (offloading_function_p): New function. Factor out
5774 * omp-offload.c (pass_omp_target_link::gate): ... here.
5775 * omp-general.h (offloading_function_p): Declare.
5776 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5777 with attribute omp declare target for offloading functions.
5779 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5781 PR tree-optimization/84005
5782 * tree-data-ref.h (get_base_for_alignment): Declare.
5783 * tree-data-ref.c (get_base_for_alignment_1): New function.
5784 (get_base_for_alignment): Likewise.
5785 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5786 get_base_for_alignment to find a suitable base object, instead
5787 of always using drb->base_address.
5789 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5792 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5793 known size by default.
5795 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5798 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5799 and non BLKmode operands.
5801 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5804 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5805 Clean up attributes.
5807 2018-03-23 Richard Biener <rguenther@suse.de>
5810 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5811 we are going to emit early debug for LTO.
5813 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5816 * function.c (match_asm_constraints_1): Don't optimize if input
5817 doesn't satisfy general_operand predicate for output's mode.
5820 * alias.c (write_dependence_p): Don't require for x_canonicalized
5821 non-VOIDmode if x has VOIDmode.
5824 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5825 just don't try to optimize it rather than assert it never happens.
5827 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5829 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5830 macro expansions for definition of ST_INTERNAL_<mode> and
5831 LD_INTERNAL_<mode> builtins.
5832 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5834 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5836 (altivec_expand_st_builtin): Likewise.
5837 (altivec_expand_builtin): Remove calls to deleted functions.
5838 (rs6000_address_for_altivec): Delete this function.
5839 * config/rs6000/vector.md: Remove expands for
5840 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5842 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5845 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5846 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5847 re-computing once computed.
5848 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5849 (arm_init_machine_status): Initialize
5850 machine->static_chain_stack_bytes.
5852 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5855 * doc/extend.texi: Add four new prototypes for vec_ld.
5856 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5857 definitions for more logical presentation.
5858 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5859 entries for V1TI variants of __builtin_altivec_ld builtin.
5860 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5861 handling of V1TI variant of LVX icode pattern.
5862 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5863 (rs6000_gimple_fold_builtin): Likewise.
5864 (altivec_init_builtins): Add code to define
5865 __builtin_altivec_lvx_v1ti function.
5867 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5870 * function.c (match_asm_constraints_1): Don't do the optimization
5871 if input isn't a REG, SUBREG, MEM or constant.
5873 2018-03-22 Tom de Vries <tom@codesourcery.com>
5875 PR tree-optimization/84956
5876 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5877 bb_has_abnormal_pred.
5879 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5882 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5883 DECL_INITIAL (decl) to decl at the end.
5884 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5887 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5889 * doc/extend.texi (__builtin_tgmath): Document when complex
5890 integer types are treated as _Complex _Float64.
5892 2018-03-21 Tom de Vries <tom@codesourcery.com>
5894 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5896 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5898 PR tree-optimization/84960
5899 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5900 if it is ENTRY block, move them into single succ of ENTRY in that case.
5902 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5904 PR tree-optimization/84811
5905 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5906 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5907 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5908 (known_subrange_p): Update accordingly. Cast each value involved
5909 in the size comparison, rather than casting the result of the
5912 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5914 PR tree-optimization/84982
5915 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5916 by flipping the least significant bit rather than all bits from
5917 bitpos to bitpos + bitsize - 1.
5919 2018-03-21 Nathan Sidwell <nathan@acm.org>
5921 * doc/extend.texi (Deprecated Features): Remove mention of
5922 long-deleted deprecations.
5924 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5927 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5928 * configure: Regenerate.
5930 2018-03-21 Tom de Vries <tom@codesourcery.com>
5932 PR tree-optimization/83126
5933 * tree-parloops.c (num_phis): New function.
5934 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5936 2018-03-21 Nathan Sidwell <nathan@acm.org>
5938 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5939 mention anon-struct/union members and trailing attributes.
5941 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5943 PR tree-optimization/84969
5944 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5945 builtin memset partitions if they set different rhs values.
5947 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5949 PR rtl-optimization/84989
5950 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5951 VEC_DUPLICATE with scalar result mode.
5953 2018-03-21 Martin Liska <mliska@suse.cz>
5956 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5957 not intended return statement.
5959 2018-03-21 Martin Liska <mliska@suse.cz>
5962 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5963 (chkp_find_bound_slots_1): Limit number of iterations.
5965 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5968 * Minor grammar fixes for x86 options.
5970 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5973 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5974 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5976 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5979 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5980 (altivec_lvx_<mode>_1op): Likewise.
5981 (altivec_stvx_<mode>_2op): Likewise.
5982 (altivec_stvx_<mode>_1op): Likewise.
5983 (altivec_lvx_<VM2:mode>): New define_expand.
5984 (altivec_stvx_<VM2:mode>): Likewise.
5985 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5986 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5987 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5988 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5989 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5990 (rs6000_gen_lvx): Likewise.
5991 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5992 (altivec_expand_stv_builtin): Likewise.
5993 (altivec_expand_builtin): Likewise.
5994 * config/rs6000/vector.md: Likewise.
5996 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5999 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
6002 2018-03-20 Richard Biener <rguenther@suse.de>
6005 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6006 sign-conversions as zero, fall back to standard scalar_stmt
6009 2018-03-20 Martin Liska <mliska@suse.cz>
6012 * predict.c (rebuild_frequencies): Handle case when we have
6013 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6015 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6018 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6019 flag_section_anchors.
6020 * varasm.c (use_blocks_for_decl_p): Remove hack for
6021 dw2_force_const_mem.
6024 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6026 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6027 be created, use lowpart_subreg of operands[0] rather than operands[0]
6029 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6030 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6031 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6032 and n constraint instead of aarch64_shift_imm_di and Usd.
6033 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6034 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6036 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6039 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6040 to favor GPR over NEON registers.
6041 (<shift>di3_neon): Likewise.
6043 2018-03-20 Tom de Vries <tom@codesourcery.com>
6046 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6047 (nvptx_process_pars): Emit bar.sync asap and alap.
6049 2018-03-20 Tom de Vries <tom@codesourcery.com>
6052 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6053 seen_label if seen_label is already set.
6055 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6058 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6059 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6060 Use 1U instead of 1. Formatting fixes.
6063 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6064 instead of TREE_TYPE (s1) for the return value.
6066 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6068 PR tree-optimization/84946
6069 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6070 bitsize + bitsize in poly_uint64 rather than poly_int64.
6073 * dwarf2asm.c: Include fold-const.c.
6074 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6075 of decl rather than decl itself.
6077 PR rtl-optimization/84643
6078 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6080 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6083 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6084 calling assemble_variable.
6086 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6089 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6090 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6092 2018-03-19 Jim Wilson <jimw@sifive.com>
6095 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6096 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6097 (riscv_first_stack_step): Likewise.
6098 (riscv_option_override): Use STACK_BOUNDARY instead of
6100 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6102 (BIGGEST_ALIGNMENT): Set to 128.
6103 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6104 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6107 2018-03-19 Richard Biener <rguenther@suse.de>
6109 PR tree-optimization/84933
6110 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6111 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6113 2018-03-19 Richard Biener <rguenther@suse.de>
6115 PR tree-optimization/84859
6116 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6117 (cond_if_else_store_replacement): Perform sinking operation on
6118 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6119 Generalize what a BB with a single eligible store is.
6121 2018-03-19 Richard Biener <rguenther@suse.de>
6123 PR tree-optimization/84929
6124 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6125 chrec_is_positive against non-chrec arg.
6127 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6130 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6132 2018-03-18 Martin Liska <mliska@suse.cz>
6134 PR rtl-optimization/84635
6135 * regrename.c (build_def_use): Use matches_mode only when
6138 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6140 PR tree-optimization/84913
6141 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6142 vectorize chains of COND_EXPRs.
6144 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6146 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6148 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6150 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6152 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6154 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6156 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6157 Kito Cheng <kito.cheng@gmail.com>
6159 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6160 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6161 (nds32_adjust_reg_alloc_order): New function.
6162 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6164 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6166 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6167 nds32_print_operand, nds32_print_operand_address): Use
6168 HOST_WIDE_INT_PRINT_DEC instead.
6170 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6172 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6174 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6177 * config/i386/i386.c (initial_ix86_tune_features,
6178 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6180 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6181 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6182 rather than 1u << ix86_tune. Formatting fix.
6183 (ix86_option_override_internal): Change ix86_arch_mask from
6184 unsigned int to unsigned HOST_WIDE_INT, initialize to
6185 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6186 (ix86_function_specific_restore): Likewise.
6188 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6191 * postreload.c (reload_combine_recognize_pattern): Perform
6192 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6193 truncate_int_for_mode the result for the destination's mode.
6196 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6197 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6199 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6202 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6205 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6206 regno_allocno_class_array and sorted_pseudos.
6207 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6208 insns where regno is used.
6210 2018-03-16 Martin Liska <mliska@suse.cz>
6213 * multiple_target.c (create_dispatcher_calls): Redirect
6214 reference in the symbol table.
6216 2018-03-16 Martin Liska <mliska@suse.cz>
6219 * multiple_target.c (create_dispatcher_calls): Redirect also
6222 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6226 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6227 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6228 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6230 2018-03-16 Julia Koval <julia.koval@intel.com>
6232 * doc/invoke.texi (Skylake Server): Add CLWB.
6233 Cannonlake): Remove CLWB.
6235 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6237 PR tree-optimization/84841
6238 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6240 (FLOAT_ONE_CONST_TYPE): Define.
6241 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6242 (sort_by_operand_rank): Put entries with higher constant_type last
6243 rather than first to match comments.
6245 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6247 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6250 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6253 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6254 check and use address of target always.
6256 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6259 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6260 (indirect_thunk_bnd_needed): Likewise.
6261 (indirect_thunks_used): Likewise.
6262 (indirect_thunks_bnd_used): Likewise.
6263 (indirect_return_needed): New.
6264 (indirect_return_bnd_needed): Likewise.
6265 (output_indirect_thunk_function): Add a bool argument for
6267 (output_indirect_thunk_function): Don't generate alias for
6268 function return thunk.
6269 (ix86_code_end): Call output_indirect_thunk_function to generate
6270 function return thunks.
6271 (ix86_output_function_return): Set indirect_return_bnd_needed
6272 and indirect_return_needed instead of indirect_thunk_bnd_needed
6273 and indirect_thunk_needed.
6275 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6277 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6278 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6279 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6281 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6282 Paul Hua <paul.hua.gm@gmail.com>
6285 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6287 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6289 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6290 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6293 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6296 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6297 instead of GET_MODE_SIZE when comparing Units.
6299 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6302 * varasm.c (hash_section): Return an unchangeble hash value
6303 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6304 Return !aarch64_can_use_per_function_literal_pools_p ().
6306 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6309 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6310 rather than address of cmode as last argument to prepare_cmp_insn.
6312 2018-03-15 Julia Koval <julia.koval@intel.com>
6314 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6315 F_AVX512VNNI, F_AVX512BITALG): New.
6317 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6320 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6321 insn for floating-point loads and stores.
6323 2018-03-14 Carl Love <cel@us.ibm.com>
6325 * config/rs6000/rs6000-c.c: Add macro definitions for
6326 ALTIVEC_BUILTIN_VEC_PERMXOR.
6327 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6328 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6329 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6330 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6332 * config/doc/extend.texi: Add prototypes for vec_permxor.
6334 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6337 * diagnostic-show-locus.c (class layout_point): Convert m_line
6338 from int to linenum_type.
6339 (line_span::comparator): Use linenum "compare" function when
6340 comparing line numbers.
6341 (test_line_span): New function.
6342 (layout_range::contains_point): Convert param "row" from int to
6344 (layout_range::intersects_line_p): Likewise.
6345 (layout::will_show_line_p): Likewise.
6346 (layout::print_source_line): Likewise.
6347 (layout::should_print_annotation_line_p): Likewise.
6348 (layout::print_annotation_line): Likewise.
6349 (layout::print_leading_fixits): Likewise.
6350 (layout::annotation_line_showed_range_p): Likewise.
6351 (struct line_corrections): Likewise for field m_row.
6352 (line_corrections::line_corrections): Likewise for param "row".
6353 (layout::print_trailing_fixits): Likewise.
6354 (layout::get_state_at_point): Likewise.
6355 (layout::get_x_bound_for_row): Likewise.
6356 (layout::print_line): Likewise.
6357 (diagnostic_show_locus): Likewise for locals "last_line" and
6359 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6360 * input.c (selftest::test_linenum_comparisons): New function.
6361 (selftest::input_c_tests): Call it.
6362 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6363 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6364 * selftest.h (ASSERT_GT): New macro.
6365 (ASSERT_GT_AT): New macro.
6366 (ASSERT_LT): New macro.
6367 (ASSERT_LT_AT): New macro.
6369 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6371 PR rtl-optimization/84780
6372 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6374 2018-03-14 Martin Liska <mliska@suse.cz>
6376 * tree.c (record_node_allocation_statistics): Use
6377 get_stats_node_kind.
6378 (get_stats_node_kind): New function extracted from
6379 record_node_allocation_statistics.
6380 (free_node): Use get_stats_node_kind.
6382 2018-03-14 Richard Biener <rguenther@suse.de>
6384 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6385 that the value-set of ANTIC_IN doesn't grow.
6388 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6390 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6391 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6393 2018-03-14 Julia Koval <julia.koval@intel.com>
6395 * config.gcc (icelake-client, icelake-server): New.
6397 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6398 (initial_ix86_arch_features): Ditto.
6399 (PTA_SKYLAKE): Add SGX.
6400 (PTA_ICELAKE): Remove.
6401 (PTA_ICELAKE_CLIENT): New.
6402 (PTA_ICELAKE_SERVER): New.
6403 (ix86_option_override_internal): Split up icelake on icelake client and
6405 (get_builtin_code_for_version): Ditto.
6406 (fold_builtin_cpu): Ditto.
6407 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6408 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6409 * config/i386/i386.h (processor_type): Ditto.
6410 * doc/invoke.texi: Ditto.
6412 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6415 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6416 INTEGER_CST offset, add it together with bitpos / 8 and
6417 sign extend based on POINTER_SIZE.
6421 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6424 * config/i386/constraints.md (Yc): New register constraint.
6425 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6426 Use Yc constraint for alternative 2 of operand 0. Remove
6427 preferred_for_speed attribute.
6429 2018-03-14 Richard Biener <rguenther@suse.de>
6431 PR tree-optimization/84830
6432 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6433 with the old one to avoid oscillations.
6435 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6438 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6440 (assign_by_spills): Return a flag of reload assignment failure.
6441 Do not process the reload assignment failures. Do not spill other
6442 reload pseudos if they has the same reg class. Update n if
6444 (lra_assign): Add a return arg. Set up from the result of
6445 assign_by_spills call.
6446 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6447 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6448 usage_insns if it is not NULL.
6449 (spill_hard_reg_in_range): New function.
6450 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6451 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6452 function prototypes.
6453 (lra_assign): Change prototype.
6454 * lra.c (lra): Add code to deal with fails by splitting hard reg
6457 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6459 * config/riscv/riscv.opt (mrelax): New option.
6460 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6461 "norelax" when riscv_mrelax is disabled.
6462 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6464 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6467 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6468 reassociation for int modes.
6470 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6472 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6473 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6475 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6476 * config/aarch64/aarch64-sve.md
6477 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6478 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6479 (*extend<mode><Vwide>2): Rename to...
6480 (aarch64_sve_extend<mode><Vwide>2): ...this.
6481 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6482 renaming the old pattern to...
6483 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6485 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6486 define_expand, renaming the old pattern to...
6487 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6488 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6489 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6490 account when deciding which SVE instruction the optab should use.
6491 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6493 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6495 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6496 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6497 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6498 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6499 (tlsdesc_small_<mode>): Turn a define_expand and use
6500 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6501 (tlsdesc_small_advsimd_<mode>): ...this.
6502 (tlsdesc_small_sve_<mode>): New pattern.
6504 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6506 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6507 (UNSPEC_UMUL_HIGHPART): New constants.
6508 (MUL_HIGHPART): New int iteraor.
6509 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6510 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6512 (*<su>mul<mode>3_highpart): New define_insn.
6514 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6517 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6520 2018-03-13 Martin Liska <mliska@suse.cz>
6523 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6525 (sem_item_optimizer::~sem_item_optimizer): Release it.
6526 (sem_item_optimizer::merge_classes): Register variable aliases.
6527 (sem_item_optimizer::fixup_pt_set): New function.
6528 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6529 * ipa-icf.h: Declare new variables and functions.
6531 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6534 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6535 integer_pow2p@2 and test integer_pow2p in condition.
6536 (A < 0 ? C : 0): Similarly for @1.
6539 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6540 characters starting at p contain '\0' character, don't look beyond
6544 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6545 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6548 * reg-stack.c (change_stack): Change update_end var from int to
6549 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6550 also call set_block_for_insn on the newly added insns and rescan.
6553 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6554 on the last operand.
6557 * tree.c (stabilize_reference_1): Return save_expr (e) for
6558 STATEMENT_LIST even if it doesn't have side-effects.
6560 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6562 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6564 2018-03-12 Renlin Li <renlin.li@arm.com>
6566 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6567 aarch64_output_scalar_simd_mov_immediate.
6569 2018-03-12 Martin Sebor <msebor@redhat.com>
6571 PR tree-optimization/83456
6572 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6573 for perfectly overlapping calls to memcpy.
6574 (gimple_fold_builtin_memory_chk): Same.
6575 (gimple_fold_builtin_strcpy): Handle no-warning.
6576 (gimple_fold_builtin_stxcpy_chk): Same.
6577 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6579 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6581 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6582 parameter. Use it for SFmode.
6583 (rs6000_function_arg_advance_1): Adjust.
6584 (rs6000_function_arg): Adjust.
6585 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6587 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6589 PR rtl-optimization/84169
6590 PR rtl-optimization/84780
6591 * combine.c (can_combine_p): Check for a 2-insn combination whether
6592 the destination register is used between the two insns, too.
6594 2018-03-12 Richard Biener <rguenther@suse.de>
6596 PR tree-optimization/84803
6597 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6598 for refs DR analysis didn't process.
6600 2018-03-12 Richard Biener <rguenther@suse.de>
6602 PR tree-optimization/84777
6603 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6604 force-vectorize loops ignore whether we are optimizing for size.
6606 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6608 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6609 (TARGET_MD_ASM_ADJUST): Define.
6611 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6612 Kito Cheng <kito.cheng@gmail.com>
6613 Chung-Ju Wu <jasonwucj@gmail.com>
6615 * config/nds32/nds32.c (nds32_compute_stack_frame,
6616 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6617 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6618 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6619 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6620 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6621 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6622 * config/nds32/nds32.md (prologue, epilogue): Use macro
6623 NDS32_V3PUSH_AVAILABLE_P to do checking.
6625 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6628 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6629 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6630 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6631 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6632 addition of most attributes on !orig_type_die or the attribute not
6633 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6635 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6636 Chung-Ju Wu <jasonwucj@gmail.com>
6638 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6640 * config/nds32/nds32.opt (mvh): New option.
6642 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6643 Chung-Ju Wu <jasonwucj@gmail.com>
6645 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6647 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6648 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6651 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6652 Chung-Ju Wu <jasonwucj@gmail.com>
6654 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6656 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6657 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6659 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6660 Kito Cheng <kito.cheng@gmail.com>
6661 Chung-Ju Wu <jasonwucj@gmail.com>
6663 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6664 UNSPEC_FFMISM and UNSPEC_FLMISM.
6665 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6666 for ffb, ffmism and flmism.
6667 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6668 (unspec_ffmism): Ditto.
6669 (unspec_flmism): Ditto.
6670 (nds32_expand_builtin_impl): Check if string extension is available.
6671 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6672 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6674 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6677 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6680 * lra-assigns.c (assign_by_spills): Return a flag of reload
6681 assignment failure. Do not process the reload assignment
6682 failures. Do not spill other reload pseudos if they has the same
6684 (lra_assign): Add a return arg. Set up from the result of
6685 assign_by_spills call.
6686 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6687 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6688 usage_insns if it is not NULL.
6689 (spill_hard_reg_in_range): New function.
6690 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6691 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6692 function prototypes.
6693 (lra_assign): Change prototype.
6694 * lra.c (lra): Add code to deal with fails by splitting hard reg
6697 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6700 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6702 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6705 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6706 (dw_val_node): Add val_symbolic_view.
6707 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6708 (symview_upper_bound): New.
6709 (new_line_info_table): Initialize symviews_since_reset.
6710 (dwarf2out_source_line): Count symviews_since_reset and set
6711 symview_upper_bound.
6712 (dw_val_equal_p): Handle symview.
6713 (add_AT_symview): New.
6714 (print_dw_val): Handle symview.
6715 (attr_checksum, attr_checksum_ordered): Likewise.
6716 (same_dw_val_p, size_of_die): Likewise.
6717 (value_format, output_die): Likewise.
6718 (add_high_low_attributes): Use add_AT_symview for entry_view.
6719 (dwarf2out_finish): Reset symview_upper_bound, clear
6722 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6725 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6726 Add strict argument and use it.
6727 (rs6000_split_multireg_move): Update for new strict argument.
6728 (mem_operand_gpr): Disallow all non-offsettable addresses.
6729 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6731 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6734 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6735 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6736 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6739 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6740 decl, use remap_type if we want to use the type.
6742 2018-03-09 Martin Sebor <msebor@redhat.com>
6744 PR tree-optimization/84526
6745 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6747 (builtin_access::generic_overlap): Be prepared to handle non-array
6750 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6752 PR rtl-optimization/84682
6753 * lra-constraints.c (process_address_1): Check is_address flag
6754 for address constraints.
6755 (process_alt_operands): Likewise.
6756 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6757 preprocess_constraints.
6758 * recog.h (preprocess_constraints): Add oploc parameter.
6760 * recog.c (preprocess_constraints): Test address_operand for
6761 CT_ADDRESS constraints.
6763 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6766 * lra-assigns.c (assign_by_spills): Return a flag of reload
6767 assignment failure. Do not process the reload assignment
6768 failures. Do not spill other reload pseudos if they has the same
6770 (lra_assign): Add a return arg. Set up from the result of
6771 assign_by_spills call.
6772 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6773 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6774 usage_insns if it is not NULL.
6775 (spill_hard_reg_in_range): New function.
6776 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6777 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6778 function prototypes.
6779 (lra_assign): Change prototype.
6780 * lra.c (lra): Add code to deal with fails by splitting hard reg
6783 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6786 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6787 Accept complain bool parameter. Only emit errors if it is true.
6788 (arm_parse_cpu_option_name): Likewise.
6789 (arm_target_thumb_only): Adjust callers of the above.
6790 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6791 prototype to take a default true bool parameter.
6792 (arm_parse_arch_option_name): Likewise.
6794 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6795 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6799 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6800 * configure: Regenerate.
6801 * configure.ac ("linker --version-script option"): New.
6802 ("linker soname option"): New.
6804 2018-03-09 Richard Biener <rguenther@suse.de>
6806 PR tree-optimization/84775
6807 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6808 immediate uses of predicate stmts and mark them modified.
6811 PR tree-optimization/84178
6812 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6814 (version_loop_for_if_conversion): Delay update_ssa call.
6815 (tree_if_conversion): Delay update_ssa until after predicate
6818 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6821 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6822 when the function accesses prior frames.
6824 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6827 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6828 gen_llsym, otherwise call maybe_gen_llsym.
6831 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6832 has ',' character inside of it.
6834 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6837 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6838 as clobbering CC_REGNUM.
6840 2018-03-08 Richard Biener <rguenther@suse.de>
6843 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6844 (follow_copies_to_constant): Do not follow SSA names registered
6847 2018-03-08 Richard Biener <rguenther@suse.de>
6849 PR tree-optimization/84178
6850 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6852 (version_loop_for_if_conversion): Delay update_ssa call.
6853 (tree_if_conversion): Delay update_ssa until after predicate
6856 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6858 PR tree-optimization/84178
6859 * tree-if-conv.c (release_bb_predicate): Remove the
6860 the assertion that the stmts have NULL use_ops.
6861 Discard the statements, asserting that they haven't
6862 yet been added to a BB.
6864 2018-03-08 Richard Biener <rguenther@suse.de>
6866 PR tree-optimization/84746
6867 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6868 (phi_translate): Pass in destination ANTIC_OUT set.
6869 (phi_translate_1): Likewise. For a simplified result lookup
6870 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6871 (phi_translate_set): Adjust.
6872 (do_pre_regular_insertion): Likewise.
6873 (do_pre_partial_partial_insertion): Likewise.
6875 2018-03-08 Martin Liska <mliska@suse.cz>
6877 PR gcov-profile/84735
6878 * doc/gcov.texi: Document usage of profile files.
6879 * gcov-io.h: Document changes in the format.
6881 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6885 * dwarf2out.c (struct dw_line_info_table): Update comments for
6887 (FORCE_RESET_NEXT_VIEW): New.
6888 (FORCE_RESETTING_VIEW_P): New.
6889 (RESETTING_VIEW_P): Check for -1 too.
6890 (ZERO_VIEW_P): Likewise.
6891 (new_line_info_table): Force-reset next view.
6892 (dwarf2out_begin_function): Likewise.
6893 (dwarf2out_source_line): Simplify zero_view_p initialization.
6894 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6895 view directly. Omit view when omitting .loc at line 0.
6897 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6899 PR tree-optimization/84740
6900 * tree-switch-conversion.c (process_switch): Call build_constructors
6901 only if info.phi_count is non-zero.
6903 PR tree-optimization/84739
6904 * tree-tailcall.c (find_tail_calls): Check call arguments against
6905 DECL_ARGUMENTS (current_function_decl) rather than
6906 DECL_ARGUMENTS (func) when checking for tail recursion.
6908 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6910 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6911 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6912 Volker Reichelt's entry and add entries for people that perform
6913 GCC fuzzy testing and report numerous bugs.
6915 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6918 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6919 readonly data in sdata, if that is disabled.
6920 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6921 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6922 -mreadonly-in-sdata option.
6924 2018-03-07 Martin Sebor <msebor@redhat.com>
6926 PR tree-optimization/84468
6927 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6928 basic block when looking for nul assignment.
6930 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6933 * except.h (output_function_exception_table): Adjust prototype.
6934 * except.c (output_function_exception_table): Remove FNNAME parameter
6935 and add SECTION parameter. Ouput one part of the table at a time.
6936 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6937 the first part of the exception table and emit unwind directives.
6938 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6939 (i386_pe_seh_cold_init): Likewise.
6940 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6941 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6942 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6943 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6944 * config/i386/winnt.c: Include except.h.
6945 (struct seh_frame_state): Add reg_offset, after_prologue and
6946 in_cold_section fields.
6947 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6948 (i386_pe_seh_cold_init): New function.
6949 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6950 to seh->in_cold_section.
6951 (seh_emit_push): Record the offset of the push.
6952 (seh_emit_save): Record the offet of the save.
6953 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6954 Test seh->after_prologue to disregard the epilogue.
6955 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6956 (i386_pe_end_cold_function): New function.
6958 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6961 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6962 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6965 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6967 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6968 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6971 * multiple_target.c: Include tree-inline.h and intl.h.
6972 (expand_target_clones): Diagnose and fail if node->definition and
6973 !tree_versionable_function_p (node->decl).
6975 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6977 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6979 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6980 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6981 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6983 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6986 * combine.c (try_combine): Use reg_or_subregno instead of handling
6987 just paradoxical SUBREGs and REGs.
6989 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6991 * config/arc/arc.c (arc_finalize_pic): Remove function.
6992 (arc_must_save_register): We use single base PIC register, remove
6993 checks to save/restore the PIC register.
6994 (arc_expand_prologue): Likewise.
6995 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6997 (arc_verify_short): Likewise.
6998 (arc_attr_type): Likewise.
6999 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
7000 (walk_stores): Likewise.
7001 (arc_address_cost): Make it static.
7002 (arc_verify_short): Likewise.
7003 (branch_dest): Likewise.
7004 (arc_attr_type): Likewise.
7005 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7006 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7007 (arc_final_prescan_insn): Remove inserting the nops due to
7008 hardware hazards. It is done in reorg step.
7009 (insn_length_variant_t): Remove.
7010 (insn_length_parameters_t): Likewise.
7011 (arc_insn_length_parameters): Likewise.
7012 (arc_get_insn_variants): Likewise.
7013 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7015 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7018 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7021 PR tree-optimization/84687
7022 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7024 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7026 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7028 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7029 Rename to ppc_speculation_barrier.
7030 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7031 __builtin_ppc_speculation_barrier.
7033 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7036 * combine.c (combine_simplify_rtx): Don't try to simplify if
7037 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7040 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7042 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7043 to 32 bytes when compiling for POWER9.
7045 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7048 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7049 regparm >= 3 with no arg reg available also for calls with
7050 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7053 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7055 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7057 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7060 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7062 2018-03-05 Richard Biener <rguenther@suse.de>
7064 PR tree-optimization/84486
7065 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7066 When inserting a __builtin_assume_aligned call set the LHS
7067 SSA name alignment info accordingly.
7069 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7071 PR tree-optimization/84114
7072 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7073 Avoid reassociation of FLOAT_MODE addition.
7075 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7077 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7078 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7079 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7080 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7081 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7082 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7083 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7085 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7086 __builtin_ia32_wbinvd): New builtins.
7087 (SPECIAL_ARGS2): New.
7088 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7089 (SPECIAL_ARGS2): New.
7090 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7091 (ix86_valid_target_attribute_inner_p): Ditto.
7092 (ix86_init_mmx_sse_builtins): Add special_args2.
7093 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7094 TARGET_WBNOINVD_P): New.
7095 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7096 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7097 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7098 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7099 * config/i386/pconfigintrin.h: New file.
7100 * config/i386/wbnoinvdintrin.h: Ditto.
7101 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7102 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7104 2018-03-05 Richard Biener <rguenther@suse.de>
7106 PR tree-optimization/84670
7107 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7109 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7110 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7111 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7112 doesn't grow if all successors (recursively) were visited at least
7115 2018-03-05 Richard Biener <rguenther@suse.de>
7117 PR tree-optimization/84650
7118 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7119 if executed in the loop pipeline.
7121 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7123 * doc/configfiles.texi (Configuration Files): Move info about
7124 conditionalizing $target-protos.h to...
7125 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7126 differs from $target-protos.h.
7128 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7129 Chung-Ju Wu <jasonwucj@gmail.com>
7131 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7132 * config/nds32/nds32-multiple.md (setmemsi): Define.
7133 * config/nds32/nds32-memory-manipulation.c
7134 (nds32_gen_dup_4_byte_to_word_value): New.
7135 (emit_setmem_word_loop): New.
7136 (emit_setmem_byte_loop): New.
7137 (nds32_expand_setmem_loop): New.
7138 (nds32_expand_setmem_loop_v3m): New.
7139 (nds32_expand_setmem_unroll): New.
7140 (nds32_expand_setmem): New.
7142 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7143 Chung-Ju Wu <jasonwucj@gmail.com>
7145 * config/nds32/nds32-memory-manipulation.c
7146 (nds32_emit_load_store): New.
7147 (nds32_emit_post_inc_load_store): New.
7148 (nds32_emit_mem_move): New.
7149 (nds32_emit_mem_move_block): New.
7150 (nds32_expand_movmemsi_loop_unknown_size): New.
7151 (nds32_expand_movmemsi_loop_known_size): New.
7152 (nds32_expand_movmemsi_loop): New.
7153 (nds32_expand_movmemsi_unroll): New.
7154 (nds32_expand_movmemqi): Rename ...
7155 (nds32_expand_movmemsi): ... to this.
7156 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7157 (movmemsi): ... to this.
7158 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7159 (nds32_expand_movmemsi): ... to this.
7161 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7162 Monk Chiang <sh.chiang04@gmail.com>
7163 Chung-Ju Wu <jasonwucj@gmail.com>
7165 * config/nds32/nds32-protos.h
7166 (nds32_expand_load_multiple): New arguments.
7167 (nds32_expand_store_multiple): Ditto.
7168 (nds32_valid_multiple_load_store): Rename ...
7169 (nds32_valid_multiple_load_store_p): ... to this.
7170 * config/nds32/nds32-memory-manipulation.c
7171 (nds32_expand_load_multiple): Refine implementation.
7172 (nds32_expand_store_multiple): Ditto.
7173 * config/nds32/nds32-multiple.md
7174 (load_multiple): Update nds32_expand_load_multiple interface.
7175 (store_multiple): Update nds32_expand_store_multiple interface.
7176 * config/nds32/nds32-predicates.c
7177 (nds32_valid_multiple_load_store): Rename ...
7178 (nds32_valid_multiple_load_store_p): ... to this and refine
7180 * config/nds32/predicates.md
7181 (nds32_load_multiple_and_update_address_operation): New predicate.
7182 (nds32_store_multiple_and_update_address_operation): New predicate.
7184 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7185 Chung-Ju Wu <jasonwucj@gmail.com>
7187 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7188 (combo): New attribute.
7189 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7191 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7193 * config/nds32/nds32.opt: Change -mcmodel= default value.
7195 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7196 Monk Chiang <sh.chiang04@gmail.com>
7197 Chung-Ju Wu <jasonwucj@gmail.com>
7199 * config/nds32/constants.md (unspec_element): New enum.
7200 * config/nds32/constraints.md (Umw): New constraint.
7201 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7202 * config/nds32/nds32-intrinsic.md: Likewise.
7203 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7204 (nds32_valid_smw_lwm_base_p): New.
7205 (nds32_output_smw_single_word): New.
7206 (nds32_output_lmw_single_word): New.
7207 (nds32_expand_unaligned_load): New.
7208 (nds32_expand_unaligned_store): New.
7209 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7210 (nds32_output_smw_single_word): Declare.
7211 (nds32_output_lmw_single_word): Declare.
7212 (nds32_expand_unaligned_load): Declare.
7213 (nds32_expand_unaligned_store): Declare.
7214 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7215 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7216 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7217 NDS32_BUILTIN_UASTORE_DW.
7218 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7221 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7222 Kito Cheng <kito.cheng@gmail.com>
7223 Chung-Ju Wu <jasonwucj@gmail.com>
7225 * config/nds32/nds32-intrinsic.c
7226 (nds32_expand_builtin_null_ftype_reg): Delete.
7227 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7228 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7229 (nds32_read_argument): New.
7230 (nds32_legitimize_target): Ditto.
7231 (nds32_legitimize_argument): Ditto.
7232 (nds32_check_constant_argument): Ditto.
7233 (nds32_expand_unop_builtin): Ditto.
7234 (nds32_expand_unopimm_builtin): Ditto.
7235 (nds32_expand_binop_builtin): Ditto.
7236 (nds32_builtin_decl_impl): Ditto.
7237 (builtin_description): Ditto.
7238 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7239 (nds32_init_builtins_impl): Ditto.
7240 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7241 (nds32_builtin_decl): New.
7242 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7243 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7245 2018-03-02 Jeff Law <law@redhat.com>
7247 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7248 (redundant_insn, fill_simple_delay_slots): Likewise.
7249 (fill_slots_from_thread): Likewise.
7250 * resource.c (mark_referenced_resources): Likewise.
7251 (mark_set_resources, find_dead_or_set_registers): Likewise.
7253 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7255 * substring-locations.h (format_warning_va): Formatting fix for
7257 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7259 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7260 * substring-locations.c: Include intl.h.
7261 (format_warning_va): Turned into small wrapper around
7262 format_warning_n_va, renamed to ...
7263 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7264 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7266 (format_warning_at_substring_n): New function.
7267 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7268 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7269 format_warning_at_substring with just a shorter name instead of
7270 const function pointer.
7271 (fmtwarn_n): New function.
7272 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7273 appropriate, get rid of all the fmtstr temporaries, move conditionals
7274 with G_() wrapped string literals directly into fmtwarn arguments,
7275 cast dir.len to (int), formatting fixes.
7277 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7279 * doc/invoke.texi: Remove "Cilk Plus" references.
7281 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7282 Richard Biener <rguenther@suse.de>
7285 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7286 for error or warning attributes if CALL_FROM_THUNK_P is set.
7289 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7292 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7293 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7296 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7297 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7299 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7300 instead of -1U in last predictors element's probability member.
7302 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7305 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7306 arguments if they are comparable.
7308 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7310 PR tree-optimization/84634
7311 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7312 masks and masked_loop_p with a single loop_masks, making sure it's
7313 null for bb vectorization.
7315 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7317 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7318 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7319 loop->force_vectorize to check whether there is no alias.
7321 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7324 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7326 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7328 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7329 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7330 instead of a loop around prev_real_insn.
7331 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7335 * config/i386/i386.c (ix86_print_operand): Use conditional
7336 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7339 2018-03-02 Richard Biener <rguenther@suse.de>
7341 PR tree-optimization/84427
7342 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7343 (bitmap_set_subtract_values): Rewrite to handle multiple
7346 (prune_clobbered_mems): Likewise.
7347 (phi_translate): Take edge instead of pred/phiblock.
7348 (phi_translate_1): Likewise.
7349 (phi_translate_set): Likewise. Insert all translated
7350 exprs for a value into the set, keeping possibly multiple
7351 expressions per value.
7352 (compute_antic_aux): Adjust for phi_translate changes.
7353 When intersecting union the expressions and prune those
7354 not in the final value set, keeping possibly multiple
7355 expressions per value. Do not use value-insertion
7356 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7357 all expressions. Add verification that the value-sets
7358 only shrink during iteration.
7359 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7360 (do_pre_regular_insertion): Likewise.
7361 (do_pre_partial_partial_insertion): Likewise.
7363 2018-03-02 Richard Biener <rguenther@suse.de>
7366 * config/darwin.c (saved_debug_info_level): New static global.
7367 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7368 (darwin_asm_lto_end): Restore debug info generation settings.
7370 2018-03-01 Martin Liska <mliska@suse.cz>
7373 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7376 2018-03-01 Richard Biener <rguenther@suse.de>
7379 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7380 type annotation with LTO when debug was disabled at compile-time.
7382 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7384 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7386 (mips_final_postscan_insn): Likewise.
7388 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7390 PR rtl-optimization/84528
7391 * alias.c (init_alias_target): Add commentary.
7392 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7393 a unique base value if the frame pointer is not eliminated
7394 to the stack pointer.
7396 2018-03-01 Tom de Vries <tom@codesourcery.com>
7398 PR rtl-optimization/83327
7399 * lra-int.h (hard_regs_spilled_into): Declare.
7400 * lra.c (hard_regs_spilled_into): Define.
7401 (init_reg_info): Init hard_regs_spilled_into.
7402 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7403 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7404 (process_bb_lives): Handle hard_regs_spilled_into.
7405 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7406 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7408 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7410 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7411 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7412 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7413 * config/rs6000/aix72.h: New file.
7415 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7417 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7418 instead of warning_at with conditional singular and plural messages
7422 * stor-layout.c (update_alignment_for_field): For
7423 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7424 && !DECL_PACKED (field), do the alignment update, just use
7425 only desired_align instead of MAX (type_align, desired_align)
7427 (place_field): Don't do known_align < desired_align handling
7428 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7429 is non-NULL, instead do it after rli->prev_field handling and
7430 only if not within a bitfield word. For DECL_PACKED (field)
7431 use type_align of BITS_PER_UNIT.
7433 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7435 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7436 superfluous parentheses and trailing spaces.
7438 2018-02-28 Richard Biener <rguenther@suse.de>
7440 PR tree-optimization/84584
7441 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7442 SCoPs with fake exit edge.
7444 2018-02-28 Martin Liska <mliska@suse.cz>
7447 * timevar.c (timer::print): Fix format to properly print 100%
7450 2018-02-28 Richard Biener <rguenther@suse.de>
7453 * genmatch.c (capture_info::walk_match): Do not mark
7454 captured expressions without operands as expr_p given
7455 they act more like predicates and should be subject to
7456 "lost tail" side-effect preserving.
7458 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7460 PR rtl-optimization/81611
7461 * auto-inc-dec.c (attempt_change): Move dead note from
7462 mem_insn if it's the next use of regno
7463 (find_address): Take address use of reg holding
7464 non-incremented value. Add parm to limit search to the named
7466 (merge_in_block): Attempt to use a mem insn that is the next
7467 use of the original regno.
7469 2018-02-27 Martin Sebor <msebor@redhat.com>
7472 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7473 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7475 2018-02-27 Martin Sebor <msebor@redhat.com>
7477 PR translation/84207
7478 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7479 n argument to unsigned HOST_WIDE_INT.
7480 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7481 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7482 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7483 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7485 2018-02-27 Richard Biener <rguenther@suse.de>
7487 PR tree-optimization/84512
7488 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7489 Do not use the estimate returned from record_stmt_cost for
7490 the scalar iteration cost but sum properly using add_stmt_cost.
7492 2018-02-27 Richard Biener <rguenther@suse.de>
7494 PR tree-optimization/84466
7495 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7496 Adjust last change to less strictly validate use operands.
7498 2018-02-27 Martin Liska <mliska@suse.cz>
7500 PR gcov-profile/84548
7501 * gcov.c (process_file): Allow partial overlap and consider it
7502 also as group functions.
7503 (output_lines): Properly calculate range of lines for a group.
7505 2018-02-27 Martin Liska <mliska@suse.cz>
7507 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7508 'ggc' suffixes. Change first column width.
7509 (timer::print): Fix formatting of the column.
7511 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7513 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7514 preserve inline entry blocks for the sake of debug inline
7515 entry point markers alone.
7516 (remove_unused_locals): Suggest in comments a better place to
7517 force the preservation of inline entry blocks that are
7518 otherwise unused, but do not preserve them.
7520 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7522 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7524 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7527 * config/i386/constraints.md (Bs): Replace
7528 ix86_indirect_branch_register with
7529 TARGET_INDIRECT_BRANCH_REGISTER.
7531 * config/i386/i386.md (indirect_jump): Likewise.
7532 (tablejump): Likewise.
7533 (*sibcall_memory): Likewise.
7534 (*sibcall_value_memory): Likewise.
7535 Peepholes of indirect call and jump via memory: Likewise.
7536 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7537 (*sibcall_value_GOT_32): Likewise.
7538 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7539 (GOT_memory_operand): Likewise.
7540 (call_insn_operand): Likewise.
7541 (sibcall_insn_operand): Likewise.
7542 (GOT32_symbol_operand): Likewise.
7543 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7545 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7547 PR rtl-optimization/83496
7548 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7549 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7550 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7551 redundant insn, if any.
7552 (relax_delay_slots): Likewise.
7553 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7555 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7557 PR tree-optimization/83965
7558 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7559 that grouped statements are part of a reduction chain. Return
7560 true if the statement is not marked as a reduction itself but
7562 (vect_recog_dot_prod_pattern): Don't check whether the statement
7563 is part of a group here.
7564 (vect_recog_sad_pattern): Likewise.
7565 (vect_recog_widen_sum_pattern): Likewise.
7567 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7570 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7572 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7575 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7577 (ix86_output_indirect_function_return): New prototype.
7578 (ix86_split_simple_return_pop_internal): Likewise.
7579 * config/i386/i386.c (indirect_return_via_cx): New.
7580 (indirect_return_via_cx_bnd): Likewise.
7581 (indirect_thunk_name): Handle return va CX_REG.
7582 (output_indirect_thunk_function): Create alias for
7583 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7584 (ix86_output_indirect_jmp): Remove the bool argument.
7585 (ix86_output_indirect_function_return): New function.
7586 (ix86_split_simple_return_pop_internal): Likewise.
7587 * config/i386/i386.md (*indirect_jump): Don't pass false
7588 to ix86_output_indirect_jmp.
7589 (*tablejump_1): Likewise.
7590 (simple_return_pop_internal): Change it to define_insn_and_split.
7591 Call ix86_split_simple_return_pop_internal to split it for
7593 (simple_return_indirect_internal): Call
7594 ix86_output_indirect_function_return instead of
7595 ix86_output_indirect_jmp.
7597 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7600 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7601 memset and value initialization afterwards.
7603 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7605 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7607 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7610 * common/config/aarch64/aarch64-common.c
7611 (aarch_option_optimization_table[]): Switch
7612 off fomit-frame-pointer
7614 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7615 Chung-Ju Wu <jasonwucj@gmail.com>
7617 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7619 (store_multiple): Ditto.
7621 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7623 * config.gcc: Add --with-cpu support for nds32 target.
7624 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7625 * config/nds32/nds32.opt: Add -mcpu= option.
7627 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7629 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7630 isel=yes): Warn for these deprecated options.
7632 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7634 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7635 ISA_2_5_MASKS_EMBEDDED.
7637 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7639 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7640 p->max as pointers rather than using iterative_hash_expr.
7642 2018-02-23 Carl Love <cel@us.ibm.com>
7644 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7645 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7647 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7648 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7649 P8V_BUILTIN_VEC_VUNSIGNED2.
7651 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7654 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7655 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7657 * lra-constraints.c (curr_insn_transform): Set up
7658 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7660 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7661 * lra-eliminations.c (spill_pseudos): Ditto.
7662 (process_insn_for_elimination): Ditto.
7663 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7664 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7665 LRA_NON_CLOBBERED_ALT.
7667 2018-02-22 Martin Sebor <msebor@redhat.com>
7669 PR tree-optimization/84480
7670 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7671 to maybe_diag_stxncpy_trunc. Call it.
7672 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7673 from gimple_fold_builtin_strcpy. Print inlining stack.
7674 (handle_builtin_stxncpy): Print inlining stack.
7675 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7677 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7680 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7681 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7682 and -fcheck-pointer-bounds are used together.
7683 (indirect_thunk_prefix): New enum.
7684 (indirect_thunk_need_prefix): New function.
7685 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7686 "_nt" instead of "_bnd" for NOTRACK prefix.
7687 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7688 (output_indirect_thunk_function): Likewise.
7690 (ix86_code_end): Update output_indirect_thunk_function calls.
7691 (ix86_output_indirect_branch_via_reg): Replace
7692 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7693 (ix86_output_indirect_branch_via_push): Likewise.
7694 (ix86_output_function_return): Likewise.
7695 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7696 incompatible with -fcf-protection=branch and
7697 -fcheck-pointer-bounds.
7699 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7702 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7703 Change gcc_assert call to output_operand_lossage.
7705 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7707 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7709 2018-02-22 DJ Delorie <dj@redhat.com>
7710 Sebastian Perta <sebastian.perta@renesas.com>
7711 Oleg Endo <olegendo@gcc.gnu.org>
7713 * config/rx/rx.c (rx_rtx_costs): New function.
7714 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7716 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7718 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7720 2018-02-22 Martin Liska <mliska@suse.cz>
7723 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7724 Add "native" as a possible value.
7726 2018-02-22 Martin Liska <mliska@suse.cz>
7729 * config/i386/i386.c (ix86_option_override_internal):
7730 Add "native" as a possible value for -march and -mtune.
7732 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7735 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7736 to all type variants.
7738 PR tree-optimization/84503
7739 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7740 width as info->bitpos + info->bitsize - start.
7741 (merged_store_group::merge_overlapping): Simplify width computation.
7742 (check_no_overlap): New function.
7743 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7744 start + width and last_order of the group, fail if check_no_overlap
7746 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7747 to group if check_no_overlap fails.
7749 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7751 * config/rs6000/altivec.md: Delete contraint arguments to
7752 define_expand, define_split, and define_peephole2, and in
7753 define_insn_and_split if always unused.
7754 * config/rs6000/darwin.md: Ditto.
7755 * config/rs6000/dfp.md: Ditto.
7756 * config/rs6000/rs6000.md: Ditto.
7757 * config/rs6000/sync.md: Ditto.
7758 * config/rs6000/vector.md: Ditto.
7759 * config/rs6000/vsx.md: Ditto.
7761 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7763 * config/rs6000/altivec.md: Write output control strings as braced
7764 blocks instead of double-quoted strings.
7765 * config/rs6000/darwin.md: Ditto.
7766 * config/rs6000/rs6000.md: Ditto.
7767 * config/rs6000/vector.md: Ditto.
7768 * config/rs6000/vsx.md: Ditto.
7770 2018-02-21 Jason Merrill <jason@redhat.com>
7772 PR c++/84314 - ICE with templates and fastcall attribute.
7773 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7775 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7777 * ipa-cp.c (determine_versionability): Fix comment typos.
7779 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7782 * ipa-cp.c (determine_versionability): Do not version functions caling
7785 2018-02-21 Martin Liska <mliska@suse.cz>
7788 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7789 Add "native" as a possible value.
7790 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7791 the macro when native cpu detection is available.
7793 2018-02-21 Martin Liska <mliska@suse.cz>
7796 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7797 Add "native" as a possible value.
7798 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7799 when native cpu detection is available.
7801 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7802 Martin Sebor <msebor@redhat.com>
7804 PR tree-optimization/84478
7805 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7807 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7808 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7809 support which is conservatively correct, for 2 only stay conservative
7810 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7811 argument to the 2 argument get_range_strlen, adjust 6 arg
7812 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7814 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7815 (gimple_fold_builtin_strlen): Pass true as last argument to
7818 2018-02-20 Martin Sebor <msebor@redhat.com>
7821 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7822 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7823 (builtin_memref::builtin_memref): Factor out parts into
7824 set_base_and_offset and call it.
7826 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7829 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7830 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7831 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7832 search at the associated MODE_INT.
7834 2018-02-20 Jeff Law <law@redhat.com>
7837 PR tree-optimization/81592
7839 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7840 for range data rather than using global data.
7841 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7842 range data rather than using global data.
7843 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7844 pass it to children as needed.
7845 (struct directive::fmtresult): Similarly.
7846 (struct directive::set_width): Similarly.
7847 (struct directive::set_precision): Similarly.
7848 (format_integer, format_directive, parse_directive): Similarly.
7849 (format_none): Accept unnamed vr_values parameter.
7850 (format_percent, format_floating, format_character): Similarly.
7851 (format_string, format_plain): Similarly.
7852 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7853 the EVRP range analyzer for range data rather than using global data.
7854 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7855 gimple-ssa-evrp-analyze.h
7856 (class sprintf_dom_walker): Add after_dom_children member function.
7857 Add evrp_range_analyzer member.
7858 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7859 range analyzer as needed.
7860 (sprintf_dom_walker::after_dom_children): New member function.
7861 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7863 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7864 (evrp_range_analyzer::pop_to_marker): Likewise.
7866 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7868 PR tree-optimization/84419
7869 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7870 with the required type if its current type is compatible but
7873 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7876 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7877 after vectorization.
7879 2018-02-20 Martin Liska <mliska@suse.cz>
7882 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7883 possible values if we don't have a hint.
7885 2018-02-20 Martin Liska <mliska@suse.cz>
7889 * final.c (shorten_branches): Build align_tab array with one
7891 * opts.c (finish_options): Add alignment option limit check.
7892 (MAX_CODE_ALIGN): Likewise.
7893 (MAX_CODE_ALIGN_VALUE): Likewise.
7894 * doc/invoke.texi: Document maximum allowed option value for
7895 all -falign-* options.
7897 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7900 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7901 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7902 * var-tracking.c (emit_note_insn_var_location): Remove all references
7903 to NOTE_INSN_CALL_ARG_LOCATION.
7904 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7905 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7906 Use copy_rtx_if_shared.
7907 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7908 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7909 (dwarf2out_var_location): Remove handling of
7910 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7912 * final.c (final_scan_insn): Remove all references to
7913 NOTE_INSN_CALL_ARG_LOCATION.
7914 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7915 before dumping final insns.
7916 * except.c (emit_note_eh_region_end): Remove all references to
7917 NOTE_INSN_CALL_ARG_LOCATION.
7918 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7919 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7920 * config/arc/arc.c (hwloop_optimize): Likewise.
7921 * config/arm/arm.c (create_fix_barrier): Likewise.
7922 * config/s390/s390.c (s390_chunkify_start): Likewise.
7923 * config/sh/sh.c (find_barrier): Likewise.
7924 * config/i386/i386.c (rest_of_insert_endbranch,
7925 ix86_seh_fixup_eh_fallthru): Likewise.
7926 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7927 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7928 * config/frv/frv.c (frv_function_prologue): Likewise.
7929 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7931 (note_outside_basic_block_p): Remove all references to
7932 NOTE_INSN_CALL_ARG_LOCATION.
7933 * gengtype.c (adjust_field_rtx_def): Likewise.
7934 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7936 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7937 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7940 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7943 PR tree-optimization/84452
7944 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7945 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7948 2018-02-19 Martin Liska <mliska@suse.cz>
7951 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7953 2018-02-19 Martin Liska <mliska@suse.cz>
7954 Richard Sandiford <richard.sandiford@linaro.org>
7956 PR tree-optimization/82491
7957 * gimple-fold.c (get_base_constructor): Make earlier bail out
7960 2018-02-19 Carl Love <cel@us.ibm.com>
7962 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7963 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7965 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7966 P8V_BUILTIN_VEC_NEG.
7968 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7970 * config/rl78/rl78.md (movdf): New define expand.
7972 2018-02-19 Martin Liska <mliska@suse.cz>
7975 * doc/invoke.texi: Fix typo.
7976 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7978 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7980 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7981 handle rs6000_single_float and rs6000_double_float specially for
7984 2018-02-16 Jeff Law <law@redhat.com>
7986 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7989 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7992 * ipa-inline.c (inline_small_functions): Fix a typo.
7994 2018-02-16 Nathan Sidwell <nathan@acm.org>
7996 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7998 2018-02-16 Carl Love <cel@us.ibm.com>
8000 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
8001 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
8002 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
8003 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
8004 expansion to P8V_BUILTIN_VEC_FLOAT2.
8006 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8008 PR rtl-optimization/70023
8009 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8010 src_regno into account.
8012 2018-02-16 Carl Love <cel@us.ibm.com>
8014 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8015 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8016 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8017 * config/rs6000/rs6000.c: Remove case statements for
8018 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8019 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8020 and P9V_BUILTIN_VEC_VINSERT4B.
8021 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8022 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8023 * config/rs6000/vsx.md:
8024 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8027 2018-02-16 Carl Love <cel@us.ibm.com>
8029 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8031 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8033 * config/rs6000/rs6000-c.c: Add the definitions for
8034 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8035 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8036 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8037 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8038 definition for insert4b and define insn *insert3b_internal.
8039 * doc/extend.texi: Add documentation for vec_extract4b.
8041 2018-02-16 Nathan Sidwell <nathan@acm.org>
8043 * doc/extend.texi (Backwards Compatibility): Mention friend
8044 injection. Note for-scope is deprecated.
8045 * doc/invoke.texi (-ffriend-injection): Deprecate.
8047 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8049 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8050 that moved to I2, also allow destinations that are a paradoxical
8051 subreg (instead of a normal reg).
8053 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8056 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8059 2018-02-16 Richard Biener <rguenther@suse.de>
8061 PR tree-optimization/84037
8062 PR tree-optimization/84016
8064 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8065 Adjust vec_construct for the fact we need additional higher latency
8066 128bit inserts for AVX256 and AVX512 vector builds.
8067 (ix86_add_stmt_cost): Scale vector construction cost for
8070 2018-02-16 Richard Biener <rguenther@suse.de>
8072 PR tree-optimization/84417
8073 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8074 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8075 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8077 2018-02-16 Martin Liska <mliska@suse.cz>
8080 * internal-fn.def (ASAN_CHECK): Set proper flags.
8081 (ASAN_MARK): Likewise.
8083 2018-02-16 Julia Koval <julia.koval@intel.com>
8085 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8086 from PTA_CANNONLAKE.
8088 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8091 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8092 Use ++iter rather than iter++ for std::list iterators.
8093 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8094 defer deleting them until all nodes in the forest are processed. Do
8095 free even leaf nodes. Change to_process into auto_vec.
8098 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8099 * vec.h (vec_default_construct): Use memset instead of placement new
8100 if BROKEN_VALUE_INITIALIZATION is defined.
8101 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8102 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8105 PR rtl-optimization/83723
8106 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8107 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8108 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8110 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8112 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8114 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8116 PR rtl-optimization/81443
8117 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8118 from inner REGs to paradoxical SUBREGs.
8120 2018-02-16 Richard Biener <rguenther@suse.de>
8122 PR tree-optimization/84399
8123 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8124 For operands we can analyze at their definition make sure we can
8125 analyze them at each use as well.
8127 2018-02-16 Richard Biener <rguenther@suse.de>
8129 PR tree-optimization/84190
8130 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8131 volatile accesses if the decl isn't volatile.
8133 2018-02-15 Jason Merrill <jason@redhat.com>
8135 PR c++/84314 - ICE with templates and fastcall attribute.
8136 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8137 TYPE_CANONICAL on an existing type.
8139 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8141 PR tree-optimization/84383
8142 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8143 dstoff nor call operand_equal_p if dstbase is NULL.
8145 PR tree-optimization/84334
8146 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8147 also a CONSTANT_CLASS_P, punt.
8149 2018-02-14 Jim Wilson <jimw@sifive.com>
8151 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8152 first SMALL_OPERAND check. New local min_second_step. Move assert
8153 to where locals are set. Add TARGET_RVC support.
8154 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8156 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8158 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8160 2018-02-14 Martin Sebor <msebor@redhat.com>
8162 PR tree-optimization/83698
8163 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8164 arrays constrain the offset range to their bounds.
8165 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8166 (builtin_access::overlap): Avoid setting the size of overlap if it's
8168 (maybe_diag_overlap): Also consider arrays when deciding what values
8169 of offsets to include in diagnostics.
8171 2018-02-14 Martin Sebor <msebor@redhat.com>
8174 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8175 that correspond to the kind of a declaration.
8177 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8180 * config/pa/pa.md: Load address of PIC label using the linkage table
8181 if the label is nonlocal.
8183 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8185 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8186 warning message if user requests -maltivec=be.
8187 * doc/invoke.texi: Document deprecation of -maltivec=be.
8189 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8192 * config/rs6000/rs6000-c.c: Update definitions for
8193 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8194 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8196 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8199 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8200 add _get_ssp intrinsics. Remove argument from
8201 __builtin_ia32_rdssp[d|q].
8202 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8203 * config/i386/i386-builtin.def: Remove argument from
8204 __builtin_ia32_rdssp[d|q].
8205 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8206 ix86_expand_special_args_builtin for _rdssp[d|q].
8207 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8208 Clear register before usage.
8209 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8210 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8212 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8214 PR tree-optimization/84357
8215 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8216 operand 1 of an ARRAY_REF too.
8218 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8221 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8222 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8224 (set_of_reg): New struct.
8225 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8226 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8227 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8229 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8230 Split into bitclr, bitset, bitinvert patterns if appropriate.
8231 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8232 use rx_fuse_in_memory_bitop.
8233 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8234 to named insn, correct maximum insn length.
8236 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8239 * machmode.def: Define a complex mode for PARTIAL_INT.
8240 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8242 * doc/rtl.texi: Document CSPImode.
8243 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8245 (msp430_hard_regno_nregs_with_padding): Likewise.
8247 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8250 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8252 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8254 PR rtl-optimization/84169
8255 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8256 we generated a parallel as new i3 and we split that to new i2 and i3
8257 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8258 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8259 those to i2, not i1. Partially rewrite this scan code.
8261 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8264 * stor-layout.c (place_field): For variable length fields, adjust
8265 offset_align afterwards not just based on the field's alignment,
8266 but also on the size.
8269 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8270 of exps and logs in the use_exp2 case.
8272 2018-02-13 Jeff Law <law@redhat.com>
8274 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8277 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8280 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8284 * common.opt (gas-loc-support, gas-locview-support): New.
8285 (ginline-points, ginternal-reset-location-views): New.
8286 * doc/invoke.texi: Document them. Use @itemx where intended.
8287 (gvariable-location-views): Adjust.
8288 * target.def (reset_location_view): New.
8289 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8290 (TARGET_RESET_LOCATION_VIEW): New.
8291 * doc/tm.texi: Rebuilt.
8292 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8293 (dwarf2out_default_as_locview_support): New.
8294 (output_asm_line_debug_info): Use option variables.
8295 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8296 (output_loc_list): Likewise.
8297 (add_high_low_attributes): Check option variables.
8298 Don't output entry view attribute in strict mode.
8299 (gen_inlined_subroutine_die): Check option variables.
8300 (dwarf2out_inline_entry): Likewise.
8301 (init_sections_and_labels): Likewise.
8302 (dwarf2out_early_finish): Likewise.
8303 (maybe_reset_location_view): New, from...
8304 (dwarf2out_var_location): ... here. Call it.
8305 * debug.h (dwarf2out_default_as_loc_support): Declare.
8306 (dwarf2out_default_as_locview_support): Declare.
8307 * hooks.c (hook_int_rtx_insn_0): New.
8308 * hooks.h (hook_int_rtx_insn_0): Declare.
8309 * toplev.c (process_options): Take -gas-loc-support and
8310 -gas-locview-support from dwarf2out. Enable
8311 -gvariable-location-views by default only with locview
8312 assembler support. Enable -ginternal-reset-location-views by
8313 default only if the target defines the corresponding hook.
8314 Enable -ginline-points by default if location views are
8315 enabled; force it disabled if statement frontiers are
8317 * tree-inline.c (expand_call_inline): Check option variables.
8318 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8320 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8322 PR tree-optimization/84321
8323 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8324 handling. Also check whether the anti-range contains any values
8325 that satisfy the mask; switch to a VR_RANGE if not.
8327 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8330 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8332 2018-02-13 Martin Jambor <mjambor@suse.cz>
8335 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8336 of call statements, also set location of a load to a temporary.
8338 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8340 * config/rl78/rl78.c (add_vector_labels): New function.
8341 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8342 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8343 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8344 which checks that no arguments are passed.
8345 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8346 * doc/extend.texi: Documentation for the new attribute.
8348 2018-02-13 Andreas Schwab <schwab@suse.de>
8350 * config/riscv/linux.h (CPP_SPEC): Define.
8352 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8355 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8356 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8357 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8358 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8359 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8360 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8361 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8362 temporarily for AES and PCLMUL builtins.
8364 PR tree-optimization/84339
8365 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8366 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8370 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8371 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8372 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8374 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8376 * omp-simd-clone.h: New file.
8377 * omp-simd-clone.c: Include omp-simd-clone.h.
8378 (expand_simd_clones): No longer static.
8379 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8380 cgraph.h and omp-simd-clone.h.
8381 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8382 (vect_recog_widen_shift_pattern): Formatting fix.
8383 (vect_pattern_recog_1): Don't check optab for calls.
8386 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8387 operands[2] into a REG before using gen_lowpart on it.
8389 2018-02-12 Jeff Law <law@redhat.com>
8392 * config/sh/sh.c (find_barrier): Consider a sibling call
8395 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8396 successfully back substituting a reg.
8398 2018-02-12 Richard Biener <rguenther@suse.de>
8400 PR tree-optimization/84037
8401 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8402 parameter, move visited init to caller.
8403 (vect_slp_analyze_operations): Separate cost from validity
8404 check, initialize visited once for all instances.
8405 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8407 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8409 (vect_model_promotion_demotion_cost): Likewise.
8410 (vectorizable_bswap): Guard cost modeling with !slp_node
8411 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8413 (vectorizable_call): Likewise.
8414 (vectorizable_conversion): Likewise.
8415 (vectorizable_assignment): Likewise.
8416 (vectorizable_shift): Likewise.
8417 (vectorizable_operation): Likewise.
8418 (vectorizable_store): Likewise.
8419 (vectorizable_load): Likewise.
8420 (vectorizable_condition): Likewise.
8421 (vectorizable_comparison): Likewise.
8423 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8426 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8427 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8428 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8430 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8433 * params.def (inline-min-speedup): Increase from 8 to 15.
8434 (max-inline-insns-auto): Decrease from 40 to 30.
8435 * ipa-split.c (consider_split): Add some buffer for function to
8436 be considered inlining candidate.
8437 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8440 2018-02-12 Richard Biener <rguenther@suse.de>
8442 PR tree-optimization/84037
8443 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8444 matched stmts if we cannot swap the non-matched ones.
8446 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8448 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8449 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8450 _mm_maskz_scalef_round_ss): New intrinsics.
8451 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8452 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8453 __builtin_ia32_scalefss_round): Remove.
8454 (__builtin_ia32_scalefsd_mask_round,
8455 __builtin_ia32_scalefss_mask_round): New intrinsics.
8456 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8457 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8458 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8459 "<round_constraint>")): Changed to ...
8460 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8461 "<round_scalar_constraint>")): ... this.
8462 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8463 %0, %1, %2<round_op3>}"): Changed to ...
8464 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8465 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8466 %2<round_scalar_mask_op3>}"): ... this.
8467 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8469 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8471 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8472 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8473 (_mm_maskz_sqrt_round_ss): New intrinsics.
8474 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8475 (__builtin_ia32_sqrtsd_mask_round)
8476 (__builtin_ia32_sqrtss_mask_round): New builtins.
8477 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8478 (__builtin_ia32_sqrtss_round): Remove.
8479 (__builtin_ia32_sqrtsd_mask_round)
8480 (__builtin_ia32_sqrtss_mask_round): New builtins.
8481 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8482 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8483 ((match_operand:VF_128 1 "vector_operand"
8484 "xBm,<round_constraint>")): Changed to ...
8485 ((match_operand:VF_128 1 "vector_operand"
8486 "xBm,<round_scalar_constraint>")): ... this.
8487 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8488 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8489 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8490 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8491 %<iptr>1<round_scalar_mask_op3>}): ... this.
8492 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8493 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8495 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8498 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8499 Cast vec_cmpeq result to correct type.
8500 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8501 Cast vec_cmpgt result to correct type.
8503 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8505 * final.c (final_scan_insn_1): Renamed from...
8506 (final_scan_insn): ... this. New wrapper, to recover
8507 seen from the outermost call in recursive ones.
8508 * config/sparc/sparc.c (output_return): Drop seen from call.
8509 (output_sibcall): Likewise.
8510 * config/visium/visium.c (output_branch): Likewise.
8512 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8514 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8517 2018-02-10 Alan Modra <amodra@gmail.com>
8520 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8521 Specify LR as an input.
8523 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8526 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8527 remove_member_access_dummy_vars): New functions.
8528 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8529 lower_omp_1, execute_lower_omp): Use them.
8531 PR rtl-optimization/84308
8532 * shrink-wrap.c (spread_components): Release todo vector.
8534 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8536 PR rtl-optimization/57193
8537 * ira-color.c (struct allocno_color_data): Add member
8538 conflict_allocno_hard_prefs.
8539 (update_conflict_allocno_hard_prefs): New.
8540 (bucket_allocno_compare_func): Add a preference based on
8541 conflict_allocno_hard_prefs.
8542 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8543 (color_allocnos): Remove a dead code. Initiate
8544 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8546 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8549 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8550 constraint from =wa to wa. Avoid a subreg on the output operand,
8551 instead use a pseudo and subreg it in a move.
8552 (p9_xxbrd_<mode>): Changed to ...
8553 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8554 (p9_xxbrd_v2df): New expander.
8555 (p9_xxbrw_<mode>): Changed to ...
8556 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8557 (p9_xxbrw_v4sf): New expander.
8559 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8561 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8563 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8566 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8567 multiply in 32-bit mode.
8568 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8569 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8572 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8574 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8575 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8576 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8577 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8579 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8582 * dwarf2out.c (is_trivial_indirect_ref): New function.
8583 (dwarf2out_late_global_decl): Do not generate a location
8584 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8585 and that are not defined in the current unit.
8587 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8589 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8590 instead of a libcall for UNORDERED.
8592 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8595 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8596 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8598 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8601 * config/s390/s390.c (s390_set_current_function): Invoke
8602 s390_indirect_branch_settings also if fndecl didn't change.
8604 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8606 * config/rs6000/rs6000.md (blockage): Set length to zero.
8608 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8610 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8612 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8615 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8616 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8620 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8621 PARALLEL incoming that failed vt_get_decl_and_offset check.
8624 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8625 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8627 (get_variable_section): For decls in named .bss* sections pass true as
8628 second argument to bss_initializer_p.
8630 2018-02-09 Marek Polacek <polacek@redhat.com>
8631 Jakub Jelinek <jakub@redhat.com>
8634 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8635 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8636 Sync some changes from cxx_fold_indirect_ref.
8638 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8640 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8642 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8643 (BLOCK_INLINE_ENTRY_LABEL): New.
8644 (dwarf2out_var_location): Disregard inline entry markers.
8645 (inline_entry_data): New struct.
8646 (inline_entry_data_hasher): New hashtable type.
8647 (inline_entry_data_hasher::hash): New.
8648 (inline_entry_data_hasher::equal): New.
8649 (inline_entry_data_table): New variable.
8650 (add_high_low_attributes): Add DW_AT_entry_pc and
8651 DW_AT_GNU_entry_view attributes if a pending entry is found
8652 in inline_entry_data_table. Add old entry_pc attribute only
8653 if debug nonbinding markers are disabled.
8654 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8655 markers are enabled.
8656 (block_within_block_p, dwarf2out_inline_entry): New.
8657 (dwarf2out_finish): Check that no entries remained in
8658 inline_entry_data_table.
8659 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8660 (final_scan_insn, notice_source_line): Likewise.
8661 (rest_of_clean_state): Skip inline entry markers.
8662 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8664 * gimple.c (gimple_build_debug_inline_entry): New.
8665 * gimple.h (enum gimple_debug_subcode): Add
8666 GIMPLE_DEBUG_INLINE_ENTRY.
8667 (gimple_build_debug_inline_entry): Declare.
8668 (gimple_debug_inline_entry_p): New.
8669 (gimple_debug_nonbind_marker_p): Adjust.
8670 * insn-notes.def (INLINE_ENTRY): New.
8671 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8672 inline entry marker notes.
8673 (print_insn): Likewise.
8674 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8675 (INSN_DEBUG_MARKER_KIND): Likewise.
8676 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8677 * tree-inline.c (expand_call_inline): Build and insert
8678 debug_inline_entry stmt.
8679 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8680 inline entry blocks early, if nonbind markers are enabled.
8681 (dump_scope_block): Dump fragment info.
8682 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8683 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8684 (gimple_build_debug_inline_entry): New.
8685 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8686 Enable/disable inline entry points too.
8687 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8688 (DEBUG_INSN): Describe inline entry markers.
8690 * common.opt (gvariable-location-views): New.
8691 (gvariable-location-views=incompat5): New.
8692 * config.in: Rebuilt.
8693 * configure: Rebuilt.
8694 * configure.ac: Test assembler for view support.
8695 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8696 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8697 * dwarf2out.c (var_loc_view): New typedef.
8698 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8699 (dwarf2out_locviews_in_attribute): New.
8700 (dwarf2out_locviews_in_loclist): New.
8701 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8702 (enum dw_line_info_opcode): Add LI_adv_address.
8703 (struct dw_line_info_table): Add view.
8704 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8705 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8706 (zero_view_p): New variable.
8707 (ZERO_VIEW_P): New macro.
8708 (output_asm_line_debug_info): New.
8709 (struct var_loc_node): Add view.
8710 (add_AT_view_list, AT_loc_list): New.
8711 (add_var_loc_to_decl): Add view param. Test it against last.
8712 (new_loc_list): Add view params. Record them.
8713 (AT_loc_list_ptr): Handle loc and view lists.
8714 (view_list_to_loc_list_val_node): New.
8715 (print_dw_val): Handle dw_val_class_view_list.
8716 (size_of_die): Likewise.
8717 (value_format): Likewise.
8718 (loc_list_has_views): New.
8719 (gen_llsym): Set vl_symbol too.
8720 (maybe_gen_llsym, skip_loc_list_entry): New.
8721 (dwarf2out_maybe_output_loclist_view_pair): New.
8722 (output_loc_list): Output view list or entries too.
8723 (output_view_list_offset): New.
8724 (output_die): Handle dw_val_class_view_list.
8725 (output_dwarf_version): New.
8726 (output_compilation_unit_header): Use it.
8727 (output_skeleton_debug_sections): Likewise.
8728 (output_rnglists, output_line_info): Likewise.
8729 (output_pubnames, output_aranges): Update version comments.
8730 (output_one_line_info_table): Output view numbers in asm comments.
8731 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8732 Call maybe_gen_llsym.
8733 (loc_list_from_tree_1): Adjust.
8734 (add_AT_location_description): Create view list attribute if
8735 needed, check it's absent otherwise.
8736 (convert_cfa_to_fb_loc_list): Adjust.
8737 (maybe_emit_file): Call output_asm_line_debug_info for test.
8738 (dwarf2out_var_location): Reset views as needed. Precompute
8739 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8740 attribute. Set view.
8741 (new_line_info_table): Reset next view.
8742 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8743 (dwarf2out_source_line): Likewise. Output view resets and labels to
8744 the assembler, or select appropriate line info opcodes.
8745 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8746 (optimize_string_length): Catch it. Adjust.
8747 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8748 dw_val_class_view_list, and remove it if no longer needed.
8749 (hash_loc_list): Hash view numbers.
8750 (loc_list_hasher::equal): Compare them.
8751 (optimize_location_lists): Check whether a view list symbol is
8752 needed, and whether the locview attribute is present, and
8753 whether they match. Remove the locview attribute if no longer
8755 (index_location_lists): Call skip_loc_list_entry for test.
8756 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8757 Use output_dwarf_version.
8758 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8759 (struct dw_val_node): Add val_view_list.
8760 * final.c (SEEN_NEXT_VIEW): New.
8761 (set_next_view_needed): New.
8762 (clear_next_view_needed): New.
8763 (maybe_output_next_view): New.
8764 (final_start_function): Rename to...
8765 (final_start_function_1): ... this. Take pointer to FIRST,
8766 add SEEN parameter. Emit param bindings in the initial view.
8767 (final_start_function): Reintroduce SEEN-less interface.
8768 (final): Rename to...
8769 (final_1): ... this. Take SEEN parameter. Output final pending
8770 next view at the end.
8771 (final): Reintroduce seen-less interface.
8772 (final_scan_insn): Output pending next view before switching
8773 sections or ending a block. Mark the next view as needed when
8774 outputting variable locations. Notify debug backend of section
8775 changes, and of location view changes.
8776 (rest_of_handle_final): Adjust.
8777 * toplev.c (process_options): Autodetect value for debug variable
8778 location views option. Warn on incompat5 without -gdwarf-5.
8779 * doc/invoke.texi (gvariable-location-views): New.
8780 (gvariable-location-views=incompat5): New.
8781 (gno-variable-location-views): New.
8783 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8785 PR tree-optimization/84136
8786 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8787 that the result of find_edge is non-NULL.
8789 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8792 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8793 storing integer register in SImode. Fix cost of 256 and 512
8794 byte aligned SSE register store.
8796 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8798 * config/i386/i386.c (ix86_multiplication_cost): Fix
8799 multiplication cost for TARGET_AVX512DQ.
8801 2018-02-08 Marek Polacek <polacek@redhat.com>
8803 PR tree-optimization/84238
8804 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8807 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8809 PR tree-optimization/84265
8810 * tree-vect-stmts.c (vectorizable_store): Don't treat
8811 VMAT_CONTIGUOUS accesses as grouped.
8812 (vectorizable_load): Likewise.
8814 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8816 PR tree-optimization/81635
8817 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8818 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8819 (test_round_for_mask): New functions.
8820 (wide_int_cc_tests): Call test_round_for_mask.
8821 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8822 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8823 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8824 range returned by get_range_info.
8826 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8829 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8830 * symtab.c: Include builtins.h
8831 (symtab_node::output_to_lto_symbol_table_p): Move here
8832 from lto-streamer-out.c:output_symbol_p.
8833 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8834 (output_symbol_p): Move all logic to symtab.c
8835 (produce_symtab): Update.
8837 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8839 * config/s390/s390-opts.h (enum indirect_branch): Define.
8840 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8841 (s390_indirect_branch_via_thunk)
8842 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8843 (enum s390_indirect_branch_type): Define.
8844 * config/s390/s390.c (struct s390_frame_layout, struct
8845 machine_function): Remove.
8846 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8847 (indirect_branch_table_label_no, indirect_branch_table_name):
8849 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8850 (enum s390_indirect_branch_option): Define.
8851 (s390_return_addr_from_memory): New function.
8852 (s390_handle_string_attribute): New function.
8853 (s390_attribute_table): Add new attribute handler.
8854 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8855 (s390_indirect_branch_via_thunk): New function.
8856 (s390_indirect_branch_via_inline_thunk): New function.
8857 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8858 sibling call optimization for non z10 compiles.
8859 (s390_emit_call): Force indirect branch target to be a single
8860 register. Add r1 clobber for non-z10 compiles.
8861 (s390_emit_epilogue): Emit return jump via return_use expander.
8862 (s390_reorg): Handle JUMP_INSNs as execute targets.
8863 (s390_option_override_internal): Perform validity checks for the
8864 new command line options.
8865 (s390_indirect_branch_attrvalue): New function.
8866 (s390_indirect_branch_settings): New function.
8867 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8868 (s390_output_indirect_thunk_function): New function.
8869 (s390_code_end): Implement target hook.
8870 (s390_case_values_threshold): Implement target hook.
8871 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8873 * config/s390/s390.h (struct s390_frame_layout)
8874 (struct machine_function): Move here from s390.c.
8875 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8876 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8877 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8878 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8879 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8880 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8881 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8882 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8883 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8884 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8885 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8886 (mnemonic attribute): Add values which aren't recognized
8888 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8889 pattern for branch conversion. Fix mnemonic attribute.
8890 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8891 indirect branch via thunk if requested.
8892 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8893 ("*indirect_jump"): Disable for branch conversion using out of
8895 ("indirect_jump_via_thunk<mode>_z10")
8896 ("indirect_jump_via_thunk<mode>")
8897 ("indirect_jump_via_inlinethunk<mode>_z10")
8898 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8899 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8900 ("casesi_jump_via_inlinethunk<mode>_z10")
8901 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8902 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8903 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8904 ("*indirect2_jump"): Disable for branch conversion.
8905 ("casesi_jump"): Turn into expander and expand patterns for branch
8907 ("return_use"): New expander.
8908 ("*return"): Emit return via thunk and rename it to ...
8909 ("*return<mode>"): ... this one.
8910 * config/s390/s390.opt: Add new options and and enum for the
8913 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8915 * lra-constraints.c (match_reload): Unconditionally use
8916 gen_lowpart_SUBREG, rather than selecting between that
8917 and equivalent gen_rtx_SUBREG code.
8919 2018-02-08 Richard Biener <rguenther@suse.de>
8921 PR tree-optimization/84233
8922 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8923 changed flag instead of boguously re-using phi_inserted.
8925 2018-02-08 Martin Jambor <mjambor@suse.cz>
8927 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8928 static local variables.
8930 2018-02-08 Richard Biener <rguenther@suse.de>
8932 PR tree-optimization/84278
8933 * tree-vect-stmts.c (vectorizable_store): When looking for
8934 smaller vector types to perform grouped strided loads/stores
8935 make sure the mode is supported by the target.
8936 (vectorizable_load): Likewise.
8938 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8940 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8941 Increase LDP/STP opportunities by adding adjacent callee-saves.
8943 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8945 PR rtl-optimization/84068
8946 PR rtl-optimization/83459
8947 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8949 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8951 PR tree-optimization/84224
8952 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8953 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8956 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8959 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8960 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8961 count, remove one USE.
8963 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8965 * doc/install.texi (Configuration): Document the
8966 --with-long-double-format={ibm,ieee} PowerPC configuration
8970 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8971 Convert from define_expand to be define_insn_and_split. Rework
8972 float/double/_Float128 conversions to QI/HI/SImode to work with
8973 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8974 conversions to QI/HImode types did a store and then a load to
8975 truncate the value. For conversions to VSX registers, don't split
8976 the insn, instead emit the code directly. Use the code iterator
8977 any_fix to combine signed and unsigned conversions.
8978 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8979 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8980 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8981 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8982 (fix_<mode>di2_hw): Likewise.
8983 (fixuns_<mode>di2_hw): Likewise.
8984 (fix_<mode>si2_hw): Likewise.
8985 (fixuns_<mode>si2_hw): Likewise.
8986 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8987 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8988 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8989 fix<uns>_trunc<SFDF:mode>si2_p8.
8990 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8992 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8993 (fix<uns>_<mode>_mem): Likewise.
8994 (fctiw<u>z_<mode>_mem): Likewise.
8995 (fix<uns>_<mode>_mem): Likewise.
8996 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8997 the register allocator from doing a direct move to the GPRs to do
8998 a store, and instead use the ISA 3.0 store byte/half-word from
8999 vector register instruction. For IEEE 128-bit floating point,
9000 also optimize stores of 32-bit ints.
9001 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
9003 2018-02-07 Alan Hayward <alan.hayward@arm.com>
9005 * genextract.c (push_pathstr_operand): New function to support
9007 (walk_rtx): Call push_pathstr_operand.
9008 (print_path): Support [a-zA-Z].
9010 2018-02-07 Richard Biener <rguenther@suse.de>
9012 PR tree-optimization/84037
9013 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9014 (cse_and_gimplify_to_preheader): Declare.
9015 (vect_get_place_in_interleaving_chain): Likewise.
9016 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9018 (_loop_vec_info::~_loop_vec_info): Delete it.
9019 (cse_and_gimplify_to_preheader): New function.
9020 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9021 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9022 (vectorizable_load): Likewise. For grouped stores always base
9023 the IV on the first element.
9024 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9025 condition before gimplifying.
9027 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9029 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9030 *DIV_EXPR and *MOD_EXPR.
9032 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9035 * config/i386/i386.c (ix86_option_override_internal): Mask out
9036 the CF_SET bit when checking -fcf-protection.
9038 2018-02-07 Tom de Vries <tom@codesourcery.com>
9041 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9044 2018-02-07 Richard Biener <rguenther@suse.de>
9046 PR tree-optimization/84204
9047 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9050 PR tree-optimization/84205
9051 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9052 special-case isl_ast_op_zdiv_r.
9054 PR tree-optimization/84223
9055 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9056 Only add conditions from within the region.
9057 (gather_bbs::after_dom_children): Adjust.
9059 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9062 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9063 * config/avr/avr.md: Only post-reload split REG-REG moves if
9064 either register is GENERAL_REG_P.
9066 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9068 PR tree-optimization/84235
9069 * tree-ssa-scopedtables.c
9070 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9071 if the subtraction is performed in floating point type where NaNs are
9072 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9073 build 1. Formatting fix.
9075 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9078 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9079 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9080 and skip it regardless of bb boundaries. Use CALL_P macro,
9081 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9082 unnecessarily, formatting fix.
9084 2018-02-06 Michael Collison <michael.collison@arm.com>
9086 * config/arm/thumb2.md:
9087 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9088 (*thumb_mov_notscc): Ditto.
9090 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9093 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9094 unsigned_fix, not "s".
9096 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9098 * configure.ac (gcc_fn_eh_frame_ro): New function.
9099 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9100 correct .eh_frame permissions.
9101 * configure: Regenerate.
9103 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9105 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9108 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9110 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9111 Display warning message for -mno-speculate-indirect-jumps.
9113 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9115 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9117 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9119 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9121 PR tree-optimization/84225
9122 * tree-eh.c (find_trapping_overflow): Only call
9123 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9125 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9128 * config/i386/i386.c: Reimplement the check of possible options
9129 -mibt/-mshstk conbination. Change error messages.
9130 * doc/invoke.texi: Fix a typo: remove extra '='.
9132 2018-02-06 Marek Polacek <polacek@redhat.com>
9134 PR tree-optimization/84228
9135 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9137 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9140 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9141 emitted arch directives.
9142 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9143 __ARM_FEATURE_COPROC before changing architectures.
9145 2018-02-06 Richard Biener <rguenther@suse.de>
9147 * config/i386/i386.c (print_reg): Fix typo.
9148 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9150 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9152 * configure: Regenerate.
9154 2018-02-05 Martin Sebor <msebor@redhat.com>
9156 PR tree-optimization/83369
9157 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9160 2018-02-05 Martin Liska <mliska@suse.cz>
9162 * doc/invoke.texi: Cherry-pick upstream r323995.
9164 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9166 * ira.c (ira_init_register_move_cost): Adjust comment.
9168 2018-02-05 Martin Liska <mliska@suse.cz>
9170 PR gcov-profile/84137
9171 * doc/gcov.texi: Fix typo in documentation.
9173 2018-02-05 Martin Liska <mliska@suse.cz>
9175 PR gcov-profile/83879
9176 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9177 using dlopen functionality.
9179 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9181 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9182 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9183 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9184 _mm_maskz_range_round_ss): New intrinsics.
9185 (__builtin_ia32_rangesd128_round)
9186 (__builtin_ia32_rangess128_round): Remove.
9187 (__builtin_ia32_rangesd128_mask_round,
9188 __builtin_ia32_rangess128_mask_round): New builtins.
9189 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9190 __builtin_ia32_rangess128_round): Remove.
9191 (__builtin_ia32_rangesd128_mask_round,
9192 __builtin_ia32_rangess128_mask_round): New builtins.
9193 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9194 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9195 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9196 "<round_saeonly_constraint>")): Changed to ...
9197 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9198 "<round_saeonly_scalar_constraint>")): ... this.
9199 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9200 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9201 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9202 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9203 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9205 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9207 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9209 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9210 Remove all values except native, 8540 and 8548.
9212 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9214 * config/i386/i386.c (ix86_output_function_return): Pass
9215 INVALID_REGNUM, instead of -1, as invalid register number to
9216 indirect_thunk_name and output_indirect_thunk.
9218 2018-02-02 Julia Koval <julia.koval@intel.com>
9220 * config.gcc: Add -march=icelake.
9221 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9222 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9223 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9224 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9225 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9226 (processor_target_table): Add icelake.
9227 (ix86_option_override_internal): Handle new PTAs.
9228 (get_builtin_code_for_version): Handle icelake.
9229 (M_INTEL_COREI7_ICELAKE): New.
9230 (fold_builtin_cpu): Handle icelake.
9231 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9232 * doc/invoke.texi: Add -march=icelake.
9234 2018-02-02 Julia Koval <julia.koval@intel.com>
9236 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9237 to wide_int_bitmask.
9238 * wide-int-bitmask.h: New.
9240 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9243 * config/i386/i386.md: Replace Pmode with word_mode in
9244 builtin_setjmp_setup and builtin_longjmp to support x32.
9246 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9250 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9252 (rs6000_supported_cpu_names): New static variable.
9253 (linux_cpu_translation_table): Likewise.
9254 (elf_platform) <cpu>: Define new static variable and use it.
9255 Translate kernel AT_PLATFORM name to canonical name if needed.
9256 Error if platform name is unknown.
9258 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9261 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9263 2018-02-01 Jeff Law <law@redhat.com>
9266 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9267 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9268 the scratch if RELEASE_VIA_POP is false.
9269 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9270 If we have to save a temporary register, decrement SIZE appropriately.
9271 Pass new arguments to release_scratch_register_on_entry.
9272 (ix86_adjust_stack_and_probe): Likewise.
9273 (ix86_emit_probe_stack_range): Pass new arguments to
9274 release_scratch_register_on_entry.
9276 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9278 PR rtl-optimization/84157
9279 * combine.c (change_zero_ext): Use REG_P predicate in
9280 front of HARD_REGISTER_P predicate.
9282 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9284 * config/avr/avr.c (avr_option_override): Move disabling of
9285 -fdelete-null-pointer-checks to...
9286 * common/config/avr/avr-common.c (avr_option_optimization_table):
9289 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9291 PR tree-optimization/81635
9292 * tree-data-ref.c (split_constant_offset_1): For types that
9293 wrap on overflow, try to use range info to prove that wrapping
9296 2018-02-01 Renlin Li <renlin.li@arm.com>
9299 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9301 (aarch64_register_move_cost): Likewise.
9302 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9304 (REG_CLASS_NAMES): Likewise.
9305 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9306 TAILCALL_ADDR_REGS. Remove IP registers.
9307 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9309 2018-02-01 Richard Biener <rguenther@suse.de>
9311 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9312 for specifying RPO order and allow NULL for that.
9313 * domwalk.c (dom_walker::dom_walker): Likewise.
9314 (dom_walker::walk): Handle NULL RPO order.
9315 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9317 (rewrite_update_dom_walker): Likewise.
9318 (mark_def_dom_walker): Likewise.
9320 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9322 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9323 (aarch64_maybe_expand_sve_subreg_move): Declare.
9324 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9325 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9327 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9328 that are semantically a reverse operation.
9329 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9330 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9331 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9333 (aarch64_can_change_mode_class): For big-endian, forbid changes
9334 between two SVE modes if they have different element sizes.
9336 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9338 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9339 the TImode handling for big-endian targets.
9341 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9343 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9344 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9346 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9347 Remove BSWAP handing for big-endian targets and use the form of
9348 LD1RQ appropariate for the mode.
9350 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9352 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9353 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9356 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9359 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9360 check for operands that need to go through aarch64_sve_reload_be.
9362 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9364 PR tree-optimization/81661
9365 PR tree-optimization/84117
9366 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9367 * tree-eh.c: Include gimplify.h.
9368 (find_trapping_overflow, replace_trapping_overflow,
9369 rewrite_to_non_trapping_overflow): New functions.
9370 * tree-vect-loop.c: Include tree-eh.h.
9371 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9372 * tree-data-ref.c: Include tree-eh.h.
9373 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9375 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9377 PR rtl-optimization/84123
9378 * combine.c (change_zero_ext): Check if hard register satisfies
9379 can_change_dest_mode before calling gen_lowpart_SUBREG.
9381 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9384 * ira.c (ira_init_register_move_cost): Remove assert.
9386 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9388 PR rtl-optimization/84071
9389 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9390 * doc/tm.texi: Regenerate.
9392 2018-01-31 Richard Biener <rguenther@suse.de>
9394 PR tree-optimization/84132
9395 * tree-data-ref.c (analyze_miv_subscript): Properly
9396 check whether evolution_function_is_affine_multivariate_p
9397 before calling gcd_of_steps_may_divide_p.
9399 2018-01-31 Julia Koval <julia.koval@intel.com>
9402 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9403 * config/i386/i386.md (rdpid_rex64) New.
9404 (rdpid): Make 32bit only.
9406 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9409 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9410 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9412 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9415 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9417 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9419 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9421 PR rtl-optimization/84071
9422 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9423 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9425 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9427 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9428 (arc_attribute_table): Add 'aux' attribute.
9429 (arc_in_small_data_p): Consider aux like variables.
9430 (arc_is_aux_reg_p): New function.
9431 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9432 (arc_get_aux_arg): New function.
9433 (prepare_move_operands): Handle aux-register access.
9434 (arc_handle_aux_attribute): New function.
9435 * doc/extend.texi (ARC Variable attributes): Add subsection.
9437 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9439 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9440 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9441 (arc_attribute_table): Add 'uncached' attribute.
9442 (arc_print_operand): Print '.di' flag for uncached memory
9444 (arc_in_small_data_p): Do not consider for small data the uncached
9446 (arc_is_uncached_mem_p): New function.
9447 * config/arc/predicates.md (compact_store_memory_operand): Check
9448 for uncached memory accesses.
9449 (nonvol_nonimm_operand): Likewise.
9450 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9452 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9455 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9456 falign-loops=): Add Optimization flag.
9458 2018-01-30 Jeff Law <law@redhat.com>
9461 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9462 INT_REGISTERS_SAVED. Check it prior to calling
9463 get_scratch_register_on_entry.
9464 (ix86_adjust_stack_and_probe): Similarly.
9465 (ix86_emit_probe_stack_range): Similarly.
9466 (ix86_expand_prologue): Corresponding changes.
9468 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9471 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9472 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9474 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9477 * lra-constraints.c (curr_insn_transform): Process AND in the
9480 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9482 PR rtl-optimization/83986
9483 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9484 dependence against last_pending_memory_flush in addition to
9487 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9489 PR tree-optimization/81611
9490 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9493 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9496 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9499 2018-01-30 Richard Biener <rguenther@suse.de>
9500 Jakub Jelinek <jakub@redhat.com>
9502 PR tree-optimization/84111
9503 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9504 inner loops added during recursion, as they don't have up-to-date
9507 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9510 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9511 (can_inline_edge_by_limits_p): ... here.
9512 (can_early_inline_edge_p, check_callers,
9513 update_caller_keys, update_callee_keys, recursive_inlining,
9514 add_new_edges_to_heap, speculation_useful_p,
9515 inline_small_functions,
9516 inline_small_functions, flatten_function,
9517 inline_to_all_callers_1): Update.
9519 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9521 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9524 2018-01-30 Richard Biener <rguenther@suse.de>
9526 PR tree-optimization/83008
9527 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9528 invariant and constant vector uses in stmts when they need
9531 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9534 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9535 * configure: Regenerate.
9537 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9539 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9541 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9542 Use gen_rtx_REG rather than gen_lowpart.
9544 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9546 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9547 rather than 0 when creating partial subregs.
9549 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9551 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9554 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9557 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9558 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9559 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9560 flags. This restores the settings used before the 2017-07-24.
9561 Turning off pre increment/decrement/modify allows IVOPTS to
9562 optimize DF/SF loops where the index is an int.
9564 2018-01-29 Richard Biener <rguenther@suse.de>
9565 Kelvin Nilsen <kelvin@gcc.gnu.org>
9568 * tree-vect-stmts.c (vectorizable_call): Don't call
9569 targetm.vectorize_builtin_md_vectorized_function if callee is
9572 2018-01-22 Carl Love <cel@us.ibm.com>
9574 * doc/extend.tex: Fix typo in second arg in
9575 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9577 2018-01-29 Richard Biener <rguenther@suse.de>
9579 PR tree-optimization/84086
9580 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9581 (flush_ssaname_freelist): When SSA names were released reset
9582 the SCEV hash table.
9584 2018-01-29 Richard Biener <rguenther@suse.de>
9586 PR tree-optimization/84057
9587 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9588 removed paths when removing edges.
9590 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9592 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9593 -mfunction-return=@var{choice}.
9595 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9598 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9600 (layout::print_source_line): Likewise.
9601 (test_get_line_width_without_trailing_whitespace): Add test cases.
9603 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9606 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9609 2018-01-26 Jim Wilson <jimw@sifive.com>
9611 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9613 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9616 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9618 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9619 and CMP + SUB-immediate -> SUBS.
9621 2018-01-26 Martin Sebor <msebor@redhat.com>
9623 PR tree-optimization/83896
9624 * tree-ssa-strlen.c (get_string_len): Rename...
9625 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9626 Avoid assuming length is constant.
9627 (handle_char_store): Use HOST_WIDE_INT for string length.
9629 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9632 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9633 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9635 2018-01-26 Richard Biener <rguenther@suse.de>
9637 PR rtl-optimization/84003
9638 * dse.c (record_store): Only record redundant stores when
9639 the earlier store aliases at least all accesses the later one does.
9641 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9643 PR rtl-optimization/83985
9644 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9645 REG_CFA_RESTORE insns.
9646 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9647 REG_CFA_RESTORE insns here.
9650 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9651 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9653 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9655 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9656 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9657 (arc_init): Likewise.
9658 (arc_override_options): Likewise.
9659 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9661 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9663 * config/arc/arc.h (TARGET_DBNZ): Define.
9664 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9665 properly set the tune attribute.
9666 (dbnz): Use TARGET_DBNZ guard.
9667 * config/arc/arc.opt (mtune): Add core3 option.
9669 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9671 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9672 recognize new pic like addresses.
9673 (arc_delegitimize_address): Clean up.
9675 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9677 * config/arc/arc-arches.def: Option mrf16 valid for all
9679 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9680 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9681 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9682 * config/arc/arc-tables.opt: Regenerate.
9683 * config/arc/arc.c (arc_conditional_register_usage): Handle
9684 reduced register file case.
9685 (arc_file_start): Set must have build attributes.
9686 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9688 * config/arc/arc.opt (mrf16): Add new option.
9689 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9690 * config/arc/genmultilib.awk: Handle new mrf16 option.
9691 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9692 * config/arc/t-multilib: Regenerate.
9693 * doc/invoke.texi (ARC Options): Document mrf16 option.
9695 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9697 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9698 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9699 (arc_attribute_table): Add 'secure_call' attribute.
9700 (arc_print_operand): Print secure call operand.
9701 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9703 (arc_is_secure_call_p): New function. * config/arc/arc.md
9704 (call_i): Add support for sjli instruction.
9705 (call_value_i): Likewise.
9706 * config/arc/constraints.md (Csc): New constraint.
9708 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9709 John Eric Martin <John.Martin@emmicro-us.com>
9711 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9712 * config/arc/arc.c (_arc_jli_section): New struct.
9713 (arc_jli_section): New type.
9714 (rc_jli_sections): New static variable.
9715 (arc_handle_jli_attribute): New function.
9716 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9717 (arc_file_end): New function.
9718 (TARGET_ASM_FILE_END): Define.
9719 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9720 (arc_add_jli_section): New function.
9721 (jli_call_scan): Likewise.
9722 (arc_reorg): Call jli_call_scan.
9723 (arc_output_addsi): Remove 'S' from printing asm operand.
9724 (arc_is_jli_call_p): New function.
9725 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9727 (movhi_insn): Likewise.
9728 (movsi_insn): Likewise.
9729 (movsi_set_cc_insn): Likewise.
9730 (loadqi_update): Likewise.
9731 (load_zeroextendqisi_update): Likewise.
9732 (load_signextendqisi_update): Likewise.
9733 (loadhi_update): Likewise.
9734 (load_zeroextendhisi_update): Likewise.
9735 (load_signextendhisi_update): Likewise.
9736 (loadsi_update): Likewise.
9737 (loadsf_update): Likewise.
9738 (movsicc_insn): Likewise.
9739 (bset_insn): Likewise.
9740 (bxor_insn): Likewise.
9741 (bclr_insn): Likewise.
9742 (bmsk_insn): Likewise.
9743 (bicsi3_insn): Likewise.
9744 (cmpsi_cc_c_insn): Likewise.
9745 (movsi_ne): Likewise.
9746 (movsi_cond_exec): Likewise.
9747 (clrsbsi2): Likewise.
9758 (call_i): Remove 'S' asm letter, add jli instruction.
9759 (call_value_i): Likewise.
9760 * config/arc/arc.op (mjli-always): New option.
9761 * config/arc/constraints.md (Cji): New constraint.
9762 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9764 (subsf3_fpx): Likewise.
9765 (mulsf3_fpx): Likewise.
9766 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9768 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9770 * doc/invoke.texi (ARC): Document mjli-always option.
9772 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9774 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9775 avoid addition with 0 and use incw and decw where possible.
9777 2018-01-26 Richard Biener <rguenther@suse.de>
9779 PR tree-optimization/81082
9780 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9781 association if it requires casting to unsigned.
9782 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9783 from fold_plusminus_mult_expr to catch important cases late when
9784 range info is available.
9786 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9788 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9789 * configure.ac (hidden_linkonce): New test.
9790 * configure: Regenerate.
9791 * config.in: Regenerate.
9793 2018-01-26 Julia Koval <julia.koval@intel.com>
9795 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9796 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9797 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9798 _mm_mask_bitshuffle_epi64_mask): Fix type.
9799 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9800 USI_FTYPE_V4DI_V4DI_USI): Remove.
9801 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9802 __builtin_ia32_vpshufbitqmb256_mask,
9803 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9804 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9805 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9807 2018-01-26 Alan Modra <amodra@gmail.com>
9810 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9811 UNSPEC_VBPERMQ. Sort other unspecs.
9813 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9815 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9817 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9820 * predict.c (drop_profile): Do not push/pop cfun; update also
9822 (handle_missing_profiles): Fix logic looking for zero profiles.
9824 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9827 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9828 on functions with #pragma omp declare simd or functions with simd
9830 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9831 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9832 Remove trailing \n from warning_at calls.
9834 2018-01-25 Tom de Vries <tom@codesourcery.com>
9837 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9838 for neutered workers.
9840 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9843 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9844 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9846 2018-01-24 Jeff Law <law@redhat.com>
9849 * i386.c (get_probe_interval): Move to earlier point.
9850 (ix86_compute_frame_layout): If -fstack-clash-protection and
9851 the frame is larger than the probe interval, then use pushes
9852 to save registers rather than reg->mem moves.
9853 (ix86_expand_prologue): Remove conditional for int_registers_saved
9856 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9859 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9860 min/max for never referenced object.
9862 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9865 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9867 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9868 attributes from DECL_ATTRIBUTES (decl) without affecting
9869 DECL_ATTRIBUTES (current_function_decl).
9870 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9871 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9873 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9875 PR tree-optimization/83979
9876 * fold-const.c (fold_comparison): Use constant_boolean_node
9877 instead of boolean_{true,false}_node.
9879 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9881 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9884 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9886 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9887 Simplify the clause that sets the length attribute.
9888 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9889 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9890 clause that sets the length attribute.
9891 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9893 2018-01-24 Tom de Vries <tom@codesourcery.com>
9896 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9897 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9898 Add strict parameter.
9899 (prevent_branch_around_nothing): Insert dummy insn between branch to
9900 label and label with no ptx insn inbetween.
9901 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9903 2018-01-24 Tom de Vries <tom@codesourcery.com>
9906 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9907 for neutered threads in warp.
9908 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9910 2018-01-24 Richard Biener <rguenther@suse.de>
9912 PR tree-optimization/83176
9913 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9916 2018-01-24 Richard Biener <rguenther@suse.de>
9918 PR tree-optimization/82819
9919 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9920 code generating pluses that are no-ops in the target precision.
9922 2018-01-24 Richard Biener <rguenther@suse.de>
9925 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9927 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9929 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9930 to merge probabilities.
9931 * predict.c (probably_never_executed): Also mark as cold functions
9932 with global 0 profile and guessed local profile.
9933 * profile-count.c (profile_probability::combine_with_count): New
9935 * profile-count.h (profile_probability::operator*,
9936 profile_probability::operator*=, profile_probability::operator/,
9937 profile_probability::operator/=): Reduce precision to adjusted
9938 and set value to guessed on contradictory divisions.
9939 (profile_probability::combine_with_freq): Remove.
9940 (profile_probability::combine_wiht_count): Declare.
9941 (profile_count::force_nonzero):: Set to adjusted.
9942 (profile_count::probability_in):: Set quality to adjusted.
9943 * tree-ssa-tail-merge.c (replace_block_by): Use
9946 2018-01-23 Andrew Waterman <andrew@sifive.com>
9947 Jim Wilson <jimw@sifive.com>
9949 * config/riscv/riscv.c (riscv_stack_boundary): New.
9950 (riscv_option_override): Set riscv_stack_boundary. Handle
9951 riscv_preferred_stack_boundary_arg.
9952 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9953 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9954 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9955 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9956 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9957 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9959 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9962 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9963 of struct ix86_frame.
9964 (ix86_expand_epilogue): Likewise. Add a local variable for
9965 the reg_save_offset field in struct ix86_frame.
9967 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9969 PR tree-optimization/82604
9970 * tree-loop-distribution.c (enum partition_kind): New enum item
9971 PKIND_PARTIAL_MEMSET.
9972 (partition_builtin_p): Support above new enum item.
9973 (generate_code_for_partition): Ditto.
9974 (compute_access_range): Differentiate cases that equality can be
9975 proven at all loops, the innermost loops or no loops.
9976 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9977 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9978 (finalize_partitions, distribute_loop): Don't fuse partition of
9979 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9980 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9983 2018-01-23 Martin Liska <mliska@suse.cz>
9985 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9986 order to ignore the predictor.
9987 (PRED_POLYMORPHIC_CALL): Likewise.
9988 (PRED_RECURSIVE_CALL): Likewise.
9990 2018-01-23 Martin Liska <mliska@suse.cz>
9992 * tree-profile.c (tree_profiling): Print function header to
9993 aware reader which function we are working on.
9994 * value-prof.c (gimple_find_values_to_profile): Do not print
9995 not interesting value histograms.
9997 2018-01-23 Martin Liska <mliska@suse.cz>
9999 * profile-count.h (enum profile_quality): Add
10000 profile_uninitialized as the first value. Do not number values
10001 as they are zero based.
10002 (profile_count::verify): Update sanity check.
10003 (profile_probability::verify): Likewise.
10005 2018-01-23 Nathan Sidwell <nathan@acm.org>
10007 * doc/invoke.texi (ffor-scope): Deprecate.
10009 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10011 PR tree-optimization/83510
10012 * domwalk.c (set_all_edges_as_executable): New function.
10013 (dom_walker::dom_walker): Convert bool param
10014 "skip_unreachable_blocks" to enum reachability. Move setup of
10015 edge flags to set_all_edges_as_executable and only do it when
10016 reachability is REACHABLE_BLOCKS.
10017 * domwalk.h (enum dom_walker::reachability): New enum.
10018 (dom_walker::dom_walker): Convert bool param
10019 "skip_unreachable_blocks" to enum reachability.
10020 (set_all_edges_as_executable): New decl.
10021 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10022 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10024 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10025 but converting true to REACHABLE_BLOCKS.
10026 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10028 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10029 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10030 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10032 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10033 if check_all_array_refs will be called.
10035 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10037 * tree.c (selftest::test_location_wrappers): Add more test
10040 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10042 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10043 (selftest::test_bit_in_range): Likewise.
10045 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10048 * doc/sourcebuild.texi (vect_float): Say that the selector
10049 only describes the situation when -funsafe-math-optimizations is on.
10050 (vect_float_strict): Document.
10052 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10054 PR tree-optimization/83965
10055 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10056 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10057 instead of checking only for a reduction.
10058 (vect_recog_widen_sum_pattern): Likewise.
10060 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10062 * predict.c (probably_never_executed): Only use precise profile info.
10063 (compute_function_frequency): Skip after inlining hack since we now
10064 have quality checking.
10066 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10068 * profile-count.h (profile_probability::very_unlikely,
10069 profile_probability::unlikely, profile_probability::even): Set
10070 precision to guessed.
10072 2018-01-23 Richard Biener <rguenther@suse.de>
10074 PR tree-optimization/83963
10075 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10076 Properly terminate dominator walk when crossing the exit edge not
10077 when visiting its source block.
10079 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10082 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10083 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10085 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10087 PR tree-optimization/83957
10088 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10089 semicolon after for body surrounded by braces.
10091 PR tree-optimization/83081
10092 * profile-count.h (profile_probability::split): New method.
10093 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10094 Use profile_probability::split.
10095 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10096 when splitting a single conditional jump into 2.
10098 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10100 PR tree-optimization/69452
10101 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10104 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10106 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10107 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10108 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10110 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10112 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10113 * config/rl78/rl78.md (movdi): New define_expand.
10114 * config/rl78/rl78.c (rl78_split_movdi): New function.
10116 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10119 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10121 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10122 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10123 128-bit to produce an UNSPEC move to get the double word with the
10124 signbit and then a shift directly to do signbit.
10125 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10126 implementation with a new version that just does either a direct
10127 move or a regular move. Move memory interface to separate insns.
10128 Move insns so they are next to the expander.
10129 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10130 with signbit move. Split big and little endian case.
10131 (signbit<mode>2_dm_mem_le): Likewise.
10132 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10133 (signbit<mode>2_dm2): Likewise.
10135 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10137 * config/rl78/rl78.md (anddi3): New define_expand.
10139 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10141 * config/rl78/rl78.md (umindi3): New define_expand.
10143 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10145 * config/rl78/rl78.md (smindi3): New define_expand.
10147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10149 * config/rl78/rl78.md (smaxdi3): New define_expand.
10151 2018-01-22 Carl Love <cel@us.ibm.com>
10153 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10154 LVX_V1TI): Add macro expansion.
10155 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10156 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10157 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10158 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10159 Change check to determine if the instruction is a byte reversing
10160 entry. Fix typo in comment.
10161 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10162 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10163 Add def_builtin calls for new builtins.
10164 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10165 Add define_insn expansion.
10167 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10169 * config/rl78/rl78.md (umaxdi3): New define_expand.
10171 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10173 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10174 for non-QImode registers.
10176 2018-01-22 Richard Biener <rguenther@suse.de>
10178 PR tree-optimization/83963
10179 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10180 including the loop exit block.
10181 (scop_detection::merge_sese): Likewise.
10182 (scop_detection::add_scop): Do it here instead.
10184 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10186 * doc/sourcebuild.texi (arm_softfloat): Document.
10188 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10191 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10192 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10193 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10195 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10196 David Edelsohn <dje.gcc@gmail.com>
10199 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10200 Change "crset eq" to "crset 2".
10201 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10202 (*call_indirect_aix<mode>_nospec): Likewise.
10203 (*call_value_indirect_aix<mode>_nospec): Likewise.
10204 (*call_indirect_elfv2<mode>_nospec): Likewise.
10205 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10206 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10207 change assembly output from . to $.
10208 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10209 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10210 (*tablejump<mode>_internal1_nospec): Likewise.
10212 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10215 * config/sh/sh_optimize_sett_clrt.cc:
10216 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10218 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10220 PR tree-optimization/83940
10221 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10222 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10223 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10224 use it to pass back the definition type.
10225 (vect_check_store_rhs): Likewise rhs_dt_out.
10226 (vect_build_gather_load_calls): Add a mask_dt argument and use
10227 it instead of a call to vect_is_simple_use.
10228 (vectorizable_store): Update calls to vect_check_load_store_mask
10229 and vect_check_store_rhs. Use the dt returned by the latter instead
10230 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10231 instead of calls to vect_is_simple_use. Pass the scalar rather
10232 than the vector operand to vect_is_simple_use when handling
10233 second and subsequent copies of an rhs value.
10234 (vectorizable_load): Update calls to vect_check_load_store_mask
10235 and vect_build_gather_load_calls. Use the cached mask_dt and
10236 gs_info.offset_dt instead of calls to vect_is_simple_use.
10238 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10240 PR middle-end/83945
10241 * tree-emutls.c: Include gimplify.h.
10242 (lower_emutls_2): New function.
10243 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10244 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10245 it before further processing.
10248 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10249 UINTVAL (trueop1) instead of INTVAL (op1).
10251 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10255 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10256 INCOMING_FRAME_SP_OFFSET if not defined.
10257 (scan_trace): Add ENTRY argument. If true and
10258 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10259 emit a note to adjust the CFA offset.
10260 (create_cfi_notes): Adjust scan_trace callers.
10261 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10262 INCOMING_FRAME_SP_OFFSET in the CIE.
10263 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10264 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10266 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10267 * doc/tm.texi: Regenerated.
10269 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10271 PR rtl-optimization/83147
10272 * lra-constraints.c (remove_inheritance_pseudos): Use
10273 lra_substitute_pseudo_within_insn.
10275 2018-01-19 Tom de Vries <tom@codesourcery.com>
10276 Cesar Philippidis <cesar@codesourcery.com>
10279 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10281 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10284 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10285 spaces for function labels.
10287 2018-01-19 Martin Liska <mliska@suse.cz>
10289 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10290 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10291 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10292 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10293 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10294 (PRED_CONST_RETURN): Change from 69 to 65.
10295 (PRED_NULL_RETURN): Change from 91 to 71.
10296 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10297 (PRED_LOOP_GUARD): Change from 66 to 73.
10299 2018-01-19 Martin Liska <mliska@suse.cz>
10301 * predict.c (predict_insn_def): Add new assert.
10302 (struct branch_predictor): Change type to signed integer.
10303 (test_prediction_value_range): Amend test to cover
10304 PROB_UNINITIALIZED.
10305 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10306 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10307 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10308 (PRED_LOOP_IV_COMPARE): Likewise.
10309 * predict.h (PROB_UNINITIALIZED): Define new constant.
10311 2018-01-19 Martin Liska <mliska@suse.cz>
10313 * predict.c (dump_prediction): Add new format for
10314 analyze_brprob.py script which is enabled with -details
10316 * profile-count.h (precise_p): New function.
10318 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10320 PR tree-optimization/83922
10321 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10322 there are no statements that need masking.
10323 (vect_active_double_reduction_p): New function.
10324 (vect_analyze_loop_operations): Use it when handling phis that
10325 are not in the loop header.
10327 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10329 PR tree-optimization/83914
10330 * tree-vect-loop.c (vectorizable_induction): Don't convert
10331 init_expr or apply the peeling adjustment for inductions
10332 that are nested within the vectorized loop.
10334 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10336 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10339 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10343 * function.h (gimplify_parameters): Add gimple_seq * argument.
10344 * function.c: Include gimple.h and options.h.
10345 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10346 for the added local temporaries if needed.
10347 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10348 if there are any parameter cleanups, wrap whole body into a
10349 try/finally with the cleanups.
10351 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10354 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10355 Use GET_MODE_CLASS for scalar floating point.
10357 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10361 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10362 Fix call of call_cgraph_insertion_hooks.
10364 2018-01-18 Martin Sebor <msebor@redhat.com>
10366 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10368 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10371 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10374 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10377 * common.opt: (-ffile-prefix-map): New option.
10378 * opts.c (common_handle_option): Defer it.
10379 * opts-global.c (handle_common_deferred_options): Handle it.
10380 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10381 * file-prefix-map.h: New file.
10382 (remap_debug_filename, add_debug_prefix_map): ...here.
10383 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10384 * final.c (debug_prefix_map, add_debug_prefix_map
10385 remap_debug_filename): Move to...
10386 * file-prefix-map.c: New file.
10387 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10388 generalize, get rid of alloca(), use strrchr() instead of strchr().
10389 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10390 Implement in terms of add_prefix_map().
10391 (remap_macro_filename, remap_debug_filename): Implement in term of
10393 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10394 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10395 * dbxout.c: Include file-prefix-map.h.
10396 * varasm.c: Likewise.
10397 * vmsdbgout.c: Likewise.
10398 * xcoffout.c: Likewise.
10399 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10400 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10401 * doc/invoke.texi (-ffile-prefix-map): Document.
10402 (-fdebug-prefix-map): Update description.
10404 2018-01-18 Martin Liska <mliska@suse.cz>
10406 * config/i386/i386.c (indirect_thunk_name): Document that also
10408 (output_indirect_thunk): Document why both instructions
10409 (pause and lfence) are generated.
10411 2018-01-18 Richard Biener <rguenther@suse.de>
10413 PR tree-optimization/83887
10414 * graphite-scop-detection.c
10415 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10416 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10417 (scop_detection::merge_sese): Re-implement with a flood-fill
10418 algorithm that properly finds a SESE region if it exists.
10420 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10423 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10424 pointer_diff optimizations use view_convert instead of convert.
10426 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10428 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10429 Generate different code for -mno-speculate-indirect-jumps.
10430 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10431 (*call_indirect_aix<mode>): Disable for
10432 -mno-speculate-indirect-jumps.
10433 (*call_indirect_aix<mode>_nospec): New define_insn.
10434 (*call_value_indirect_aix<mode>): Disable for
10435 -mno-speculate-indirect-jumps.
10436 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10437 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10438 -mno-speculate-indirect-jumps.
10439 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10441 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10443 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10444 long double type, set the flags for noting the default long double
10445 type, even if we don't pass or return a long double type.
10447 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10450 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10453 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10455 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10456 support for merge[hl].
10457 (fold_mergehl_helper): New helper function.
10458 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10459 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10460 (altivec_vmrglw_direct): Add xxmrglw insn.
10462 2018-01-17 Andrew Waterman <andrew@sifive.com>
10464 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10465 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10467 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10470 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10471 call the lto_location_cache before reading the
10472 DECL_SOURCE_LOCATION of the types.
10474 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10475 Richard Sandiford <richard.sandiford@linaro.org>
10477 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10478 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10479 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10480 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10481 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10483 * config/aarch64/constraints.md (aarch64_movti_operand):
10485 * config/aarch64/predicates.md (Uti): Add new constraint.
10487 2018-01-17 Carl Love <cel@us.ibm.com>
10489 * config/rs6000/vsx.md (define_expand xl_len_r,
10490 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10491 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10493 (define_expand, define_insn): Move the shift left from the
10494 define_insn to the define_expand for lxvl and stxvl instructions.
10495 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10496 and XL_LEN_R definitions to PURE.
10498 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10500 * config/i386/i386.c (indirect_thunk_name): Declare regno
10501 as unsigned int. Compare regno with INVALID_REGNUM.
10502 (output_indirect_thunk): Ditto.
10503 (output_indirect_thunk_function): Ditto.
10504 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10505 in the call to output_indirect_thunk_function.
10507 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10509 PR middle-end/83884
10510 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10511 rather than the size of inner_type to determine the stack slot size
10512 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10514 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10517 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10520 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10522 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10523 endian Linux systems to optionally enable multilibs for selecting
10524 the long double type if the user configured an explicit type.
10525 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10526 have no long double multilibs if not defined.
10527 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10528 warn if the user used -mabi={ieee,ibm}longdouble and we built
10529 multilibs for long double.
10530 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10531 appropriate multilib option.
10532 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10534 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10535 for building long double multilibs.
10536 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10538 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10540 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10543 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10545 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10548 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10551 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10554 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10556 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10557 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10559 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10561 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10562 different rtl trees depending on TARGET_64BIT.
10563 (rs6000_gen_lvx): Likewise.
10565 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10567 * config/visium/visium.md (nop): Tweak comment.
10568 (hazard_nop): Likewise.
10570 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10572 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10573 -mspeculate-indirect-jumps.
10574 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10575 for -mno-speculate-indirect-jumps.
10576 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10577 (*call_value_indirect_elfv2<mode>): Disable for
10578 -mno-speculate-indirect-jumps.
10579 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10580 (indirect_jump): Emit different RTL for
10581 -mno-speculate-indirect-jumps.
10582 (*indirect_jump<mode>): Disable for
10583 -mno-speculate-indirect-jumps.
10584 (*indirect_jump<mode>_nospec): New define_insn.
10585 (tablejump): Emit different RTL for
10586 -mno-speculate-indirect-jumps.
10587 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10588 (tablejumpsi_nospec): New define_expand.
10589 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10590 (tablejumpdi_nospec): New define_expand.
10591 (*tablejump<mode>_internal1): Disable for
10592 -mno-speculate-indirect-jumps.
10593 (*tablejump<mode>_internal1_nospec): New define_insn.
10594 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10597 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10599 * caller-save.c (insert_save): Drop unnecessary parameter. All
10602 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10603 Richard Biener <rguenth@suse.de>
10606 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10607 return early, inline manually is_gimple_sizepos. Make sure if we
10608 call gimplify_expr we don't end up with a gimple constant.
10609 * tree.c (variably_modified_type_p): Don't return true for
10610 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10611 * gimplify.h (is_gimple_sizepos): Remove.
10613 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10615 PR tree-optimization/83857
10616 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10617 vectorizable_live_operation for pure SLP statements.
10618 (vectorizable_live_operation): Handle PHIs.
10620 2018-01-16 Richard Biener <rguenther@suse.de>
10622 PR tree-optimization/83867
10623 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10624 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10626 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10629 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10630 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10631 If off is not INTEGER_CST, issue a may not be aligned warning
10632 rather than isn't aligned. Use isn%'t rather than isn't.
10633 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10635 <case MULT_EXPR>: Improve the case when bottom and one of the
10636 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10637 operand, in that case check if the other operand is multiple of
10638 bottom divided by the INTEGER_CST operand.
10640 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10643 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10644 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10645 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10646 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10647 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10648 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10649 (pa_function_arg_size): New function.
10651 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10653 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10654 in a separate statement.
10656 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10658 PR tree-optimization/83847
10659 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10660 group gathers and scatters.
10662 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10664 PR rtl-optimization/86620
10665 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10667 PR rtl-optimization/83213
10668 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10669 to last if both are JUMP_INSNs.
10671 PR tree-optimization/83843
10672 * gimple-ssa-store-merging.c
10673 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10674 store_immediate_info for bswap/nop orig_stores.
10676 2018-01-15 Andrew Waterman <andrew@sifive.com>
10678 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10680 <UDIV>: Increase cost if !TARGET_DIV.
10682 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10684 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10685 (define_attr "cr_logical_3op"): New.
10686 (cceq_ior_compare): Adjust.
10687 (cceq_ior_compare_complement): Adjust.
10688 (*cceq_rev_compare): Adjust.
10689 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10690 (is_cracked_insn): Adjust.
10691 (insn_must_be_first_in_group): Adjust.
10692 * config/rs6000/40x.md: Adjust.
10693 * config/rs6000/440.md: Adjust.
10694 * config/rs6000/476.md: Adjust.
10695 * config/rs6000/601.md: Adjust.
10696 * config/rs6000/603.md: Adjust.
10697 * config/rs6000/6xx.md: Adjust.
10698 * config/rs6000/7450.md: Adjust.
10699 * config/rs6000/7xx.md: Adjust.
10700 * config/rs6000/8540.md: Adjust.
10701 * config/rs6000/cell.md: Adjust.
10702 * config/rs6000/e300c2c3.md: Adjust.
10703 * config/rs6000/e500mc.md: Adjust.
10704 * config/rs6000/e500mc64.md: Adjust.
10705 * config/rs6000/e5500.md: Adjust.
10706 * config/rs6000/e6500.md: Adjust.
10707 * config/rs6000/mpc.md: Adjust.
10708 * config/rs6000/power4.md: Adjust.
10709 * config/rs6000/power5.md: Adjust.
10710 * config/rs6000/power6.md: Adjust.
10711 * config/rs6000/power7.md: Adjust.
10712 * config/rs6000/power8.md: Adjust.
10713 * config/rs6000/power9.md: Adjust.
10714 * config/rs6000/rs64.md: Adjust.
10715 * config/rs6000/titan.md: Adjust.
10717 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10719 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10720 ix86_indirect_branch_register logic.
10722 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10724 * config/i386/constraints.md (Bs): Update
10725 ix86_indirect_branch_register check. Don't check
10726 ix86_indirect_branch_register with GOT_memory_operand.
10728 * config/i386/predicates.md (GOT_memory_operand): Don't check
10729 ix86_indirect_branch_register here.
10730 (GOT32_symbol_operand): Likewise.
10732 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10734 * config/i386/predicates.md (constant_call_address_operand):
10735 Rewrite ix86_indirect_branch_register logic.
10736 (sibcall_insn_operand): Likewise.
10738 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10740 * config/i386/constraints.md (Bs): Replace
10741 ix86_indirect_branch_thunk_register with
10742 ix86_indirect_branch_register.
10744 * config/i386/i386.md (indirect_jump): Likewise.
10745 (tablejump): Likewise.
10746 (*sibcall_memory): Likewise.
10747 (*sibcall_value_memory): Likewise.
10748 Peepholes of indirect call and jump via memory: Likewise.
10749 * config/i386/i386.opt: Likewise.
10750 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10751 (GOT_memory_operand): Likewise.
10752 (call_insn_operand): Likewise.
10753 (sibcall_insn_operand): Likewise.
10754 (GOT32_symbol_operand): Likewise.
10756 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10758 PR middle-end/83837
10759 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10760 type rather than type addr's type points to.
10761 (expand_omp_atomic_mutex): Likewise.
10762 (expand_omp_atomic): Likewise.
10764 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10767 * config/i386/i386.c (output_indirect_thunk_function): Use
10768 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10769 for __x86_return_thunk.
10771 2018-01-15 Richard Biener <rguenther@suse.de>
10773 PR middle-end/83850
10774 * expmed.c (extract_bit_field_1): Fix typo.
10776 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10779 * config/arm/iterators.md (VF): New mode iterator.
10780 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10781 Remove integer-related logic from pattern.
10782 (neon_vabd<mode>_3): Likewise.
10784 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10786 PR middle-end/82694
10787 * common.opt (fstrict-overflow): No longer an alias.
10788 (fwrapv-pointer): New option.
10789 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10790 also for pointer types based on flag_wrapv_pointer.
10791 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10792 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10793 opts->x_flag_wrapv got set.
10794 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10795 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10796 POINTER_TYPE_OVERFLOW_UNDEFINED.
10797 * match.pd: Likewise in address comparison pattern.
10798 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10800 2018-01-15 Richard Biener <rguenther@suse.de>
10803 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10804 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10805 Reset type names to their identifier if their TYPE_DECL doesn't
10806 have linkage (and thus is used for ODR and devirt).
10807 (save_debug_info_for_decl): Remove.
10808 (save_debug_info_for_type): Likewise.
10809 (add_tree_to_fld_list): Adjust.
10810 * tree-pretty-print.c (dump_generic_node): Make dumping of
10811 type names more robust.
10813 2018-01-15 Richard Biener <rguenther@suse.de>
10815 * BASE-VER: Bump to 8.0.1.
10817 2018-01-14 Martin Sebor <msebor@redhat.com>
10820 * builtins.c (check_access): Avoid warning when the no-warning bit
10823 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10825 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10826 * ira-color (allocno_hard_regs_compare): Likewise.
10828 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10831 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10832 Use .pushsection/.popsection.
10834 2018-01-14 Martin Sebor <msebor@redhat.com>
10837 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10839 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10841 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10842 entry from extra_headers.
10843 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10844 extra_headers, make the list bitwise identical to the i?86-*-* one.
10846 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10848 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10849 -mcmodel=large with -mindirect-branch=thunk,
10850 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10851 -mfunction-return=thunk-extern.
10852 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10853 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10854 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10856 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10858 * config/i386/i386.c (print_reg): Print the name of the full
10859 integer register without '%'.
10860 (ix86_print_operand): Handle 'V'.
10861 * doc/extend.texi: Document 'V' modifier.
10863 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10865 * config/i386/constraints.md (Bs): Disallow memory operand for
10866 -mindirect-branch-register.
10868 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10869 (GOT_memory_operand): Likewise.
10870 (call_insn_operand): Likewise.
10871 (sibcall_insn_operand): Likewise.
10872 (GOT32_symbol_operand): Likewise.
10873 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10874 for -mindirect-branch-register.
10875 (tablejump): Likewise.
10876 (*sibcall_memory): Likewise.
10877 (*sibcall_value_memory): Likewise.
10878 Disallow peepholes of indirect call and jump via memory for
10879 -mindirect-branch-register.
10880 (*call_pop): Replace m with Bw.
10881 (*call_value_pop): Likewise.
10882 (*sibcall_pop_memory): Replace m with Bs.
10883 * config/i386/i386.opt (mindirect-branch-register): New option.
10884 * doc/invoke.texi: Document -mindirect-branch-register option.
10886 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10888 * config/i386/i386-protos.h (ix86_output_function_return): New.
10889 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10890 set function_return_type.
10891 (indirect_thunk_name): Add ret_p to indicate thunk for function
10893 (output_indirect_thunk_function): Pass false to
10894 indirect_thunk_name.
10895 (ix86_output_indirect_branch_via_reg): Likewise.
10896 (ix86_output_indirect_branch_via_push): Likewise.
10897 (output_indirect_thunk_function): Create alias for function
10898 return thunk if regno < 0.
10899 (ix86_output_function_return): New function.
10900 (ix86_handle_fndecl_attribute): Handle function_return.
10901 (ix86_attribute_table): Add function_return.
10902 * config/i386/i386.h (machine_function): Add
10903 function_return_type.
10904 * config/i386/i386.md (simple_return_internal): Use
10905 ix86_output_function_return.
10906 (simple_return_internal_long): Likewise.
10907 * config/i386/i386.opt (mfunction-return=): New option.
10908 (indirect_branch): Mention -mfunction-return=.
10909 * doc/extend.texi: Document function_return function attribute.
10910 * doc/invoke.texi: Document -mfunction-return= option.
10912 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10914 * config/i386/i386-opts.h (indirect_branch): New.
10915 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10916 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10917 with local indirect jump when converting indirect call and jump.
10918 (ix86_set_indirect_branch_type): New.
10919 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10920 (indirectlabelno): New.
10921 (indirect_thunk_needed): Likewise.
10922 (indirect_thunk_bnd_needed): Likewise.
10923 (indirect_thunks_used): Likewise.
10924 (indirect_thunks_bnd_used): Likewise.
10925 (INDIRECT_LABEL): Likewise.
10926 (indirect_thunk_name): Likewise.
10927 (output_indirect_thunk): Likewise.
10928 (output_indirect_thunk_function): Likewise.
10929 (ix86_output_indirect_branch_via_reg): Likewise.
10930 (ix86_output_indirect_branch_via_push): Likewise.
10931 (ix86_output_indirect_branch): Likewise.
10932 (ix86_output_indirect_jmp): Likewise.
10933 (ix86_code_end): Call output_indirect_thunk_function if needed.
10934 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10936 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10937 (ix86_attribute_table): Add indirect_branch.
10938 * config/i386/i386.h (machine_function): Add indirect_branch_type
10939 and has_local_indirect_jump.
10940 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10942 (tablejump): Likewise.
10943 (*indirect_jump): Use ix86_output_indirect_jmp.
10944 (*tablejump_1): Likewise.
10945 (simple_return_indirect_internal): Likewise.
10946 * config/i386/i386.opt (mindirect-branch=): New option.
10947 (indirect_branch): New.
10950 (thunk-inline): Likewise.
10951 (thunk-extern): Likewise.
10952 * doc/extend.texi: Document indirect_branch function attribute.
10953 * doc/invoke.texi: Document -mindirect-branch= option.
10955 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10958 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10960 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10962 * ipa-inline.c (want_inline_small_function_p): Return false if
10963 inlining has already failed with CIF_FINAL_ERROR.
10964 (update_caller_keys): Call want_inline_small_function_p before
10966 (update_callee_keys): Likewise.
10968 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10970 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10972 (rs6000_quadword_masked_address_p): Likewise.
10973 (quad_aligned_load_p): Likewise.
10974 (quad_aligned_store_p): Likewise.
10975 (const_load_sequence_p): Add comment to describe the outer-most loop.
10976 (mimic_memory_attributes_and_flags): New function.
10977 (rs6000_gen_stvx): Likewise.
10978 (replace_swapped_aligned_store): Likewise.
10979 (rs6000_gen_lvx): Likewise.
10980 (replace_swapped_aligned_load): Likewise.
10981 (replace_swapped_load_constant): Capitalize argument name in
10982 comment describing this function.
10983 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10984 and stores that access quad-word aligned addresses and replace
10985 with stvx or lvx instructions when appropriate.
10986 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10987 New function prototype.
10988 (rs6000_quadword_masked_address_p): Likewise.
10989 (rs6000_gen_lvx): Likewise.
10990 (rs6000_gen_stvx): Likewise.
10991 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10992 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10993 when memory address is aligned.
10994 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10995 this split to select lvx instruction when memory address is aligned.
10996 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10997 instruction when memory address is aligned.
10998 (*vsx_le_perm_load_v16qi): Likewise.
10999 (four unnamed splitters): Modify to select the stvx instruction
11000 when memory is aligned.
11002 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
11004 * predict.c (determine_unlikely_bbs): Handle correctly BBs
11005 which appears in the queue multiple times.
11007 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11008 Alan Hayward <alan.hayward@arm.com>
11009 David Sherwood <david.sherwood@arm.com>
11011 * tree-vectorizer.h (vec_lower_bound): New structure.
11012 (_loop_vec_info): Add check_nonzero and lower_bounds.
11013 (LOOP_VINFO_CHECK_NONZERO): New macro.
11014 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11015 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11016 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11017 fields. Make seg_len the distance travelled, not including the
11019 (dr_direction_indicator): Declare.
11020 (dr_zero_step_indicator): Likewise.
11021 (dr_known_forward_stride_p): Likewise.
11022 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11024 (runtime_alias_check_p): Allow runtime alias checks with
11026 (operator ==): Compare access_size and align.
11027 (prune_runtime_alias_test_list): Rework for new distinction between
11028 the access_size and seg_len.
11029 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11031 (get_segment_min_max): New function.
11032 (create_intersect_range_checks): Use it.
11033 (dr_step_indicator): New function.
11034 (dr_direction_indicator): Likewise.
11035 (dr_zero_step_indicator): Likewise.
11036 (dr_known_forward_stride_p): Likewise.
11037 * tree-loop-distribution.c (data_ref_segment_size): Return
11038 DR_STEP * (niters - 1).
11039 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11041 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11042 (vect_preserves_scalar_order_p): New function, split out from...
11043 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11044 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11045 (vect_vfa_access_size): New function.
11046 (vect_vfa_align): Likewise.
11047 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11048 (dump_lower_bound): New function.
11049 (vect_check_lower_bound): Likewise.
11050 (vect_small_gap_p): Likewise.
11051 (vectorizable_with_step_bound_p): Likewise.
11052 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11053 depencies if the vectorization factor is 1. Convert the checks
11054 for nonzero steps into checks on the bounds of DR_STEP. Try using
11055 a bunds check for variable steps if the minimum required step is
11056 relatively small. Update calls to the dr_with_seg_len
11057 constructor and to vect_compile_time_alias.
11058 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11060 (vect_loop_versioning): Call it.
11061 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11063 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11065 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11066 Alan Hayward <alan.hayward@arm.com>
11067 David Sherwood <david.sherwood@arm.com>
11069 * doc/sourcebuild.texi (vect_scatter_store): Document.
11070 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11072 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11074 * genopinit.c (main): Add supports_vec_scatter_store and
11075 supports_vec_scatter_store_cached to target_optabs.
11076 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11077 IFN_MASK_SCATTER_STORE.
11078 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11080 * internal-fn.h (internal_store_fn_p): Declare.
11081 (internal_fn_stored_value_index): Likewise.
11082 * internal-fn.c (scatter_store_direct): New macro.
11083 (expand_scatter_store_optab_fn): New function.
11084 (direct_scatter_store_optab_supported_p): New macro.
11085 (internal_store_fn_p): New function.
11086 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11087 IFN_MASK_SCATTER_STORE.
11088 (internal_fn_mask_index): Likewise.
11089 (internal_fn_stored_value_index): New function.
11090 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11091 for scatter stores.
11092 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11093 * optabs-query.c (supports_vec_scatter_store_p): New function.
11094 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11095 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11096 true for scatter stores.
11097 (vect_gather_scatter_fn_p): Handle scatter stores too.
11098 (vect_check_gather_scatter): Consider using scatter stores if
11099 supports_vec_scatter_store_p.
11100 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11101 scatter stores too.
11102 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11103 internal_fn_stored_value_index.
11104 (check_load_store_masking): Handle scatter stores too.
11105 (vect_get_store_rhs): Make public.
11106 (vectorizable_call): Use internal_store_fn_p.
11107 (vectorizable_store): Handle scatter store internal functions.
11108 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11109 when deciding whether the end of the group has been reached.
11110 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11111 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11112 (mask_scatter_store<mode>): New insns.
11114 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11115 Alan Hayward <alan.hayward@arm.com>
11116 David Sherwood <david.sherwood@arm.com>
11118 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11119 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11120 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11122 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11123 Use vect_truncate_gather_scatter_offset if we can't treat the
11124 operation as a normal gather load or scatter store.
11125 (get_group_load_store_type): Take the gather_scatter_info
11126 as argument. Try using a gather load or scatter store for
11127 single-element groups.
11128 (get_load_store_type): Update calls to get_group_load_store_type
11129 and vect_use_strided_gather_scatters_p.
11131 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11132 Alan Hayward <alan.hayward@arm.com>
11133 David Sherwood <david.sherwood@arm.com>
11135 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11136 optional tree argument.
11137 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11139 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11140 but continue to use the current value as a fallback.
11141 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11142 to compare the updates.
11143 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11144 (get_load_store_type): Use it when handling a strided access.
11145 (vect_get_strided_load_store_ops): New function.
11146 (vect_get_data_ptr_increment): Likewise.
11147 (vectorizable_load): Handle strided gather loads. Always pass
11148 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11151 Alan Hayward <alan.hayward@arm.com>
11152 David Sherwood <david.sherwood@arm.com>
11154 * doc/md.texi (gather_load@var{m}): Document.
11155 (mask_gather_load@var{m}): Likewise.
11156 * genopinit.c (main): Add supports_vec_gather_load and
11157 supports_vec_gather_load_cached to target_optabs.
11158 * optabs-tree.c (init_tree_optimization_optabs): Use
11159 ggc_cleared_alloc to allocate target_optabs.
11160 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11161 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11163 * internal-fn.h (internal_load_fn_p): Declare.
11164 (internal_gather_scatter_fn_p): Likewise.
11165 (internal_fn_mask_index): Likewise.
11166 (internal_gather_scatter_fn_supported_p): Likewise.
11167 * internal-fn.c (gather_load_direct): New macro.
11168 (expand_gather_load_optab_fn): New function.
11169 (direct_gather_load_optab_supported_p): New macro.
11170 (direct_internal_fn_optab): New function.
11171 (internal_load_fn_p): Likewise.
11172 (internal_gather_scatter_fn_p): Likewise.
11173 (internal_fn_mask_index): Likewise.
11174 (internal_gather_scatter_fn_supported_p): Likewise.
11175 * optabs-query.c (supports_at_least_one_mode_p): New function.
11176 (supports_vec_gather_load_p): Likewise.
11177 * optabs-query.h (supports_vec_gather_load_p): Declare.
11178 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11179 and memory_type field.
11180 (NUM_PATTERNS): Bump to 15.
11181 * tree-vect-data-refs.c: Include internal-fn.h.
11182 (vect_gather_scatter_fn_p): New function.
11183 (vect_describe_gather_scatter_call): Likewise.
11184 (vect_check_gather_scatter): Try using internal functions for
11185 gather loads. Recognize existing calls to a gather load function.
11186 (vect_analyze_data_refs): Consider using gather loads if
11187 supports_vec_gather_load_p.
11188 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11189 (vect_get_gather_scatter_offset_type): Likewise.
11190 (vect_convert_mask_for_vectype): Likewise.
11191 (vect_add_conversion_to_patterm): Likewise.
11192 (vect_try_gather_scatter_pattern): Likewise.
11193 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11194 (vect_vect_recog_func_ptrs): Add it.
11195 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11196 internal_fn_mask_index and internal_gather_scatter_fn_p.
11197 (check_load_store_masking): Take the gather_scatter_info as an
11198 argument and handle gather loads.
11199 (vect_get_gather_scatter_ops): New function.
11200 (vectorizable_call): Check internal_load_fn_p.
11201 (vectorizable_load): Likewise. Handle gather load internal
11203 (vectorizable_store): Update call to check_load_store_masking.
11204 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11205 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11206 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11207 (aarch64_gather_scale_operand_d): New predicates.
11208 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11209 (mask_gather_load<mode>): New insns.
11211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11212 Alan Hayward <alan.hayward@arm.com>
11213 David Sherwood <david.sherwood@arm.com>
11215 * optabs.def (fold_left_plus_optab): New optab.
11216 * doc/md.texi (fold_left_plus_@var{m}): Document.
11217 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11218 * internal-fn.c (fold_left_direct): Define.
11219 (expand_fold_left_optab_fn): Likewise.
11220 (direct_fold_left_optab_supported_p): Likewise.
11221 * fold-const-call.c (fold_const_fold_left): New function.
11222 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11223 * tree-parloops.c (valid_reduction_p): New function.
11224 (gather_scalar_reductions): Use it.
11225 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11226 (vect_finish_replace_stmt): Declare.
11227 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11228 (needs_fold_left_reduction_p): New function, split out from...
11229 (vect_is_simple_reduction): ...here. Accept reductions that
11230 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11231 (vect_force_simple_reduction): Also store the reduction type in
11232 the assignment's STMT_VINFO_REDUC_TYPE.
11233 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11234 (merge_with_identity): New function.
11235 (vect_expand_fold_left): Likewise.
11236 (vectorize_fold_left_reduction): Likewise.
11237 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11238 scalar phi in place for it. Check for target support and reject
11239 cases that would reassociate the operation. Defer the transform
11240 phase to vectorize_fold_left_reduction.
11241 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11242 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11243 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11247 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11248 call to ifc_temp_var.
11250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11251 Alan Hayward <alan.hayward@arm.com>
11252 David Sherwood <david.sherwood@arm.com>
11254 * target.def (legitimize_address_displacement): Take the original
11255 offset as a poly_int.
11256 * targhooks.h (default_legitimize_address_displacement): Update
11258 * targhooks.c (default_legitimize_address_displacement): Likewise.
11259 * doc/tm.texi: Regenerate.
11260 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11261 as an argument, moving assert of ad->disp == ad->disp_term to...
11262 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11263 Try calling targetm.legitimize_address_displacement before expanding
11264 the address rather than afterwards, and adjust for the new interface.
11265 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11266 Match the new hook interface. Handle SVE addresses.
11267 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11268 new hook interface.
11270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11272 * Makefile.in (OBJS): Add early-remat.o.
11273 * target.def (select_early_remat_modes): New hook.
11274 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11275 * doc/tm.texi: Regenerate.
11276 * targhooks.h (default_select_early_remat_modes): Declare.
11277 * targhooks.c (default_select_early_remat_modes): New function.
11278 * timevar.def (TV_EARLY_REMAT): New timevar.
11279 * passes.def (pass_early_remat): New pass.
11280 * tree-pass.h (make_pass_early_remat): Declare.
11281 * early-remat.c: New file.
11282 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11284 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11287 Alan Hayward <alan.hayward@arm.com>
11288 David Sherwood <david.sherwood@arm.com>
11290 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11291 vfm1 with a bound_epilog parameter.
11292 (vect_do_peeling): Update calls accordingly, and move the prologue
11293 call earlier in the function. Treat the base bound_epilog as 0 for
11294 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11295 this base when peeling for gaps.
11296 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11297 with fully-masked loops.
11298 (vect_estimate_min_profitable_iters): Handle the single peeled
11299 iteration in that case.
11301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11302 Alan Hayward <alan.hayward@arm.com>
11303 David Sherwood <david.sherwood@arm.com>
11305 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11306 single-element interleaving even if the size is not a power of 2.
11307 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11308 accesses for single-element interleaving if the group size is
11311 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11312 Alan Hayward <alan.hayward@arm.com>
11313 David Sherwood <david.sherwood@arm.com>
11315 * doc/md.texi (fold_extract_last_@var{m}): Document.
11316 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11317 * optabs.def (fold_extract_last_optab): New optab.
11318 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11319 * internal-fn.c (fold_extract_direct): New macro.
11320 (expand_fold_extract_optab_fn): Likewise.
11321 (direct_fold_extract_optab_supported_p): Likewise.
11322 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11323 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11324 EXTRACT_LAST_REDUCTION.
11325 (get_initial_def_for_reduction): Do not create an initial vector
11326 for EXTRACT_LAST_REDUCTION reductions.
11327 (vectorizable_reduction): Leave the scalar phi in place for
11328 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11329 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11330 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11331 transform phase to vectorizable_condition.
11332 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11334 (vect_finish_stmt_generation): ...here.
11335 (vect_finish_replace_stmt): New function.
11336 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11337 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11339 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11341 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11342 Alan Hayward <alan.hayward@arm.com>
11343 David Sherwood <david.sherwood@arm.com>
11345 * doc/md.texi (extract_last_@var{m}): Document.
11346 * optabs.def (extract_last_optab): New optab.
11347 * internal-fn.def (EXTRACT_LAST): New internal function.
11348 * internal-fn.c (cond_unary_direct): New macro.
11349 (expand_cond_unary_optab_fn): Likewise.
11350 (direct_cond_unary_optab_supported_p): Likewise.
11351 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11352 loops using EXTRACT_LAST.
11353 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11354 (extract_last_<mode>): ...this optab.
11355 (vec_extract<mode><Vel>): Update accordingly.
11357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11358 Alan Hayward <alan.hayward@arm.com>
11359 David Sherwood <david.sherwood@arm.com>
11361 * target.def (empty_mask_is_expensive): New hook.
11362 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11363 * doc/tm.texi: Regenerate.
11364 * targhooks.h (default_empty_mask_is_expensive): Declare.
11365 * targhooks.c (default_empty_mask_is_expensive): New function.
11366 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11367 if the target says that empty masks are expensive.
11368 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11370 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11373 Alan Hayward <alan.hayward@arm.com>
11374 David Sherwood <david.sherwood@arm.com>
11376 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11377 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11378 (vect_use_loop_mask_for_alignment_p): New function.
11379 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11380 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11381 niters_skip argument. Make sure that the first niters_skip elements
11382 of the first iteration are inactive.
11383 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11384 Update call to vect_set_loop_masks_directly.
11385 (get_misalign_in_elems): New function, split out from...
11386 (vect_gen_prolog_loop_niters): ...here.
11387 (vect_update_init_of_dr): Take a code argument that specifies whether
11388 the adjustment should be added or subtracted.
11389 (vect_update_init_of_drs): Likewise.
11390 (vect_prepare_for_masked_peels): New function.
11391 (vect_do_peeling): Skip prologue peeling if we're using a mask
11392 instead. Update call to vect_update_inits_of_drs.
11393 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11395 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11396 alignment. Do not include the number of peeled iterations in
11397 the minimum threshold in that case.
11398 (vectorizable_induction): Adjust the start value down by
11399 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11400 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11401 Take the number of skipped iterations into account when calculating
11403 * tree-vect-stmts.c (vect_gen_while_not): New function.
11405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11406 Alan Hayward <alan.hayward@arm.com>
11407 David Sherwood <david.sherwood@arm.com>
11409 * doc/sourcebuild.texi (vect_fully_masked): Document.
11410 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11411 default value to 0.
11412 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11414 (vect_analyze_loop_2): ...here. Don't check the vectorization
11415 factor against the number of loop iterations if the loop is
11418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11419 Alan Hayward <alan.hayward@arm.com>
11420 David Sherwood <david.sherwood@arm.com>
11422 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11423 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11424 (dump_groups): Update accordingly.
11425 (iv_use::mem_type): New member variable.
11426 (address_p): New function.
11427 (record_use): Add a mem_type argument and initialize the new
11429 (record_group_use): Add a mem_type argument. Use address_p.
11430 Remove obsolete null checks of base_object. Update call to record_use.
11431 (find_interesting_uses_op): Update call to record_group_use.
11432 (find_interesting_uses_cond): Likewise.
11433 (find_interesting_uses_address): Likewise.
11434 (get_mem_type_for_internal_fn): New function.
11435 (find_address_like_use): Likewise.
11436 (find_interesting_uses_stmt): Try find_address_like_use before
11437 calling find_interesting_uses_op.
11438 (addr_offset_valid_p): Use the iv mem_type field as the type
11439 of the addressed memory.
11440 (add_autoinc_candidates): Likewise.
11441 (get_address_cost): Likewise.
11442 (split_small_address_groups_p): Use address_p.
11443 (split_address_groups): Likewise.
11444 (add_iv_candidate_for_use): Likewise.
11445 (autoinc_possible_for_pair): Likewise.
11446 (rewrite_groups): Likewise.
11447 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11448 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11449 (get_alias_ptr_type_for_ptr_address): New function.
11450 (rewrite_use_address): Rewrite address uses in calls that were
11451 identified by find_address_like_use.
11453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11454 Alan Hayward <alan.hayward@arm.com>
11455 David Sherwood <david.sherwood@arm.com>
11457 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11459 * gimple-expr.h (is_gimple_addressable: Likewise.
11460 * gimple-expr.c (is_gimple_address): Likewise.
11461 * internal-fn.c (expand_call_mem_ref): New function.
11462 (expand_mask_load_optab_fn): Use it.
11463 (expand_mask_store_optab_fn): Likewise.
11465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11466 Alan Hayward <alan.hayward@arm.com>
11467 David Sherwood <david.sherwood@arm.com>
11469 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11470 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11471 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11472 (cond_umax@var{mode}): Document.
11473 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11474 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11475 (cond_umin_optab, cond_umax_optab): New optabs.
11476 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11477 (COND_IOR, COND_XOR): New internal functions.
11478 * internal-fn.h (get_conditional_internal_fn): Declare.
11479 * internal-fn.c (cond_binary_direct): New macro.
11480 (expand_cond_binary_optab_fn): Likewise.
11481 (direct_cond_binary_optab_supported_p): Likewise.
11482 (get_conditional_internal_fn): New function.
11483 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11484 Cope with reduction statements that are vectorized as calls rather
11486 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11487 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11488 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11489 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11490 (UNSPEC_COND_EOR): New unspecs.
11491 (optab): Add mappings for them.
11492 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11493 (sve_int_op, sve_fp_op): New int attributes.
11495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11496 Alan Hayward <alan.hayward@arm.com>
11497 David Sherwood <david.sherwood@arm.com>
11499 * optabs.def (while_ult_optab): New optab.
11500 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11501 * internal-fn.def (WHILE_ULT): New internal function.
11502 * internal-fn.h (direct_internal_fn_supported_p): New override
11503 that takes two types as argument.
11504 * internal-fn.c (while_direct): New macro.
11505 (expand_while_optab_fn): New function.
11506 (convert_optab_supported_p): Likewise.
11507 (direct_while_optab_supported_p): New macro.
11508 * wide-int.h (wi::udiv_ceil): New function.
11509 * tree-vectorizer.h (rgroup_masks): New structure.
11510 (vec_loop_masks): New typedef.
11511 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11512 and fully_masked_p.
11513 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11514 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11515 (vect_max_vf): New function.
11516 (slpeel_make_loop_iterate_ntimes): Delete.
11517 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11518 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11519 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11520 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11521 internal-fn.h, stor-layout.h and optabs-query.h.
11522 (vect_set_loop_mask): New function.
11523 (add_preheader_seq): Likewise.
11524 (add_header_seq): Likewise.
11525 (interleave_supported_p): Likewise.
11526 (vect_maybe_permute_loop_masks): Likewise.
11527 (vect_set_loop_masks_directly): Likewise.
11528 (vect_set_loop_condition_masked): Likewise.
11529 (vect_set_loop_condition_unmasked): New function, split out from
11530 slpeel_make_loop_iterate_ntimes.
11531 (slpeel_make_loop_iterate_ntimes): Rename to..
11532 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11533 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11534 (vect_do_peeling): Update call accordingly.
11535 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11537 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11538 mask_compare_type, can_fully_mask_p and fully_masked_p.
11539 (release_vec_loop_masks): New function.
11540 (_loop_vec_info): Use it to free the loop masks.
11541 (can_produce_all_loop_masks_p): New function.
11542 (vect_get_max_nscalars_per_iter): Likewise.
11543 (vect_verify_full_masking): Likewise.
11544 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11545 retries, and free the mask rgroups before retrying. Check loop-wide
11546 reasons for disallowing fully-masked loops. Make the final decision
11547 about whether use a fully-masked loop or not.
11548 (vect_estimate_min_profitable_iters): Do not assume that peeling
11549 for the number of iterations will be needed for fully-masked loops.
11550 (vectorizable_reduction): Disable fully-masked loops.
11551 (vectorizable_live_operation): Likewise.
11552 (vect_halve_mask_nunits): New function.
11553 (vect_double_mask_nunits): Likewise.
11554 (vect_record_loop_mask): Likewise.
11555 (vect_get_loop_mask): Likewise.
11556 (vect_transform_loop): Handle the case in which the final loop
11557 iteration might handle a partial vector. Call vect_set_loop_condition
11558 instead of slpeel_make_loop_iterate_ntimes.
11559 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11560 (check_load_store_masking): New function.
11561 (prepare_load_store_mask): Likewise.
11562 (vectorizable_store): Handle fully-masked loops.
11563 (vectorizable_load): Likewise.
11564 (supportable_widening_operation): Use vect_halve_mask_nunits for
11566 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11567 (vect_gen_while): New function.
11568 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11569 (aarch64_uqdec<mode>): New insn.
11571 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11572 Alan Hayward <alan.hayward@arm.com>
11573 David Sherwood <david.sherwood@arm.com>
11575 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11576 (reduc_xor_scal_optab): New optabs.
11577 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11578 (reduc_xor_scal_@var{m}): Document.
11579 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11580 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11581 internal functions.
11582 * fold-const-call.c (fold_const_call): Handle them.
11583 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11584 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11585 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11586 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11587 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11588 (UNSPEC_XORV): New unspecs.
11589 (optab): Add entries for them.
11590 (BITWISEV): New int iterator.
11591 (bit_reduc_op): New int attributes.
11593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11594 Alan Hayward <alan.hayward@arm.com>
11595 David Sherwood <david.sherwood@arm.com>
11597 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11598 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11599 * optabs.def (vec_shl_insert_optab): New optab.
11600 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11601 (duplicate_and_interleave): Likewise.
11602 * tree-vect-loop.c: Include internal-fn.h.
11603 (neutral_op_for_slp_reduction): New function, split out from
11604 get_initial_defs_for_reduction.
11605 (get_initial_def_for_reduction): Handle option 2 for variable-length
11606 vectors by loading the neutral value into a vector and then shifting
11607 the initial value into element 0.
11608 (get_initial_defs_for_reduction): Replace the code argument with
11609 the neutral value calculated by neutral_op_for_slp_reduction.
11610 Use gimple_build_vector for constant-length vectors.
11611 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11612 but the first group_size elements have a neutral value.
11613 Use duplicate_and_interleave otherwise.
11614 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11615 Update call to get_initial_defs_for_reduction. Handle SLP
11616 reductions for variable-length vectors by creating one vector
11617 result for each scalar result, with the elements associated
11618 with other scalar results stubbed out with the neutral value.
11619 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11620 Require IFN_VEC_SHL_INSERT for double reductions on
11621 variable-length vectors, or SLP reductions that have
11622 a neutral value. Require can_duplicate_and_interleave_p
11623 support for variable-length unchained SLP reductions if there
11624 is no neutral value, such as for MIN/MAX reductions. Also require
11625 the number of vector elements to be a multiple of the number of
11626 SLP statements when doing variable-length unchained SLP reductions.
11627 Update call to vect_create_epilog_for_reduction.
11628 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11629 and remove initial values.
11630 (duplicate_and_interleave): Make public.
11631 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11632 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11635 Alan Hayward <alan.hayward@arm.com>
11636 David Sherwood <david.sherwood@arm.com>
11638 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11639 (can_duplicate_and_interleave_p): New function.
11640 (vect_get_and_check_slp_defs): Take the vector of statements
11641 rather than just the current one. Remove excess parentheses.
11642 Restriction rejectinon of vect_constant_def and vect_external_def
11643 for variable-length vectors to boolean types, or types for which
11644 can_duplicate_and_interleave_p is false.
11645 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11646 (duplicate_and_interleave): New function.
11647 (vect_get_constant_vectors): Use gimple_build_vector for
11648 constant-length vectors and suitable variable-length constant
11649 vectors. Use duplicate_and_interleave for other variable-length
11650 vectors. Don't defer the update when inserting new statements.
11652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11653 Alan Hayward <alan.hayward@arm.com>
11654 David Sherwood <david.sherwood@arm.com>
11656 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11657 min_profitable_iters doesn't go negative.
11659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11660 Alan Hayward <alan.hayward@arm.com>
11661 David Sherwood <david.sherwood@arm.com>
11663 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11664 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11665 * optabs.def (vec_mask_load_lanes_optab): New optab.
11666 (vec_mask_store_lanes_optab): Likewise.
11667 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11668 (MASK_STORE_LANES): Likewise.
11669 * internal-fn.c (mask_load_lanes_direct): New macro.
11670 (mask_store_lanes_direct): Likewise.
11671 (expand_mask_load_optab_fn): Handle masked operations.
11672 (expand_mask_load_lanes_optab_fn): New macro.
11673 (expand_mask_store_optab_fn): Handle masked operations.
11674 (expand_mask_store_lanes_optab_fn): New macro.
11675 (direct_mask_load_lanes_optab_supported_p): Likewise.
11676 (direct_mask_store_lanes_optab_supported_p): Likewise.
11677 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11679 (vect_load_lanes_supported): Likewise.
11680 * tree-vect-data-refs.c (strip_conversion): New function.
11681 (can_group_stmts_p): Likewise.
11682 (vect_analyze_data_ref_accesses): Use it instead of checking
11683 for a pair of assignments.
11684 (vect_store_lanes_supported): Take a masked_p parameter.
11685 (vect_load_lanes_supported): Likewise.
11686 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11687 vect_store_lanes_supported and vect_load_lanes_supported.
11688 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11689 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11690 parameter. Don't allow gaps for masked accesses.
11691 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11692 and vect_load_lanes_supported.
11693 (get_load_store_type): Take a masked_p parameter and update
11694 call to get_group_load_store_type.
11695 (vectorizable_store): Update call to get_load_store_type.
11696 Handle IFN_MASK_STORE_LANES.
11697 (vectorizable_load): Update call to get_load_store_type.
11698 Handle IFN_MASK_LOAD_LANES.
11700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11701 Alan Hayward <alan.hayward@arm.com>
11702 David Sherwood <david.sherwood@arm.com>
11704 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11706 * config/aarch64/aarch64-protos.h
11707 (aarch64_sve_struct_memory_operand_p): Declare.
11708 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11709 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11710 (VPRED, vpred): Handle SVE structure modes.
11711 * config/aarch64/constraints.md (Utx): New constraint.
11712 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11713 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11714 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11715 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11716 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11717 structure modes. Split into pieces after RA.
11718 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11719 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11721 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11722 SVE structure modes.
11723 (aarch64_classify_address): Likewise.
11724 (sizetochar): Move earlier in file.
11725 (aarch64_print_operand): Handle SVE register lists.
11726 (aarch64_array_mode): New function.
11727 (aarch64_sve_struct_memory_operand_p): Likewise.
11728 (TARGET_ARRAY_MODE): Redefine.
11730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11731 Alan Hayward <alan.hayward@arm.com>
11732 David Sherwood <david.sherwood@arm.com>
11734 * target.def (array_mode): New target hook.
11735 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11736 * doc/tm.texi: Regenerate.
11737 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11738 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11739 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11740 targetm.array_mode.
11741 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11745 Alan Hayward <alan.hayward@arm.com>
11746 David Sherwood <david.sherwood@arm.com>
11748 * fold-const.c (fold_binary_loc): Check the argument types
11749 rather than the result type when testing for a vector operation.
11751 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11753 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11754 * doc/tm.texi: Regenerate.
11756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11757 Alan Hayward <alan.hayward@arm.com>
11758 David Sherwood <david.sherwood@arm.com>
11760 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11761 (sve): Document new AArch64 extension.
11762 * doc/md.texi (w): Extend the description of the AArch64
11763 constraint to include SVE vectors.
11764 (Upl, Upa): Document new AArch64 predicate constraints.
11765 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11767 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11768 (msve-vector-bits=): New option.
11769 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11770 SVE when these are disabled.
11771 (sve): New extension.
11772 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11773 modes. Adjust their number of units based on aarch64_sve_vg.
11774 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11775 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11776 aarch64_addr_query_type.
11777 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11778 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11779 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11780 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11781 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11782 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11783 (aarch64_simd_imm_zero_p): Delete.
11784 (aarch64_check_zero_based_sve_index_immediate): Declare.
11785 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11786 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11787 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11788 (aarch64_sve_float_mul_immediate_p): Likewise.
11789 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11790 rather than an rtx.
11791 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11792 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11793 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11794 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11795 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11796 (aarch64_regmode_natural_size): Likewise.
11797 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11798 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11800 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11801 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11802 for VG and the SVE predicate registers.
11803 (V_ALIASES): Add a "z"-prefixed alias.
11804 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11805 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11806 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11807 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11808 (REG_CLASS_NAMES): Add entries for them.
11809 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11810 and the predicate registers.
11811 (aarch64_sve_vg): Declare.
11812 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11813 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11814 (REGMODE_NATURAL_SIZE): Define.
11815 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11817 * config/aarch64/aarch64.c: Include cfgrtl.h.
11818 (simd_immediate_info): Add a constructor for series vectors,
11819 and an associated step field.
11820 (aarch64_sve_vg): New variable.
11821 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11822 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11823 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11824 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11825 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11826 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11827 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11828 (aarch64_get_mask_mode): New functions.
11829 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11830 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11831 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11832 predicate modes and predicate registers. Explicitly restrict
11833 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11834 to store a vector mode if it is recognized by
11835 aarch64_classify_vector_mode.
11836 (aarch64_regmode_natural_size): New function.
11837 (aarch64_hard_regno_caller_save_mode): Return the original mode
11839 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11840 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11841 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11842 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11844 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11845 does not overlap dest if the function is frame-related. Handle
11847 (aarch64_split_add_offset): New function.
11848 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11849 them aarch64_add_offset.
11850 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11851 and update call to aarch64_sub_sp.
11852 (aarch64_add_cfa_expression): New function.
11853 (aarch64_expand_prologue): Pass extra temporary registers to the
11854 functions above. Handle the case in which we need to emit new
11855 DW_CFA_expressions for registers that were originally saved
11856 relative to the stack pointer, but now have to be expressed
11857 relative to the frame pointer.
11858 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11860 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11861 IP0 and IP1 values for SVE frames.
11862 (aarch64_expand_vec_series): New function.
11863 (aarch64_expand_sve_widened_duplicate): Likewise.
11864 (aarch64_expand_sve_const_vector): Likewise.
11865 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11866 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11867 into the register, rather than emitting a SET directly.
11868 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11869 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11870 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11871 (offset_9bit_signed_scaled_p): New functions.
11872 (aarch64_replicate_bitmask_imm): New function.
11873 (aarch64_bitmask_imm): Use it.
11874 (aarch64_cannot_force_const_mem): Reject expressions involving
11875 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11876 (aarch64_classify_index): Handle SVE indices, by requiring
11877 a plain register index with a scale that matches the element size.
11878 (aarch64_classify_address): Handle SVE addresses. Assert that
11879 the mode of the address is VOIDmode or an integer mode.
11880 Update call to aarch64_classify_symbol.
11881 (aarch64_classify_symbolic_expression): Update call to
11882 aarch64_classify_symbol.
11883 (aarch64_const_vec_all_in_range_p): New function.
11884 (aarch64_print_vector_float_operand): Likewise.
11885 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11886 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11887 and the FP immediates 1.0 and 0.5.
11888 (aarch64_print_address_internal): Handle SVE addresses.
11889 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11890 (aarch64_regno_regclass): Handle predicate registers.
11891 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11893 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11894 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11895 (aarch64_convert_sve_vector_bits): New function.
11896 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11897 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11898 rather than an rtx.
11899 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11900 Handle SVE vector and predicate modes. Accept VL-based constants
11901 that need only one temporary register, and VL offsets that require
11902 no temporary registers.
11903 (aarch64_conditional_register_usage): Mark the predicate registers
11904 as fixed if SVE isn't available.
11905 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11906 Return true for SVE vector and predicate modes.
11907 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11908 rather than an unsigned int. Handle SVE modes.
11909 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11911 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11913 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11914 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11915 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11916 (aarch64_sve_float_mul_immediate_p): New functions.
11917 (aarch64_sve_valid_immediate): New function.
11918 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11919 Explicitly reject structure modes. Check for INDEX constants.
11920 Handle PTRUE and PFALSE constants.
11921 (aarch64_check_zero_based_sve_index_immediate): New function.
11922 (aarch64_simd_imm_zero_p): Delete.
11923 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11924 vector modes. Accept constants in the range of CNT[BHWD].
11925 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11926 ask for an Advanced SIMD mode.
11927 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11928 (aarch64_simd_vector_alignment): Handle SVE predicates.
11929 (aarch64_vectorize_preferred_vector_alignment): New function.
11930 (aarch64_simd_vector_alignment_reachable): Use it instead of
11932 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11933 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11935 (MAX_VECT_LEN): Delete.
11936 (expand_vec_perm_d): Add a vec_flags field.
11937 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11938 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11939 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11941 (aarch64_evpc_rev): Rename to...
11942 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11943 (aarch64_evpc_rev_global): New function.
11944 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11945 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11947 (aarch64_evpc_sve_tbl): New function.
11948 (aarch64_expand_vec_perm_const_1): Update after rename of
11949 aarch64_evpc_rev. Handle SVE permutes too, trying
11950 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11951 than aarch64_evpc_tbl.
11952 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11953 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11954 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11955 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11956 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11957 (aarch64_expand_sve_vcond): New functions.
11958 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11959 of aarch64_vector_mode_p.
11960 (aarch64_dwarf_poly_indeterminate_value): New function.
11961 (aarch64_compute_pressure_classes): Likewise.
11962 (aarch64_can_change_mode_class): Likewise.
11963 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11964 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11965 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11966 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11967 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11968 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11969 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11970 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11972 (Dn, Dl, Dr): Accept const as well as const_vector.
11973 (Dz): Likewise. Compare against CONST0_RTX.
11974 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11975 of "vector" where appropriate.
11976 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11977 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11978 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11979 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11980 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11981 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11982 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11983 (v_int_equiv): Extend to SVE modes.
11984 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11986 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11987 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11988 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11989 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11990 (SVE_COND_FP_CMP): New int iterators.
11991 (perm_hilo): Handle the new unpack unspecs.
11992 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11994 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11995 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11996 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11997 (aarch64_equality_operator, aarch64_constant_vector_operand)
11998 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11999 (aarch64_sve_nonimmediate_operand): Likewise.
12000 (aarch64_sve_general_operand): Likewise.
12001 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
12002 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
12003 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
12004 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
12005 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12006 (aarch64_sve_float_arith_immediate): Likewise.
12007 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12008 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12009 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12010 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12011 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12012 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12013 (aarch64_sve_float_arith_operand): Likewise.
12014 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12015 (aarch64_sve_float_mul_operand): Likewise.
12016 (aarch64_sve_vec_perm_operand): Likewise.
12017 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12018 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12019 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12020 as well as const_vector.
12021 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12022 in file. Use CONST0_RTX and CONSTM1_RTX.
12023 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12024 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12025 Use aarch64_simd_imm_zero.
12026 * config/aarch64/aarch64-sve.md: New file.
12027 * config/aarch64/aarch64.md: Include it.
12028 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12029 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12030 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12031 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12032 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12033 (sve): New attribute.
12034 (enabled): Disable instructions with the sve attribute unless
12036 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12037 aarch64_expand_mov_immediate.
12038 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12039 CNT[BHSD] immediates.
12040 (movti): Split CONST_POLY_INT moves into two halves.
12041 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12042 Split additions that need a temporary here if the destination
12043 is the stack pointer.
12044 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12045 (*add<mode>3_poly_1): New instruction.
12046 (set_clobber_cc): New expander.
12048 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12050 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12051 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12052 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12053 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12054 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12055 Change innermode from fixed_mode_size to machine_mode.
12056 (simplify_subreg): Update call accordingly. Handle a constant-sized
12057 subreg of a variable-length CONST_VECTOR.
12059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12060 Alan Hayward <alan.hayward@arm.com>
12061 David Sherwood <david.sherwood@arm.com>
12063 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12064 (add_offset_to_base): New function, split out from...
12065 (create_mem_ref): ...here. When handling a scale other than 1,
12066 check first whether the address is valid without the offset.
12067 Add it into the base if so, leaving the index and scale as-is.
12069 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12072 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12073 fold_for_warn before checking if arg2 is INTEGER_CST.
12075 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12077 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12078 (store_multiple_operation): Delete.
12079 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12080 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12081 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12082 guarded by TARGET_STRING.
12083 (rs6000_output_load_multiple): Delete.
12084 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12085 OPTION_MASK_STRING / TARGET_STRING handling.
12086 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12087 (const rs6000_opt_masks) <"string">: Change mask to 0.
12088 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12089 (MASK_STRING): Delete.
12090 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12092 (load_multiple): Delete.
12099 (store_multiple): Delete.
12106 (movmemsi_8reg): Delete.
12107 (corresponding unnamed define_insn): Delete.
12108 (movmemsi_6reg): Delete.
12109 (corresponding unnamed define_insn): Delete.
12110 (movmemsi_4reg): Delete.
12111 (corresponding unnamed define_insn): Delete.
12112 (movmemsi_2reg): Delete.
12113 (corresponding unnamed define_insn): Delete.
12114 (movmemsi_1reg): Delete.
12115 (corresponding unnamed define_insn): Delete.
12116 * config/rs6000/rs6000.opt (mno-string): New.
12117 (mstring): Replace by deprecation warning stub.
12118 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12120 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12122 * regrename.c (regrename_do_replace): If replacing the same
12123 reg multiple times, try to reuse last created gen_raw_REG.
12126 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12127 main to workaround a bug in GDB.
12129 2018-01-12 Tom de Vries <tom@codesourcery.com>
12132 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12134 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12136 PR rtl-optimization/80481
12137 * ira-color.c (get_cap_member): New function.
12138 (allocnos_conflict_by_live_ranges_p): Use it.
12139 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12140 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12142 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12145 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12146 (*saddl_se_1): Ditto.
12147 (*ssubsi_1): Ditto.
12148 (*ssubl_se_1): Ditto.
12150 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12152 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12153 rather than wi::to_widest for DR_INITs.
12154 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12155 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12156 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12158 (vect_analyze_group_access_1): Note that here.
12160 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12162 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12163 polynomial type sizes.
12165 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12167 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12168 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12169 (gimple_add_tmp_var): Likewise.
12171 2018-01-12 Martin Liska <mliska@suse.cz>
12173 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12174 (gimple_alloc_sizes): Likewise.
12175 (dump_gimple_statistics): Use PRIu64 in printf format.
12176 * gimple.h: Change uint64_t to int.
12178 2018-01-12 Martin Liska <mliska@suse.cz>
12180 * tree-core.h: Use uint64_t instead of int.
12181 * tree.c (tree_node_counts): Likewise.
12182 (tree_node_sizes): Likewise.
12183 (dump_tree_statistics): Use PRIu64 in printf format.
12185 2018-01-12 Martin Liska <mliska@suse.cz>
12187 * Makefile.in: As qsort_chk is implemented in vec.c, add
12188 vec.o to linkage of gencfn-macros.
12189 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12190 passing the info to record_node_allocation_statistics.
12191 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12193 * ggc-common.c (struct ggc_usage): Add operator== and use
12194 it in operator< and compare function.
12195 * mem-stats.h (struct mem_usage): Likewise.
12196 * vec.c (struct vec_usage): Remove operator< and compare
12197 function. Can be simply inherited.
12199 2018-01-12 Martin Jambor <mjambor@suse.cz>
12202 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12203 * tree-ssa-math-opts.c: Include domwalk.h.
12204 (convert_mult_to_fma_1): New function.
12205 (fma_transformation_info): New type.
12206 (fma_deferring_state): Likewise.
12207 (cancel_fma_deferring): New function.
12208 (result_of_phi): Likewise.
12209 (last_fma_candidate_feeds_initial_phi): Likewise.
12210 (convert_mult_to_fma): Added deferring logic, split actual
12211 transformation to convert_mult_to_fma_1.
12212 (math_opts_dom_walker): New type.
12213 (math_opts_dom_walker::after_dom_children): New method, body moved
12214 here from pass_optimize_widening_mul::execute, added deferring logic
12216 (pass_optimize_widening_mul::execute): Moved most of code to
12217 math_opts_dom_walker::after_dom_children.
12218 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12219 * config/i386/i386.c (ix86_option_override_internal): Added
12220 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12222 2018-01-12 Richard Biener <rguenther@suse.de>
12225 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12226 inline instance vars.
12228 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12231 * config/rx/rx.c (rx_is_restricted_memory_address):
12232 Handle SUBREG case.
12234 2018-01-12 Richard Biener <rguenther@suse.de>
12236 PR tree-optimization/80846
12237 * target.def (split_reduction): New target hook.
12238 * targhooks.c (default_split_reduction): New function.
12239 * targhooks.h (default_split_reduction): Declare.
12240 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12241 target requests first reduce vectors by combining low and high
12243 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12244 (get_vectype_for_scalar_type_and_size): Export.
12245 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12246 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12247 * doc/tm.texi: Regenerate.
12248 * config/i386/i386.c (ix86_split_reduction): Implement
12249 TARGET_VECTORIZE_SPLIT_REDUCTION.
12251 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12254 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12255 in PIC mode except for TARGET_VXWORKS_RTP.
12256 * config/sparc/sparc.c: Include cfgrtl.h.
12257 (TARGET_INIT_PIC_REG): Define.
12258 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12259 (sparc_pic_register_p): New predicate.
12260 (sparc_legitimate_address_p): Use it.
12261 (sparc_legitimize_pic_address): Likewise.
12262 (sparc_delegitimize_address): Likewise.
12263 (sparc_mode_dependent_address_p): Likewise.
12264 (gen_load_pcrel_sym): Remove 4th parameter.
12265 (load_got_register): Adjust call to above. Remove obsolete stuff.
12266 (sparc_expand_prologue): Do not call load_got_register here.
12267 (sparc_flat_expand_prologue): Likewise.
12268 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12269 (sparc_use_pseudo_pic_reg): New function.
12270 (sparc_init_pic_reg): Likewise.
12271 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12272 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12274 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12276 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12277 Add item for branch_cost.
12279 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12281 PR rtl-optimization/83565
12282 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12283 not extend the result to a larger mode for rotate operations.
12284 (num_sign_bit_copies1): Likewise.
12286 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12289 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12291 Use values-Xc.o for -pedantic.
12292 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12294 2018-01-12 Martin Liska <mliska@suse.cz>
12297 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12299 (possible_polymorphic_call_targets): Use it.
12300 (ipa_devirt): Likewise.
12302 2018-01-12 Martin Liska <mliska@suse.cz>
12304 * profile-count.h (enum profile_quality): Use 0 as invalid
12305 enum value of profile_quality.
12307 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12309 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12310 -mext-string options.
12312 2018-01-12 Richard Biener <rguenther@suse.de>
12314 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12315 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12316 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12318 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12320 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12322 * configure.ac (--with-long-double-format): Add support for the
12323 configuration option to change the default long double format on
12325 * config.gcc (powerpc*-linux*-*): Likewise.
12326 * configure: Regenerate.
12327 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12328 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12329 used without modification.
12331 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12333 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12334 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12335 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12336 MISC_BUILTIN_SPEC_BARRIER.
12337 (rs6000_init_builtins): Likewise.
12338 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12340 (speculation_barrier): New define_insn.
12341 * doc/extend.texi: Document __builtin_speculation_barrier.
12343 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12346 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12347 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12348 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12350 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12351 integral modes instead of "ss" and "sd".
12352 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12353 vectors with 32-bit and 64-bit elements.
12354 (vecdupssescalarmodesuffix): New mode attribute.
12355 (vec_dup<mode>): Use it.
12357 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12360 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12361 frame if argument is passed on stack.
12363 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12366 * ree.c (combine_reaching_defs): Optimize also
12367 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12368 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12370 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12372 PR middle-end/83189
12373 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12375 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12377 PR middle-end/83718
12378 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12379 after they are computed.
12381 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12383 PR tree-optimization/83695
12384 * gimple-loop-linterchange.cc
12385 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12386 reset cached scev information after interchange.
12387 (pass_linterchange::execute): Remove call to scev_reset_htab.
12389 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12391 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12392 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12393 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12394 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12395 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12396 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12397 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12398 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12399 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12400 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12401 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12402 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12403 (V_lane_reg): Likewise.
12404 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12406 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12407 (vfmal_lane_low<mode>_intrinsic,
12408 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12409 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12410 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12411 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12412 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12413 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12415 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12417 * config/arm/arm-cpus.in (fp16fml): New feature.
12418 (ALL_SIMD): Add fp16fml.
12419 (armv8.2-a): Add fp16fml as an option.
12420 (armv8.3-a): Likewise.
12421 (armv8.4-a): Add fp16fml as part of fp16.
12422 * config/arm/arm.h (TARGET_FP16FML): Define.
12423 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12425 * config/arm/arm-modes.def (V2HF): Define.
12426 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12427 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12428 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12429 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12430 vfmsl_low, vfmsl_high): New set of builtins.
12431 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12432 (vfml_op): New code attribute.
12433 (VFMLHALVES): New int iterator.
12434 (VFML, VFMLSEL): New mode attributes.
12435 (V_reg): Define mapping for V2HF.
12436 (V_hi, V_lo): New mode attributes.
12437 (VF_constraint): Likewise.
12438 (vfml_half, vfml_half_selector): New int attributes.
12439 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12441 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12442 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12444 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12445 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12446 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12447 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12449 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12450 Document new effective target and option set.
12452 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12454 * config/arm/arm-cpus.in (armv8_4): New feature.
12455 (ARMv8_4a): New fgroup.
12456 (armv8.4-a): New arch.
12457 * config/arm/arm-tables.opt: Regenerate.
12458 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12459 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12460 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12461 Add matching rules for -march=armv8.4-a and extensions.
12462 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12464 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12467 * config/rx/rx.md (BW): New mode attribute.
12468 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12470 2018-01-11 Richard Biener <rguenther@suse.de>
12472 PR tree-optimization/83435
12473 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12474 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12475 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12477 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12478 Alan Hayward <alan.hayward@arm.com>
12479 David Sherwood <david.sherwood@arm.com>
12481 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12483 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12484 (aarch64_print_address_internal): Use it to check for a zero offset.
12486 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12487 Alan Hayward <alan.hayward@arm.com>
12488 David Sherwood <david.sherwood@arm.com>
12490 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12491 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12492 Return a poly_int64 rather than a HOST_WIDE_INT.
12493 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12494 rather than a HOST_WIDE_INT.
12495 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12496 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12497 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12498 final_offset from HOST_WIDE_INT to poly_int64.
12499 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12500 to_constant when getting the number of units in an Advanced SIMD
12502 (aarch64_builtin_vectorized_function): Check for a constant number
12504 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12506 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12507 attribute instead of GET_MODE_NUNITS.
12508 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12509 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12510 GET_MODE_SIZE for fixed-size registers.
12511 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12512 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12513 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12514 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12515 (aarch64_print_operand, aarch64_print_address_internal)
12516 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12517 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12518 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12519 Handle polynomial GET_MODE_SIZE.
12520 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12521 wider than SImode without modification.
12522 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12523 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12524 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12525 passing and returning SVE modes.
12526 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12527 rather than GEN_INT.
12528 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12529 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12530 (aarch64_allocate_and_probe_stack_space): Likewise.
12531 (aarch64_layout_frame): Cope with polynomial offsets.
12532 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12533 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12534 polynomial offsets.
12535 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12536 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12537 poly_int64 rather than a HOST_WIDE_INT.
12538 (aarch64_get_separate_components, aarch64_process_components)
12539 (aarch64_expand_prologue, aarch64_expand_epilogue)
12540 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12541 (aarch64_anchor_offset): New function, split out from...
12542 (aarch64_legitimize_address): ...here.
12543 (aarch64_builtin_vectorization_cost): Handle polynomial
12544 TYPE_VECTOR_SUBPARTS.
12545 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12547 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12548 number of elements from the PARALLEL rather than the mode.
12549 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12550 rather than GET_MODE_BITSIZE.
12551 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12552 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12553 (aarch64_expand_vec_perm_const_1): Handle polynomial
12554 d->perm.length () and d->perm elements.
12555 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12556 Apply to_constant to d->perm elements.
12557 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12558 polynomial CONST_VECTOR_NUNITS.
12559 (aarch64_move_pointer): Take amount as a poly_int64 rather
12561 (aarch64_progress_pointer): Avoid temporary variable.
12562 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12563 the mode attribute instead of GET_MODE.
12565 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12566 Alan Hayward <alan.hayward@arm.com>
12567 David Sherwood <david.sherwood@arm.com>
12569 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12570 x exists before using it.
12571 (aarch64_add_constant_internal): Rename to...
12572 (aarch64_add_offset_1): ...this. Replace regnum with separate
12573 src and dest rtxes. Handle the case in which they're different,
12574 including when the offset is zero. Replace scratchreg with an rtx.
12575 Use 2 additions if there is no spare register into which we can
12576 move a 16-bit constant.
12577 (aarch64_add_constant): Delete.
12578 (aarch64_add_offset): Replace reg with separate src and dest
12579 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12580 Use aarch64_add_offset_1.
12581 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12582 an rtx rather than an int. Take the delta as a poly_int64
12583 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12584 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12585 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12586 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12587 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12588 and aarch64_add_sp.
12589 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12590 aarch64_add_constant.
12592 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12594 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12595 Use scalar_float_mode.
12597 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12599 * config/aarch64/aarch64-simd.md
12600 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12601 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12602 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12603 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12604 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12605 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12606 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12607 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12608 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12609 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12611 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12614 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12615 targ_options->x_arm_arch_string is non NULL.
12617 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12619 * config/aarch64/aarch64.h
12620 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12622 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12625 * expmed.c (emit_store_flag_force): Swap if const op0
12626 and change VOIDmode to mode of op0.
12628 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12630 PR rtl-optimization/83761
12631 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12632 than bytes to mode_for_size.
12634 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12636 PR middle-end/83189
12637 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12638 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12641 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12643 PR middle-end/83575
12644 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12645 when in layout mode.
12646 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12647 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12650 2018-01-10 Michael Collison <michael.collison@arm.com>
12652 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12653 * config/aarch64/aarch64-option-extension.def: Add
12654 AARCH64_OPT_EXTENSION of 'fp16fml'.
12655 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12656 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12657 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12658 * config/aarch64/constraints.md (Ui7): New constraint.
12659 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12660 (VFMLA_SEL_W): Ditto.
12663 (VFMLA16_LOW): New int iterator.
12664 (VFMLA16_HIGH): Ditto.
12665 (UNSPEC_FMLAL): New unspec.
12666 (UNSPEC_FMLSL): Ditto.
12667 (UNSPEC_FMLAL2): Ditto.
12668 (UNSPEC_FMLSL2): Ditto.
12669 (f16mac): New code attribute.
12670 * config/aarch64/aarch64-simd-builtins.def
12671 (aarch64_fmlal_lowv2sf): Ditto.
12672 (aarch64_fmlsl_lowv2sf): Ditto.
12673 (aarch64_fmlalq_lowv4sf): Ditto.
12674 (aarch64_fmlslq_lowv4sf): Ditto.
12675 (aarch64_fmlal_highv2sf): Ditto.
12676 (aarch64_fmlsl_highv2sf): Ditto.
12677 (aarch64_fmlalq_highv4sf): Ditto.
12678 (aarch64_fmlslq_highv4sf): Ditto.
12679 (aarch64_fmlal_lane_lowv2sf): Ditto.
12680 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12681 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12682 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12683 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12684 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12685 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12686 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12687 (aarch64_fmlal_lane_highv2sf): Ditto.
12688 (aarch64_fmlsl_lane_highv2sf): Ditto.
12689 (aarch64_fmlal_laneq_highv2sf): Ditto.
12690 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12691 (aarch64_fmlalq_lane_highv4sf): Ditto.
12692 (aarch64_fmlsl_lane_highv4sf): Ditto.
12693 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12694 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12695 * config/aarch64/aarch64-simd.md:
12696 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12697 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12698 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12699 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12700 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12701 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12702 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12703 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12704 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12705 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12706 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12707 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12708 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12709 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12710 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12711 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12712 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12713 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12714 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12715 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12716 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12717 (vfmlsl_low_u32): Ditto.
12718 (vfmlalq_low_u32): Ditto.
12719 (vfmlslq_low_u32): Ditto.
12720 (vfmlal_high_u32): Ditto.
12721 (vfmlsl_high_u32): Ditto.
12722 (vfmlalq_high_u32): Ditto.
12723 (vfmlslq_high_u32): Ditto.
12724 (vfmlal_lane_low_u32): Ditto.
12725 (vfmlsl_lane_low_u32): Ditto.
12726 (vfmlal_laneq_low_u32): Ditto.
12727 (vfmlsl_laneq_low_u32): Ditto.
12728 (vfmlalq_lane_low_u32): Ditto.
12729 (vfmlslq_lane_low_u32): Ditto.
12730 (vfmlalq_laneq_low_u32): Ditto.
12731 (vfmlslq_laneq_low_u32): Ditto.
12732 (vfmlal_lane_high_u32): Ditto.
12733 (vfmlsl_lane_high_u32): Ditto.
12734 (vfmlal_laneq_high_u32): Ditto.
12735 (vfmlsl_laneq_high_u32): Ditto.
12736 (vfmlalq_lane_high_u32): Ditto.
12737 (vfmlslq_lane_high_u32): Ditto.
12738 (vfmlalq_laneq_high_u32): Ditto.
12739 (vfmlslq_laneq_high_u32): Ditto.
12740 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12741 (AARCH64_FL_FOR_ARCH8_4): New.
12742 (AARCH64_ISA_F16FML): New ISA flag.
12743 (TARGET_F16FML): New feature flag for fp16fml.
12744 (doc/invoke.texi): Document new fp16fml option.
12746 2018-01-10 Michael Collison <michael.collison@arm.com>
12748 * config/aarch64/aarch64-builtins.c:
12749 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12750 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12751 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12752 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12753 (AARCH64_ISA_SHA3): New ISA flag.
12754 (TARGET_SHA3): New feature flag for sha3.
12755 * config/aarch64/iterators.md (sha512_op): New int attribute.
12756 (CRYPTO_SHA512): New int iterator.
12757 (UNSPEC_SHA512H): New unspec.
12758 (UNSPEC_SHA512H2): Ditto.
12759 (UNSPEC_SHA512SU0): Ditto.
12760 (UNSPEC_SHA512SU1): Ditto.
12761 * config/aarch64/aarch64-simd-builtins.def
12762 (aarch64_crypto_sha512hqv2di): New builtin.
12763 (aarch64_crypto_sha512h2qv2di): Ditto.
12764 (aarch64_crypto_sha512su0qv2di): Ditto.
12765 (aarch64_crypto_sha512su1qv2di): Ditto.
12766 (aarch64_eor3qv8hi): Ditto.
12767 (aarch64_rax1qv2di): Ditto.
12768 (aarch64_xarqv2di): Ditto.
12769 (aarch64_bcaxqv8hi): Ditto.
12770 * config/aarch64/aarch64-simd.md:
12771 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12772 (aarch64_crypto_sha512su0qv2di): Ditto.
12773 (aarch64_crypto_sha512su1qv2di): Ditto.
12774 (aarch64_eor3qv8hi): Ditto.
12775 (aarch64_rax1qv2di): Ditto.
12776 (aarch64_xarqv2di): Ditto.
12777 (aarch64_bcaxqv8hi): Ditto.
12778 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12779 (vsha512h2q_u64): Ditto.
12780 (vsha512su0q_u64): Ditto.
12781 (vsha512su1q_u64): Ditto.
12782 (veor3q_u16): Ditto.
12783 (vrax1q_u64): Ditto.
12784 (vxarq_u64): Ditto.
12785 (vbcaxq_u16): Ditto.
12786 * config/arm/types.md (crypto_sha512): New type attribute.
12787 (crypto_sha3): Ditto.
12788 (doc/invoke.texi): Document new sha3 option.
12790 2018-01-10 Michael Collison <michael.collison@arm.com>
12792 * config/aarch64/aarch64-builtins.c:
12793 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12794 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12795 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12796 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12797 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12798 (AARCH64_ISA_SM4): New ISA flag.
12799 (TARGET_SM4): New feature flag for sm4.
12800 * config/aarch64/aarch64-simd-builtins.def
12801 (aarch64_sm3ss1qv4si): Ditto.
12802 (aarch64_sm3tt1aq4si): Ditto.
12803 (aarch64_sm3tt1bq4si): Ditto.
12804 (aarch64_sm3tt2aq4si): Ditto.
12805 (aarch64_sm3tt2bq4si): Ditto.
12806 (aarch64_sm3partw1qv4si): Ditto.
12807 (aarch64_sm3partw2qv4si): Ditto.
12808 (aarch64_sm4eqv4si): Ditto.
12809 (aarch64_sm4ekeyqv4si): Ditto.
12810 * config/aarch64/aarch64-simd.md:
12811 (aarch64_sm3ss1qv4si): Ditto.
12812 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12813 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12814 (aarch64_sm4eqv4si): Ditto.
12815 (aarch64_sm4ekeyqv4si): Ditto.
12816 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12817 (sm3part_op): Ditto.
12818 (CRYPTO_SM3TT): Ditto.
12819 (CRYPTO_SM3PART): Ditto.
12820 (UNSPEC_SM3SS1): New unspec.
12821 (UNSPEC_SM3TT1A): Ditto.
12822 (UNSPEC_SM3TT1B): Ditto.
12823 (UNSPEC_SM3TT2A): Ditto.
12824 (UNSPEC_SM3TT2B): Ditto.
12825 (UNSPEC_SM3PARTW1): Ditto.
12826 (UNSPEC_SM3PARTW2): Ditto.
12827 (UNSPEC_SM4E): Ditto.
12828 (UNSPEC_SM4EKEY): Ditto.
12829 * config/aarch64/constraints.md (Ui2): New constraint.
12830 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12831 * config/arm/types.md (crypto_sm3): New type attribute.
12832 (crypto_sm4): Ditto.
12833 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12834 (vsm3tt1aq_u32): Ditto.
12835 (vsm3tt1bq_u32): Ditto.
12836 (vsm3tt2aq_u32): Ditto.
12837 (vsm3tt2bq_u32): Ditto.
12838 (vsm3partw1q_u32): Ditto.
12839 (vsm3partw2q_u32): Ditto.
12840 (vsm4eq_u32): Ditto.
12841 (vsm4ekeyq_u32): Ditto.
12842 (doc/invoke.texi): Document new sm4 option.
12844 2018-01-10 Michael Collison <michael.collison@arm.com>
12846 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12847 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12848 (AARCH64_FL_FOR_ARCH8_4): New.
12849 (AARCH64_FL_V8_4): New flag.
12850 (doc/invoke.texi): Document new armv8.4-a option.
12852 2018-01-10 Michael Collison <michael.collison@arm.com>
12854 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12855 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12856 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12857 * config/aarch64/aarch64-option-extension.def: Add
12858 AARCH64_OPT_EXTENSION of 'sha2'.
12859 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12860 (crypto): Disable sha2 and aes if crypto disabled.
12861 (crypto): Enable aes and sha2 if enabled.
12862 (simd): Disable sha2 and aes if simd disabled.
12863 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12865 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12866 (TARGET_SHA2): New feature flag for sha2.
12867 (TARGET_AES): New feature flag for aes.
12868 * config/aarch64/aarch64-simd.md:
12869 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12870 conditional on TARGET_AES.
12871 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12872 (aarch64_crypto_sha1hsi): Make pattern conditional
12874 (aarch64_crypto_sha1hv4si): Ditto.
12875 (aarch64_be_crypto_sha1hv4si): Ditto.
12876 (aarch64_crypto_sha1su1v4si): Ditto.
12877 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12878 (aarch64_crypto_sha1su0v4si): Ditto.
12879 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12880 (aarch64_crypto_sha256su0v4si): Ditto.
12881 (aarch64_crypto_sha256su1v4si): Ditto.
12882 (doc/invoke.texi): Document new aes and sha2 options.
12884 2018-01-10 Martin Sebor <msebor@redhat.com>
12886 PR tree-optimization/83781
12887 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12890 2018-01-11 Martin Sebor <msebor@gmail.com>
12891 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12893 PR tree-optimization/83501
12894 PR tree-optimization/81703
12896 * tree-ssa-strlen.c (get_string_cst): Rename...
12897 (get_string_len): ...to this. Handle global constants.
12898 (handle_char_store): Adjust.
12900 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12901 Jim Wilson <jimw@sifive.com>
12903 * config/riscv/riscv-protos.h (riscv_output_return): New.
12904 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12905 (riscv_attribute_table, riscv_output_return),
12906 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12907 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12908 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12910 (riscv_expand_prologue): Add early return for naked function.
12911 (riscv_expand_epilogue): Likewise.
12912 (riscv_function_ok_for_sibcall): Return false for naked function.
12913 (riscv_set_current_function): New.
12914 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12915 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12916 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12917 * doc/extend.texi (RISC-V Function Attributes): New.
12919 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12921 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12922 check for 128-bit long double before checking TCmode.
12923 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12924 128-bit long doubles before checking TFmode or TCmode.
12925 (FLOAT128_IBM_P): Likewise.
12927 2018-01-10 Martin Sebor <msebor@redhat.com>
12929 PR tree-optimization/83671
12930 * builtins.c (c_strlen): Unconditionally return zero for the empty
12932 Use -Warray-bounds for warnings.
12933 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12934 for non-constant array indices with COMPONENT_REF, arrays of
12935 arrays, and pointers to arrays.
12936 (gimple_fold_builtin_strlen): Determine and set length range for
12937 non-constant character arrays.
12939 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12941 PR middle-end/81897
12942 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12945 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12947 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12949 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12952 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12953 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12954 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12955 indexed_or_indirect_operand predicate.
12956 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12957 (*vsx_le_perm_load_v8hi): Likewise.
12958 (*vsx_le_perm_load_v16qi): Likewise.
12959 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12960 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12961 (*vsx_le_perm_store_v8hi): Likewise.
12962 (*vsx_le_perm_store_v16qi): Likewise.
12963 (eight unnamed splitters): Likewise.
12965 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12967 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12968 * config/rs6000/emmintrin.h: Likewise.
12969 * config/rs6000/mmintrin.h: Likewise.
12970 * config/rs6000/xmmintrin.h: Likewise.
12972 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12975 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12977 * tree.c (tree_nop_conversion): Return true for location wrapper
12979 (maybe_wrap_with_location): New function.
12980 (selftest::check_strip_nops): New function.
12981 (selftest::test_location_wrappers): New function.
12982 (selftest::tree_c_tests): Call it.
12983 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12984 (maybe_wrap_with_location): New decl.
12985 (EXPR_LOCATION_WRAPPER_P): New macro.
12986 (location_wrapper_p): New inline function.
12987 (tree_strip_any_location_wrapper): New inline function.
12989 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12992 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12993 stack_realign_offset for the largest alignment of stack slot
12995 (ix86_find_max_used_stack_alignment): New function.
12996 (ix86_finalize_stack_frame_flags): Use it. Set
12997 max_used_stack_alignment if we don't realign stack.
12998 * config/i386/i386.h (machine_function): Add
12999 max_used_stack_alignment.
13001 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
13003 * config/arm/arm.opt (-mbranch-cost): New option.
13004 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13007 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13010 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13011 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13013 2018-01-10 Richard Biener <rguenther@suse.de>
13016 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13017 early out so it also covers the case where we have a non-NULL
13020 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13022 PR tree-optimization/83753
13023 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13024 for non-strided grouped accesses if the number of elements is 1.
13026 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13029 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13030 * i386.h (TARGET_USE_GATHER): Define.
13031 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13033 2018-01-10 Martin Liska <mliska@suse.cz>
13036 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13037 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13039 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13040 CLEANUP_NO_PARTITIONING is not set.
13042 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13044 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13045 for vectors, as a partial revert of r254296.
13046 * rtl.h (const_vec_p): Delete.
13047 (const_vec_duplicate_p): Don't test for vector CONSTs.
13048 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13049 * expmed.c (make_tree): Likewise.
13052 * common.md (E, F): Use CONSTANT_P instead of checking for
13054 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13055 checking for CONST_VECTOR.
13057 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13059 PR middle-end/83575
13060 * predict.c (force_edge_cold): Handle in more sane way edges
13061 with no prediction.
13063 2018-01-09 Carl Love <cel@us.ibm.com>
13065 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13067 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13068 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13069 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13070 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13071 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13072 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13073 * config/rs6000/rs6000-protos.h: Add extern defition for
13074 rs6000_generate_float2_double_code.
13075 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13077 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13078 (float2_v2df): Add define_expand.
13080 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13083 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13084 op_mode in the force_to_mode call.
13086 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13088 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13089 instead of checking each element individually.
13090 (aarch64_evpc_uzp): Likewise.
13091 (aarch64_evpc_zip): Likewise.
13092 (aarch64_evpc_ext): Likewise.
13093 (aarch64_evpc_rev): Likewise.
13094 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13095 instead of checking each element individually. Return true without
13097 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13098 whether all selected elements come from the same input, instead of
13099 checking each element individually. Remove calls to gen_rtx_REG,
13100 start_sequence and end_sequence and instead assert that no rtl is
13103 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13105 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13106 order of HIGH and CONST checks.
13108 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13110 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13111 if the destination isn't an SSA_NAME.
13113 2018-01-09 Richard Biener <rguenther@suse.de>
13115 PR tree-optimization/83668
13116 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13118 (canonicalize_loop_form): ... here, renamed from ...
13119 (canonicalize_loop_closed_ssa_form): ... this and amended to
13120 swap successor edges for loop exit blocks to make us use
13121 the RPO order we need for initial schedule generation.
13123 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13125 PR tree-optimization/64811
13126 * match.pd: When optimizing comparisons with Inf, avoid
13127 introducing or losing exceptions from comparisons with NaN.
13129 2018-01-09 Martin Liska <mliska@suse.cz>
13132 * asan.c (shadow_mem_size): Add gcc_assert.
13134 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13136 Don't save registers in main().
13139 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13140 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13141 * config/avr/avr.c (avr_set_current_function): Don't error if
13142 naked, OS_task or OS_main are specified at the same time.
13143 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13145 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13147 * common/config/avr/avr-common.c (avr_option_optimization_table):
13148 Switch on -mmain-is-OS_task for optimizing compilations.
13150 2018-01-09 Richard Biener <rguenther@suse.de>
13152 PR tree-optimization/83572
13153 * graphite.c: Include cfganal.h.
13154 (graphite_transform_loops): Connect infinite loops to exit
13155 and remove fake edges at the end.
13157 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13159 * ipa-inline.c (edge_badness): Revert accidental checkin.
13161 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13164 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13165 symbols; not inline clones.
13167 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13170 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13171 hard registers. Formatting fixes.
13173 PR preprocessor/83722
13174 * gcc.c (try_generate_repro): Pass
13175 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13176 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13179 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13180 Kito Cheng <kito.cheng@gmail.com>
13182 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13183 (riscv_leaf_function_p): Delete.
13184 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13186 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13188 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13190 (do_ifelse): New function.
13191 (do_isel): New function.
13192 (do_sub3): New function.
13193 (do_add3): New function.
13194 (do_load_mask_compare): New function.
13195 (do_overlap_load_compare): New function.
13196 (expand_compare_loop): New function.
13197 (expand_block_compare): Call expand_compare_loop() when appropriate.
13198 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13199 option description.
13200 (-mblock-compare-inline-loop-limit): New option.
13202 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13205 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13206 Reverse order of second and third operands in first alternative.
13207 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13208 of first and second elements in UNSPEC_VPERMR vector.
13209 (altivec_expand_vec_perm_le): Likewise.
13211 2018-01-08 Jeff Law <law@redhat.com>
13213 PR rtl-optimizatin/81308
13214 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13215 (process_switch): If group_case_labels makes a change, then set
13217 (pass_convert_switch::execute): If a switch is converted, then
13218 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13220 PR rtl-optimization/81308
13221 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13224 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13226 PR target/83663 - Revert r255946
13227 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13228 generation for cases where splatting a value is not useful.
13229 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13230 across a vec_duplicate and a paradoxical subreg forming a vector
13231 mode to a vec_concat.
13233 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13235 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13236 -march=armv8.3-a variants.
13237 * config/arm/t-multilib: Likewise.
13238 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13240 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13242 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13244 (cceq_ior_compare_complement): Give it a name so I can use it, and
13245 change boolean_or_operator predicate to boolean_operator so it can
13246 be used to generate a crand.
13247 (eqne): New code iterator.
13248 (bd/bd_neg): New code_attrs.
13249 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13250 a single define_insn.
13251 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13252 decrement (bdnzt/bdnzf/bdzt/bdzf).
13253 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13254 with the new names of the branch decrement patterns, and added the
13255 names of the branch decrement conditional patterns.
13257 2018-01-08 Richard Biener <rguenther@suse.de>
13259 PR tree-optimization/83563
13260 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13263 2018-01-08 Richard Biener <rguenther@suse.de>
13265 PR middle-end/83713
13266 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13268 2018-01-08 Richard Biener <rguenther@suse.de>
13270 PR tree-optimization/83685
13271 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13272 references to abnormals.
13274 2018-01-08 Richard Biener <rguenther@suse.de>
13277 * dwarf2out.c (output_indirect_strings): Handle empty
13278 skeleton_debug_str_hash.
13279 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13281 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13283 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13284 (emit_store_direct): Likewise.
13285 (arc_trampoline_adjust_address): Likewise.
13286 (arc_asm_trampoline_template): New function.
13287 (arc_initialize_trampoline): Use asm_trampoline_template.
13288 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13289 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13290 * config/arc/arc.md (flush_icache): Delete pattern.
13292 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13294 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13295 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13298 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13301 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13302 by not USED_FOR_TARGET.
13303 (make_pass_resolve_sw_modes): Likewise.
13305 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13307 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13310 2018-01-08 Richard Biener <rguenther@suse.de>
13312 PR middle-end/83580
13313 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13315 2018-01-08 Richard Biener <rguenther@suse.de>
13317 PR middle-end/83517
13318 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13320 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13322 PR middle-end/81897
13323 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13324 basic blocks with a small number of successors.
13325 (convert_control_dep_chain_into_preds): Improve handling of
13327 (dump_predicates): Split apart into...
13328 (dump_pred_chain): ...here...
13329 (dump_pred_info): ...and here.
13330 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13331 (can_chain_union_be_invalidated_p): Improve check for invalidation
13333 (uninit_uses_cannot_happen): Avoid unnecessary if
13334 convert_control_dep_chain_into_preds yielded nothing.
13336 2018-01-06 Martin Sebor <msebor@redhat.com>
13338 PR tree-optimization/83640
13339 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13340 subtracting negative offset from size.
13341 (builtin_access::overlap): Adjust offset bounds of the access to fall
13342 within the size of the object if possible.
13344 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13346 PR rtl-optimization/83699
13347 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13348 extract_bit_field_as_subreg to cases in which the extracted
13349 value is also a vector.
13351 * lra-constraints.c (process_alt_operands): Test for the equivalence
13352 substitutions when detecting a possible reload cycle.
13354 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13357 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13358 by default if flag_selective_schedling{,2}. Formatting fixes.
13360 PR rtl-optimization/83682
13361 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13362 if it has non-VECTOR_MODE element mode.
13363 (vec_duplicate_p): Likewise.
13365 PR middle-end/83694
13366 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13367 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13369 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13372 * config/i386/i386-builtin.def
13373 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13374 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13375 Require also OPTION_MASK_ISA_AVX512F in addition to
13376 OPTION_MASK_ISA_GFNI.
13377 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13378 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13379 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13380 to OPTION_MASK_ISA_GFNI.
13381 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13382 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13383 OPTION_MASK_ISA_AVX512BW.
13384 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13385 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13386 addition to OPTION_MASK_ISA_GFNI.
13387 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13388 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13389 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13390 to OPTION_MASK_ISA_GFNI.
13391 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13392 a requirement for all ISAs rather than any of them with a few
13394 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13396 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13397 bitmasks to be enabled with 3 exceptions, instead of requiring any
13398 enabled ISA with lots of exceptions.
13399 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13400 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13401 Change avx512bw in isa attribute to avx512f.
13402 * config/i386/sgxintrin.h: Add license boilerplate.
13403 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13404 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13405 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13406 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13408 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13409 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13410 temporarily sse2 rather than sse if not enabled already.
13413 * config/i386/sse.md (VI248_VLBW): Rename to ...
13414 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13415 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13416 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13417 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13418 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13419 mode iterator instead of VI248_VLBW.
13421 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13423 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13424 (record_modified): Skip clobbers; add debug output.
13425 (param_change_prob): Use sreal frequencies.
13427 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13429 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13430 punt for user-aligned variables.
13432 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13434 * tree-chrec.c (chrec_contains_symbols): Return true for
13437 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13440 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13441 of (x|y) == x for BICS pattern.
13443 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13445 PR tree-optimization/83605
13446 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13447 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13450 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13452 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13453 * config/epiphany/rtems.h: New file.
13455 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13456 Uros Bizjak <ubizjak@gmail.com>
13459 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13460 QIreg_operand instead of register_operand predicate.
13461 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13462 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13463 comments instead of -fmitigate[-_]rop.
13465 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13468 * cgraphunit.c (symbol_table::compile): Switch to text_section
13469 before calling assembly_start debug hook.
13470 * run-rtl-passes.c (run_rtl_passes): Likewise.
13473 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13475 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13476 range_int_cst_p rather than !symbolic_range_p before calling
13477 extract_range_from_multiplicative_op_1.
13479 2018-01-04 Jeff Law <law@redhat.com>
13481 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13482 redundant test in assertion.
13484 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13486 * doc/rtl.texi: Document machine_mode wrapper classes.
13488 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13490 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13491 using tree_to_uhwi.
13493 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13495 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13496 the VEC_PERM_EXPR fold to fail.
13498 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13501 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13502 to switched_sections.
13504 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13507 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13508 test for d.testing.
13510 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13513 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13514 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13516 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13519 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13520 is BLKmode and bitpos not zero or mode change is needed.
13522 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13525 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13528 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13531 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13532 instead of MULT rtx. Update all corresponding splitters.
13533 (*saddl_se): Ditto.
13534 (*ssub<modesuffix>): Ditto.
13535 (*ssubl_se): Ditto.
13536 (*cmp_sadd_di): Update split patterns.
13537 (*cmp_sadd_si): Ditto.
13538 (*cmp_sadd_sidi): Ditto.
13539 (*cmp_ssub_di): Ditto.
13540 (*cmp_ssub_si): Ditto.
13541 (*cmp_ssub_sidi): Ditto.
13542 * config/alpha/predicates.md (const23_operand): New predicate.
13543 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13544 Look for ASHIFT, not MULT inner operand.
13545 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13547 2018-01-04 Martin Liska <mliska@suse.cz>
13549 PR gcov-profile/83669
13550 * gcov.c (output_intermediate_file): Add version to intermediate
13552 * doc/gcov.texi: Document new field 'version' in intermediate
13553 file format. Fix location of '-k' option of gcov command.
13555 2018-01-04 Martin Liska <mliska@suse.cz>
13558 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13560 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13562 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13564 2018-01-03 Martin Sebor <msebor@redhat.com>
13566 PR tree-optimization/83655
13567 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13568 checking calls with invalid arguments.
13570 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13572 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13573 (vectorizable_mask_load_store): Delete.
13574 (vectorizable_call): Return false for masked loads and stores.
13575 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13576 instead of gimple_assign_rhs1.
13577 (vectorizable_load): Handle IFN_MASK_LOAD.
13578 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13582 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13584 (vectorizable_mask_load_store): ...here.
13585 (vectorizable_load): ...and here.
13587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13589 * tree-vect-stmts.c (vect_build_all_ones_mask)
13590 (vect_build_zero_merge_argument): New functions, split out from...
13591 (vectorizable_load): ...here.
13593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13595 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13597 (vectorizable_mask_load_store): ...here.
13598 (vectorizable_store): ...and here.
13600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13602 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13604 (vectorizable_mask_load_store): ...here.
13606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13608 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13609 (vect_model_store_cost): Take a vec_load_store_type instead of a
13611 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13612 (vect_model_store_cost): Take a vec_load_store_type instead of a
13614 (vectorizable_mask_load_store): Update accordingly.
13615 (vectorizable_store): Likewise.
13616 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13620 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13621 IFN_MASK_LOAD calls here rather than...
13622 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13625 Alan Hayward <alan.hayward@arm.com>
13626 David Sherwood <david.sherwood@arm.com>
13628 * expmed.c (extract_bit_field_1): For vector extracts,
13629 fall back to extract_bit_field_as_subreg if vec_extract
13632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13633 Alan Hayward <alan.hayward@arm.com>
13634 David Sherwood <david.sherwood@arm.com>
13636 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13637 they are variable or constant sized.
13638 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13639 slots for constant-sized data.
13641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13642 Alan Hayward <alan.hayward@arm.com>
13643 David Sherwood <david.sherwood@arm.com>
13645 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13646 handling COND_EXPRs with boolean comparisons, try to find a better
13647 basis for the mask type than the boolean itself.
13649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13651 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13652 is calculated and how it can be overridden.
13653 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13654 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13656 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13660 Alan Hayward <alan.hayward@arm.com>
13661 David Sherwood <david.sherwood@arm.com>
13663 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13664 Remove the mode argument.
13665 (aarch64_simd_valid_immediate): Remove the mode and inverse
13667 * config/aarch64/iterators.md (bitsize): New iterator.
13668 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13669 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13670 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13671 aarch64_simd_valid_immediate.
13672 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13673 (aarch64_reg_or_bic_imm): Likewise.
13674 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13675 with an insn_type enum and msl with a modifier_type enum.
13676 Replace element_width with a scalar_mode. Change the shift
13677 to unsigned int. Add constructors for scalar_float_mode and
13678 scalar_int_mode elements.
13679 (aarch64_vect_float_const_representable_p): Delete.
13680 (aarch64_can_const_movi_rtx_p)
13681 (aarch64_simd_scalar_immediate_valid_for_move)
13682 (aarch64_simd_make_constant): Update call to
13683 aarch64_simd_valid_immediate.
13684 (aarch64_advsimd_valid_immediate_hs): New function.
13685 (aarch64_advsimd_valid_immediate): Likewise.
13686 (aarch64_simd_valid_immediate): Remove mode and inverse
13687 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13688 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13689 and aarch64_float_const_representable_p on the result.
13690 (aarch64_output_simd_mov_immediate): Remove mode argument.
13691 Update call to aarch64_simd_valid_immediate and use of
13692 simd_immediate_info.
13693 (aarch64_output_scalar_simd_mov_immediate): Update call
13696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13697 Alan Hayward <alan.hayward@arm.com>
13698 David Sherwood <david.sherwood@arm.com>
13700 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13701 (mode_nunits): Likewise CONST_MODE_NUNITS.
13702 * machmode.def (ADJUST_NUNITS): Document.
13703 * genmodes.c (mode_data::need_nunits_adj): New field.
13704 (blank_mode): Update accordingly.
13705 (adj_nunits): New variable.
13706 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13708 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13709 listed in adj_nunits.
13710 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13711 listed in adj_nunits. Don't emit case statements for such modes.
13712 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13713 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13714 nothing if adj_nunits is nonnull.
13715 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13716 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13717 (emit_mode_fbit): Update use of print_maybe_const_decl.
13718 (emit_move_size): Likewise. Treat the array as non-const
13720 (emit_mode_adjustments): Handle adj_nunits.
13722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13724 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13725 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13726 (VECTOR_MODES): Use it.
13727 (make_vector_modes): Take the prefix as an argument.
13729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13730 Alan Hayward <alan.hayward@arm.com>
13731 David Sherwood <david.sherwood@arm.com>
13733 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13734 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13735 for MODE_VECTOR_BOOL.
13736 * machmode.def (VECTOR_BOOL_MODE): Document.
13737 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13738 (make_vector_bool_mode): New function.
13739 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13741 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13742 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13744 * stor-layout.c (int_mode_for_mode): Likewise.
13745 * tree.c (build_vector_type_for_mode): Likewise.
13746 * varasm.c (output_constant_pool_2): Likewise.
13747 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13748 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13749 for MODE_VECTOR_BOOL.
13750 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13751 of mode class checks.
13752 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13753 instead of a list of mode class checks.
13754 (expand_vector_scalar_condition): Likewise.
13755 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13758 Alan Hayward <alan.hayward@arm.com>
13759 David Sherwood <david.sherwood@arm.com>
13761 * machmode.h (mode_size): Change from unsigned short to
13763 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13764 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13765 or if measurement_type is not polynomial.
13766 (fixed_size_mode::includes_p): Check for constant-sized modes.
13767 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13768 return a poly_uint16 rather than an unsigned short.
13769 (emit_mode_size): Change the type of mode_size from unsigned short
13770 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13771 (emit_mode_adjustments): Cope with polynomial vector sizes.
13772 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13774 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13776 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13777 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13778 * caller-save.c (setup_save_areas): Likewise.
13779 (replace_reg_with_saved_mem): Likewise.
13780 * calls.c (emit_library_call_value_1): Likewise.
13781 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13782 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13783 (gen_lowpart_for_combine): Likewise.
13784 * convert.c (convert_to_integer_1): Likewise.
13785 * cse.c (equiv_constant, cse_insn): Likewise.
13786 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13787 (cselib_subst_to_values): Likewise.
13788 * dce.c (word_dce_process_block): Likewise.
13789 * df-problems.c (df_word_lr_mark_ref): Likewise.
13790 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13791 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13792 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13793 (rtl_for_decl_location): Likewise.
13794 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13795 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13796 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13797 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13798 (expand_expr_real_1): Likewise.
13799 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13800 (pad_below): Likewise.
13801 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13802 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13803 * ira.c (get_subreg_tracking_sizes): Likewise.
13804 * ira-build.c (ira_create_allocno_objects): Likewise.
13805 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13806 (ira_sort_regnos_for_alter_reg): Likewise.
13807 * ira-costs.c (record_operand_costs): Likewise.
13808 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13809 (resolve_simple_move): Likewise.
13810 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13811 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13812 (lra_constraints): Likewise.
13813 (CONST_POOL_OK_P): Reject variable-sized modes.
13814 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13815 (add_pseudo_to_slot, lra_spill): Likewise.
13816 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13817 * optabs-query.c (get_best_extraction_insn): Likewise.
13818 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13819 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13820 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13821 * recog.c (offsettable_address_addr_space_p): Likewise.
13822 * regcprop.c (maybe_mode_change): Likewise.
13823 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13824 * regrename.c (build_def_use): Likewise.
13825 * regstat.c (dump_reg_info): Likewise.
13826 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13827 (find_reloads, find_reloads_subreg_address): Likewise.
13828 * reload1.c (eliminate_regs_1): Likewise.
13829 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13830 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13831 (simplify_binary_operation_1, simplify_subreg): Likewise.
13832 * targhooks.c (default_function_arg_padding): Likewise.
13833 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13834 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13835 (verify_gimple_assign_ternary): Likewise.
13836 * tree-inline.c (estimate_move_cost): Likewise.
13837 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13838 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13839 (get_address_cost_ainc): Likewise.
13840 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13841 (vect_supportable_dr_alignment): Likewise.
13842 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13843 (vectorizable_reduction): Likewise.
13844 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13845 (vectorizable_operation, vectorizable_load): Likewise.
13846 * tree.c (build_same_sized_truth_vector_type): Likewise.
13847 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13848 * var-tracking.c (emit_note_insn_var_location): Likewise.
13849 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13850 (ADDR_VEC_ALIGN): Likewise.
13852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13853 Alan Hayward <alan.hayward@arm.com>
13854 David Sherwood <david.sherwood@arm.com>
13856 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13858 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13859 or if measurement_type is polynomial.
13860 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13861 * combine.c (make_extraction): Likewise.
13862 * dse.c (find_shift_sequence): Likewise.
13863 * dwarf2out.c (mem_loc_descriptor): Likewise.
13864 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13865 (extract_bit_field, extract_low_bits): Likewise.
13866 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13867 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13868 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13869 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13870 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13871 * reload.c (find_reloads): Likewise.
13872 * reload1.c (alter_reg): Likewise.
13873 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13874 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13875 * tree-if-conv.c (predicate_mem_writes): Likewise.
13876 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13877 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13878 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13879 * valtrack.c (dead_debug_insert_temp): Likewise.
13880 * varasm.c (mergeable_constant_section): Likewise.
13881 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13884 Alan Hayward <alan.hayward@arm.com>
13885 David Sherwood <david.sherwood@arm.com>
13887 * expr.c (expand_assignment): Cope with polynomial mode sizes
13888 when assigning to a CONCAT.
13890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13891 Alan Hayward <alan.hayward@arm.com>
13892 David Sherwood <david.sherwood@arm.com>
13894 * machmode.h (mode_precision): Change from unsigned short to
13896 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13898 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13899 or if measurement_type is not polynomial.
13900 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13901 in which the mode is already known to be a scalar_int_mode.
13902 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13903 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13905 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13906 for GET_MODE_PRECISION.
13907 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13908 for GET_MODE_PRECISION.
13909 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13911 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13912 (expand_field_assignment, make_extraction): Likewise.
13913 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13914 (get_last_value): Likewise.
13915 * convert.c (convert_to_integer_1): Likewise.
13916 * cse.c (cse_insn): Likewise.
13917 * expr.c (expand_expr_real_1): Likewise.
13918 * lra-constraints.c (simplify_operand_subreg): Likewise.
13919 * optabs-query.c (can_atomic_load_p): Likewise.
13920 * optabs.c (expand_atomic_load): Likewise.
13921 (expand_atomic_store): Likewise.
13922 * ree.c (combine_reaching_defs): Likewise.
13923 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13924 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13925 * tree.h (type_has_mode_precision_p): Likewise.
13926 * ubsan.c (instrument_si_overflow): Likewise.
13928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13929 Alan Hayward <alan.hayward@arm.com>
13930 David Sherwood <david.sherwood@arm.com>
13932 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13933 polynomial numbers of units.
13934 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13935 (valid_vector_subparts_p): New function.
13936 (build_vector_type): Remove temporary shim and take the number
13937 of units as a poly_uint64 rather than an int.
13938 (build_opaque_vector_type): Take the number of units as a
13939 poly_uint64 rather than an int.
13940 * tree.c (build_vector_from_ctor): Handle polynomial
13941 TYPE_VECTOR_SUBPARTS.
13942 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13943 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13944 (build_vector_from_val): If the number of units is variable,
13945 use build_vec_duplicate_cst for constant operands and
13946 VEC_DUPLICATE_EXPR otherwise.
13947 (make_vector_type): Remove temporary is_constant ().
13948 (build_vector_type, build_opaque_vector_type): Take the number of
13949 units as a poly_uint64 rather than an int.
13950 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13952 * cfgexpand.c (expand_debug_expr): Likewise.
13953 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13954 (store_constructor, expand_expr_real_1): Likewise.
13955 (const_scalar_mask_from_tree): Likewise.
13956 * fold-const-call.c (fold_const_reduction): Likewise.
13957 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13958 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13959 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13960 (fold_relational_const): Likewise.
13961 (native_interpret_vector): Likewise. Change the size from an
13962 int to an unsigned int.
13963 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13964 TYPE_VECTOR_SUBPARTS.
13965 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13966 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13967 duplicating a non-constant operand into a variable-length vector.
13968 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13969 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13970 * ipa-icf.c (sem_variable::equals): Likewise.
13971 * match.pd: Likewise.
13972 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13973 * print-tree.c (print_node): Likewise.
13974 * stor-layout.c (layout_type): Likewise.
13975 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13976 * tree-cfg.c (verify_gimple_comparison): Likewise.
13977 (verify_gimple_assign_binary): Likewise.
13978 (verify_gimple_assign_ternary): Likewise.
13979 (verify_gimple_assign_single): Likewise.
13980 * tree-pretty-print.c (dump_generic_node): Likewise.
13981 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13982 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13983 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13984 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13985 (vect_shift_permute_load_chain): Likewise.
13986 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13987 (expand_vector_condition, optimize_vector_constructor): Likewise.
13988 (lower_vec_perm, get_compute_type): Likewise.
13989 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13990 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13991 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13992 (vect_recog_mask_conversion_pattern): Likewise.
13993 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13994 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13995 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13996 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13997 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13998 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13999 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
14000 (supportable_widening_operation): Likewise.
14001 (supportable_narrowing_operation): Likewise.
14002 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
14004 * varasm.c (output_constant): Likewise.
14006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14007 Alan Hayward <alan.hayward@arm.com>
14008 David Sherwood <david.sherwood@arm.com>
14010 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14011 so that both the length == 3 and length != 3 cases set up their
14012 own permute vectors. Add comments explaining why we know the
14013 number of elements is constant.
14014 (vect_permute_load_chain): Likewise.
14016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14017 Alan Hayward <alan.hayward@arm.com>
14018 David Sherwood <david.sherwood@arm.com>
14020 * machmode.h (mode_nunits): Change from unsigned char to
14022 (ONLY_FIXED_SIZE_MODES): New macro.
14023 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14024 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14025 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14027 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14028 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14029 or if measurement_type is not polynomial.
14030 * genmodes.c (ZERO_COEFFS): New macro.
14031 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14033 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14034 Use ZERO_COEFFS when emitting initializers.
14035 * data-streamer.h (bp_pack_poly_value): New function.
14036 (bp_unpack_poly_value): Likewise.
14037 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14038 for GET_MODE_NUNITS.
14039 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14040 for GET_MODE_NUNITS.
14041 * tree.c (make_vector_type): Remove temporary shim and make
14042 the real function take the number of units as a poly_uint64
14043 rather than an int.
14044 (build_vector_type_for_mode): Handle polynomial nunits.
14045 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14046 * emit-rtl.c (const_vec_series_p_1): Likewise.
14047 (gen_rtx_CONST_VECTOR): Likewise.
14048 * fold-const.c (test_vec_duplicate_folding): Likewise.
14049 * genrecog.c (validate_pattern): Likewise.
14050 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14051 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14052 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14053 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14054 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14055 * rtlanal.c (subreg_get_info): Likewise.
14056 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14057 (vect_grouped_load_supported): Likewise.
14058 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14059 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14060 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14061 (simplify_const_unary_operation, simplify_binary_operation_1)
14062 (simplify_const_binary_operation, simplify_ternary_operation)
14063 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14064 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14065 instead of CONST_VECTOR_NUNITS.
14066 * varasm.c (output_constant_pool_2): Likewise.
14067 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14068 explicit-encoded elements in the XVEC for variable-length vectors.
14070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14072 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14075 Alan Hayward <alan.hayward@arm.com>
14076 David Sherwood <david.sherwood@arm.com>
14078 * coretypes.h (fixed_size_mode): Declare.
14079 (fixed_size_mode_pod): New typedef.
14080 * builtins.h (target_builtins::x_apply_args_mode)
14081 (target_builtins::x_apply_result_mode): Change type to
14082 fixed_size_mode_pod.
14083 * builtins.c (apply_args_size, apply_result_size, result_vector)
14084 (expand_builtin_apply_args_1, expand_builtin_apply)
14085 (expand_builtin_return): Update accordingly.
14087 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14089 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14090 * cselib.c (cselib_hash_rtx): Likewise.
14091 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14092 CONST_VECTOR encoding.
14094 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14095 Jeff Law <law@redhat.com>
14098 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14099 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14100 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14101 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14104 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14105 explicitly probe *sp in a noreturn function if there were any callee
14106 register saves or frame pointer is needed.
14108 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14111 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14112 BLKmode for ternary, binary or unary expressions.
14115 * var-tracking.c (delete_vta_debug_insn): New inline function.
14116 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14117 insns from get_insns () to NULL instead of each bb separately.
14118 Use delete_vta_debug_insn. No longer static.
14119 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14120 delete_vta_debug_insns callers.
14121 * rtl.h (delete_vta_debug_insns): Declare.
14122 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14123 instead of variable_tracking_main.
14125 2018-01-03 Martin Sebor <msebor@redhat.com>
14127 PR tree-optimization/83603
14128 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14129 arguments past the endof the argument list in functions declared
14130 without a prototype.
14131 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14132 Avoid checking when arguments are null.
14134 2018-01-03 Martin Sebor <msebor@redhat.com>
14137 * doc/extend.texi (attribute const): Fix a typo.
14138 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14139 issuing -Wsuggest-attribute for void functions.
14141 2018-01-03 Martin Sebor <msebor@redhat.com>
14143 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14144 offset_int::from instead of wide_int::to_shwi.
14145 (maybe_diag_overlap): Remove assertion.
14146 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14147 * gimple-ssa-sprintf.c (format_directive): Same.
14148 (parse_directive): Same.
14149 (sprintf_dom_walker::compute_format_length): Same.
14150 (try_substitute_return_value): Same.
14152 2018-01-03 Jeff Law <law@redhat.com>
14154 PR middle-end/83654
14155 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14156 non-constant residual for zero at runtime and avoid probing in
14157 that case. Reorganize code for trailing problem to mirror handling
14160 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14162 PR tree-optimization/83501
14163 * tree-ssa-strlen.c (get_string_cst): New.
14164 (handle_char_store): Call get_string_cst.
14166 2018-01-03 Martin Liska <mliska@suse.cz>
14168 PR tree-optimization/83593
14169 * tree-ssa-strlen.c: Include tree-cfg.h.
14170 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14171 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14172 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14174 (strlen_dom_walker::before_dom_children): Call
14175 gimple_purge_dead_eh_edges. Dump tranformation with details
14177 (strlen_dom_walker::before_dom_children): Update call by adding
14178 new argument cleanup_eh.
14179 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14181 2018-01-03 Martin Liska <mliska@suse.cz>
14184 * cif-code.def (VARIADIC_THUNK): New enum value.
14185 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14188 2018-01-03 Jan Beulich <jbeulich@suse.com>
14190 * sse.md (mov<mode>_internal): Tighten condition for when to use
14191 vmovdqu<ssescalarsize> for TI and OI modes.
14193 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14195 Update copyright years.
14197 2018-01-03 Martin Liska <mliska@suse.cz>
14200 * ipa-visibility.c (function_and_variable_visibility): Skip
14201 functions with noipa attribure.
14203 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14205 * gcc.c (process_command): Update copyright notice dates.
14206 * gcov-dump.c (print_version): Ditto.
14207 * gcov.c (print_version): Ditto.
14208 * gcov-tool.c (print_version): Ditto.
14209 * gengtype.c (create_file): Ditto.
14210 * doc/cpp.texi: Bump @copying's copyright year.
14211 * doc/cppinternals.texi: Ditto.
14212 * doc/gcc.texi: Ditto.
14213 * doc/gccint.texi: Ditto.
14214 * doc/gcov.texi: Ditto.
14215 * doc/install.texi: Ditto.
14216 * doc/invoke.texi: Ditto.
14218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14220 * vector-builder.h (vector_builder::m_full_nelts): Change from
14221 unsigned int to poly_uint64.
14222 (vector_builder::full_nelts): Update prototype accordingly.
14223 (vector_builder::new_vector): Likewise.
14224 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14225 (vector_builder::operator ==): Likewise.
14226 (vector_builder::finalize): Likewise.
14227 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14228 Take the number of elements as a poly_uint64 rather than an
14230 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14231 from unsigned int to poly_uint64.
14232 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14233 (vec_perm_indices::new_vector): Likewise.
14234 (vec_perm_indices::length): Likewise.
14235 (vec_perm_indices::nelts_per_input): Likewise.
14236 (vec_perm_indices::input_nelts): Likewise.
14237 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14238 number of elements per input as a poly_uint64 rather than an
14239 unsigned int. Use the original encoding for variable-length
14240 vectors, rather than clamping each individual element.
14241 For the second and subsequent elements in each pattern,
14242 clamp the step and base before clamping their sum.
14243 (vec_perm_indices::series_p): Handle polynomial element counts.
14244 (vec_perm_indices::all_in_range_p): Likewise.
14245 (vec_perm_indices_to_tree): Likewise.
14246 (vec_perm_indices_to_rtx): Likewise.
14247 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14248 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14249 (tree_vector_builder::new_binary_operation): Handle polynomial
14250 element counts. Return false if we need to know the number
14251 of elements at compile time.
14252 * fold-const.c (fold_vec_perm): Punt if the number of elements
14253 isn't known at compile time.
14255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14257 * vec-perm-indices.h (vec_perm_builder): Change element type
14258 from HOST_WIDE_INT to poly_int64.
14259 (vec_perm_indices::element_type): Update accordingly.
14260 (vec_perm_indices::clamp): Handle polynomial element_types.
14261 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14262 (vec_perm_indices::all_in_range_p): Likewise.
14263 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14265 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14266 polynomial vec_perm_indices element types.
14267 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14268 * fold-const.c (fold_vec_perm): Likewise.
14269 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14270 * tree-vect-generic.c (lower_vec_perm): Likewise.
14271 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14272 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14273 element type to HOST_WIDE_INT.
14275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14276 Alan Hayward <alan.hayward@arm.com>
14277 David Sherwood <david.sherwood@arm.com>
14279 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14280 rather than an int. Use plus_constant.
14281 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14282 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14285 Alan Hayward <alan.hayward@arm.com>
14286 David Sherwood <david.sherwood@arm.com>
14288 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14289 a HOST_WIDE_INT to a poly_int64.
14291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14292 Alan Hayward <alan.hayward@arm.com>
14293 David Sherwood <david.sherwood@arm.com>
14295 * calls.c (load_register_parameters): Cope with polynomial
14296 mode sizes. Require a constant size for BLKmode parameters
14297 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14298 forces a parameter to be padded at the lsb end in order to
14299 fill a complete number of words, require the parameter size
14300 to be ordered wrt UNITS_PER_WORD.
14302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14303 Alan Hayward <alan.hayward@arm.com>
14304 David Sherwood <david.sherwood@arm.com>
14306 * reload1.c (spill_stack_slot_width): Change element type
14307 from unsigned int to poly_uint64_pod.
14308 (alter_reg): Treat mode sizes as polynomial.
14310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14311 Alan Hayward <alan.hayward@arm.com>
14312 David Sherwood <david.sherwood@arm.com>
14314 * reload.c (complex_word_subreg_p): New function.
14315 (reload_inner_reg_of_subreg, push_reload): Use it.
14317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14318 Alan Hayward <alan.hayward@arm.com>
14319 David Sherwood <david.sherwood@arm.com>
14321 * lra-constraints.c (process_alt_operands): Reject matched
14322 operands whose sizes aren't ordered.
14323 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14330 that the mode size is in the set {1, 2, 4, 8, 16}.
14332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14333 Alan Hayward <alan.hayward@arm.com>
14334 David Sherwood <david.sherwood@arm.com>
14336 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14337 Use plus_constant instead of gen_rtx_PLUS.
14339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14340 Alan Hayward <alan.hayward@arm.com>
14341 David Sherwood <david.sherwood@arm.com>
14343 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14344 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14345 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14346 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14347 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14348 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14349 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14350 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14351 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14352 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14354 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14355 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14356 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14357 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14358 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14359 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14360 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14361 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14362 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14363 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14365 * expr.c (emit_move_resolve_push): Treat the input and result
14366 of PUSH_ROUNDING as a poly_int64.
14367 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14368 (emit_push_insn): Likewise.
14369 * lra-eliminations.c (mark_not_eliminable): Likewise.
14370 * recog.c (push_operand): Likewise.
14371 * reload1.c (elimination_effects): Likewise.
14372 * rtlanal.c (nonzero_bits1): Likewise.
14373 * calls.c (store_one_arg): Likewise. Require the padding to be
14374 known at compile time.
14376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14377 Alan Hayward <alan.hayward@arm.com>
14378 David Sherwood <david.sherwood@arm.com>
14380 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14381 Use plus_constant instead of gen_rtx_PLUS.
14383 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14384 Alan Hayward <alan.hayward@arm.com>
14385 David Sherwood <david.sherwood@arm.com>
14387 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14388 rather than an int.
14390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14391 Alan Hayward <alan.hayward@arm.com>
14392 David Sherwood <david.sherwood@arm.com>
14394 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14395 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14396 via stack temporaries. Treat the mode size as polynomial too.
14398 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14399 Alan Hayward <alan.hayward@arm.com>
14400 David Sherwood <david.sherwood@arm.com>
14402 * expr.c (expand_expr_real_2): When handling conversions involving
14403 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14404 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14405 as a poly_uint64 too.
14407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14408 Alan Hayward <alan.hayward@arm.com>
14409 David Sherwood <david.sherwood@arm.com>
14411 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14414 Alan Hayward <alan.hayward@arm.com>
14415 David Sherwood <david.sherwood@arm.com>
14417 * combine.c (can_change_dest_mode): Handle polynomial
14418 REGMODE_NATURAL_SIZE.
14419 * expmed.c (store_bit_field_1): Likewise.
14420 * expr.c (store_constructor): Likewise.
14421 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14422 and polynomial REGMODE_NATURAL_SIZE.
14423 (gen_lowpart_common): Likewise.
14424 * reginfo.c (record_subregs_of_mode): Likewise.
14425 * rtlanal.c (read_modify_subreg_p): Likewise.
14427 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14428 Alan Hayward <alan.hayward@arm.com>
14429 David Sherwood <david.sherwood@arm.com>
14431 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14432 numbers of elements.
14434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14435 Alan Hayward <alan.hayward@arm.com>
14436 David Sherwood <david.sherwood@arm.com>
14438 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14445 in a POINTER_PLUS_EXPR.
14447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14448 Alan Hayward <alan.hayward@arm.com>
14449 David Sherwood <david.sherwood@arm.com>
14451 * omp-simd-clone.c (simd_clone_subparts): New function.
14452 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14453 (ipa_simd_modify_function_body): Likewise.
14455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14456 Alan Hayward <alan.hayward@arm.com>
14457 David Sherwood <david.sherwood@arm.com>
14459 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14460 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14461 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14462 (expand_vector_condition, vector_element): Likewise.
14463 (subparts_gt): New function.
14464 (get_compute_type): Use subparts_gt.
14465 (count_type_subparts): Delete.
14466 (expand_vector_operations_1): Use subparts_gt instead of
14467 count_type_subparts.
14469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14470 Alan Hayward <alan.hayward@arm.com>
14471 David Sherwood <david.sherwood@arm.com>
14473 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14474 (vect_compile_time_alias): ...this new function. Do the calculation
14475 on poly_ints rather than trees.
14476 (vect_prune_runtime_alias_test_list): Update call accordingly.
14478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14479 Alan Hayward <alan.hayward@arm.com>
14480 David Sherwood <david.sherwood@arm.com>
14482 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14484 (vect_schedule_slp_instance): Likewise.
14486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14487 Alan Hayward <alan.hayward@arm.com>
14488 David Sherwood <david.sherwood@arm.com>
14490 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14491 constant and extern definitions for variable-length vectors.
14492 (vect_get_constant_vectors): Note that the number of units
14493 is known to be constant.
14495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14496 Alan Hayward <alan.hayward@arm.com>
14497 David Sherwood <david.sherwood@arm.com>
14499 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14500 of units as polynomial. Choose between WIDE and NARROW based
14503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14504 Alan Hayward <alan.hayward@arm.com>
14505 David Sherwood <david.sherwood@arm.com>
14507 * tree-vect-stmts.c (simd_clone_subparts): New function.
14508 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14511 Alan Hayward <alan.hayward@arm.com>
14512 David Sherwood <david.sherwood@arm.com>
14514 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14515 vectors as polynomial. Use build_index_vector for
14516 IFN_GOMP_SIMD_LANE.
14518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14519 Alan Hayward <alan.hayward@arm.com>
14520 David Sherwood <david.sherwood@arm.com>
14522 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14523 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14524 for variable-length vectors.
14525 (vectorizable_mask_load_store): Treat the number of units as
14526 polynomial, asserting that it is constant if the condition has
14527 already been enforced.
14528 (vectorizable_store, vectorizable_load): Likewise.
14530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14531 Alan Hayward <alan.hayward@arm.com>
14532 David Sherwood <david.sherwood@arm.com>
14534 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14535 of units as polynomial. Punt if we can't tell at compile time
14536 which vector contains the final result.
14538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14539 Alan Hayward <alan.hayward@arm.com>
14540 David Sherwood <david.sherwood@arm.com>
14542 * tree-vect-loop.c (vectorizable_induction): Treat the number
14543 of units as polynomial. Punt on SLP inductions. Use an integer
14544 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14545 cast of such a series for variable-length floating-point
14548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14549 Alan Hayward <alan.hayward@arm.com>
14550 David Sherwood <david.sherwood@arm.com>
14552 * tree.h (build_index_vector): Declare.
14553 * tree.c (build_index_vector): New function.
14554 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14555 of units as polynomial, forcibly converting it to a constant if
14556 vectorizable_reduction has already enforced the condition.
14557 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14558 to create a {1,2,3,...} vector.
14559 (vectorizable_reduction): Treat the number of units as polynomial.
14560 Choose vectype_in based on the largest scalar element size rather
14561 than the smallest number of units. Enforce the restrictions
14564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14565 Alan Hayward <alan.hayward@arm.com>
14566 David Sherwood <david.sherwood@arm.com>
14568 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14569 number of units as polynomial.
14571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14572 Alan Hayward <alan.hayward@arm.com>
14573 David Sherwood <david.sherwood@arm.com>
14575 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14576 * target.def (autovectorize_vector_sizes): Return the vector sizes
14577 by pointer, using vector_sizes rather than a bitmask.
14578 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14579 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14580 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14582 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14583 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14584 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14585 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14586 * omp-general.c (omp_max_vf): Likewise.
14587 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14588 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14589 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14590 * tree-vect-slp.c (vect_slp_bb): Likewise.
14591 * doc/tm.texi: Regenerate.
14592 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14594 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14595 the vector size as a poly_uint64 rather than an unsigned int.
14596 (current_vector_size): Change from an unsigned int to a poly_uint64.
14597 (get_vectype_for_scalar_type): Update accordingly.
14598 * tree.h (build_truth_vector_type): Take the size and number of
14599 units as a poly_uint64 rather than an unsigned int.
14600 (build_vector_type): Add a temporary overload that takes
14601 the number of units as a poly_uint64 rather than an unsigned int.
14602 * tree.c (make_vector_type): Likewise.
14603 (build_truth_vector_type): Take the number of units as a poly_uint64
14604 rather than an unsigned int.
14606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14607 Alan Hayward <alan.hayward@arm.com>
14608 David Sherwood <david.sherwood@arm.com>
14610 * target.def (get_mask_mode): Take the number of units and length
14611 as poly_uint64s rather than unsigned ints.
14612 * targhooks.h (default_get_mask_mode): Update accordingly.
14613 * targhooks.c (default_get_mask_mode): Likewise.
14614 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14615 * doc/tm.texi: Regenerate.
14617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14618 Alan Hayward <alan.hayward@arm.com>
14619 David Sherwood <david.sherwood@arm.com>
14621 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14622 * omp-general.c (omp_max_vf): Likewise.
14623 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14624 (expand_omp_simd): Handle polynomial safelen.
14625 * omp-low.c (omplow_simd_context): Add a default constructor.
14626 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14627 (lower_rec_simd_input_clauses): Update accordingly.
14628 (lower_rec_input_clauses): Likewise.
14630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14631 Alan Hayward <alan.hayward@arm.com>
14632 David Sherwood <david.sherwood@arm.com>
14634 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14635 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14636 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14637 (vect_analyze_slp_cost): Likewise.
14638 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14639 (vect_model_load_cost): Likewise.
14641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14642 Alan Hayward <alan.hayward@arm.com>
14643 David Sherwood <david.sherwood@arm.com>
14645 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14646 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14647 from an unsigned int * to a poly_uint64_pod *.
14648 (calculate_unrolling_factor): New function.
14649 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14652 Alan Hayward <alan.hayward@arm.com>
14653 David Sherwood <david.sherwood@arm.com>
14655 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14656 from an unsigned int to a poly_uint64.
14657 (_loop_vec_info::slp_unrolling_factor): Likewise.
14658 (_loop_vec_info::vectorization_factor): Change from an int
14660 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14661 (vect_get_num_vectors): New function.
14662 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14663 (vect_get_num_copies): Use vect_get_num_vectors.
14664 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14665 to an unsigned int *.
14666 (vect_analyze_data_refs): Change min_vf from an int * to a
14668 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14669 than an unsigned HOST_WIDE_INT.
14670 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14671 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14672 to an unsigned int *.
14673 (vect_analyze_data_ref_dependences): Likewise.
14674 (vect_compute_data_ref_alignment): Handle polynomial vf.
14675 (vect_enhance_data_refs_alignment): Likewise.
14676 (vect_prune_runtime_alias_test_list): Likewise.
14677 (vect_shift_permute_load_chain): Likewise.
14678 (vect_supportable_dr_alignment): Likewise.
14679 (dependence_distance_ge_vf): Take the vectorization factor as a
14680 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14681 (vect_analyze_data_refs): Change min_vf from an int * to a
14683 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14684 vfm1 as a poly_uint64 rather than an int. Make the same change
14685 for the returned bound_scalar.
14686 (vect_gen_vector_loop_niters): Handle polynomial vf.
14687 (vect_do_peeling): Likewise. Update call to
14688 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14689 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14691 * tree-vect-loop.c (vect_determine_vectorization_factor)
14692 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14693 (vect_get_known_peeling_cost): Likewise.
14694 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14695 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14696 (vect_transform_loop): Likewise. Use the lowest possible VF when
14697 updating the upper bounds of the loop.
14698 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14699 rather than an int.
14700 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14701 polynomial unroll factors.
14702 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14703 (vect_make_slp_decision): Likewise.
14704 (vect_supported_load_permutation_p): Likewise, and polynomial
14706 (vect_analyze_slp_cost): Handle polynomial vf.
14707 (vect_slp_analyze_node_operations): Likewise.
14708 (vect_slp_analyze_bb_1): Likewise.
14709 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14710 than an unsigned HOST_WIDE_INT.
14711 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14712 (vectorizable_load): Handle polynomial vf.
14713 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14715 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14718 Alan Hayward <alan.hayward@arm.com>
14719 David Sherwood <david.sherwood@arm.com>
14721 * match.pd: Handle bit operations involving three constants
14722 and try to fold one pair.
14724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14726 * tree-vect-loop-manip.c: Include gimple-fold.h.
14727 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14728 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14729 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14730 Add a path that uses a step of VF instead of 1, but disable it
14732 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14733 and niters_no_overflow parameters. Update calls to
14734 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14735 Create a new SSA name if the latter choses to use a ste other
14736 than zero, and return it via niters_vector_mult_vf_var.
14737 * tree-vect-loop.c (vect_transform_loop): Update calls to
14738 vect_do_peeling, vect_gen_vector_loop_niters and
14739 slpeel_make_loop_iterate_ntimes.
14740 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14741 (vect_gen_vector_loop_niters): Update declarations after above changes.
14743 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14745 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14746 128-bit round to integer instructions.
14747 (ceil<mode>2): Likewise.
14748 (btrunc<mode>2): Likewise.
14749 (round<mode>2): Likewise.
14751 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14753 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14754 unaligned VSX load/store on P8/P9.
14755 (expand_block_clear): Allow the use of unaligned VSX
14756 load/store on P8/P9.
14758 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14760 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14762 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14763 swap associated with both a load and a store.
14765 2018-01-02 Andrew Waterman <andrew@sifive.com>
14767 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14768 * config/riscv/riscv.md (clear_cache): Use it.
14770 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14772 * web.c: Remove out-of-date comment.
14774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14776 * expr.c (fixup_args_size_notes): Check that any existing
14777 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14778 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14779 (emit_single_push_insn): ...here.
14781 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14783 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14784 (const_vector_encoded_nelts): New function.
14785 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14786 (const_vector_int_elt, const_vector_elt): Declare.
14787 * emit-rtl.c (const_vector_int_elt_1): New function.
14788 (const_vector_elt): Likewise.
14789 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14790 of CONST_VECTOR_ELT.
14792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14794 * expr.c: Include rtx-vector-builder.h.
14795 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14796 directly on the tree encoding.
14797 (const_vector_from_tree): Likewise.
14798 * optabs.c: Include rtx-vector-builder.h.
14799 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14800 sequence of "u" values.
14801 * vec-perm-indices.c: Include rtx-vector-builder.h.
14802 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14803 directly on the vec_perm_indices encoding.
14805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14807 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14808 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14809 * rtx-vector-builder.h: New file.
14810 * rtx-vector-builder.c: Likewise.
14811 * rtl.h (rtx_def::u2): Add a const_vector field.
14812 (CONST_VECTOR_NPATTERNS): New macro.
14813 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14814 (CONST_VECTOR_DUPLICATE_P): Likewise.
14815 (CONST_VECTOR_STEPPED_P): Likewise.
14816 (CONST_VECTOR_ENCODED_ELT): Likewise.
14817 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14818 (unwrap_const_vec_duplicate): Likewise.
14819 (const_vec_series_p): Check for a non-duplicated vector encoding.
14820 Say that the function only returns true for integer vectors.
14821 * emit-rtl.c: Include rtx-vector-builder.h.
14822 (gen_const_vec_duplicate_1): Delete.
14823 (gen_const_vector): Call gen_const_vec_duplicate instead of
14824 gen_const_vec_duplicate_1.
14825 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14826 (gen_const_vec_duplicate): Use rtx_vector_builder.
14827 (gen_const_vec_series): Likewise.
14828 (gen_rtx_CONST_VECTOR): Likewise.
14829 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14830 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14831 Build a new vector rather than modifying a CONST_VECTOR in-place.
14832 (handle_special_swappables): Update call accordingly.
14833 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14834 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14835 Build a new vector rather than modifying a CONST_VECTOR in-place.
14836 (handle_special_swappables): Update call accordingly.
14838 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14840 * simplify-rtx.c (simplify_const_binary_operation): Use
14841 CONST_VECTOR_ELT instead of XVECEXP.
14843 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14845 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14846 the selector elements to be different from the data elements
14847 if the selector is a VECTOR_CST.
14848 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14849 ssizetype for the selector.
14851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14853 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14854 before testing each element individually.
14855 * tree-vect-generic.c (lower_vec_perm): Likewise.
14857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14859 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14860 * selftest-run-tests.c (selftest::run_tests): Call it.
14861 * vector-builder.h (vector_builder::operator ==): New function.
14862 (vector_builder::operator !=): Likewise.
14863 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14864 (vec_perm_indices::all_from_input_p): New function.
14865 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14866 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14867 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14868 instead of reading the VECTOR_CST directly. Detect whether both
14869 vector inputs are the same before constructing the vec_perm_indices,
14870 and update the number of inputs argument accordingly. Use the
14871 utility functions added above. Only construct sel2 if we need to.
14873 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14875 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14876 the broadcast of the low byte.
14877 (expand_mult_highpart): Use an explicit encoding for the permutes.
14878 * optabs-query.c (can_mult_highpart_p): Likewise.
14879 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14880 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14881 (vectorizable_bswap): Likewise.
14882 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14883 explicit encoding for the power-of-2 permutes.
14884 (vect_permute_store_chain): Likewise.
14885 (vect_grouped_load_supported): Likewise.
14886 (vect_permute_load_chain): Likewise.
14888 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14890 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14891 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14892 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14893 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14894 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14895 (vect_gen_perm_mask_any): Likewise.
14897 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14899 * int-vector-builder.h: New file.
14900 * vec-perm-indices.h: Include int-vector-builder.h.
14901 (vec_perm_indices): Redefine as an int_vector_builder.
14902 (auto_vec_perm_indices): Delete.
14903 (vec_perm_builder): Redefine as a stand-alone class.
14904 (vec_perm_indices::vec_perm_indices): New function.
14905 (vec_perm_indices::clamp): Likewise.
14906 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14907 (vec_perm_indices::new_vector): New function.
14908 (vec_perm_indices::new_expanded_vector): Update for new
14909 vec_perm_indices class.
14910 (vec_perm_indices::rotate_inputs): New function.
14911 (vec_perm_indices::all_in_range_p): Operate directly on the
14912 encoded form, without computing elided elements.
14913 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14914 encoding. Update for new vec_perm_indices class.
14915 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14916 the given vec_perm_builder.
14917 (expand_vec_perm_var): Update vec_perm_builder constructor.
14918 (expand_mult_highpart): Use vec_perm_builder instead of
14919 auto_vec_perm_indices.
14920 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14921 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14922 or double series encoding as appropriate.
14923 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14924 vec_perm_indices instead of auto_vec_perm_indices.
14925 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14926 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14927 (vect_permute_store_chain): Likewise.
14928 (vect_grouped_load_supported): Likewise.
14929 (vect_permute_load_chain): Likewise.
14930 (vect_shift_permute_load_chain): Likewise.
14931 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14932 (vect_transform_slp_perm_load): Likewise.
14933 (vect_schedule_slp_instance): Likewise.
14934 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14935 (vectorizable_mask_load_store): Likewise.
14936 (vectorizable_bswap): Likewise.
14937 (vectorizable_store): Likewise.
14938 (vectorizable_load): Likewise.
14939 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14940 vec_perm_indices instead of auto_vec_perm_indices. Use
14941 tree_to_vec_perm_builder to read the vector from a tree.
14942 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14943 vec_perm_builder instead of a vec_perm_indices.
14944 (have_whole_vector_shift): Use vec_perm_builder and
14945 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14946 truncation to calc_vec_perm_mask_for_shift.
14947 (vect_create_epilog_for_reduction): Likewise.
14948 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14949 from auto_vec_perm_indices to vec_perm_indices.
14950 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14951 instead of changing individual elements.
14952 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14953 the vector in d.perm.
14954 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14955 from auto_vec_perm_indices to vec_perm_indices.
14956 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14957 instead of changing individual elements.
14958 (arm_vectorize_vec_perm_const): Use new_vector to install
14959 the vector in d.perm.
14960 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14961 Update vec_perm_builder constructor.
14962 (rs6000_expand_interleave): Likewise.
14963 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14964 (rs6000_expand_interleave): Likewise.
14966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14968 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14969 to qimode could truncate the indices.
14970 * optabs.c (expand_vec_perm_var): Likewise.
14972 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14974 * Makefile.in (OBJS): Add vec-perm-indices.o.
14975 * vec-perm-indices.h: New file.
14976 * vec-perm-indices.c: Likewise.
14977 * target.h (vec_perm_indices): Replace with a forward class
14979 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14980 * optabs.h: Include vec-perm-indices.h.
14981 (expand_vec_perm): Delete.
14982 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14983 (expand_vec_perm_const): Declare.
14984 * target.def (vec_perm_const_ok): Replace with...
14985 (vec_perm_const): ...this new hook.
14986 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14987 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14988 * doc/tm.texi: Regenerate.
14989 * optabs.def (vec_perm_const): Delete.
14990 * doc/md.texi (vec_perm_const): Likewise.
14991 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14992 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14993 expand_vec_perm for constant permutation vectors. Assert that
14994 the mode of variable permutation vectors is the integer equivalent
14995 of the mode that is being permuted.
14996 * optabs-query.h (selector_fits_mode_p): Declare.
14997 * optabs-query.c: Include vec-perm-indices.h.
14998 (selector_fits_mode_p): New function.
14999 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
15000 is defined, instead of checking whether the vec_perm_const_optab
15001 exists. Use targetm.vectorize.vec_perm_const instead of
15002 targetm.vectorize.vec_perm_const_ok. Check whether the indices
15003 fit in the vector mode before using a variable permute.
15004 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
15005 vec_perm_indices instead of an rtx.
15006 (expand_vec_perm): Replace with...
15007 (expand_vec_perm_const): ...this new function. Take the selector
15008 as a vec_perm_indices rather than an rtx. Also take the mode of
15009 the selector. Update call to shift_amt_for_vec_perm_mask.
15010 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15011 Use vec_perm_indices::new_expanded_vector to expand the original
15012 selector into bytes. Check whether the indices fit in the vector
15013 mode before using a variable permute.
15014 (expand_vec_perm_var): Make global.
15015 (expand_mult_highpart): Use expand_vec_perm_const.
15016 * fold-const.c: Includes vec-perm-indices.h.
15017 * tree-ssa-forwprop.c: Likewise.
15018 * tree-vect-data-refs.c: Likewise.
15019 * tree-vect-generic.c: Likewise.
15020 * tree-vect-loop.c: Likewise.
15021 * tree-vect-slp.c: Likewise.
15022 * tree-vect-stmts.c: Likewise.
15023 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15025 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15026 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15027 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15028 (aarch64_vectorize_vec_perm_const): ...this new function.
15029 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15030 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15031 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15032 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15033 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15034 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15035 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15037 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15038 check for NEON modes.
15039 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15040 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15041 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15042 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15044 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15045 the old VEC_PERM_CONST conditions.
15046 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15047 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15048 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15049 (ia64_vectorize_vec_perm_const_ok): Merge into...
15050 (ia64_vectorize_vec_perm_const): ...this new function.
15051 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15052 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15053 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15054 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15055 * config/mips/mips.c (mips_expand_vec_perm_const)
15056 (mips_vectorize_vec_perm_const_ok): Merge into...
15057 (mips_vectorize_vec_perm_const): ...this new function.
15058 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15059 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15060 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15061 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15062 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15063 (rs6000_expand_vec_perm_const): Delete.
15064 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15066 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15067 (altivec_expand_vec_perm_const_le): Take each operand individually.
15068 Operate on constant selectors rather than rtxes.
15069 (altivec_expand_vec_perm_const): Likewise. Update call to
15070 altivec_expand_vec_perm_const_le.
15071 (rs6000_expand_vec_perm_const): Delete.
15072 (rs6000_vectorize_vec_perm_const_ok): Delete.
15073 (rs6000_vectorize_vec_perm_const): New function.
15074 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15075 an element count and rtx array.
15076 (rs6000_expand_extract_even): Update call accordingly.
15077 (rs6000_expand_interleave): Likewise.
15078 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15079 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15080 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15081 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15082 (rs6000_expand_vec_perm_const): Delete.
15083 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15084 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15085 (altivec_expand_vec_perm_const_le): Take each operand individually.
15086 Operate on constant selectors rather than rtxes.
15087 (altivec_expand_vec_perm_const): Likewise. Update call to
15088 altivec_expand_vec_perm_const_le.
15089 (rs6000_expand_vec_perm_const): Delete.
15090 (rs6000_vectorize_vec_perm_const_ok): Delete.
15091 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15092 reference to the SPE evmerge intructions.
15093 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15094 an element count and rtx array.
15095 (rs6000_expand_extract_even): Update call accordingly.
15096 (rs6000_expand_interleave): Likewise.
15097 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15098 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15100 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15102 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15104 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15105 vector mode and that that mode matches the mode of the data
15107 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15108 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15109 directly using expand_vec_perm_1 when forcing selectors into
15111 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15113 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15115 * optabs-query.h (can_vec_perm_p): Delete.
15116 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15117 * optabs-query.c (can_vec_perm_p): Split into...
15118 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15119 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15120 particular selector is valid.
15121 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15122 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15123 (vect_grouped_load_supported): Likewise.
15124 (vect_shift_permute_load_chain): Likewise.
15125 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15126 (vect_transform_slp_perm_load): Likewise.
15127 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15128 (vectorizable_bswap): Likewise.
15129 (vect_gen_perm_mask_checked): Likewise.
15130 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15131 implementations of variable permutation vectors into account
15132 when deciding which selector to use.
15133 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15134 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15135 with a false third argument.
15136 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15137 to test whether the constant selector is valid and can_vec_perm_var_p
15138 to test whether a variable selector is valid.
15140 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15142 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15143 * optabs-query.c (can_vec_perm_p): Likewise.
15144 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15145 instead of vec_perm_indices.
15146 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15147 (vect_gen_perm_mask_checked): Likewise,
15148 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15149 (vect_gen_perm_mask_checked): Likewise,
15151 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15153 * optabs-query.h (qimode_for_vec_perm): Declare.
15154 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15155 (qimode_for_vec_perm): ...this new function.
15156 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15158 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15160 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15161 does not have a conditional at the top.
15163 2018-01-02 Richard Biener <rguenther@suse.de>
15165 * ipa-inline.c (big_speedup_p): Fix expression.
15167 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15170 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15173 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15177 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15178 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15179 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15180 cond_taken_branch_cost 3->4.
15182 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15184 PR tree-optimization/83581
15185 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15186 TODO_cleanup_cfg if any changes have been made.
15188 PR middle-end/83608
15189 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15190 convert_modes if target mode has the right side, but different mode
15193 PR middle-end/83609
15194 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15195 last argument when extracting from CONCAT. If either from_real or
15196 from_imag is NULL, use expansion through memory. If result is not
15197 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15198 the parts directly to inner mode, if even that fails, use expansion
15201 PR middle-end/83623
15202 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15203 check for bswap in mode rather than HImode and use that in expand_unop
15206 Copyright (C) 2018 Free Software Foundation, Inc.
15208 Copying and distribution of this file, with or without modification,
15209 are permitted in any medium without royalty provided the copyright
15210 notice and this notice are preserved.