1 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
3 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4 bit_insertion field and declare can_be_merged_into method.
5 (merged_store_group::can_be_merged_into): New method.
6 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
7 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
8 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10 2018-06-04 Richard Biener <rguenther@suse.de>
12 PR tree-optimization/85955
13 * builtins.c (fold_builtin_sincos): Convert pointers to
14 destination to appropriate type before dereferencing.
16 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
18 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
20 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
22 * expr.c (expand_expr_real_1): Force the operand into memory if
23 its TYPE_MODE is BLKmode and if there is no integer mode for
24 the number of bits being extracted.
26 2018-06-04 Jakub Jelinek <jakub@redhat.com>
30 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
31 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
33 2018-06-04 Richard Biener <rguenther@suse.de>
35 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
36 (cleanup_tree_cfg_noloop): ... single caller. Do
37 start_recording_case_labels later.
39 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
41 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
42 to _IMMINTRIN_H_INCLUDED.
43 * config/i386/pconfigintrin.h: Ditto.
44 * config/i386/waitpkgintrin.h: Ditto.
45 * config/i386/immintrin.h: Add includes for sgxintrin.h,
46 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
47 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
48 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
49 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
50 waitpkgintrin.h and cldemoteintrin.h.
52 2018-06-04 Richard Biener <rguenther@suse.de>
54 PR tree-optimization/86038
55 * tracer.c (find_best_successor): Check probability for
56 being initialized, bail out if not.
58 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
61 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
62 of bits to ignore when comparing architectures.
64 2018-06-04 Jakub Jelinek <jakub@redhat.com>
66 PR tree-optimization/69615
67 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
68 maximum or minimum of the type, try to merge it also as if
69 range1 is + [-, x - 1] or + [x + 1, -].
72 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
74 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
76 PR tree-optimization/86034
77 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
78 the unsigned bitfield type in a bit insertion sequence if it does not
79 have a larger precision than the bitfield size.
80 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
82 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
84 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
86 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
88 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
89 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
90 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
91 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
93 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
95 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
96 Disable -fdelete-null-pointer-checks for ELF toolchain.
98 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
99 Kito Cheng <kito.cheng@gmail.com>
101 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
102 (nds32le-*-*, nds32be-*-*): Integrate checking process.
103 (nds32*-*-*): Add glibc and uclibc conditions.
104 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
105 (TARGET_EXCEPT_UNWIND_INFO): Define.
106 * config/nds32/elf.h: New file.
107 * config/nds32/linux.h: New file.
108 * config/nds32/nds32-elf.opt: New file.
109 * config/nds32/nds32-linux.opt: New file.
110 * config/nds32/nds32-fp-as-gp.c
111 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
112 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
114 (nds32_asm_file_end): Ditto.
115 (nds32_print_operand): Ditto.
116 (nds32_insert_attributes): Ditto.
117 (nds32_init_libfuncs): New function.
118 (TARGET_HAVE_TLS): Define.
119 (TARGET_INIT_LIBFUNCS): Define.
120 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
122 (TARGET_ELF): Apply different mcmodel setting.
123 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
124 been migrated into elf.h and linux.h files.
125 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
126 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
127 (mcmodel): The content has been migrated into nds32-elf.opt and
128 nds32-linux.opt files.
129 * config/nds32/t-elf: New file.
130 * config/nds32/t-linux: New file.
132 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
133 Shiva Chen <shiva0217@gmail.com>
135 * config/nds32/constants.md (unspec_volatile_element): Add
136 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
137 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
139 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
140 (make_pass_nds32_fp_as_gp): Declare.
141 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
143 (nds32_asm_function_end_prologue): Remove unused asm output.
144 (nds32_asm_function_begin_epilogue): Remove unused asm output.
145 (nds32_asm_file_start): Output necessary fp_as_gp information.
146 (nds32_option_override): Adjust register usage.
147 (nds32_expand_prologue): Consider fp_as_gp situation.
148 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
149 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
152 (simple_return): Ditto.
153 (omit_fp_begin): Output special directive for fp_as_gp.
154 (omit_fp_end): Output special directive for fp_as_gp.
155 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
156 mforbid-fp-as-gp): New options.
158 2018-06-01 Mark Wielaard <mark@klomp.org>
160 * dwarf2out.c (dwarf2out_finish): Remove generation of
163 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
165 * gimple-ssa-store-merging.c: Include gimple-fold.h.
166 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
167 (struct merged_store_group): Add bit_insertion field.
168 (dump_char_array): Use standard hexadecimal format.
169 (merged_store_group::merged_store_group): Set bit_insertion to false.
170 (merged_store_group::apply_stores): Use optimal buffer size. Deal
171 with BIT_INSERT_EXPR stores. Move up code updating the mask and
172 also print the mask in the dump file.
173 (pass_store_merging::gate): Minor tweak.
174 (imm_store_chain_info::coalesce_immediate): Fix wrong association
175 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
176 stores with INTEGER_CST stores.
177 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
178 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
179 and use it throughout. Generate bit insertion sequences if need be.
180 (pass_store_merging::process_store): Remove redundant condition.
181 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
183 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
185 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
186 the 128-bit floating point types. Fix function comment.
188 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
190 * config/aarch64/aarch64-simd.md
191 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
193 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
196 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
198 PR tree-optimization/85989
199 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
201 (backprop::intersect_uses): Check it when deciding whether this
202 is a backedge reference.
203 (backprop::process_block): Add each phi to m_visited_phis
204 after visiting it, then clear it at the end.
206 2018-06-01 Richard Biener <rguenther@suse.de>
208 * tree-vectorizer.h (vect_dr_stmt): New function.
209 (vect_get_load_cost): Adjust.
210 (vect_get_store_cost): Likewise.
211 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
212 Use vect_dr_stmt instead of DR_SMTT.
213 (vect_record_base_alignments): Likewise.
214 (vect_calculate_target_alignment): Likewise.
215 (vect_compute_data_ref_alignment): Likewise and make static.
216 (vect_update_misalignment_for_peel): Likewise.
217 (vect_verify_datarefs_alignment): Likewise.
218 (vector_alignment_reachable_p): Likewise.
219 (vect_get_data_access_cost): Likewise. Pass down
220 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
221 (vect_get_peeling_costs_all_drs): Likewise.
222 (vect_peeling_hash_get_lowest_cost): Likewise.
223 (vect_enhance_data_refs_alignment): Likewise.
224 (vect_find_same_alignment_drs): Likewise.
225 (vect_analyze_data_refs_alignment): Likewise.
226 (vect_analyze_group_access_1): Likewise.
227 (vect_analyze_group_access): Likewise.
228 (vect_analyze_data_ref_access): Likewise.
229 (vect_analyze_data_ref_accesses): Likewise.
230 (vect_vfa_segment_size): Likewise.
231 (vect_small_gap_p): Likewise.
232 (vectorizable_with_step_bound_p): Likewise.
233 (vect_prune_runtime_alias_test_list): Likewise.
234 (vect_analyze_data_refs): Likewise.
235 (vect_supportable_dr_alignment): Likewise.
236 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
237 (vect_gen_prolog_loop_niters): Likewise.
238 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
239 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
241 (vect_recog_mask_conversion_pattern): Likewise.
242 (vect_try_gather_scatter_pattern): Likewise.
243 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
244 to vect_get_store_cost.
245 (vect_get_store_cost): Get stmt_info instead of DR.
246 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
247 (vect_get_load_cost): Get stmt_info instead of DR.
249 2018-06-01 Richard Biener <rguenther@suse.de>
252 * gimple-fold.c (var_decl_component_p): Also allow offsetted
253 vars wrapped in MEM_REFs.
255 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
257 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
258 Fix subreg tests so that we only return a choice between
259 GENERAL_REGS and FP_REGS if the original classes included both.
261 2018-06-01 Richard Biener <rguenther@suse.de>
264 * tree-ssa-structalias.c (get_function_part_constraint):
265 Handle NULL fi->decl.
266 (find_func_aliases_for_call): Properly handle indirect
268 (find_func_clobbers): Likewise.
269 (ipa_pta_execute): Likewise.
270 (create_variable_info_for): For functions that are ifunc_resolver
271 resolve to a varinfo that contains the result of the resolver
273 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
276 2018-05-31 Michael Collison <michael.collison@arm.com>
278 * config/aarch64/aarch64.md:
279 (*fix_to_zero_extenddfdi2): New pattern.
280 * gcc.target/aarch64/fix_extend1.c: New testcase.
282 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
286 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
287 and BUILT_IN_STRNCMP_EQ.
288 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
290 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
291 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
292 (gimple_fold_builtin): Likewise.
293 * tree-ssa-strlen.c (compute_string_length): New function.
294 (determine_min_obsize): New function.
295 (handle_builtin_string_cmp): New function to handle calls to
296 string compare functions.
297 (strlen_optimize_stmt): Add handling to builtin string compare
299 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
300 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
301 * tree.c (build_common_builtin_nodes): Add new defines of
302 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
304 2018-05-31 Jakub Jelinek <jakub@redhat.com>
307 * bb-reorder.c (pass_partition_blocks::gate): Return false for
308 functions with naked attribute.
310 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
312 * config/i386/sse.md (avx_vec_concat<mode>):
313 Substitute concat_tg_mode mode attribute with xtg_mode.
314 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
315 (concat_tg_mode): Remove mode attribute.
317 2018-05-31 Martin Sebor <msebor@redhat.com>
320 * calls.c (alloc_max_size): Correct a logic error/typo.
321 Treat excessive arguments as infinite. Warn for invalid arguments.
322 * doc/invoke.texi (-Walloc-size-larger-than): Update.
324 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
327 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
328 and movx for Haswell.
330 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
331 Cesar Philippidis <cesar@codesourcery.com>
334 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
335 when emitting error on private/firstprivate reductions.
336 * omp-low.c (lower_omp_target): Avoid reference-type processing
337 on pointers for firstprivate clause.
339 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
341 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
344 * config/aarch64/aarch64-simd.md
345 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
346 (aarch64_ld1_x3_<mode>): Likewise
347 (aarch64_st1x2<VALLDIF:mode>): Likewise
348 (aarch64_st1_x2_<mode>): Likewise
349 (aarch64_st1x3<VALLDIF:mode>): Likewise
350 (aarch64_st1_x3_<mode>): Likewise
351 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
352 (vld1_s8_x3): Likewise.
353 (vld1_u16_x3): Likewise.
354 (vld1_s16_x3): Likewise.
355 (vld1_u32_x3): Likewise.
356 (vld1_s32_x3): Likewise.
357 (vld1_u64_x3): Likewise.
358 (vld1_s64_x3): Likewise.
359 (vld1_f16_x3): Likewise.
360 (vld1_f32_x3): Likewise.
361 (vld1_f64_x3): Likewise.
362 (vld1_p8_x3): Likewise.
363 (vld1_p16_x3): Likewise.
364 (vld1_p64_x3): Likewise.
365 (vld1q_u8_x3): Likewise.
366 (vld1q_s8_x3): Likewise.
367 (vld1q_u16_x3): Likewise.
368 (vld1q_s16_x3): Likewise.
369 (vld1q_u32_x3): Likewise.
370 (vld1q_s32_x3): Likewise.
371 (vld1q_u64_x3): Likewise.
372 (vld1q_s64_x3): Likewise.
373 (vld1q_f16_x3): Likewise.
374 (vld1q_f32_x3): Likewise.
375 (vld1q_f64_x3): Likewise.
376 (vld1q_p8_x3): Likewise.
377 (vld1q_p16_x3): Likewise.
378 (vld1q_p64_x3): Likewise.
379 (vst1_s64_x2): Likewise.
380 (vst1_u64_x2): Likewise.
381 (vst1_f64_x2): Likewise.
382 (vst1_s8_x2): Likewise.
383 (vst1_p8_x2): Likewise.
384 (vst1_s16_x2): Likewise.
385 (vst1_p16_x2): Likewise.
386 (vst1_s32_x2): Likewise.
387 (vst1_u8_x2): Likewise.
388 (vst1_u16_x2): Likewise.
389 (vst1_u32_x2): Likewise.
390 (vst1_f16_x2): Likewise.
391 (vst1_f32_x2): Likewise.
392 (vst1_p64_x2): Likewise.
393 (vst1q_s8_x2): Likewise.
394 (vst1q_p8_x2): Likewise.
395 (vst1q_s16_x2): Likewise.
396 (vst1q_p16_x2): Likewise.
397 (vst1q_s32_x2): Likewise.
398 (vst1q_s64_x2): Likewise.
399 (vst1q_u8_x2): Likewise.
400 (vst1q_u16_x2): Likewise.
401 (vst1q_u32_x2): Likewise.
402 (vst1q_u64_x2): Likewise.
403 (vst1q_f16_x2): Likewise.
404 (vst1q_f32_x2): Likewise.
405 (vst1q_f64_x2): Likewise.
406 (vst1q_p64_x2): Likewise.
407 (vst1_s64_x3): Likewise.
408 (vst1_u64_x3): Likewise.
409 (vst1_f64_x3): Likewise.
410 (vst1_s8_x3): Likewise.
411 (vst1_p8_x3): Likewise.
412 (vst1_s16_x3): Likewise.
413 (vst1_p16_x3): Likewise.
414 (vst1_s32_x3): Likewise.
415 (vst1_u8_x3): Likewise.
416 (vst1_u16_x3): Likewise.
417 (vst1_u32_x3): Likewise.
418 (vst1_f16_x3): Likewise.
419 (vst1_f32_x3): Likewise.
420 (vst1_p64_x3): Likewise.
421 (vst1q_s8_x3): Likewise.
422 (vst1q_p8_x3): Likewise.
423 (vst1q_s16_x3): Likewise.
424 (vst1q_p16_x3): Likewise.
425 (vst1q_s32_x3): Likewise.
426 (vst1q_s64_x3): Likewise.
427 (vst1q_u8_x3): Likewise.
428 (vst1q_u16_x3): Likewise.
429 (vst1q_u32_x3): Likewise.
430 (vst1q_u64_x3): Likewise.
431 (vst1q_f16_x3): Likewise.
432 (vst1q_f32_x3): Likewise.
433 (vst1q_f64_x3): Likewise.
434 (vst1q_p64_x3): Likewise.
436 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
438 * config/msp430/msp430.c (msp430_output_labelref): Prepend
439 user_label_prefix to name.
441 * tree-core.h: Update comment about the format of NAME string
442 passed to handler in attribute_spec.
444 * config/msp430/msp430.md: Remove erroneous subreg expression from
445 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
446 zero_extend{q,h}isi2.
448 2018-05-30 Borislav Petkov <bp@suse.de>
450 * doc/extend.texi: Document some architecture specific
451 constraints and sort entries.
453 2018-05-30 Martin Sebor <msebor@redhat.com>
456 * builtins.c (expand_builtin_stpcpy_1): New function.
457 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
458 only if the former succeeds.
460 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
462 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
465 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
467 * doc/invoke.texi (-flinker-output): Document
469 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
471 * passes.c (ipa_write_summaries): Only modify statements if body
473 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
474 incrementally linking.
475 (ipa_passes): Likewise.
476 * lto-cgraph.c (lto_output_node): When incrementally linking do not
477 pass down resolution info.
478 * common.opt (flag_incremental_link): Update info.
479 * gcc.c (plugin specs): Turn flinker-output=* to
480 -plugin-opt=-linker-output-known
481 * toplev.c (compile_file): Also cut compilation when doing incremental
483 * flag-types. (enum lto_partition_model): Add
484 LTO_LINKER_OUTPUT_NOLTOREL.
485 (invoke.texi): Add -flinker-output docs.
486 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
487 link same way as WPA; do not stream in dead initializers.
489 * dwarf2out.c (dwarf2out_die_ref_for_decl,
490 darf2out_register_external_decl): Support incremental link.
492 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
494 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
496 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
498 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
499 it down to simple_object_copy_lto_debug_sections.
500 (run_gcc): Determine incremental LTO link time and configure
501 lto1 into non-wpa mode, disable renaming of debug sections.
503 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
505 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
506 descriptions of various incorrectly documented functions.
508 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
511 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
512 address check not strict.
514 2018-05-30 Richard Biener <rguenther@suse.de>
516 PR tree-optimization/85964
517 * tracer.c (better_p): Drop initialized count check, we only
518 call the function with initialized counts now.
519 (find_best_successor): Do find a best edge if one
520 has uninitialized count.
521 (find_best_predecessor): Likewise. Do BB frequency check only
522 if count is initialized.
524 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
526 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
527 (aarch64_ldrstr_offset_compare): New.
528 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
529 load/store orderings.
530 (aarch64_gen_adjusted_ldpstp): Likewise.
532 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
534 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
535 Check for subset of GENERAL_REGS and FP_REGS.
536 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
539 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
541 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
542 and wi::to_poly_offset. Add the current offset and then check
543 whether the sum fits, rather than using an unchecked addition of
544 a checked term. Check for a shwi rather than a uhwi.
545 * expr.c (get_bit_range): Use tree_to_poly_uint64.
546 (store_constructor): Use poly_int_tree_p.
547 (expand_expr_real_1): Likewise.
548 * function.c (assign_temp): Likewise.
549 * fold-const.c (const_binop): Use poly_int_tree_p and
551 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
553 * ipa-icf-gimple.c (func_checker::compare_operand): Use
554 to_poly_offset for MEM offsets.
555 * ipa-icf.c (sem_variable::equals): Likewise.
556 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
557 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
558 wi::to_poly_offset for BIT_FIELD_REF offsets.
559 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
561 * var-tracking.c (emit_note_insn_var_location): Use
564 2018-05-29 Jim Wilson <jimw@sifive.com>
566 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
568 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
571 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
572 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
574 (sse4_1_round<mode>2): Use nonimmediate_operand
575 for operand 1 predicate.
577 2018-05-29 Martin Sebor <msebor@redhat.com>
578 Richard Biener <rguenther@suse.de>
581 * calls.c (get_size_range): Call determine_value_range instead
583 * tree-vrp.h (determine_value_range): Declared new function.
584 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
586 2018-05-29 Richard Biener <rguenther@suse.de>
588 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
589 sure to use non-pattern stmts for get_earlier_stmt arguments.
590 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
591 called on pattern stmts.
592 (get_later_stmt): Likewise.
594 2018-05-29 Martin Liska <mliska@suse.cz>
596 PR gcov-profile/85759
597 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
600 2018-05-29 Jakub Jelinek <jakub@redhat.com>
602 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
604 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
608 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
609 VEC_PACK_FLOAT_EXPR): New tree codes.
610 * tree-pretty-print.c (op_code_prio): Handle
611 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
612 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
613 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
614 * tree-inline.c (estimate_operator_cost): Likewise.
615 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
616 * fold-const.c (const_binop): Likewise.
617 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
618 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
619 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
620 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
621 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
622 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
623 * expr.c (expand_expr_real_2): Likewise.
624 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
625 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
626 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
628 * optabs.c (expand_widen_pattern_expr): For
629 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
630 sign from result type rather than operand's type.
631 (expand_binop_directly): For vec_packu_float_optab and
632 vec_packs_float_optab allow result type to be different from operand's
634 * optabs-tree.c (optab_for_tree_code): Handle
635 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
636 VEC_PACK_FLOAT_EXPR. Formatting fixes.
637 * tree-vect-generic.c (expand_vector_operations_1): Handle
638 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
640 * tree-vect-stmts.c (supportable_widening_operation): Handle
642 (supportable_narrowing_operation): Handle FLOAT_EXPR.
643 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
644 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
645 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
646 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
648 (vec_pack<floatprefix>_float_<mode>): New expander.
649 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
651 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
652 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
653 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
654 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
655 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
657 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
658 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
659 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
660 VEC_PACK_FLOAT_EXPR): Document.
662 2018-05-29 Richard Biener <rguenther@suse.de>
664 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
666 (stmt_vec_info_vec): Make pointer.
667 (init_stmt_vec_info_vec): Remove.
668 (free_stmt_vec_info_vec): Likewise.
669 (set_stmt_vec_info_vec): New function.
670 (free_stmt_vec_infos): Likewise.
671 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
672 (set_vinfo_for_stmt): Likewise.
673 (get_earlier_stmt): Likewise.
674 (get_later_stmt): Likewise.
675 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
676 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
677 (vec_info::~vec_info): Free stmt_vec_infos.
678 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
679 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
680 (pass_slp_vectorize::execute): Likewise.
681 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
682 (free_stmt_vec_info_vec): Likewise.
683 (set_stmt_vec_info_vec): New function.
684 (free_stmt_vec_infos): Likewise.
685 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
686 the global stmt_vec_info_vec.
687 * tree-parloops.c (gather_scalar_reductions): Use
688 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
691 2018-05-29 Richard Biener <rguenther@suse.de>
693 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
695 2018-05-29 Martin Liska <mliska@suse.cz>
696 David Malcolm <dmalcolm@redhat.com>
698 * vec.c (test_reverse): New.
699 (vec_c_tests): Add new test.
700 * vec.h (vl_ptr>::reverse): New function.
702 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
704 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
706 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
709 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
711 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
713 2018-05-28 Richard Biener <rguenther@suse.de>
715 PR tree-optimization/85933
716 * tree-vect-data-refs.c (vect_record_base_alignments): Only
717 look at stmts marked as vectorizable.
719 2018-05-28 Richard Biener <rguenther@suse.de>
721 PR tree-optimization/85934
722 * tree-vect-generic.c (expand_vector_operations_1): Hoist
723 vector boolean check before scalar optimization.
725 2018-05-28 Jakub Jelinek <jakub@redhat.com>
727 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
730 2018-05-28 Mark Wielaard <mark@klomp.org>
732 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
733 if it is an expression containing a minus sign.
735 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
737 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
739 2018-05-27 Paul Koning <ni1d@arrl.net>
741 * config/pdp11/pdp11.md (truncsihi2): Remove.
743 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
744 Chung-Ju Wu <jasonwucj@gmail.com>
746 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
748 (unaligned_store_dw): Ditto.
749 * config/nds32/nds32-memory-manipulation.c
750 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
751 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
752 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
753 (emit_setmem_word_loop): Rename to ...
754 (emit_setmem_doubleword_loop): ... this.
755 (nds32_gen_dup_4_byte_to_word_value): New function.
756 (nds32_gen_dup_8_byte_to_double_word_value): New function.
757 (nds32_expand_setmem_loop): Refine implementation.
758 (nds32_expand_setmem_loop_v3m): Ditto.
759 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
762 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
764 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
766 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
768 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
769 (nds32_init_machine_status): Initialize machine->attr_naked_p and
770 machine->attr_no_prologue_p.
771 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
772 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
773 (nds32_expand_epilogue): Consider attr_naked_p.
774 (nds32_expand_epilogue_v3pop): Likewise.
775 (nds32_can_use_return_insn): Likewise.
776 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
777 attr_no_prologue_p fields.
778 * config/nds32/nds32.opt (mret-in-naked-func): New option.
780 2018-05-27 Jakub Jelinek <jakub@redhat.com>
783 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
786 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
788 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
790 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
792 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
794 (*<floatsuffix>floatv2div2sf2): Rename to ...
795 (*float<floatunssuffix>v2div2sf2): ... this.
796 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
797 (float<floatunssuffix>v2div2sf2_mask): ... this.
798 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
799 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
800 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
802 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
804 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
806 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
808 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
810 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
812 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
813 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
814 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
815 gen_ufix_truncv8dfv8si2.
816 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
817 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
818 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
819 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
820 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
821 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
822 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
823 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
825 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
829 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
831 2018-05-25 Jim Wilson <jimw@sifive.com>
833 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
834 * config/riscv/riscv.c (struct machine_function): Add
835 interrupt_handler_p and attribute_checked_p fields.
836 (riscv_attribute_table): Add interrupt.
837 (riscv_interrupt_type_p): New.
838 (riscv_save_reg_p): Save extra regs for interrupt handler.
839 (riscv_use_save_libcall): Return false for interrupt handler.
840 (riscv_first_stack_step): Add forward declaration.
841 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
842 for interrupt handler with large frame. Use it for saved reg list.
843 (riscv_expand_prologue): Move flag_stack_usage_info support to
844 eliminate duplication.
845 (riscv_expand_epilogue): Generate mret for interrupt handler.
846 (riscv_epilogue_uses): New.
847 (riscv_can_use_return_insn): Return false for interrupt handler.
848 (riscv_function_ok_for_sibcall): Likewise.
849 (riscv_set_current_function): Add interrupt handler support.
850 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
851 * config/riscv/riscv.md (UNSPECV_MRET): New.
853 (riscv_frflags, riscv_fsflags): Use tab after opcode.
855 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
857 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
859 PR tree-optimization/85712
860 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
861 this candidate has already been replaced in-situ by a copy.
863 2018-05-25 Jason Merrill <jason@redhat.com>
865 PR c++/80485 - inline function non-zero address.
866 * symtab.c (nonzero_address): Check DECL_COMDAT.
868 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
871 * config/alpha/alpha.md (ashlsi3): New insn pattern.
872 (*ashlsi_se): Rename from *ashldi_se. Define as sign
873 extension of SImode operation. Use const123_operand predicate.
875 (*saddl_se_1): Ditto.
877 (*ssubl_se_1): Ditto.
878 * config/alpha/predicates.md (const123_operand): New predicate.
879 * config/alpha/constraints.md (P): Use IN_RANGE.
881 2018-05-25 Richard Biener <rguenther@suse.de>
883 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
885 (ref_maybe_used_by_stmt_p): Likewise.
886 (stmt_may_clobber_ref_p): Likewise.
887 (stmt_may_clobber_ref_p_1): Likewise.
888 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
890 (ref_maybe_used_by_stmt_p): Likewise.
891 (stmt_may_clobber_ref_p): Likewise.
892 (stmt_may_clobber_ref_p_1): Likewise.
893 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
894 the alias oracle to disambiguate DRs with stmts DR analysis
896 (vect_analyze_data_refs): Do not give up on not analyzable
897 DRs for BB vectorization. Remove code truncating the dataref
900 2018-05-25 Jakub Jelinek <jakub@redhat.com>
903 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
904 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
905 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
907 2018-05-25 Richard Biener <rguenther@suse.de>
909 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
910 function, combining stmt data ref gathering and fatal analysis
912 (vect_analyze_data_refs): Remove now redudnant code and simplify.
913 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
914 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
915 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
916 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
918 2018-05-25 Bin Cheng <bin.cheng@arm.com>
920 PR tree-optimization/85720
921 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
922 SCC if all partitions are builtins.
923 (version_loop_by_alias_check): New parameter. Generate cancelable
924 runtime alias check if all partitions are builtins.
925 (distribute_loop): Update call to above function.
927 2018-05-25 Bin Cheng <bin.cheng@arm.com>
929 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
930 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
931 (parm_default_def_partition_arg): Ditto.
932 (set_parm_default_def_partition): Ditto.
933 (get_parm_default_def_partitions): Ditto and make it static.
934 (get_undefined_value_partitions): Ditto and make it static.
935 (remove_ssa_form): Refactor call to init_var_map here.
936 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
937 computation for loop region.
938 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
939 (register_default_def): Delete.
940 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
941 (parm_default_def_partition_arg): Ditto.
942 (set_parm_default_def_partition): Ditto.
943 (get_parm_default_def_partitions): Ditto and make it static.
944 (get_undefined_value_partitions): Ditto and make it static.
945 (coalesce_with_default, coalesce_with_default): Update comment.
946 (create_coalesce_list_for_region): New func factored out from
947 create_outofssa_var_map.
948 (populate_coalesce_list_for_outofssa): New func factored out from
949 create_outofssa_var_map and coalesce_ssa_name.
950 (create_outofssa_var_map): Delete.
951 (coalesce_ssa_name): Refactor to support live range computation.
952 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
953 (get_parm_default_def_partitions): Delete.
954 (get_undefined_value_partitions): Ditto.
955 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
956 computation for loop region.
957 (new_tree_live_info, loe_visit_block): Ditto.
958 (live_worklist, set_var_live_on_entry): Ditto.
959 (calculate_live_on_exit, verify_live_on_entry): Ditto.
960 * tree-ssa-live.h (struct _var_map): New fields.
961 (init_var_map): Change decl.
962 (region_contains_p): New.
964 2018-05-25 Bin Cheng <bin.cheng@arm.com>
966 * tree-ssa-live.h (live_merge_and_clear): Delete.
968 2018-05-25 Richard Biener <rguenther@suse.de>
971 * tree-dump.c (dequeue_and_dump): Remove access to removed
972 operand 2 of a SWITCH_EXPR.
974 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
976 * doc/sourcebuild.texi (vect_double_cond_arith): Include
977 multiplication and division.
978 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
979 (cond_udiv@var{m}, cond_umod@var{m}): Document.
980 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
981 (cond_udiv_optab, cond_umod_optab): New optabs.
982 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
983 (IFN_COND_RDIV): New internal functions.
984 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
985 TRUNC_MOD_EXPR and RDIV_EXPR.
986 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
987 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
989 (SVE_INT_BINARY): Include mult.
990 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
991 (optab, sve_int_op): Handle mult.
992 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
994 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
995 for SVE_INT_BINARY_SD.
997 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
999 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1000 (optab, sve_int_op): Handle div and udiv.
1001 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1002 for SVE_INT_BINARY_SD.
1003 (*<optab><mode>3): New insn for the same.
1005 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1007 * tree-vect-patterns.c: Include predict.h.
1008 (vect_recog_divmod_pattern): Restrict check for division support
1009 to when optimizing for size.
1011 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1013 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1014 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1015 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1016 (gimple_match_op::set_op): Likewise.
1017 (gimple_resimplify4): Declare.
1018 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1019 (expr::gen_transform): Likewise.
1020 (decision_tree::gen): Generate a simplification routine for 4 operands.
1021 * gimple-match-head.c (gimple_simplify): Add an overload for
1022 4 operands. In the top-level function, handle up to 4 call
1023 arguments and call gimple_resimplify4.
1024 (gimple_resimplify4): New function.
1025 (build_call_internal): Pass a fourth operand.
1026 (maybe_push_to_seq): Likewise.
1027 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1028 Fold VEC_COND_EXPRs of an operation and a default value into
1029 an IFN_COND_* function if possible.
1030 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1032 (SVE_COND_FP_BINARY): Include them.
1033 (optab, sve_fp_op): Handle them.
1034 (SVE_INT_BINARY_REV): New code iterator.
1035 (SVE_COND_FP_BINARY_REV): New int iterator.
1036 (commutative): New int attribute.
1037 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1039 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1041 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1042 (*cond_<optab><mode>): New patterns for reversed operands.
1044 2018-05-25 Richard Biener <rguenther@suse.de>
1046 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1047 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1048 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1049 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1050 * tree-vect-data-refs.c (everywhere): Adjust users.
1051 * tree-vect-loop.c (everywhere): Likewise.
1052 * tree-vect-slp.c (everywhere): Likewise.
1053 * tree-vect-stmts.c (everywhere): Likewise.
1054 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1056 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1058 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1060 (gcc_cv_as_section_exclude): ... this.
1061 Try Solaris as #exclude syntax.
1062 * configure: Regenerate.
1063 * config.in: Regenerate.
1064 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1066 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1067 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1069 * varasm.c (default_elf_asm_named_section): Don't check if
1070 HAVE_GAS_SECTION_EXCLUDE is defined.
1072 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1074 * doc/md.texi: Update the documentation of the cond_* optabs
1075 to mention the new final operand. Fix GET_MODE_NUNITS call.
1076 Describe the scalar case too.
1077 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1078 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1080 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1081 (get_conditional_internal_fn): Update comment.
1082 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1083 accumulator value as a final argument to conditional functions.
1084 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1085 a define_expand and add an "else" operand. Assert for now that
1086 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1087 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1088 (*cond_<optab><mode>): New patterns.
1089 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1090 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1091 (UNSPEC_COND_EOR): Delete.
1092 (optab): Remove associated mappings.
1093 (SVE_INT_BINARY): New code iterator.
1094 (sve_int_op): Remove int attribute and add "minus" to the code
1096 (SVE_COND_INT_OP): Delete.
1097 (SVE_COND_FP_OP): Rename to...
1098 (SVE_COND_FP_BINARY): ...this.
1100 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1102 * optabs.c (can_reuse_operands_p): New function.
1103 (maybe_legitimize_operands): Try to reuse the results for
1106 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1108 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1109 Add {q} suffix to insn mnemonic.
1111 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1113 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1114 (msp430_warn_func_return): New.
1116 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1118 * fold-const.c (tree_nonzero_bits): New function.
1119 * fold-const.h (tree_nonzero_bits): Likewise.
1120 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1121 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1123 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1127 * varasm.c (assemble_alias): Check ifunc_resolver only on
1130 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1133 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1134 when memory input operand is handled.
1136 2018-05-24 Luis Machado <luis.machado@linaro.org>
1138 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1140 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1142 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1144 * match.pd: Delay FMA folds until after vectorization.
1146 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1149 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1150 address check not strict.
1152 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1154 * gimple-match.h (gimple_match_op): New class.
1155 (mprts_hook): Replace parameters with a gimple_match_op *.
1156 (maybe_build_generic_op): Likewise.
1157 (gimple_simplified_result_is_gimple_val): Replace parameters with
1158 a const gimple_match_op *.
1159 (gimple_simplify): Replace code_helper * and tree * parameters with
1160 a gimple_match_op * parameter.
1161 (gimple_resimplify1): Replace code_helper *, tree and tree *
1162 parameters with a gimple_match_op * parameter.
1163 (gimple_resimplify2): Likewise.
1164 (gimple_resimplify3): Likewise.
1165 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1166 parameters with a gimple_match_op * parameter.
1167 * gimple-match-head.c (gimple_simplify): Change prototypes of
1168 auto-generated functions to take a gimple_match_op * instead of
1169 separate code_helper * and tree * parameters. Make the same
1170 change in the top-level overload and update calls to the
1171 gimple_resimplify routines. Update calls to the auto-generated
1172 functions and to maybe_push_res_to_seq in the publicly-facing
1173 operation-specific gimple_simplify overloads.
1174 (gimple_match_op::MAX_NUM_OPS): Define.
1175 (gimple_resimplify1): Replace rcode and ops with a single res_op
1176 parameter. Update call to gimple_simplify.
1177 (gimple_resimplify2): Likewise.
1178 (gimple_resimplify3): Likewise.
1179 (mprts_hook): Replace parameters with a gimple_match_op *.
1180 (maybe_build_generic_op): Likewise.
1181 (build_call_internal): Replace type, nargs and ops with
1182 a gimple_match_op *.
1183 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1184 with a single gimple_match_op *. Update calls to mprts_hook,
1185 build_call_internal and gimple_simplified_result_is_gimple_val.
1186 Factor out code that is common to the tree_code and combined_fn cases.
1187 * genmatch.c (expr::gen_transform): Replace tem_code and
1188 tem_ops with a gimple_match_op called tem_op. Update calls
1189 to the gimple_resimplify functions and maybe_push_res_to_seq.
1190 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1191 res_ops. Update call to the gimple_resimplify functions.
1192 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1193 (decision_tree::gen): Make the functions take a gimple_match_op *
1194 called res_op instead of separate res_code and res_ops parameters.
1195 Update call accordingly.
1196 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1197 and ops with a single res_op parameter. Update calls to
1198 maybe_build_generic_op and maybe_push_res_to_seq.
1199 (fold_stmt_1): Update calls to gimple_simplify and
1200 replace_stmt_with_simplification.
1201 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1202 and gimple_simplified_result_is_gimple_val.
1203 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1205 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1206 with a gimple_match_op *.
1207 (vn_nary_build_or_lookup): Likewise. Update call to
1208 vn_nary_build_or_lookup_1.
1209 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1210 gimple_match_op *. Update calls to the gimple_resimplify routines
1211 and to gimple_simplified_result_is_gimple_val.
1212 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1213 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1214 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1215 (visit_nary_op): Likewise.
1216 (visit_reference_op_load): Likewise.
1218 2018-05-23 Luis Machado <luis.machado@linaro.org>
1220 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1221 modifier for printing the step amount.
1223 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1226 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1229 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1231 * doc/sourcebuild.texi (Endianness): New subsubsection.
1233 2018-05-23 Luis Machado <luis.machado@linaro.org>
1235 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1236 <prefetch_dynamic_strides>: New const bool field.
1237 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1238 prefetch_dynamic_strides.
1239 (exynosm1_prefetch_tune): Likewise.
1240 (thunderxt88_prefetch_tune): Likewise.
1241 (thunderx_prefetch_tune): Likewise.
1242 (thunderx2t99_prefetch_tune): Likewise.
1243 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1245 (aarch64_override_options_internal): Update to set
1246 PARAM_PREFETCH_DYNAMIC_STRIDES.
1247 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1248 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1249 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1250 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1251 prefetch-dynamic-strides setting.
1253 2018-05-23 Luis Machado <luis.machado@linaro.org>
1255 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1256 <minimum_stride>: New const int field.
1257 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1258 minimum_stride field defaulting to -1.
1259 (exynosm1_prefetch_tune): Likewise.
1260 (thunderxt88_prefetch_tune): Likewise.
1261 (thunderx_prefetch_tune): Likewise.
1262 (thunderx2t99_prefetch_tune): Likewise.
1263 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1264 <default_opt_level>: Set to 3.
1265 (aarch64_override_options_internal): Update to set
1266 PARAM_PREFETCH_MINIMUM_STRIDE.
1267 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1268 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1269 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1270 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1271 stride is constant and is below the minimum stride threshold.
1273 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1275 * config/arm/arm-cpus.in (mode26): Delete.
1276 (armv4): Delete mode26 reference.
1277 * config/arm/arm.c (arm_configure_build_target): Delete use of
1280 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1282 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1284 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1285 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1287 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1288 (floatunsdidf2): Ditto.
1290 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1292 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1293 (fixuns_trunc<mode>si2_avx512f): Ditto.
1294 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1295 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1296 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1298 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1300 PR rtl-optimization/79985
1301 * df-scan.c (df_insn_refs_collect): Remove special case for
1302 global registers and asm statements.
1304 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1306 * extend.texi (Global Register Variables): Rewrite the bullet list.
1307 Note that the register is available for allocation. Note that access
1308 via inline asm must use constraints. Add note about async-signal
1309 handlers. Remove paragraph about automagic register selection.
1311 2018-05-23 Richard Biener <rguenther@suse.de>
1313 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1314 of fixed offset from memset VN.
1316 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1318 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1320 (alloc_cand_and_find_basis): Initialize first_interp field.
1321 (slsr_process_mul): Modify first_interp field.
1322 (slsr_process_add): Likewise.
1323 (slsr_process_cast): Modify first_interp field for each new
1325 (slsr_process_copy): Likewise.
1326 (dump_candidate): Dump first_interp field.
1327 (replace_mult_candidate): Process all interpretations, not just
1329 (replace_rhs_if_not_dup): Likewise.
1330 (replace_one_candidate): Likewise.
1332 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1334 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1336 (aarch64_needs_frame_chain): New function.
1337 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1339 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1342 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1343 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1344 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1345 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1346 as true for strict-align.
1347 (aarch64_can_inline_p): Perform checks even when callee has no
1348 attributes to check for strict alignment.
1349 * doc/extend.texi (AArch64 Function Attributes): Document
1351 * doc/invoke.texi: (AArch64 Options): Likewise.
1353 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1355 PR tree-optimization/85853
1356 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1357 the handling of the root of the node to...
1358 (vect_slp_analyze_node_operations_1): ...this new function,
1359 and run the whole thing with the child nodes' def types
1360 set according to their SLP node's def type.
1362 2018-05-23 Richard Biener <rguenther@suse.de>
1365 * tree-data-ref.c (create_runtime_alias_checks): Defer
1366 and ignore overflow warnings.
1368 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1370 PR tree-optimization/85822
1371 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1374 2018-05-23 Richard Biener <rguenther@suse.de>
1376 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1377 memset constants via native_interpret_expr.
1379 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1382 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1384 (cgraph_node::create_alias): Likewise.
1385 (cgraph_node::get_availability): Check ifunc_resolver instead
1386 of looking up ifunc attribute.
1387 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1388 * varasm.c (do_assemble_alias): Likewise.
1389 (assemble_alias): Likewise.
1390 (default_binds_local_p_3): Likewise.
1391 * cgraph.h (cgraph_node): Add ifunc_resolver.
1392 (cgraph_node::only_called_directly_or_aliased_p): Return false
1394 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1396 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1397 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1398 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1399 instead of looking up ifunc attribute.
1401 2018-05-22 Luis Machado <luis.machado@linaro.org>
1403 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1405 2018-05-22 Martin Sebor <msebor@redhat.com>
1408 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1409 only when expasion succeeds.
1410 (expand_builtin_strcmp): Same.
1411 (expand_builtin_strncmp): Same.
1413 2018-05-22 Martin Sebor <msebor@redhat.com>
1415 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1417 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1418 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1420 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1421 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1423 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1424 Add check for legitimate_address.
1425 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1426 (aarch64_swap_ldrstr_operands): New.
1427 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1430 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1431 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1433 * config/aarch64/aarch64.md: New patterns to generate stp
1435 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1436 single words and double words.
1437 (load_pair_sw, load_pair_dw): Likewise.
1438 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1440 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1442 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1443 for different mode ldpstp and add peephole for merged zero stores.
1445 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1447 (aarch64_gen_store_pair): Rename calls to match new patterns.
1448 (aarch64_gen_load_pair): Rename calls to match new patterns.
1449 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1450 (load_pair<DREG:mode><DREG2:mode>): ... This.
1451 (store_pair<mode>): Rename to...
1452 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1453 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1455 (V_INT_EQUIV): Handle SImode.
1456 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1459 2018-05-22 Martin Sebor <msebor@redhat.com>
1462 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1463 or ajust the presumed bound on an operation to avoid unnecessary
1466 2018-05-22 Martin Sebor <msebor@redhat.com>
1468 PR tree-optimization/85826
1469 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1470 assuming that a DECL necesarily has a constant size.
1472 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1475 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1477 2018-05-22 Richard Biener <rguenther@suse.de>
1479 PR tree-optimization/85834
1480 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1481 non-constant and non-zero memset arguments.
1483 2018-05-22 Martin Liska <mliska@suse.cz>
1486 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1488 2018-05-22 Richard Biener <rguenther@suse.de>
1490 PR tree-optimization/85863
1491 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1492 comparisons when vectype is specified.
1493 (vectorizable_condition): Do not specify vectype for
1494 vect_is_simple_cond when SLP vectorizing.
1496 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1499 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1500 define __ibm128 as long double.
1501 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1502 as a distinct type when IEEE 128-bit support is enabled.
1503 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1504 128-bit types to use the correct functions.
1505 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1506 convert between 128-bit floating point types that have different
1507 modes but the same representation, instead of using gen_lowpart to
1509 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1511 (IFKF_reg): New attributes to give the register constraints for
1513 (extend<mode>tf2_internal): New insns to mark an explicit
1514 conversion between 128-bit floating point types that have a
1515 different mode but share the same representation.
1517 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1519 PR tree-optimization/85814
1520 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1521 a null return from get_strinfo when unsharing the next
1522 strinfo in the chain.
1524 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1527 * varasm.c (weak_finish): Clean up weak_decls.
1529 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1531 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1532 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1533 UNSPEC_UADALP values.
1534 * config/aarch64/iterators.md (ABAL): New int iterator.
1537 (sur): Add mappings for the above.
1538 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1540 (aarch64_<sur>abal<mode>_4): Likewise.
1541 (aarch64_<sur>adalp<mode>_3): Likewise.
1542 (<sur>sadv16qi): New define_expand.
1544 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1546 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1547 (*movdf_internal): Ditto.
1548 (*rcpsf2_sse): Ditto.
1549 (*rsqrtsf2_sse): Ditto.
1550 (*sqrt<mode>2_sse): Ditto.
1552 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1554 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1556 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1557 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1558 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1559 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1561 * config/aarch64/arm_neon.h: Likewise.
1562 * config/aarch64/iterators.md (VQ_I): New.
1564 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1566 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1568 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1571 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1573 * config/nds32/constraints.md (S): New constraint.
1574 * config/nds32/nds32.md (call_internal): Use constraint S.
1575 (call_value_internal): Likewise.
1576 (sibcall_internal): Likewise.
1577 (sibcall_value_internal): Likewise.
1579 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1580 Chung-Ju Wu <jasonwucj@gmail.com>
1582 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1585 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1586 Chung-Ju Wu <jasonwucj@gmail.com>
1588 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1589 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1590 (nds32_rtx_costs_impl): Simplify.
1591 (nds32_address_cost_impl): Simplify.
1592 (nds32_init_rtx_costs): New function.
1593 (nds32_rtx_costs_speed_prefer): Likewise.
1594 (nds32_rtx_costs_size_prefer): Likewise.
1595 (nds32_address_cost_speed_prefer): Likewise.
1596 (nds32_address_cost_speed_fwprop): Likewise.
1597 (nds32_address_cost_size_prefer): Likewise.
1598 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1599 * config/nds32/nds32.c (nds32_option_override): Use
1600 nds32_init_rtx_costs function.
1602 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1604 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1605 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1606 (TARGET_PIPELINE_N8): Likewise.
1607 (TARGET_PIPELINE_N10): Likewise.
1608 (TARGET_PIPELINE_N13): Likewise.
1609 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1611 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1613 * config/nds32/nds32-fpu.md: Update copyright year.
1615 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1617 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1619 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1621 * config/nds32/nds32.c
1622 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1623 * config/nds32/nds32.opt (minline-asm-r15): New option.
1625 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1627 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1629 * config/nds32/nds32.md (abssi2): New pattern.
1631 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1633 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1634 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1635 Merge insn pattern from sse_cvtsi2ss<round_name> and
1636 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1637 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1638 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1639 using SWI48 mode iterator.
1640 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1641 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1642 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1643 pattern from sse_cvttss2si<round_saeonly_name>
1644 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1645 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1646 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1647 using SWI48 mode iterator.
1648 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1649 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1650 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1651 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1652 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1653 using SWI48 mode iterator.
1654 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1655 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1656 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1657 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1658 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1659 SWI48 mode iterator.
1660 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1661 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1662 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1663 pattern from sse_cvttsd2si<round_saeonly_name>
1664 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1666 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1668 * config/nds32/nds32-md-auxiliary.c
1669 (nds32_valid_smw_lwm_base_p): Refine.
1670 (nds32_output_smw_single_word): Refine.
1671 (nds32_output_smw_double_word): New.
1672 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1674 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1676 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1677 (nds32_output_stack_pop): Refine.
1678 (nds32_expand_unaligned_load): Refine.
1679 (nds32_expand_unaligned_store): Refine.
1681 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1682 Chung-Ju Wu <jasonwucj@gmail.com>
1684 * config/nds32/constants.md: Add TP_REGNUM constant.
1685 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1686 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1688 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1689 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1690 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1691 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1692 and PIC code generation.
1693 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1695 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1697 * config/nds32/nds32.md: Support TLS and PIC.
1698 * config/nds32/nds32.c: Support TLS and PIC.
1699 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1700 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1703 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1705 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1706 mode with E_ prefix.
1708 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1709 Chung-Ju Wu <jasonwucj@gmail.com>
1711 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1712 * config/nds32/nds32-md-auxiliary.c
1713 (symbolic_reference_mentioned_p): New.
1714 (nds32_legitimize_ict_address): New.
1715 (nds32_expand_ict_move): New.
1716 (nds32_indirect_call_referenced_p): New.
1717 (nds32_symbol_binds_local_p): Delete.
1718 (nds32_long_call_p): Modify.
1719 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1720 * config/nds32/nds32-protos.h
1721 (symbolic_reference_mentioned_p): Declare.
1722 (nds32_legitimize_ict_address): Declare.
1723 (nds32_expand_ict_move): Declare.
1724 (nds32_indirect_call_referenced_p): Declare.
1725 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1726 (nds32_relax_group): Use nds32_ict_const_p as condition.
1727 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1728 (nds32_asm_file_start): Output ict_model directive in asm code.
1729 (nds32_legitimate_address_p): Consider indirect call.
1730 (nds32_print_operand): Consider indirect call.
1731 (nds32_print_operand_address): Consider indirect call.
1732 (nds32_insert_attributes): Handle "indirect_call" attribute.
1733 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1734 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1735 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1736 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1737 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1738 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1739 (TARGET_ICT_MODEL_SMALL): Define.
1740 (TARGET_ICT_MODEL_LARGE): Define.
1741 * config/nds32/nds32.md (movsi): Consider ict model.
1742 (call, call_value): Consider ict model.
1743 (sibcall, sibcall_value): Consider ict model.
1744 * config/nds32/nds32.opt (mict-model): New option.
1745 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1748 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1749 Monk Chiang <sh.chiang04@gmail.com>
1750 Jim Wilson <jimw@sifive.com>
1752 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1753 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1754 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1755 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1756 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1757 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1758 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1759 compute save_libcall_adjustment properly.
1760 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1761 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1762 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1763 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1764 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1765 (ABI_SPEC): Handle mabi=ilp32e.
1766 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1767 (RVE): Add RVE mask.
1768 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1769 <-march>: Add rv32e as an example.
1771 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1774 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1775 (intra_create_variable_infos): Handle C++ constructors.
1777 2018-05-18 Martin Liska <mliska@suse.cz>
1779 * passes.def: Remove a redundant pass.
1781 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1784 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1786 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1788 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1790 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1792 (armv2, armv2a, armv3, armv3m): Delete architectures.
1793 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1794 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1795 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1797 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1798 (*mulsidi3adddi): Likewise.
1799 (mulsidi3): Likewise.
1800 (*mulsidi3_nov6): Likewise.
1801 (umulsidi3): Likewise.
1802 (umulsidi3_nov6): Likewise.
1803 (umaddsidi4): Likewise.
1804 (*umulsidi3adddi): Likewise.
1805 (smulsi3_highpart): Likewise.
1806 (*smulsi3_highpart_nov6): Likewise.
1807 (umulsi3_highpart): Likewise.
1808 (*umulsi3_highpart_nov6): Likewise.
1809 * config/arm/arm.h (arm_arch3m): Delete.
1810 * config/arm/arm.c (arm_arch3m): Delete.
1811 (arm_option_override_internal): Update armv3-related comment.
1812 (arm_configure_build_target): Delete use of isa_bit_mode32.
1813 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1814 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1815 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1817 (mulusa3): Likewise.
1818 * config/arm/arm-protos.h (arm_arch3m): Delete.
1819 * config/arm/arm-tables.opt: Regenerate.
1820 * config/arm/arm-tune.md: Likewise.
1821 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1822 deleted architectures.
1824 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1826 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1827 (armv5t, armv5te): New features.
1828 (ARMv5, ARMv5e): Delete fgroups.
1829 (ARMv5t, ARMv5te): Adjust for above changes.
1831 (armv5, armv5e): Delete arches.
1832 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1834 (*call_reg_arm): Likewise.
1835 (*call_value_reg_armv5): Likewise.
1836 (*call_value_reg_arm): Likewise.
1837 (*call_symbol): Likewise.
1838 (*call_value_symbol): Likewise.
1839 (*sibcall_insn): Likewise.
1840 (*sibcall_value_insn): Likewise.
1842 (prefetch): Likewise.
1843 (define_split and define_peephole2 dependent on arm_arch5):
1845 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1847 (TARGET_ARM_QBIT): Likewise.
1848 (TARGET_DSP_MULTIPLY): Likewise.
1849 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1850 (arm_arch5, arm_arch5e): Delete.
1851 (arm_arch5t, arm_arch5te): Declare.
1852 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1853 (arm_arch5t): Declare.
1854 (arm_option_reconfigure_globals): Update for the above.
1855 (arm_options_perform_arch_sanity_checks): Update comment, replace
1856 use of arm_arch5 with arm_arch5t.
1857 (use_return_insn): Likewise.
1858 (arm_emit_call_insn): Likewise.
1859 (output_return_instruction): Likewise.
1860 (arm_final_prescan_insn): Likewise.
1861 (arm_coproc_builtin_available): Likewise.
1862 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1863 arm_arch5e with arm_arch5t and arm_arch5te.
1864 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1865 (arm_arch5t, arm_arch5te): Declare.
1866 * config/arm/arm-tables.opt: Regenerate.
1867 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1868 * config/arm/t-multilib: Likewise.
1869 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1870 instead of arm_arch5.
1871 (*call_reg_thumb1): Likewise.
1872 (*call_value_reg_thumb1_v5): Likewise.
1873 (*call_value_reg_thumb1): Likewise.
1874 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1876 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1878 2018-05-18 Martin Liska <mliska@suse.cz>
1880 PR gcov-profile/84846
1881 * doc/gcov.texi: Document -t option of gcov tool.
1883 2018-05-18 Martin Liska <mliska@suse.cz>
1885 PR gcov-profile/84846
1886 * gcov.c (print_usage): Add new -t option.
1887 (process_args): Handle the option.
1888 (generate_results): Use stdout as output when requested by
1891 2018-05-18 Martin Liska <mliska@suse.cz>
1893 PR gcov-profile/84846
1894 * coverage.c (coverage_init): Write PWD to .gcno file.
1895 * doc/gcov.texi: Document how working directory is printed.
1896 * gcov-dump.c (dump_gcov_file): Print PWD.
1897 * gcov.c (output_intermediate_file): Likewise.
1898 (read_graph_file): Read PWD string.
1899 (output_lines): Print PWD.
1901 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1904 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1905 for retval and return false if all args to phi are zero.
1907 2018-05-18 Richard Biener <rguenther@suse.de>
1909 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1911 (evrp_dom_walker::before_dom_children): Call it.
1913 2018-05-18 Richard Biener <rguenther@suse.de>
1915 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1916 results when processing array refs with variable index.
1918 2018-05-18 Toon Moene <toon@moene.org>
1920 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1921 directly after that of -floop-interchange. Indicate that both
1922 options are enabled by default when specifying -O3.
1924 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1926 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1927 iterator. Delete separate integer-mode vec_set<mode> expander.
1928 (aarch64_simd_vec_setv2di): Delete.
1929 (vec_setv2di): Delete.
1930 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1931 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1932 the "w, r" alternative.
1934 2018-05-18 Martin Liska <mliska@suse.cz>
1936 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1937 * tree-pass.h (make_pass_lower_switch_O0): New function.
1938 * tree-switch-conversion.c (node_has_low_bound): Remove.
1939 (node_has_high_bound): Likewise.
1940 (node_is_bounded): Likewise.
1941 (class pass_lower_switch): Make it a template type and create
1943 (pass_lower_switch::execute): Add template argument.
1944 (make_pass_lower_switch): New function.
1945 (make_pass_lower_switch_O0): New function.
1946 (do_jump_if_equal): Remove.
1947 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1948 all the hard work to tree optimization passes.
1950 2018-05-18 Martin Liska <mliska@suse.cz>
1952 * dbgcnt.c (limit_low): Renamed from limit.
1953 (limit_high): New variable.
1954 (dbg_cnt_is_enabled): Check for upper limit.
1955 (dbg_cnt): Adjust dumping.
1956 (dbg_cnt_set_limit_by_index): Add new argument for high
1958 (dbg_cnt_set_limit_by_name): Likewise.
1959 (dbg_cnt_process_single_pair): Parse new format.
1960 (dbg_cnt_process_opt): Use strtok.
1961 (dbg_cnt_list_all_counters): Remove 'value' and add
1963 * doc/invoke.texi: Document changes.
1965 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1967 * doc/sourcebuild.texi (scalar_all_fma): Document.
1968 * tree.def (FMA_EXPR): Delete.
1969 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1970 * internal-fn.c (ternary_direct): New macro.
1971 (expand_ternary_optab_fn): Likewise.
1972 (direct_ternary_optab_supported_p): Likewise.
1973 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1974 * builtins.c (fold_builtin_fma): Delete.
1975 (fold_builtin_3): Don't call it.
1976 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1977 * expr.c (expand_expr_real_2): Likewise.
1978 * fold-const.c (operand_equal_p): Likewise.
1979 (fold_ternary_loc): Likewise.
1980 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1981 * gimple.c (DEFTREECODE): Likewise.
1982 * gimplify.c (gimplify_expr): Likewise.
1983 * optabs-tree.c (optab_for_tree_code): Likewise.
1984 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1985 * tree-eh.c (operation_could_trap_p): Likewise.
1986 (stmt_could_throw_1_p): Likewise.
1987 * tree-inline.c (estimate_operator_cost): Likewise.
1988 * tree-pretty-print.c (dump_generic_node): Likewise.
1989 (op_code_prio): Likewise.
1990 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1991 * tree-ssa-operands.c (get_expr_operands): Likewise.
1992 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1993 * fold-const-call.h (fold_fma): Delete.
1994 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1995 CFN_FNMA and CFN_FNMS.
1997 * genmatch.c (combined_fn): New enum.
1998 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1999 (commutative_op): New function.
2000 (commutate): Use it. Handle more than 2 operands.
2001 (dt_operand::gen_gimple_expr): Use commutative_op.
2002 (parser::parse_expr): Allow :c to be used with non-binary
2003 operators if the commutative operand is known.
2004 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2005 CFN_FMS, CFN_FNMA and CFN_FNMS.
2006 (backprop::process_assign_use): Remove FMA_EXPR handling.
2007 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2008 (gen_hsa_fma): New function.
2009 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2010 IFN_FNMA and IFN_FNMS.
2011 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2012 * gimple-fold.h (follow_all_ssa_edges): Declare.
2013 * gimple-fold.c (follow_all_ssa_edges): New function.
2014 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2015 gimple_build interface and use follow_all_ssa_edges to fold the result.
2016 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2017 instead of checking for optabs directly.
2018 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2019 rather than FMA_EXPRs.
2020 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2021 call to IFN_FMA instead of an FMA_EXPR.
2023 2018-05-17 Jim Wilson <jimw@sifive.com>
2025 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2026 sign extended promoted subreg, and the range does not have the sign bit
2027 set, then do a sign extend.
2029 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2030 test, check for sign extended subreg and/or constant operands, and
2031 do a sign extend in that case.
2033 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2035 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2036 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2038 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2039 Change logics_shift_reg to logics_shift_imm.
2040 (thunderx2t99_fp_loadpair_basic): Delete.
2041 (thunderx2t99_fp_storepair_basic): Delete.
2042 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2043 (thunderx2t99_asimd_polynomial): Delete.
2044 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2045 and neon_fp_mul_d_scalar_q.
2046 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2047 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2048 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2049 (thunderx2t99_asimd_lut): Add missing tbl types.
2050 (thunderx2t99_asimd_ext): Delete.
2051 (thunderx2t99_asimd_load1_1_mult): Delete.
2052 (thunderx2t99_asimd_load1_2_mult): Delete.
2053 (thunderx2t99_asimd_load1_ldp): New.
2054 (thunderx2t99_asimd_load1): New.
2055 (thunderx2t99_asimd_load2): Add missing *load2* types.
2056 (thunderx2t99_asimd_load3): New.
2057 (thunderx2t99_asimd_load4): New.
2058 (thunderx2t99_asimd_store1_1_mult): Delete.
2059 (thunderx2t99_asimd_store1_2_mult): Delete.
2060 (thunderx2t99_asimd_store2_mult): Delete.
2061 (thunderx2t99_asimd_store2_onelane): Delete.
2062 (thunderx2t99_asimd_store_stp): New.
2063 (thunderx2t99_asimd_store1): New.
2064 (thunderx2t99_asimd_store2): New.
2065 (thunderx2t99_asimd_store3): New.
2066 (thunderx2t99_asimd_store4): New.
2068 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2070 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2071 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2073 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2074 Segher Boessenkool <segher@kernel.crashing.org>
2077 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2080 2018-05-17 Richard Biener <rguenther@suse.de>
2082 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2083 for pruning loop and prune defs feeding only already visited PHIs.
2085 2018-05-17 Richard Biener <rguenther@suse.de>
2087 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2089 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2090 Richard Biener <rguenther@suse.de>
2092 PR tree-optimization/85793
2093 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2094 for VMAT_ELEMENTWISE.
2096 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2098 * internal-fn.h (lookup_internal_fn): Declare
2099 * internal-fn.c (lookup_internal_fn): New function.
2100 * gimple.c (gimple_build_call_from_tree): Handle calls to
2102 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2103 internal function names.
2104 * tree-pretty-print.c (dump_generic_node): Likewise.
2105 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2107 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2109 * gimple-fold.h (gimple_build): Make the function forms take
2110 combined_fn rather than built_in_function.
2111 (gimple_simplify): Likewise.
2112 * gimple-match-head.c (gimple_simplify): Likewise.
2113 * gimple-fold.c (gimple_build): Likewise.
2114 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2115 rather than gimple_build_call_internal.
2116 (get_initial_defs_for_reduction): Likewise.
2117 (vect_create_epilog_for_reduction): Likewise.
2118 (vectorizable_live_operation): Likewise.
2120 2018-05-17 Martin Liska <mliska@suse.cz>
2122 * gimple-ssa-sprintf.c (format_directive): Do not use
2123 space in between 'G_' and '('.
2125 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2128 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2129 even if the mask is not all ones.
2132 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2134 (ix86_gimple_fold_builtin): Likewise.
2137 * config/i386/i386.c: Include tree-vector-builder.h.
2138 (ix86_vector_shift_count): New function.
2139 (ix86_fold_builtin): Fold shift builtins by scalar count.
2140 (ix86_gimple_fold_builtin): Likewise.
2142 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2143 _mm512_setzero): New intrinsics.
2145 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2146 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2148 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2149 code generation for cases where splatting a value is not useful.
2150 * simplify-rtx.c (simplify_ternary_operation): Simplify
2151 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2152 mode to a vec_concat.
2154 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2156 * config.gcc: Support "goldmont-plus".
2157 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2159 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2160 PROCESSOR_GOLDMONT_PLUS.
2161 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2162 (processor_target_table): Add "goldmont-plus".
2163 (PTA_GOLDMONT_PLUS): Define.
2164 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2165 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2166 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2167 (fold_builtin_cpu): Add "goldmont-plus".
2168 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2169 (ix86_option_override_internal): Add "goldmont-plus".
2170 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2171 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2172 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2173 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2175 2018-05-17 Richard Biener <rguenther@suse.de>
2177 PR tree-optimization/85757
2178 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2179 remove defs that only feed that PHI from further processing.
2181 2018-05-16 Jim Wilson <jimw@sifive.com>
2183 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2185 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2187 2018-05-16 Mark Wielaard <mark@klomp.org>
2189 * dwarf2out.c (count_index_strings): New function.
2190 (output_indirect_strings): Call count_index_strings and generate
2191 header for dwarf_version >= 5.
2193 2018-05-16 Mark Wielaard <mark@klomp.org>
2195 * dwarf2out.c (dwarf_FORM): New function.
2196 (set_indirect_string): Use dwarf_FORM.
2197 (reset_indirect_string): Likewise.
2198 (size_of_die): Likewise.
2199 (value_format): Likewise.
2200 (output_die): Likewise.
2201 (add_skeleton_AT_string): Likewise.
2202 (output_macinfo_op): Likewise.
2203 (index_string): Likewise.
2204 (output_index_string_offset): Likewise.
2205 (output_index_string): Likewise.
2206 (count_index_strings): Likewise.
2208 2018-05-16 Carl Love <cel@us.ibm.com>
2210 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2211 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2213 2018-05-16 Martin Jambor <mjambor@suse.cz>
2215 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2216 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2218 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2220 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2221 (fnma<mode>4): Likewise.
2222 (fms<mode>4): Likewise.
2223 (fnms<mode>4): Likewise.
2224 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2225 (aarch64_fnma<mode>4): Likewise.
2226 (aarch64_fms<mode>4): Likewise.
2227 (aarch64_fnms<mode>4): Likewise.
2228 (aarch64_fnmadd<mode>4): Likewise.
2230 2018-05-16 Jason Merrill <jason@redhat.com>
2232 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2234 2018-05-16 Richard Biener <rguenther@suse.de>
2236 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2237 (dump_stmt_cost): Declare.
2238 (add_stmt_cost): Dump cost we add.
2239 (add_stmt_costs): New function.
2240 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2242 (vect_analyze_stmt): Adjust prototype.
2243 (vectorizable_condition): Likewise.
2244 (vectorizable_live_operation): Likewise.
2245 (vectorizable_reduction): Likewise.
2246 (vectorizable_induction): Likewise.
2247 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2248 cost vector to pass to vectorizable_ and record afterwards.
2249 (vect_model_reduction_cost): Take cost vector argument and adjust.
2250 (vect_model_induction_cost): Likewise.
2251 (vectorizable_reduction): Likewise.
2252 (vectorizable_induction): Likewise.
2253 (vectorizable_live_operation): Likewise.
2254 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2255 SLP_TREE_NUMBER_OF_VEC_STMTS.
2256 (vect_analyze_slp_cost_1): Remove.
2257 (vect_analyze_slp_cost): Likewise.
2258 (vect_slp_analyze_node_operations): Take visited args and
2259 a target cost vector. Avoid processing already visited stmt sets.
2260 (vect_slp_analyze_operations): Use a local cost vector to gather
2261 costs and register those of non-discarded instances.
2262 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2263 (vect_schedule_slp_instance): Remove copying of
2264 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2266 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2267 adding cost. Record cost entry location.
2268 (vect_prologue_cost_for_slp_op): Function to compute cost of
2269 a constant or invariant generated for SLP vect in the prologue,
2270 split out from vect_analyze_slp_cost_1.
2271 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2272 (vect_model_promotion_demotion_cost): Likewise.
2273 (vect_model_store_cost): Likewise, make static.
2274 (vect_model_load_cost): Likewise.
2275 (vectorizable_bswap): Add cost vector arg and adjust.
2276 (vectorizable_call): Likewise.
2277 (vectorizable_simd_clone_call): Likewise.
2278 (vectorizable_conversion): Likewise.
2279 (vectorizable_assignment): Likewise.
2280 (vectorizable_shift): Likewise.
2281 (vectorizable_operation): Likewise.
2282 (vectorizable_store): Likewise.
2283 (vectorizable_load): Likewise.
2284 (vectorizable_condition): Likewise.
2285 (vectorizable_comparison): Likewise.
2286 (can_vectorize_live_stmts): Likewise.
2287 (vect_analyze_stmt): Likewise.
2288 (vect_transform_stmt): Adjust calls to vectorizable_*.
2289 * tree-vectorizer.c: Include gimple-pretty-print.h.
2290 (dump_stmt_cost): New function.
2292 2018-05-16 Richard Biener <rguenther@suse.de>
2294 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2295 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2296 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2297 (check_name): New callback.
2298 (dse_classify_store): Track cycles via a visited bitmap of PHI
2299 defs and simplify handling of in-loop and across loop dead stores
2300 and properly fail for loop-variant refs. Handle byte-tracking with
2301 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2304 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2306 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2307 (vect_get_mask_type_for_stmt): Likewise.
2308 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2310 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2311 to determine the statement's vector type and the vector type that
2312 should be used for calculating nunits. Deal with cases in which
2313 the type has to be deferred.
2314 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2315 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2316 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2317 (vect_determine_vf_for_stmt): New functions, split out from...
2318 (vect_determine_vectorization_factor): ...here.
2319 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2320 (vect_get_mask_type_for_stmt): New functions, split out from
2321 vect_determine_vectorization_factor.
2323 2018-05-16 Richard Biener <rguenther@suse.de>
2325 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2326 verify the [VEC_]COND_EXPR embedded comparison.
2328 2018-05-15 Martin Sebor <msebor@redhat.com>
2330 PR tree-optimization/85753
2331 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2332 RECORD_TYPE in addition to ARRAY_TYPE.
2334 2018-05-15 Martin Sebor <msebor@redhat.com>
2337 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2339 2018-05-15 Richard Biener <rguenther@suse.de>
2341 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2342 add by_clobber_p one. Change algorithm to collect all defs
2343 representing uses we need to walk and try reducing them to
2344 a single one before failing.
2345 (dse_dom_walker::dse_optimize_stmt): Adjust.
2347 2018-05-13 Mark Wielaard <mark@klomp.org>
2349 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2350 (size_of_loc_descr): Likewise.
2351 (output_loc_operands): Likewise.
2352 (output_loc_operands_raw): Likewise.
2353 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2354 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2355 (hash_loc_operands): Likewise.
2356 (compare_loc_operands): Likewise.
2358 2018-05-14 Mark Wielaard <mark@klomp.org>
2360 * dwarf2out.c (count_index_addrs): New function.
2361 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2363 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2365 PR tree-optimization/83648
2366 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2367 return value as malloc candidate.
2369 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2372 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2373 as true in call to suggest_attribute.
2375 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2377 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2378 -mreadonly-in-sdata.
2380 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2382 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2384 (aarch64_crypto_aesd_fused): Likewise.
2386 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2388 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2389 (movsi_aarch64): Likewise.
2390 (load_pairsi): Likewise.
2391 (load_pairdi): Likewise.
2392 (store_pairsi): Likewise.
2393 (store_pairdi): Likewise.
2394 (load_pairsf): Likewise.
2395 (load_pairdf): Likewise.
2396 (store_pairsf): Likewise.
2397 (store_pairdf): Likewise.
2398 (zero_extend): Likewise.
2399 (trunc): Swap alternatives.
2400 (fcvt_target): Add '?' to prefer w over r.
2402 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2405 * config/i386/i386.md: Disallow non-commutative arithmetics in
2406 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2407 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2408 in the peephole2 before it.
2410 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2412 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2413 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2414 (ix86_handle_option): Handle -mcldemote.
2415 * config.gcc: New header.
2416 * config/i386/cldemoteintrin.h: New file.
2417 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2418 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2420 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2421 OPTION_MASK_ISA_CLDEMOTE.
2422 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2423 (ix86_valid_target_attribute_inner_p): Ditto.
2424 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2425 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2426 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2427 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2428 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2430 * config/i386/i386.opt: Add -mcldemote.
2431 * config/i386/x86intrin.h: New header.
2432 * doc/invoke.texi: Add -mcldemote.
2434 2018-05-14 Richard Biener <rguenther@suse.de>
2436 * doc/match-and-simplify.texi: Adjust :s documentation.
2438 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2440 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2441 intended memcpy size.
2442 (REORDER_45): Likewise.
2444 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2446 * sort.cc: New file.
2447 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2448 * vec.c (qsort_chk): Use gcc_qsort.
2449 * Makefile.in (OBJS-libcommon): Add sort.o.
2450 (build/sort.o): New target. Use it...
2451 (BUILD_RTL): ... here, and...
2452 (build/gencfn-macros): ... here, and...
2453 (build/genmatch): ... here.
2455 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2456 Chung-Ju Wu <jasonwucj@gmail.com>
2458 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2459 * config/nds32/nds32-graywolf.md: New file.
2460 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2461 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2463 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2464 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2465 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2466 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2467 * config/nds32/pipelines.md: Include n15 settings.
2469 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2470 Chung-Ju Wu <jasonwucj@gmail.com>
2472 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2473 * config/nds32/nds32-n13.md: New file.
2474 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2475 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2477 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2478 * config/nds32/nds32.md (pipeline_model): Add n13.
2479 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2480 * config/nds32/pipelines.md: Include n13 settings.
2482 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2483 Chung-Ju Wu <jasonwucj@gmail.com>
2485 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2486 * config/nds32/nds32-n10.md: New file.
2487 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2488 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2490 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2491 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2492 * config/nds32/nds32.md (pipeline_model): Add n10.
2493 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2494 * config/nds32/pipelines.md: Include n10 settings.
2496 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2497 Kito Cheng <kito.cheng@gmail.com>
2498 Chung-Ju Wu <jasonwucj@gmail.com>
2500 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2501 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2502 Add enum values for DSP extension instructions.
2503 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2505 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2506 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2508 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2509 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2510 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2511 * config/nds32/nds32-intrinsic.md: Likewise.
2512 * config/nds32/nds32_intrinsic.h: Likewise.
2513 * config/nds32/nds32-md-auxiliary.c: Likewise.
2514 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2515 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2516 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2517 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2518 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2519 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2520 TYPE_DMAC in switch statement.
2521 * config/nds32/nds32.c: New checking and implementation for DSP
2522 extension instructions.
2523 * config/nds32/nds32.h: Likewise.
2524 * config/nds32/nds32.md: Likewise.
2525 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2526 * config/nds32/predicates.md: Implement new predicates for DSP
2529 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2531 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2532 Reformat alternatives and attributes so it is easier to identify
2533 which constraints/attributes go with which instruction.
2534 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2535 (mov<mode>_softfloat32, FMOVE64): Likewise.
2536 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2537 (mov<mode>_softfloat64, FMOVE64): Likewise.
2539 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2541 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2543 (Basic PowerPC Built-in Functions): The new name of the
2544 subsection previously known as "PowerPC Built-in Functions".
2545 (Basic PowerPC Built-in Functions Available on all Configurations):
2547 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2548 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2549 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2550 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2552 2018-05-11 Martin Jambor <mjambor@suse.cz>
2555 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2558 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2561 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2563 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2565 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2566 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2567 (ix86_handle_option): Handle -mwaitpkg.
2568 * config.gcc: New header.
2569 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2570 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2571 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2573 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2574 OPTION_MASK_ISA_WAITPKG.
2575 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2576 (ix86_option_override_internal): Add PTA_WAITPKG.
2577 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2578 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2579 IX86_BUILTIN_TPAUSE.
2580 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2581 __builtin_ia32_umwait and __builtin_ia32_tpause.
2582 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2583 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2584 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2585 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2586 UNSPECV_TPAUSE): New.
2587 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2588 * config/i386/i386.opt: Add -mwaitpkg.
2589 * config/i386/waitpkgintrin.h: New file.
2590 * config/i386/x86intrin.h: New header.
2591 * doc/invoke.texi: Add -mwaitpkg.
2593 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2596 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2598 (cortex-m0): Use armv6s-m isa.
2599 (cortex-m0plus): Likewise.
2600 (cortex-m1): Likewise.
2601 (cortex-m0.small-multiply): Likewise.
2602 (cortex-m0plus.small-multiply): Likewise.
2603 (cortex-m1.small-multiply): Likewise.
2605 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2606 Jakub Jelinek <jakub@redhat.com>
2608 PR tree-optimization/85692
2609 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2610 source permute as well.
2612 2018-05-11 Martin Liska <mliska@suse.cz>
2615 * doc/extend.texi: Document LLVM style format for no_sanitize
2618 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2620 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2621 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2622 (mode_supports_vsx_dform_quad): Likewise.
2623 (mode_supports_vmx_dform): Move these functions to be next to the
2624 other mode_supports functions.
2625 (mode_supports_dq_form): Likewise.
2626 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2627 mode_supports_dq_form.
2628 (reg_offset_addressing_ok_p): Likewise.
2629 (offsettable_ok_by_alignment): Likewise.
2630 (rs6000_legitimate_offset_address_p): Likewise.
2631 (legitimate_lo_sum_address_p): Likewise.
2632 (rs6000_legitimize_address): Likewise.
2633 (rs6000_legitimize_reload_address): Likewise.
2634 (rs6000_secondary_reload_inner): Likewise.
2635 (rs6000_preferred_reload_class): Likewise.
2636 (rs6000_output_move_128bit): Likewise.
2638 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2640 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2641 Generate SImode target register for null target.
2642 <case IX86_BUILTIN_XGETBV>: Ditto.
2643 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2644 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2646 2018-05-10 Carl Love <cel@us.ibm.com>
2648 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2649 dcbtt and dcbtstt if operands[2] is 0.
2651 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2654 * config/i386/sse.md (usadv64qi): New expander.
2656 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2658 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2659 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2660 -maltivec=be support.
2661 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2662 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2663 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2664 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2665 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2666 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2667 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2668 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2669 altivec_vsumsws): Adjust.
2670 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2671 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2672 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2674 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2675 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2676 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2677 (altivec_lve<VI_char>x): Delete expand.
2678 (*altivec_lve<VI_char>x_internal): Rename to...
2679 (altivec_lve<VI_char>x): ... this.
2680 (altivec_lvxl_<mode>): Delete expand.
2681 (*altivec_lvxl_<mode>_internal): Rename to ...
2682 (altivec_lvxl_<mode>): ... this.
2683 (altivec_stvxl_<mode>): Delete expand.
2684 (*altivec_stvxl_<mode>_internal): Rename to ...
2685 (altivec_stvxl_<mode>): ... this.
2686 (altivec_stve<VI_char>x): Delete expand.
2687 (*altivec_stve<VI_char>x_internal): Rename to ...
2688 (altivec_stve<VI_char>x): ... this.
2689 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2690 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2691 reduc_plus_scal_<mode>): Adjust.
2692 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2694 (rs6000_cpu_cpp_builtins): Adjust.
2695 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2696 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2697 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2698 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2699 -maltivec=be support.
2700 (rs6000_split_vec_extract_var): Adjust.
2701 (rs6000_split_v4si_init): Adjust.
2702 (swap_selector_for_mode): Delete.
2703 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2704 altivec_expand_stvex_be): Delete.
2705 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2706 -maltivec=be support.
2707 (rs6000_gimple_fold_builtin): Ditto.
2708 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2710 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2711 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2712 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2713 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2714 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2715 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2716 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2717 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2718 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2719 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2720 anonymous split): Adjust.
2721 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2722 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2724 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2726 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2727 when --with-gxx-include-dir is also specified.
2728 * configure: Regenerate.
2730 2018-05-09 Jim Wilson <jimw@sifive.com>
2733 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2734 * config/riscv/t-withmultilib: New.
2735 * config/riscv/withmultilib.h: New.
2736 * doc/install.texi: Document RISC-V --with-multilib-list support.
2738 2018-05-09 Richard Biener <rguenther@suse.de>
2740 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2742 (vect_bb_vectorization_profitable_p): Adjust. Compute
2743 actual scalar cost using the cost vector and the add_stmt_cost
2746 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2748 PR rtl-optimization/85645
2749 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2750 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2752 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2754 PR rtl-optimization/85645
2755 * shrink-wrap.c (spread_components): Return a boolean saying if
2756 anything was changed.
2757 (try_shrink_wrapping_separate): Iterate spread_components until
2758 nothing changes anymore.
2760 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2762 PR rtl-optimization/85645
2763 * regrename.c (build_def_use): Also kill the chains that include the
2764 destination of a REG_CFA_REGISTER note.
2766 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2768 PR rtl-optimization/85645
2769 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2770 insn that has a REG_CFA_REGISTER note.
2772 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2774 * cfgexpand.c (expand_clobber): New function.
2775 (expand_gimple_stmt_1): Use it.
2776 * tree-vect-stmts.c (vect_clobber_variable): New function,
2778 (vectorizable_simd_clone_call): ...here.
2779 (vectorizable_store): Emit a clobber either side of an
2780 IFN_STORE_LANES sequence.
2781 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2783 2018-05-09 Tom de Vries <tom@codesourcery.com>
2786 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2787 (define_insn "trap_if_false"): Add exit after trap.
2789 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2791 PR rtl-optimization/85638
2792 * bb-reorder.c: Include common/common-target.h.
2793 (create_forwarder_block): New function extracted from...
2794 (fix_up_crossing_landing_pad): ...here. Rename into...
2795 (dw2_fix_up_crossing_landing_pad): ...this.
2796 (sjlj_fix_up_crossing_landing_pad): New function.
2797 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2798 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2799 from both partitions and exit the loop after one iteration.
2801 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2804 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2806 (Basic PowerPC Built-in Functions): The new name of the
2807 subsection previously known as "PowerPC Built-in Functions".
2808 (Basic PowerPC Built-in Functions Available on all Configurations):
2810 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2812 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2814 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2816 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2819 2018-05-08 Jim Wilson <jimw@sifive.com>
2821 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2822 (LD_EMUL_SUFFIX): New.
2823 (LINK_SPEC): Use it.
2825 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2827 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2829 (Basic PowerPC Built-in Functions): The new name of the
2830 subsection previously known as "PowerPC Built-in Functions".
2831 (Basic PowerPC Built-in Functions Available on all Configurations):
2833 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2835 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2837 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2839 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2842 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2845 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2846 after cmpelim optimization.
2848 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2850 * config.gcc: Support "goldmont".
2851 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2852 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2854 * config/i386/i386.c (m_GOLDMONT): Define.
2855 (processor_target_table): Add "goldmont".
2856 (PTA_GOLDMONT): Define.
2857 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2858 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2859 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2860 (fold_builtin_cpu): Add "goldmont".
2861 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2862 (ix86_option_override_internal): Add "goldmont".
2863 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2864 (processor_type): Add PROCESSOR_GOLDMONT.
2865 * config/i386/i386.md: Add CPU "glm".
2866 * config/i386/glm.md: New file.
2867 * config/i386/x86-tune.def: Add m_GOLDMONT.
2868 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2870 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2873 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2875 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2876 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2877 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2880 * config/i386/i386.c (ix86_fold_builtin): Handle
2881 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2884 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2885 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2887 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2890 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2891 (check_arch): Likewise.
2892 (check_fpu): Return the result rather than printing it.
2893 (end arch): Fix operator precedence.
2894 (end cpu): Likewise.
2895 (END): Print the result from check_fpu.
2897 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2898 Alan Hayward <alan.hayward@arm.com>
2899 David Sherwood <david.sherwood@arm.com>
2901 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2902 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2903 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2904 (*fcmuo<mode>_and): New patterns.
2906 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2908 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2909 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2910 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2911 (cmp_op, sve_imm_con): New code attributes.
2912 (SVE_COND_INT_CMP, imm_con): Delete.
2913 (cmp_op): Remove above unspecs from int attribute.
2914 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2916 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2917 comparison-specific unspecs.
2918 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2919 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2920 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2921 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2922 (*vec_fcm<cmp_op><mode>): Rename to...
2923 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2924 (*vec_fcmuo<mode>): Rename to...
2925 (*fcmuo<mode>): ...this and adjust likewise.
2926 (*pred_fcm<cmp_op><mode>): New pattern.
2927 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2928 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2930 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2932 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2933 (aarch64_emit_sve_predicated_cond): New function.
2934 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2935 (aarch64_emit_unspec_cond_or): Replace with...
2936 (aarch64_emit_sve_or_conds): ...this new function. Use
2937 aarch64_emit_sve_ptrue_op for the individual comparisons and
2938 aarch64_emit_binop to OR them together.
2939 (aarch64_emit_inverted_unspec_cond): Replace with...
2940 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2941 aarch64_emit_sve_ptrue_op for the comparison and
2942 aarch64_emit_unop to invert the result.
2943 (aarch64_expand_sve_vec_cmp_float): Update after the above
2944 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2946 2018-05-07 Nathan Sidwell <nathan@acm.org>
2948 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2949 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2950 (Backwards Compatibility): Likewise.
2952 2018-05-07 Luis Machado <luis.machado@linaro.org>
2956 2018-05-07 Luis Machado <luis.machado@linaro.org>
2958 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2959 <prefetch_dynamic_strides>: New const bool field.
2960 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2961 prefetch_dynamic_strides.
2962 (exynosm1_prefetch_tune): Likewise.
2963 (thunderxt88_prefetch_tune): Likewise.
2964 (thunderx_prefetch_tune): Likewise.
2965 (thunderx2t99_prefetch_tune): Likewise.
2966 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2967 (aarch64_override_options_internal): Update to set
2968 PARAM_PREFETCH_DYNAMIC_STRIDES.
2969 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2970 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2971 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2972 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2973 prefetch-dynamic-strides setting.
2975 2018-05-07 Luis Machado <luis.machado@linaro.org>
2977 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2978 <minimum_stride>: New const int field.
2979 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2980 minimum_stride field.
2981 (exynosm1_prefetch_tune): Likewise.
2982 (thunderxt88_prefetch_tune): Likewise.
2983 (thunderx_prefetch_tune): Likewise.
2984 (thunderx2t99_prefetch_tune): Likewise.
2985 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2986 (aarch64_override_options_internal): Update to set
2987 PARAM_PREFETCH_MINIMUM_STRIDE.
2988 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2989 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2990 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2991 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2992 stride is constant and is below the minimum stride threshold.
2994 2018-05-07 Luis Machado <luis.machado@linaro.org>
2996 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2999 2018-05-07 Luis Machado <luis.machado@linaro.org>
3001 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3002 <prefetch_dynamic_strides>: New const bool field.
3003 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3004 prefetch_dynamic_strides.
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 prefetch_dynamic_strides to false.
3010 (aarch64_override_options_internal): Update to set
3011 PARAM_PREFETCH_DYNAMIC_STRIDES.
3012 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3013 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3014 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3015 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3016 prefetch-dynamic-strides setting.
3018 2018-05-07 Luis Machado <luis.machado@linaro.org>
3020 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3021 <minimum_stride>: New const int field.
3022 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3023 minimum_stride field.
3024 (exynosm1_prefetch_tune): Likewise.
3025 (thunderxt88_prefetch_tune): Likewise.
3026 (thunderx_prefetch_tune): Likewise.
3027 (thunderx2t99_prefetch_tune): Likewise.
3028 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3029 (aarch64_override_options_internal): Update to set
3030 PARAM_PREFETCH_MINIMUM_STRIDE.
3031 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3032 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3033 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3034 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3035 stride is constant and is below the minimum stride threshold.
3037 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3040 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3041 the type is addressable. Don't force op into register if it has
3044 2018-05-05 Roland McGrath <mcgrathr@google.com>
3047 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3048 any section for which we don't know a specific type it should have,
3049 regardless of name. Previously this was done only for the exact
3050 names ".init_array", ".fini_array", and ".preinit_array".
3051 (default_elf_asm_named_section): Add comment about
3052 relationship with default_section_type_flags and SECTION_NOTYPE.
3053 (get_section): Don't consider it a type conflict if one side has
3054 SECTION_NOTYPE and the other doesn't, as long as neither has the
3055 SECTION_BSS et al used in the default_section_type_flags logic.
3057 2018-05-05 Tom de Vries <tom@codesourcery.com>
3060 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3061 (workaround_barsyncs): New function.
3062 (nvptx_reorg): Use workaround_barsyncs.
3063 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3064 (define_expand "nvptx_membar_cta"): New define_expand.
3065 (define_insn "*nvptx_membar_cta"): New insn.
3067 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3069 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3070 To improve optimization opportunities.
3071 * builtin-types.def: The new needed builtin types for the above.
3073 2018-05-04 Richard Biener <rguenther@suse.de>
3075 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3076 * gimple-ssa-store-merging.c
3077 (imm_store_chain_info::output_merged_store): Remove redundant create,
3078 release split_store vector contents on failure.
3079 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3080 scalar stmt vector on cache hit.
3082 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3084 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3086 * config.gcc (powerpc-xilinx-eabi*): Remove.
3087 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3089 (fusion_addis_mem_combo_load): Ditto.
3090 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3092 (rs6000_cpu_cpp_builtins): Ditto.
3093 * config/rs6000/rs6000-linux.c
3094 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3095 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3096 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3098 (rs6000_setup_reg_addr_masks): Ditto.
3099 (rs6000_init_hard_regno_mode_ok): Ditto.
3100 (rs6000_option_override_internal): Ditto.
3101 (legitimate_lo_sum_address_p): Ditto.
3102 (rs6000_legitimize_address): Ditto.
3103 (rs6000_legitimize_reload_address): Ditto.
3104 (rs6000_legitimate_address_p): Ditto.
3105 (abi_v4_pass_in_fpr): Ditto.
3106 (setup_incoming_varargs): Ditto.
3107 (rs6000_gimplify_va_arg): Ditto.
3108 (rs6000_split_multireg_move): Ditto.
3109 (rs6000_savres_strategy): Ditto.
3110 (rs6000_emit_prologue_components): Ditto.
3111 (rs6000_emit_epilogue_components): Ditto.
3112 (rs6000_emit_prologue): Ditto.
3113 (rs6000_emit_epilogue): Ditto.
3114 (rs6000_elf_file_end): Ditto.
3115 (rs6000_function_value): Ditto.
3116 (rs6000_libcall_value): Ditto.
3117 * config/rs6000/rs6000.h: Ditto.
3118 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3119 (TARGET_MINMAX): ... this. New.
3120 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3121 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3122 (*movsi_internal1_single): Delete.
3123 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3124 mfpu=, mxilinx-fpu): Delete.
3125 * config/rs6000/singlefp.h: Delete.
3126 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3127 * config/rs6000/t-rs6000: Ditto.
3128 * config/rs6000/t-xilinx: Delete.
3129 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3130 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3131 (VStype_simple): Delete.
3132 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3133 * config/rs6000/xfpu.h: Delete.
3134 * config/rs6000/xfpu.md: Delete.
3135 * config/rs6000/xilinx.h: Delete.
3136 * config/rs6000/xilinx.opt: Delete.
3137 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3138 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3140 2018-05-04 Tom de Vries <tom@codesourcery.com>
3143 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3146 2018-05-04 Richard Biener <rguenther@suse.de>
3149 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3150 (expand_complex_div_wide): Likewise.
3151 (expand_complex_operations_1): Likewise.
3152 (expand_complex_libcall): Preserve EH info of the original stmt.
3153 (tree_lower_complex): Handle removed blocks.
3154 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3155 on complex multiplication and division libcall builtins.
3157 2018-05-04 Richard Biener <rguenther@suse.de>
3160 * fold-const.c (negate_expr_p): Restrict negation of operand
3161 zero of a division to when we know that can happen without
3163 (fold_negate_expr_1): Likewise.
3165 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3168 * real.h (real_nextafter): Declare.
3169 * real.c (real_nextafter): New function.
3170 * fold-const-call.c (fold_const_nextafter): New function.
3171 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3172 CASE_CFN_NEXTTOWARD.
3173 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3174 even when arg1_mode is different from arg0_mode.
3176 2018-05-03 Nathan Sidwell <nathan@acm.org>
3178 * doc/extend.texi (Deprecated Features): Remove
3180 (Backwards Compatibility): Likewise.
3181 * doc/invoke.texi (C++ Language Options): Likewise.
3182 (C++ Dialect Options): Likewise.
3184 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3187 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3188 _mm512_mask_mullox_epi64): New intrinsics.
3190 2018-05-03 Tom de Vries <tom@codesourcery.com>
3193 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3194 dump files): Add offload-tree.
3196 2018-05-03 Richard Biener <rguenther@suse.de>
3198 PR tree-optimization/85615
3199 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3200 to loops not nested in BBs loop father to avoid creating multi-entry
3203 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3205 PR tree-optimization/70291
3206 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3207 arguments. Change return type to tree. Emit libcall as a new
3208 statement rather than replacing existing one when inplace_p is true.
3209 (expand_complex_multiplication_components): New function.
3210 (expand_complex_multiplication): Expand floating-point complex
3211 multiplication using the above.
3212 (expand_complex_division): Rename inner_type parameter to type.
3213 Update expand_complex_libcall call-site.
3214 (expand_complex_operations_1): Update expand_complex_multiplication
3215 and expand_complex_division call-sites.
3217 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3220 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3221 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3222 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3223 the highest significant bit of the shift count mask is clear. In
3224 check whether and[sq]i3 is needed verify that all significant bits
3225 of the shift count other than the highest are set.
3227 2018-05-02 Tom de Vries <tom@codesourcery.com>
3230 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3231 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3232 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3233 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3234 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3235 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3236 __builtin_goacc_parlevel_size.
3238 2018-05-02 Richard Biener <rguenther@suse.de>
3240 PR tree-optimization/85597
3241 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3242 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3245 2018-05-02 Tom de Vries <tom@codesourcery.com>
3248 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3249 dump files): Add ltrans-tree.
3251 2018-05-02 Tom de Vries <tom@codesourcery.com>
3254 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3255 dump files): Add wpa-ipa.
3257 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3259 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3260 powerpc*-*-linux*paired* target.
3261 * config/rs6000/750cl.h: Delete.
3262 * config/rs6000/paired.h: Delete.
3263 * config/rs6000/paired.md: Delete.
3264 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3266 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3267 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3268 comment. Remove paired float support.
3269 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3270 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3272 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3273 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3275 * config/rs6000/rs6000.c: Remove paired float support.
3276 (paired_expand_vector_init, paired_expand_vector_move,
3277 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3278 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3279 paired_expand_builtin, paired_expand_predicate_builtin,
3280 paired_init_builtins): Delete.
3281 * config/rs6000/rs6000.h: Remove paired float support.
3282 * config/rs6000/rs6000.md: Remove paired float support.
3283 (move_from_CR_ov_bit): Delete.
3284 * config/rs6000/rs6000.opt (mpaired): Delete.
3285 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3286 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3288 2018-05-02 Richard Biener <rguenther@suse.de>
3291 * gimplify.c (gimplify_save_expr): When in SSA form allow
3292 SAVE_EXPRs to compute to SSA vars.
3294 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3297 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3298 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3299 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3300 clobber operands[2], instead use a new pseudo. Formatting fixes.
3302 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3304 PR tree-optimization/85586
3305 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3306 exit early for statements in the same group if the accesses are
3309 2018-05-02 Tom de Vries <tom@codesourcery.com>
3312 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3315 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3317 PR tree-optimization/85143
3318 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3320 2018-05-01 Tom de Vries <tom@codesourcery.com>
3323 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3324 not found" error message.
3326 2018-05-01 Tom de Vries <tom@codesourcery.com>
3329 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3330 * vec.c (test_ordered_remove_if): New function.
3331 (vec_c_tests): Call test_ordered_remove_if.
3332 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3333 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3334 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3335 VEC_ORDERED_REMOVE_IF.
3337 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3339 PR tree-optimization/82665
3340 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3341 pointer subtraction where arguments come from a memchr call.
3343 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3345 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3346 --push-state --as-needed and --pop-state instead of --as-needed and
3347 --no-as-needed if ld supports it.
3348 * configure: Regenerated.
3351 * doc/install.texi2html: Replace _002d with - and _002a with * in
3352 generated html files using sed.
3354 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3357 * gcc-rich-location.c (blank_line_before_p): New function.
3358 (use_new_line): New function.
3359 (gcc_rich_location::add_fixit_insert_formatted): New function.
3360 * gcc-rich-location.h
3361 (gcc_rich_location::add_fixit_insert_formatted): New function.
3363 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3365 * selftest.c (assert_streq): Rename "expected" and "actual" to
3366 "val1" and "val2". Extend NULL-handling to cover both inputs
3367 symmetrically, while still requiring both to be non-NULL for a pass.
3368 * selftest.h (assert_streq): Rename "expected" and "actual" to
3370 (ASSERT_EQ): Likewise.
3371 (ASSERT_EQ_AT): Likewise.
3372 (ASSERT_KNOWN_EQ): Likewise.
3373 (ASSERT_KNOWN_EQ_AT): Likewise.
3374 (ASSERT_NE): Likewise.
3375 (ASSERT_MAYBE_NE): Likewise.
3376 (ASSERT_MAYBE_NE_AT): Likewise.
3377 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3378 the assertion to pass.
3379 (ASSERT_STREQ_AT): Likewise.
3381 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3383 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3384 interaction with -pie.
3386 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3388 * selftest.h: Fix alphabetization of per-source-file selftest
3391 2018-04-30 Jason Merrill <jason@redhat.com>
3393 PR c++/61982 - dead stores to destroyed objects.
3394 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3397 2018-04-30 Jason Merrill <jason@redhat.com>
3399 * tree.c (build_clobber): New.
3400 * tree.h: Declare it.
3401 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3403 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3405 * diagnostic-show-locus.c (layout::layout): Update for
3406 location_get_source_line returning a char_span.
3407 (struct char_span): Move to input.h.
3408 (struct correction): Update for fields in char_span becoming
3410 (struct source_line): Update for location_get_source_line
3411 returning a char_span.
3412 (layout::print_line): Likewise.
3413 * edit-context.c (edited_file::print_content): Likewise.
3414 (edited_file::print_diff_hunk): Likewise.
3415 (edited_file::print_run_of_changed_lines): Likewise.
3416 (edited_file::get_num_lines): Likewise.
3417 (edited_line::edited_line): Likewise.
3418 * final.c (asm_show_source): Likewise.
3419 * input.c (location_get_source_line): Convert return type
3420 from const char * to char_span, losing the final "line_len"
3422 (dump_location_info): Update for the above.
3423 (get_substring_ranges_for_loc): Likewise. Use a char_span
3424 when handling the literal within the line.
3425 (test_reading_source_line): Update for location_get_source_line
3426 returning a char_span.
3427 * input.h (class char_span): Move here from
3428 diagnostic-show-locus.c, converting from a struct to a class.
3429 Make data members private.
3430 (char_span::operator bool): New.
3431 (char_span::length): New.
3432 (char_span::get_buffer): New.
3433 (char_span::operator[]): New.
3434 (char_span::subspan): Make const.
3435 (char_span::xstrdup): New.
3436 (location_get_source_line): Convert return type from const char *
3437 to char_span, losing the final "line_size" param.
3439 2018-04-30 Jan Hubicka <jh@suse.cz>
3441 * lto-wrapper.c (ltrans_priorities): New static var.
3442 (cmp_priority): New.
3443 (run_gcc): Read priorities and if doing parallel build order
3444 the Makefile by them.
3446 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3448 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3450 2018-04-30 Richard Biener <rguenther@suse.de>
3452 * tree-cfg.c (verify_address): Remove base argument, add
3453 flag whether to check TREE_ADDRESSABLE and do that.
3454 (verify_expr): Remove.
3455 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3456 (verify_gimple_assign_single): Likewise.
3457 (verify_gimple_switch): Likewise.
3458 (verify_expr_location_1): Dereference tp once. Add (disabled)
3459 piece from verify_expr.
3460 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3462 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3464 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3466 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3468 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3469 (small_data_pattern): Likewise.
3470 (arc_rewrite_small_data): Likewise.
3471 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3472 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3473 (get_symbol_alignment): New function.
3474 (legitimate_small_data_address_p): Likewise.
3475 (legitimate_scaled_address): Update, call
3476 legitimate_small_data_address_p.
3477 (output_sdata): New static variable.
3478 (arc_print_operand): Update how we handle small data operands.
3479 (arc_print_operand_address): Likewise.
3480 (arc_legitimate_address_p): Update, use
3481 legitimate_small_data_address_p.
3482 (arc_rewrite_small_data_p): Remove.
3483 (arc_rewrite_small_data_1): Likewise.
3484 (arc_rewrite_small_data): Likewise.
3485 (small_data_pattern): Likewise.
3486 (compact_sda_memory_operand): Update to use
3487 legitimate_small_data_address_p and get_symbol_alignment.
3488 (prepare_move_operands): Don't rewite sdata pattern.
3489 (prepare_extend_operands): Remove.
3490 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3492 (zero_extendqisi2): Likewise.
3493 (zero_extendhisi2): Likewise.
3494 (extendqihi2): Likewise.
3495 (extendqisi2): Likewise.
3496 (extendhisi2): Likewise.
3500 * config/arc/constraints.md (Usd): Change it to memory constraint.
3502 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3504 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3505 as source of std instructions.
3506 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3507 allow 6-bit constants as source for store instructions.
3508 (movdi_insn): Update instruction pattern to allow 6-bit constants
3509 as source for store instructions.
3511 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3513 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3515 2018-04-30 Nathan Sidwell <nathan@acm.org>
3516 Sandra Loosemore <sandra@codesourcery.com>
3518 * dumpfile.c (dump_open): Allow '-' for stdout.
3519 * doc/invoke.texi (Developer Options): Document dump filename
3520 determination early. Document stdin/stdout selection.
3522 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3524 Microblaze Target: PIC data text relative
3526 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3527 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3529 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3530 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3531 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3532 New addressing mode for data-text relative position indepenedent code.
3533 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3534 'ADDRESS_SYMBOLIC_TXT_REL'.
3535 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3536 (microblaze_legitimate_pic_operand): Exclude function calls from
3537 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3538 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3540 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3541 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3542 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3543 for 'address + offset'.
3544 (microblaze_expand_prologue): Add new function prologue call for
3546 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3547 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3548 table in case of TARGET_PIC_DATA_TEXT_REL.
3549 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3550 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3551 Add new macros 'UNSPEC_TEXT',
3552 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3553 + exclude function calls from 'UNSPEC_PLT' in case of data text
3555 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3556 new target hook for generating address diff vector tables in case of
3558 * doc/tm.texi : Regenerate.
3559 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3560 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3561 of addr diff vector generation.
3562 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3563 target hook definition.
3564 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3565 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3566 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3567 option for data text relative.
3569 2018-04-30 Richard Biener <rguenther@suse.de>
3571 * tree-chrec.h (evolution_function_is_constant_p): Remove
3573 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3575 2018-04-30 Richard Biener <rguenther@suse.de>
3578 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3580 2018-04-30 Richard Biener <rguenther@suse.de>
3582 PR tree-optimization/28364
3583 PR tree-optimization/85275
3584 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3585 copying first exit test.
3587 2018-04-28 Mark Wielaard <mark@klomp.org>
3589 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3591 (dwarf_AT): Handle DW_AT_addr_base.
3592 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3594 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3597 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3598 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3599 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3600 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3602 2018-04-28 Richard Biener <rguenther@suse.de>
3604 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3605 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3606 to reflect use. Only add interesting stmts.
3608 2018-04-27 Martin Jambor <mjambor@suse.cz>
3611 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3612 the jump function allows for passing through aggregate values.
3614 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3616 * input.h (in_system_header_at): Convert from macro to inline
3618 (from_macro_expansion_at): Likewise.
3619 (from_macro_definition_at): Likewise.
3621 2018-04-27 Jeff Law <law@redhat.com>
3623 * config.gcc: Mark tile* targets as deprecated/obsolete.
3625 2018-04-27 Richard Biener <rguenther@suse.de>
3627 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3630 2018-04-27 Richard Biener <rguenther@suse.de>
3632 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3634 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3636 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3637 with Yd constraint. Set "preferred_for_speed" attribute from
3638 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3640 (*movdi_internal): Ditto.
3641 (movti_interunit splitters): Remove
3642 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3643 (movdi_interunit splitters): Ditto.
3644 * config/i386/constraints.md (Ye): Remove.
3645 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3647 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3650 * config/aarch64/constraints.md (Usg): Limit to 31.
3653 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3655 PR tree-optimization/85529
3656 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3657 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3658 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3659 zero extension or masking of the MSB bit.
3660 (optimize_range_tests): Add FIRST_BB argument, pass it through
3661 to optimize_range_tests_var_bound.
3662 (maybe_optimize_range_tests, reassociate_bb): Adjust
3663 optimize_range_tests callers.
3665 2018-04-26 Richard Biener <rguenther@suse.de>
3666 Jakub Jelinek <jakub@redhat.com>
3668 * cgraph.h (symbol_table): Just declare debug method here.
3669 * symtab.c (symbol_table::debug): Define.
3671 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3673 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3675 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3677 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3678 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3679 (*movdi_internal): Substitute Yi and Yj constraint with x
3680 and Ym and Yn constraint with y constraint. Update "isa"
3681 attribute and set "preferred_for_speed" attribute from
3682 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3683 (*movsi_internal): Ditto.
3684 (*movdf_internal): Ditto.
3685 (*movsf_internal): Ditto.
3686 (*zero_extendsidi2): Ditto.
3687 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3688 (sse2_loadld): Ditto.
3689 (*vec_extract<ssevecmodelower>_0): Ditto.
3690 (*vec_extractv4si_0_zext_sse4): Ditto.
3691 (vec_concatv2di): Ditto.
3692 (*vec_dup<mode>): Ditto.
3693 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3694 * config/i386/constraints.md (Yi): Remove.
3699 2018-04-26 Nathan Sidwell <nathan@acm.org>
3701 * dumpfile.c (dump_open): New.
3702 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3703 (dump_finish): Detect stdio/stderr by value not name.
3705 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3707 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3709 2018-04-26 Tom de Vries <tom@codesourcery.com>
3712 * config/nvptx/nvptx.c (verify_neutering_jumps)
3713 (verify_neutering_labels): New function
3714 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3716 2018-04-26 Tom de Vries <tom@codesourcery.com>
3719 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3720 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3723 2018-04-26 Richard Biener <rguenther@suse.de>
3724 Tom de Vries <tom@codesourcery.com>
3727 * lto-streamer-out.c (output_function): Fixup loops if required to match
3728 discovery done in the reader.
3730 2018-04-26 Richard Biener <rguenther@suse.de>
3732 PR tree-optimization/85116
3733 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3734 have a loop exit from the single latch predecessor. Remove
3735 case of header with just condition.
3736 (ch_base::copy_headers): Exclude infinite loops from any
3738 (pass_ch::execute): Record exits.
3740 2018-04-26 Richard Biener <rguenther@suse.de>
3742 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3743 prologue cost vector and pass it to vect_get_load_cost.
3744 (vect_get_peeling_costs_all_drs): Likewise.
3745 (vect_peeling_hash_get_lowest_cost): Likewise.
3746 (vect_enhance_data_refs_alignment): Likewise.
3748 2018-04-26 Richard Biener <rguenther@suse.de>
3751 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3752 checking of integer<->pointer conversions.
3753 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3754 sign-/zero-extending pointer types.
3755 (expand_omp_for_static_chunk): Likewise.
3757 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3758 Jean Lee <xiaoyur347@gmail.com>
3760 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3761 (TARGET_ASAN_SHADOW_OFFSET): Define.
3762 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3763 true for -fsanitize=address.
3765 2018-04-25 Mark Wielaard <mark@klomp.org>
3767 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3770 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3772 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3773 than "alu", remove explicit "memory" and "imm_disp" attributes.
3774 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3777 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3778 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3779 gen_lowpart_no_emit.
3781 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3784 * config/i386/i386.c (ix86_expand_builtin): Change memory
3785 operand to XI, extend p0 to Pmode.
3786 * config/i386/i386.md: Change unspec volatile and operand
3787 1 mode to XI, change operand 0 mode to P.
3789 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3791 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3792 GET_MODE_MASK before any checking.
3793 (nds32_can_use_bset_p): Likewise.
3794 (nds32_can_use_btgl_p): Likewise.
3796 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3798 * config/nds32/nds32-doubleword.md: New define_split pattern for
3799 illegal register number.
3801 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3803 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3805 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3807 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3809 2018-04-25 Richard Biener <rguenther@suse.de>
3811 * lto-streamer.h (LTO_major_version): Bump to 8.
3813 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3815 * BASE-VER: Set to 9.0.0.
3817 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3819 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3820 in __abskf2 and __powikf2.
3822 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3825 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3826 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3827 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3828 Use the above on operand 2. Reindent.
3829 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3831 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3834 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3836 * config/i386/i386.opt (mcet): Removed.
3837 * doc/install.texi: Remove -mcet documentation.
3838 * doc/invoke.texi: Likewise.
3840 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3843 * doc/install.texi: Remove -mcet from bootstrap-cet.
3845 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3848 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3849 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3853 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3854 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3855 containing a CONST_VECTOR.
3857 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3859 * doc/install.texi: Update newlib dependency for nvptx.
3861 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3864 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3865 instead of INTVAL when shifting x left.
3867 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3869 PR tree-optimization/85478
3870 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3871 vect_grouped_store_supported for single element vectors.
3873 2018-04-24 Richard Biener <rguenther@suse.de>
3876 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3877 load cost increase to the case of non-constant step.
3879 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3882 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3883 destination if any_malformed_asm.
3885 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3888 * expr.c (store_field): In the bitfield case, if the value comes from
3889 a function call and is returned in registers by means of a PARALLEL,
3890 do not change the mode of the temporary unless BLKmode and VOIDmode.
3892 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3894 PR rtl-optimization/85423
3895 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3896 dependencies to debug insns when the previous insn is non-debug.
3898 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3900 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3901 enums into a single definition.
3902 (fls): Fix predicates and printing.
3905 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3907 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3908 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3909 and short u6 immediate.
3910 (check_if_valid_sleep_operand): Remove.
3911 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3913 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3915 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3916 flag_always_save_lp condition.
3917 * config/nds32/nds32.opt (malways-save-lp): New option.
3919 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3921 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3922 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3923 * config/nds32/nds32.h
3924 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3925 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3927 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3929 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3930 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3932 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3933 Chung-Ju Wu <jasonwucj@gmail.com>
3935 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3937 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3938 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3940 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3942 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3944 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3946 * config/nds32/nds32-protos.h (nds32_data_alignment,
3947 nds32_local_alignment): Declare.
3948 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3949 nds32_local_alignment): New functions.
3950 (TARGET_CONSTANT_ALIGNMENT): Define.
3951 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3953 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3955 * config/nds32/nds32.c
3956 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3957 (TARGET_MODES_TIEABLE_P): Likewise.
3959 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3961 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3964 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3965 Chung-Ju Wu <jasonwucj@gmail.com>
3967 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3968 for unaligned access.
3969 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3971 * config/nds32/nds32-intrinsic.md: Likewise.
3972 * config/nds32/nds32_intrinsic.h: Likewise.
3973 * config/nds32/nds32.h (nds32_builtins): Likewise.
3974 * config/nds32/nds32.opt (munaligned-access): New option.
3975 * config/nds32/nds32.c (nds32_asm_file_start): Display
3976 flag_unaligned_access status.
3978 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3980 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3981 -mno-relax is present.
3982 * config/riscv/linux.h (LINK_SPEC): Ditto.
3984 2018-04-20 Martin Sebor <msebor@redhat.com>
3987 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3989 (gimple_fold_builtin_stxcpy_chk): Same.
3990 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3992 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3995 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3996 __powikf2 when long double is IEEE 128-bit.
3998 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4000 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4001 step to make sure stack always aligned.
4003 2018-04-20 Carl Love <cel@us.ibm.com>
4006 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4007 size check for arg0.
4009 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4010 Tom de Vries <tom@codesourcery.com>
4013 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4014 Emit insns for calls too.
4015 (nvptx_find_par): Always look for worker-level predecessor insn.
4016 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4018 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4019 (nvptx_process_pars): Propagate frames for calls.
4021 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4024 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4026 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4027 (ix86_handle_option): Don't handle OPT_mibt.
4028 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4030 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4032 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4033 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4034 (ix86_target_macros): Define __CET__ with flag_cf_protection
4035 for -fcf-protection.
4036 * config/i386/i386.c (isa2_opts): Remove -mibt.
4037 * config/i386/i386.h (TARGET_IBT): Removed.
4038 (TARGET_IBT_P): Likewise.
4039 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4040 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4041 * config/i386/i386.opt (mcet): Update help message.
4044 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4045 -mcet as an alias for -mshstk.
4047 2018-04-20 Richard Biener <rguenther@suse.de>
4050 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4051 complexity by forcing a single use of the multiply operand.
4053 2018-04-20 Martin Jambor <mjambor@suse.cz>
4056 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4057 recursion dependency to only apply to non-clones.
4059 2018-04-20 Martin Jambor <mjambor@suse.cz>
4062 * ipa-cp.c (create_specialized_node): Check that clones of
4063 self-recursive edges exist during IPA-CP.
4065 2018-04-19 Toon Moene <toon@moene.org>
4067 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4070 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4072 PR tree-optimization/85467
4073 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4074 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4075 VECTOR_CST element to type.
4077 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4080 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4081 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4082 (builtin_longjmp): Likewise.
4083 (save_stack_nonlocal): New pattern.
4084 (restore_stack_nonlocal): Likewise.
4086 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4089 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4090 Replace ASM_OUTPUT_LABEL with fprintf.
4092 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4095 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4096 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4097 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4098 define __IBT__ and __SHSTK__ for -fcf-protection.
4099 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4101 (ix86_trampoline_init): Likewise.
4102 (x86_output_mi_thunk): Likewise.
4103 (ix86_notrack_prefixed_insn_p): Likewise.
4104 (ix86_option_override_internal): Don't disallow -fcf-protection.
4105 * config/i386/i386.md (rdssp<mode>): Also enable for
4107 (incssp<mode>): Likewise.
4108 (nop_endbr): Likewise.
4109 * config/i386/i386.opt (mcet): Change help message to built-in
4113 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4114 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4115 enable CET built-in functions.
4117 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4119 * common/config/i386/i386-common.c
4120 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4121 OPTION_MASK_ISA_MOVDIRI_UNSET,
4122 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4123 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4124 * config.gcc (movdirintrin.h): New header.
4125 * config/i386/cpuid.h (bit_MOVDIRI,
4126 bit_MOVDIR64B): New bits.
4127 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4129 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4130 (VOID, PVOID, PCVOID)): New function types.
4131 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4132 __builtin_ia32_directstoreu_u64,
4133 __builtin_ia32_movdir64b): New builtins.
4134 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4135 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4137 (ix86_valid_target_attribute_inner_p): Ditto.
4138 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4139 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4140 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4141 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4142 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4143 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4144 (movdiri<mode>, movdir64b_<mode>): New.
4145 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4146 * config/i386/immintrin.h: Include movdirintrin.h.
4147 * config/i386/movdirintrin.h: New file.
4148 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4150 2018-04-19 Richard Biener <rguenther@suse.de>
4153 * cfg.c (clear_bb_flags): When loop state says we have
4154 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4156 2018-04-19 Richard Biener <rguenther@suse.de>
4158 PR tree-optimization/84737
4159 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4160 copying restrict info.
4161 (vect_setup_realignment): Use it.
4162 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4163 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4164 the first DR to all generated stores.
4165 (vectorizable_load): Likewise for loads.
4167 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4169 PR tree-optimization/85446
4170 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4171 the integral and pointer types to have the same precision.
4173 * doc/install.texi: Document --disable-cet being the default and
4176 2018-04-18 Martin Liska <mliska@suse.cz>
4178 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4181 2018-04-18 Martin Liska <mliska@suse.cz>
4184 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4187 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4188 arguments if they are comparable.
4190 2018-04-18 Martin Liska <mliska@suse.cz>
4193 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4196 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4199 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4202 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4203 ENDBR after calling __morestack.
4205 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4208 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4209 by using gcc_base_ver to generate a gcc_driver_version, and use
4210 it when generating GCC_DRIVER_NAME.
4211 * configure: Regenerate.
4213 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4216 * config.gcc: Obsolete powerpc*-*-*spe*.
4218 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4221 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4222 (stabstr_D): Change type of unum from unsigned int to
4223 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4226 2018-04-17 Jim Wilson <jimw@sifive.com>
4229 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4230 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4231 Set arg_pointer_offset after using pretend_args_size.
4233 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4235 PR rtl-optimization/85431
4236 * dse.c (record_store): Ignore zero width stores.
4239 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4240 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4241 __builtin_stack_restore rather than after it.
4242 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4243 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4244 argument instead of virtual_dynamic_stack_rtx.
4246 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4248 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4250 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4251 Add note to error message to explain internal mapping of overloaded
4252 built-in function name to non-overloaded built-in function name.
4253 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4256 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4259 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4260 where the inputs overlap with the output.
4262 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4264 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4265 (=v, v) alternative and explicit "memory" attribute.
4266 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4267 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4269 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4270 "sselog1" type instead of "sselog".
4271 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4272 "sselog". Remove explicit "memory" attribute.
4273 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4274 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4276 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4277 "sselog1" type instead of "sselog". Remove explicit "memory"
4279 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4280 use "sselog1" type instead of "sselog". Remove explicit "memory"
4282 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4283 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4285 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4286 "sselog1" type instead of "sselog". Remove explicit "memory"
4288 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4289 use "sselog1" type instead of "sselog". Remove explicit "memory"
4293 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4296 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4299 2018-04-17 Martin Jambor <mjambor@suse.cz>
4302 * ipa-cp.c (create_specialized_node): Call
4303 expand_all_artificial_thunks if necessary.
4305 2018-04-17 Martin Liska <mliska@suse.cz>
4308 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4309 in message, remote space in between '_G' and '('.
4311 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4314 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4315 avx512f_vmcmp<mode>3<round_saeonly_name>,
4316 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4317 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4318 avx512f_rndscale<mode><round_saeonly_name>,
4319 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4320 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4321 Use %<iptr>2 instead of %2 for -masm=intel.
4322 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4323 avx512f_vcvttss2usi<round_saeonly_name>,
4324 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4326 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4327 avx512f_vcvttsd2usi<round_saeonly_name>,
4328 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4329 Use %q1 instead of %1 for -masm=intel.
4330 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4331 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4332 of %3 for -masm=intel.
4333 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4335 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4337 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4339 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4340 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4342 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4344 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4345 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4346 %g1 and one with %0 and %1.
4347 (avx512er_vmrcp28<mode><round_saeonly_name>,
4348 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4350 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4351 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4352 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4353 of %0 and %{%4%} for -masm=intel.
4354 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4355 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4356 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4357 order of %0 and %{%5%}%{z%} for -masm=intel.
4359 2018-04-17 Jan Hubicka <jh@suse.cz>
4362 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4364 2018-04-17 Martin Liska <mliska@suse.cz>
4367 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4368 for target_clone error message. Make default implementation
4369 clone to be a local declaration.
4370 (separate_attrs): Add new argument and check for an empty
4372 (expand_target_clones): Handle it.
4373 (ipa_target_clone): Make redirection just for target_clones
4376 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4377 Tom de Vries <tom@codesourcery.com>
4380 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4381 tiled basic blocks without omp continue statements.
4383 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4386 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4387 vec_extract expression as having side effects to make sure it gets
4390 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4393 * config/i386/i386.c (get_builtin_code_for_version): Check
4396 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4399 * gcc/config.gcc: Support "skylake".
4400 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4402 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4403 (processor_target_table): Add "skylake".
4404 (ix86_option_override_internal): Add "skylake".
4405 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4406 PROCESSOR_CANNONLAKE.
4407 (get_builtin_code_for_version): Fix priority for
4408 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4409 PROCESSOR_SKYLAKE-AVX512.
4410 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4411 (processor_type): Add PROCESSOR_SKYLAKE.
4413 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4414 Jason Merrill <jason@redhat.com>
4417 * convert.c (convert_to_integer_1): Use direct recursion for
4418 enumeral types and types with a precision less than the number
4419 of bits in their mode.
4421 2018-04-16 Julia Koval <julia.koval@intel.com>
4424 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4425 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4427 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4430 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4431 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4432 and -mno-direct-move.
4434 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4437 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4438 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4439 Ensure negative shifts result in {0}.
4441 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4443 PR rtl-optimization/79916
4444 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4445 regs (if any) to define how to gnerate SD moves when LRA is in
4448 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4450 PR rtl-optimization/85393
4451 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4452 * except.c (expand_dw2_landing_pad_for_region): Make static.
4453 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4454 a label and unconditional jump to old_bb, rather than
4455 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4458 PR rtl-optimization/85376
4459 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4460 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4461 instead of a specific value.
4463 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4464 Bin Cheng <bin.cheng@arm.com>
4466 PR tree-optimization/82965
4467 PR tree-optimization/83991
4468 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4469 by_profile_only parameter.
4470 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4471 information if the loop was predicted to iterate too many times.
4472 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4474 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4477 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4480 2018-04-13 Martin Liska <mliska@suse.cz>
4481 Jakub Jelinek <jakub@redhat.com>
4484 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4485 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4486 * builtins.c (expand_builtin_memory_copy_args): Use
4487 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4488 handle dest_addr == pc_rtx.
4490 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4493 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4494 asked to not generate direct moves.
4495 (fix_trunc<mode>si2_stfiwx): Similar.
4496 (fix_trunc<mode>si2_internal): Similar.
4498 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4501 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4502 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4503 lookup if dest in some wider mode is known to be const0_rtx and
4504 if so, record permanent equivalence for it to be ZERO_EXTEND of
4505 the narrower mode destination.
4507 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4509 * lto-streamer-out.c (output_function): Revert 259346.
4510 * omp-expand.c (expand_oacc_for): Likewise.
4512 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4514 PR rtl-optimization/85354
4515 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4516 * sel-sched.c (sel_global_init): ... here.
4518 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4521 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4522 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4523 mode for PE-COFF targets.
4524 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4525 (i386_pe_asm_lto_end): Likewise.
4526 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4527 (TARGET_ASM_LTO_END): Likewise.
4528 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4529 (i386_pe_asm_lto_start): New function.
4530 (i386_pe_asm_lto_end): Likewise.
4532 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4533 Richard Biener <rguenther@suse.de>
4536 * lto-streamer-out.c (output_function): Fix CFG loop state before
4538 * omp-expand.c (expand_oacc_for): Handle calls to internal
4539 functions like regular functions.
4541 2018-04-12 Richard Biener <rguenther@suse.de>
4544 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4545 for the early LTO debug to properly generate references to it
4546 during DIE emission. Do not re-use that for the skeleton for
4548 (dwarf2out_early_finish): Likewise.
4550 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4553 * config/i386/sse.md
4554 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4555 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4556 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4557 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4558 and output is a reg, avoid creating invalid lowpart subreg, but
4559 instead split into a 512-bit move. Don't split if not AVX512VL,
4560 input is xmm16+ reg and output is a mem.
4561 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4562 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4563 xmm16+ reg and output is a mem.
4565 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4567 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4568 also for flag_dwarf2_cfi_asm.
4570 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4572 PR rtl-optimization/85342
4573 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4574 a bool scalar var inside of the loop instead. Don't try to update
4575 recog_data.operand after failed apply_change_group.
4577 2018-04-12 Tom de Vries <tom@codesourcery.com>
4580 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4581 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4582 array with flexible array member as array without given dimension.
4583 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4584 argument for undefined param to true.
4586 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4589 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4590 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4591 from PowerPC section.
4592 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4593 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4594 help text that is too long.
4595 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4596 help text that is too long.
4597 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4598 help text that is too long.
4600 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4602 * config/alpha/alpha.md (stack_probe_internal): Rename
4603 from "probe_stack". Update all callers.
4605 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4607 PR rtl-optimization/84566
4608 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4609 sched_macro_fuse_insns.
4611 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4614 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4615 (compute_block_dependences): ... from here.
4617 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4619 PR tree-optimization/85331
4620 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4621 from int to HOST_WIDE_INT.
4623 2018-04-11 Martin Jambor <mjambor@suse.cz>
4626 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4627 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4628 not the same as the source val.
4629 (cgraph_edge_brings_value_p): New parameter.
4630 (gather_edges_for_value): Pass destination value to
4631 cgraph_edge_brings_value_p.
4632 (perhaps_add_new_callers): Likewise.
4633 (get_info_about_necessary_edges): Likewise and exclude values brought
4634 only by self-recursive edges.
4635 (create_specialized_node): Redirect only clones of self-calling edges.
4636 (+self_recursive_pass_through_p): New function.
4637 (find_more_scalar_values_for_callers_subset): Use it.
4638 (find_aggregate_values_for_callers_subset): Likewise.
4639 (known_aggs_to_agg_replacement_list): Removed.
4640 (decide_whether_version_node): Re-calculate known constants for all
4641 remaining context clones.
4643 2018-04-11 Richard Biener <rguenther@suse.de>
4646 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4647 from early DWARF output.
4648 (dwarf2out_early_finish): Output line info unconditionally into
4649 early DWARF and add reference to it.
4651 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4654 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4655 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4656 other than V2DFmode using iptr mode attribute.
4657 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4659 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4661 PR rtl-optimization/84659
4662 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4664 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4667 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4669 (output_loc_list): Pass address of a dummy size variable even in the
4670 locview handling loop.
4671 (index_location_lists): Add comment on why skip_loc_list_entry can't
4674 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4677 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4680 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4683 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4684 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4685 and -mstring-compare-inline-limit.
4687 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4690 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4691 for stack clash protection in a register whenever we need it to be in
4694 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4696 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4697 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4699 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4702 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4704 (mlong-double-): Ditto.
4705 * config/rs6000/sysv4.opt (msdata=): Ditto.
4706 (mtls-size=): Ditto.
4708 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4710 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4711 erroneous entries for
4712 "vector int vec_ldl (int, long int *)", and
4713 "vector unsigned int vec_ldl (int, unsigned long int *)".
4714 Add comments and entries for
4715 "vector bool char vec_ldl (int, bool char *)",
4716 "vector bool short vec_ldl (int, bool short *)",
4717 "vector bool int vec_ldl (int, bool int *)",
4718 "vector bool long long vec_ldl (int, bool long long *)",
4719 "vector pixel vec_ldl (int, pixel *)",
4720 "vector long long vec_ldl (int, long long *)",
4721 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4722 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4723 type tree bool_long_long_type_node and correct definition of
4724 bool_V2DI_type_node to make reference to this new type tree.
4725 (rs6000_mangle_type): Replace erroneous reference to
4726 bool_long_type_node with bool_long_long_type_node.
4727 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4728 comments to emphasize sign distinctions for char and int types and
4729 replace RS6000_BTI_bool_long constant with
4730 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4731 use of RS6000_BTI_pixel.
4732 (bool_long_type_node): Remove this macro definition.
4733 (bool_long_long_type_node): New macro definition
4735 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4737 PR rtl-optimization/85300
4738 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4739 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4740 simplify_unary_operation fails.
4742 2018-04-10 Martin Liska <mliska@suse.cz>
4744 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4745 cgraph_edge and ipa_ref.
4747 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4751 * config/i386/sse.md
4752 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4753 computation of the VEC_MERGE selector from mask.
4754 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4755 Fix decoding of the VEC_MERGE selector into mask.
4757 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4759 PR tree-optimization/85286
4760 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4762 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4764 * final.c (final_1): Set insn_last_address as well as
4765 insn_current_address.
4767 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4770 * explow.c (emit_stack_probe): Call validize_mem on memory location
4771 before passing it to gen_probe_stack. Create address operand and
4772 legitimize it for the probe_stack_address case.
4774 2018-04-09 Jan Hubicka <jh@suse.cz>
4777 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4778 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4779 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4781 (free_lang_data): Rebuild type inheritance graph.
4783 2018-04-09 Martin Sebor <msebor@redhat.com>
4785 * invoke.texi (-finline-small-functions): Mention other optimization
4787 (-findirect-inlining, -fpartial-inlining): Same.
4788 (-finline-functions-called-once): Same.
4789 (-freorder-blocks-and-partition): Same.
4791 2018-04-09 Jan Hubicka <jh@suse.cz>
4794 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4795 jumps; choose last target that matches the criteria (i.e.
4796 no partition changes for non-crossing jumps).
4797 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4798 support for redirecting crossing jumps to non-crossing.
4800 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4802 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4803 also for naked functions.
4805 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4807 * config/arc/arc.md (add_shift): New pattern.
4808 (add_shift2): Likewise.
4809 (sub_shift): Likewise.
4810 (sub_shift_cmp0_noout): Likewise.
4811 (compare_si_ashiftsi): Likewise.
4812 (xbfu_cmp0_noout): New combine pattern.
4813 (xbfu_cmp0"): Likewise.
4814 (movsi_set_cc_insn): Place the predicable variant first.
4815 (commutative_binary_cmp0_noout): Remove clobber.
4816 (commutative_binary_cmp0): New pattern.
4817 (noncommutative_binary_cmp0): Likewise.
4818 (noncommutative_binary_cmp0_noout): Likewise.
4819 (noncommutative_binary_comparison_result_used): Removed.
4820 (rsub_cmp0): New pattern.
4821 (rsub_cmp0_noout): Likewise.
4822 (extzvsi): Changed, keep only meaningful variants.
4823 (SQH, SEZ): New iterators.
4824 (SQH_postfix): New mode attribute.
4825 (SEZ_prefix): New code attribute.
4826 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4827 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4828 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4830 (noncommutative_operator): Check the availability of barrel
4833 2018-04-09 Richard Biener <rguenther@suse.de>
4835 PR tree-optimization/85284
4836 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4837 Only use the niter constraining form of simple_iv when the exit
4840 2018-04-09 Tom de Vries <tom@codesourcery.com>
4843 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4844 (define_expand "*memory_barrier"): New define_expand.
4845 (define_insn "memory_barrier"): New insn.
4847 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4849 PR rtl-optimization/80463
4850 PR rtl-optimization/83972
4851 PR rtl-optimization/83480
4853 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4854 correct producer for the insn.
4855 (tidy_control_flow): Fixup seqnos in case of debug insns.
4857 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4859 PR rtl-optimization/83913
4861 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4862 different sched-times when merging exprs.
4864 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4866 PR rtl-optimization/83962
4868 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4869 tidy_fallthru_edge and tidy_control_flow.
4871 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4873 PR rtl-optimization/83530
4875 * sel-sched.c (force_next_insn): New global variable.
4876 (remove_insn_for_debug): When force_next_insn is true, also leave only
4877 next insn in the ready list.
4878 (sel_sched_region): When the region wasn't scheduled, make another pass
4879 over it with force_next_insn set to 1.
4881 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4883 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4885 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4886 for interrupt control.
4887 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4888 functions for interrupt control.
4889 * config/nds32/nds32-intrinsic.md: Likewise.
4890 * config/nds32/nds32_intrinsic.h: Likewise.
4891 * config/nds32/nds32.h (nds32_builtins): Likewise.
4893 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4895 * config/nds32/nds32.c (nds32_init_machine_status,
4896 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4897 strict_aligned_p field.
4898 (nds32_expand_to_rtl_hook): New function.
4899 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4900 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4902 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4903 Chung-Ju Wu <jasonwucj@gmail.com>
4905 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4906 * config/nds32/nds32-n7.md: New file.
4907 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4908 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4910 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4911 * config/nds32/nds32.md (pipeline_model): Add n7.
4912 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4913 * config/nds32/pipelines.md: Include n7 settings.
4915 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4916 Chung-Ju Wu <jasonwucj@gmail.com>
4918 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4919 * config/nds32/nds32-e8.md: New file.
4920 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4921 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4923 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4924 * config/nds32/nds32.md (pipeline_model): Add e8.
4925 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4926 * config/nds32/pipelines.md: Include e8 settings.
4928 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4929 Chung-Ju Wu <jasonwucj@gmail.com>
4931 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4932 * config/nds32/nds32-n8.md: New file.
4933 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4934 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4936 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4937 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4938 * config/nds32/nds32.md (pipeline_model): Add n8.
4939 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4940 * config/nds32/pipelines.md: Include n8 settings.
4942 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4943 Chung-Ju Wu <jasonwucj@gmail.com>
4945 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4946 * config/nds32/nds32-n9-2r1w.md: New file.
4947 * config/nds32/nds32-n9-3r2w.md: New file.
4948 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4949 nds32_register_ports): New or modify for cpu n9.
4950 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4952 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4953 * config/nds32/nds32-utils.c: New file.
4954 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4955 TARGET_MUL_SLOW): Define.
4956 * config/nds32/nds32.md (pipeline_model): New attribute.
4957 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4958 New options that support cpu n9.
4959 * config/nds32/pipelines.md: Include n9 settings.
4960 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4962 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4964 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4965 information if necessary.
4966 (output_cond_branch_compare_zero): Likewise.
4967 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4968 (nds32_target_alignment): Refine for alignment.
4969 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4970 (FUNCTION_BOUNDARY): Modify.
4971 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4973 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4975 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4977 * config/nds32/constants.md (unspec_volatile_element): Add values for
4978 TLB operation and data prefetch.
4979 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4980 functions for TLB operation and data prefetch.
4981 * config/nds32/nds32-intrinsic.md: Likewise.
4982 * config/nds32/nds32_intrinsic.h: Likewise.
4983 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4984 (nds32_print_operand): Likewise.
4985 * config/nds32/nds32.h (nds32_builtins): Likewise.
4987 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4988 Andrew Pinski <pinsika@gcc.gnu.org>
4991 * match.pd: Use constant_boolean_node of correct type instead of
4992 boolean_true_node or boolean_false_node for simplifying
4993 pointer comparisons to zero.
4995 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4997 PR tree-optimization/80021
4998 * tree.c (verify_type_variant): Make error call in verify_variant_match
4999 translatable and remove final full stop.
5001 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5003 * config/nds32/constants.md (unspec_volatile_element): Add
5004 UNSPEC_VOLATILE_EH_RETURN.
5005 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5006 nds32_output_stack_pop): Support dwarf exception handling process.
5007 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5008 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5009 exception handling process.
5010 (nds32_compute_stack_frame): Likewise.
5011 (nds32_return_addr_rtx): Likewise.
5012 (nds32_initial_elimination_offset): Likewise.
5013 (nds32_expand_prologue): Likewise.
5014 (nds32_expand_epilogue): Likewise.
5015 (nds32_dynamic_chain_address): New function.
5016 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5018 (DYNAMIC_CHAIN_ADDRESS): Define.
5019 (EH_RETURN_DATA_REGNO): Define.
5020 (EH_RETURN_STACKADJ_RTX): Define.
5021 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5022 patterns for dwarf exception handling.
5024 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5026 * config/nds32/nds32.h: Clean up obsolete macros.
5028 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5030 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5031 Add enum values for particular instructions.
5032 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5033 particular intrinsic functions.
5034 * config/nds32/nds32-intrinsic.md: Likewise.
5035 * config/nds32/nds32_intrinsic.h: Likewise.
5036 * config/nds32/nds32.h (nds32_builtins): Likewise.
5037 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5038 (btst, ave): New patterns for particular instructions.
5040 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5042 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5043 Add enum values for atomic load/store and memory sync.
5044 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5046 * config/nds32/nds32-intrinsic.md: Likewise.
5047 * config/nds32/nds32_intrinsic.h: Likewise.
5048 * config/nds32/nds32.h (nds32_builtins): Likewise.
5050 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5052 PR tree-optimization/85257
5053 * fold-const.c (native_encode_vector): If not all elts could fit
5054 and off is -1, return 0 rather than offset.
5055 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5056 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5057 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5058 adjust buffer in native_interpret_expr call.
5060 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5062 * config/nds32/constants.md (unspec_volatile_element): Add cache
5063 control enum values.
5064 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5065 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5066 * config/nds32/nds32.c (nds32_cctl_names): New.
5067 (nds32_print_operand): Handle cache control register names.
5068 * config/nds32/nds32.h (nds32_builtins): New enum values.
5069 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5071 * config/nds32/nds32.md (type): Add mmu.
5072 * config/nds32/pipelines.md (simple_insn): Add mmu.
5074 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5076 * config/nds32/nds32.md (type): Remove call.
5077 * config/nds32/pipelines.md (simple_insn): Likewise.
5079 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5081 * config/nds32/constants.md (unspec_volatile_element): Add
5082 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5083 UNSPEC_VOLATILE_FMFCFG.
5084 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5085 description for fmfcfg and fmfcsr.
5086 (bdesc_1arg): Add fmtcsr.
5087 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5088 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5089 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5090 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5091 unspec_fmfcfg): New patterns.
5092 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5093 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5094 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5095 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5096 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5097 __nds32__fmfcfg): Define.
5099 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5101 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5102 intrinsic register names.
5103 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5104 intrinsic register enum values and macros.
5106 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5108 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5109 for load/store addressing form.
5110 (nds32_print_operand_address): Likewise.
5112 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5115 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5116 based on LABEL_REF. Remove useless assertion.
5117 (pic_address_needs_scratch): Fix formatting.
5118 (sparc_legitimize_pic_address): Minor tweaks.
5119 (sparc_delegitimize_address): Adjust assertion accordingly.
5120 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5121 into symbolic_operand.
5122 (movsi_high_pic_label_ref): Likewise.
5123 (movsi_lo_sum_pic_label_ref): Likewise.
5124 (movdi_pic_label_ref): Likewise.
5125 (movdi_high_pic_label_ref): Likewise.
5126 (movdi_lo_sum_pic_label_ref): Likewise.
5128 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5130 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5131 custom LIB_SPEC setup.
5133 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5134 Kito Cheng <kito.cheng@gmail.com>
5136 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5137 * config/riscv/freebsd.h: New.
5139 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5141 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5142 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5145 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5146 Kito Cheng <kito.cheng@gmail.com>
5148 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5149 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5150 * config/nds32/nds32-protos.h (nds32_output_call,
5151 nds32_output_return): Declare.
5152 * config/nds32/nds32.md: Refine all the call and return patterns.
5154 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5157 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5158 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5160 PR rtl-optimization/84872
5161 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5162 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5165 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5167 * expr.c (copy_blkmode_to_reg): Revert 254862.
5168 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5170 2018-04-06 Richard Biener <rguenther@suse.de>
5173 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5174 after seeing a component reference with an adjacent field. Treat
5175 refs to arrays at struct end of external decls similar to
5176 refs to unconstrained commons.
5178 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5181 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5182 look through SAVE_EXPRs with non-side-effects argument. Adjust
5184 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5187 2018-04-06 Richard Biener <rguenther@suse.de>
5190 * alias.c (find_base_term): New wrapper around find_base_term
5191 unwinding CSELIB_VAL_PTR changes.
5192 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5195 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5197 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5199 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5200 constant definitions.
5201 ("nop"): lr 0,0 -> nopr r0
5202 ("nop_lr0", "nop_lr1"): New insn definitions.
5204 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5206 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5207 NDS32_V3PUSH_AVAILABLE_P macro.
5209 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5210 Chung-Ju Wu <jasonwucj@gmail.com>
5212 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5213 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5214 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5215 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5216 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5217 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5218 * config/nds32/constraints.md: New constraints and checking for hard
5219 float configuration.
5220 * config/nds32/iterators.md: New mode iterator and attribute for hard
5221 float configuration.
5222 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5224 * config/nds32/nds32-fpu.md: New file.
5225 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5226 deal with hard float code generation.
5227 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5229 (abi_type, float_reg_number): New enum type.
5230 * config/nds32/nds32-predicates.c: New predicates for hard float.
5231 * config/nds32/nds32-protos.h: Declare functions for hard float.
5232 * config/nds32/nds32.c: Implementation for hard float configuration.
5233 * config/nds32/nds32.h: Definitions for hard float configuration.
5234 * config/nds32/nds32.md: Include hard float machine description and
5235 modify patterns for hard float configuration.
5236 * config/nds32/nds32.opt: New options for hard float configuration.
5237 * config/nds32/predicates.md: New predicates for hard float
5240 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5242 * common/config/nds32/nds32-common.c
5243 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5245 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5248 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5249 CONSTRUCTOR_ELT (ctor, ...)->value.
5251 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5254 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5256 2018-04-05 Tom de Vries <tom@codesourcery.com>
5259 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5262 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5263 Kito Cheng <kito.cheng@gmail.com>
5265 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5266 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5267 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5268 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5270 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5271 Kito Cheng <kito.cheng@gmail.com>
5273 * config/nds32/constraints.md (Ufe): New memory constraint.
5274 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5275 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5276 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5278 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5279 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5281 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5283 * config/nds32/nds32.md: Use optimize_size in the condition for
5284 alu-shift instructions.
5286 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5288 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5290 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5292 * config/nds32/nds32.md (negsi2): Refine pattern.
5294 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5295 Chung-Ju Wu <jasonwucj@gmail.com>
5297 * config/nds32/iterators.md (shift_rotate): New code iterator.
5298 (shift): New code attribute.
5299 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5300 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5301 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5302 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5303 bit-wise operations.
5304 (andsi3, *andsi3): Ditto.
5305 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5306 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5307 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5308 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5309 nds32_ior_operand, nds32_xor_operand): New predicates.
5311 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5313 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5314 (addsi3, subsi3): ... this.
5316 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5318 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5320 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5322 * config/nds32/nds32.md: Adjust indention.
5324 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5326 * config/nds32/nds32.md (feature): New attribute.
5328 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5330 * config/nds32/nds32.md (subtype): New attribute.
5332 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5335 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5336 expansion to perform a bitwise AND of the argument followed by a
5337 boolean negation of the result.
5339 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5341 PR rtl-optimization/84878
5342 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5343 the basic block. Assert the use reference is not artificial and that
5344 it has an associated insn.
5346 2018-04-04 Michael Matz <matz@suse.de>
5348 * builtins.c (compute_objsize): Pass correct operand
5349 to array_at_struct_end_p.
5351 2018-04-04 Richard Biener <rguenther@suse.de>
5354 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5355 from contexts for DINFO_LEVEL_TERSE and below.
5357 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5359 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5360 resiter_operand condition.
5361 * config/nds32/nds32.md (*move<mode>): Ditto.
5363 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5364 Monk Chiang <sh.chiang04@gmail.com>
5366 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5368 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5370 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5372 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5373 Kito Cheng <kito.cheng@gmail.com>
5375 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5376 nds32_cond_code_str, output_cond_branch,
5377 output_cond_branch_compare_zero, nds32_expand_cbranch,
5378 nds32_expand_cstore, nds32_expand_movcc,
5379 nds32_output_cbranchsi4_equality_zero,
5380 nds32_output_cbranchsi4_equality_reg,
5381 nds32_output_cbranchsi4_equality_reg_or_const_int,
5382 nds32_output_cbranchsi4_greater_less_zero: New functions.
5383 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5384 nds32_expand_cstore, nds32_expand_movcc,
5385 nds32_output_cbranchsi4_equality_zero,
5386 nds32_output_cbranchsi4_equality_reg,
5387 nds32_output_cbranchsi4_equality_reg_or_const_int,
5388 nds32_output_cbranchsi4_greater_less_zero): Declare.
5389 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5390 nds32_rimm11s_operand): New predicates.
5391 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5392 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5395 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5397 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5398 * config/nds32/nds32.md: Ditto.
5399 * config/nds32/pipelines.md: Ditto.
5401 2018-04-04 Richard Biener <rguenther@suse.de>
5403 PR tree-optimization/85168
5404 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5405 propagating abnormals.
5407 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5409 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5411 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5412 Kito Cheng <kito.cheng@gmail.com>
5414 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5415 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5416 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5417 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5418 * config/nds32/nds32.md (sibcall_internal): New.
5419 (sibcall_register): Remove.
5420 (sibcall_immediate): Remove.
5421 (sibcall_value_internal): New.
5422 (sibcall_value_register): Remove.
5423 (sibcall_value_immediate): Remove.
5424 * config/nds32/predicates.md (nds32_general_register_operand): New.
5425 (nds32_call_address_operand): New.
5427 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5429 PR rtl-optimization/85167
5430 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5431 bb_defs if *split_p, instead preinitialize it to NULL.
5433 PR tree-optimization/85156
5434 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5435 evaluating the argument multiple times.
5437 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5439 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5441 (_mm_cvtpd_ps): Likewise.
5442 (_mm_cvttpd_epi32): Likewise.
5443 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5444 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5445 vector, pixel, and bool following altivec.h include.
5447 2018-04-03 Martin Sebor <msebor@redhat.com>
5449 * doc/extend.texi (Common Function Attributes): Clarify.
5450 (const attribute): Likewise.
5451 (pure attribute): Likewise.
5453 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5456 * config/i386/i386.c (ix86_expand_vector_set): Use
5457 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5459 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5461 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5462 instructions when changing rounding bits to preserve precision bits
5463 in the x87 control word.
5465 2018-04-03 Martin Liska <mliska@suse.cz>
5467 PR tree-optimization/82491
5468 * rtl.h (strip_offset_and_add): Replace += suboffset with
5469 poly_uint64 () + suboffset.
5471 2018-03-29 Martin Liska <mliska@suse.cz>
5472 Martin Jambor <mjambor@suse.cz>
5475 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5476 param_type is not an integral or pointer type.
5478 2018-04-03 Richard Biener <rguenther@suse.de>
5480 * sese.h (recompute_all_dominators): Remove.
5482 2018-04-02 Martin Sebor <msebor@redhat.com>
5484 * doc/invoke.texi (-Wrestrict): Fix typos.
5486 2018-04-02 Jim Wilson <jimw@sifive.com>
5488 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5489 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5490 (<optab>di3, <optab>si3_extend): Likewise.
5491 (<optab>si3_mask, <optab>si3_mask_1): New.
5492 (<optab>di3_mask, <optab>di3_mask_1): New.
5493 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5494 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5495 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5497 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5499 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5502 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5504 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5505 (nds32_canonicalize_comparison): New function.
5507 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5508 Kito Cheng <kito.cheng@gmail.com>
5509 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5511 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5512 * config/nds32/constants.md (unspec_volatile_element): Add
5513 UNSPEC_VOLATILE_RELAX_GROUP.
5514 * config/nds32/nds32-relax-opt.c: New file.
5515 * config/nds32/nds32-predicates.c
5516 (nds32_symbol_load_store_p): New function.
5517 * config/nds32/nds32-protos.h
5518 (nds32_symbol_load_store_p): Declare function.
5519 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5520 * config/nds32/nds32.c
5521 (nds32_register_pass): New function to register pass.
5522 (nds32_register_passes): New function to register passes.
5523 * config/nds32/nds32.md (relax_group): New pattern.
5524 * config/nds32/nds32.opt (mrelax-hint): New option.
5525 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5527 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5529 * config/nds32/t-nds32: Modify files dependency.
5531 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5533 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5534 (PROFILE_HOOK): Define its implementation.
5536 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5538 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5539 type and 32-bit size.
5541 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5544 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5545 (V_128_256): New mode iterator.
5546 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5547 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5548 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5550 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5551 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5553 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5556 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5557 NaN inputs correctly.
5559 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5562 * config/rs6000/vsx.md (??r): New mode attribute.
5563 (*vsx_mov<mode>_64bit): Use it.
5564 (*vsx_mov<mode>_32bit): Likewise.
5566 2018-03-30 Martin Sebor <msebor@redhat.com>
5568 PR tree-optimization/84818
5569 * builtins.c (check_access): Use warning_n.
5571 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5574 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5576 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5579 2018-03-30 Julia Koval <julia.koval@intel.com>
5582 * x86-tune.def (movx, partial_reg_dependency): Enable for
5585 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5588 * lra-constraints.c (process_alt_operands): Move setting
5589 this_alternative_matches below.
5591 2018-03-29 Martin Liska <mliska@suse.cz>
5594 * doc/invoke.texi: Document how LTO works with debug info.
5595 Describe auto-load support of binutils. Mention 'x86-64'
5596 as valid option value of -march option.
5598 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5600 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5603 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5604 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5605 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5608 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5611 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5612 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5613 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5614 for RS6000_BTM_POWERPC64.
5615 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5616 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5617 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5622 2018-03-28 Carl Love <cel@us.ibm.com>
5625 2017-09-27 Carl Love <cel@us.ibm.com>
5627 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5628 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5629 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5632 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5634 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5635 instead of __vector bool.
5636 (_mm_max_pu8): Likewise.
5637 (_mm_min_pi16): Likewise.
5639 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5642 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5643 (DIVWEUO): Likewise.
5645 (DIVDEUO): Likewise.
5646 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5647 DIVWEUO and DIVDEUO.
5648 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5649 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5650 (div_extend): Likewise.
5651 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5653 (__builtin_divweuo): Likewise.
5654 (__builtin_divdeo): Likewise.
5655 (__builtin_divdeuo): Likewise.
5657 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5660 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5661 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5663 PR tree-optimization/82004
5664 * gimple-match-head.c (optimize_pow_to_exp): New function.
5665 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5666 Don't fold to exp if optimize_pow_to_exp is false.
5668 2018-03-28 Martin Liska <mliska@suse.cz>
5671 * calls.c (initialize_argument_information): Fix trailing space.
5672 * common.opt: Fix typo and provide better explanation for
5673 -fsanitize-coverage option.
5674 * config/i386/i386.opt: Fix typo.
5676 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5677 Martin Liska <mliska@suse.cz>
5680 * gimplify.c (asan_poison_variable): Don't do the check for
5681 gimplify_omp_ctxp here.
5682 (gimplify_decl_expr): Do it here.
5683 (gimplify_target_expr): Likewise.
5685 2018-03-28 Martin Liska <mliska@suse.cz>
5688 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5689 chkp_type_bounds_count if MPX is not enabled.
5691 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5693 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5695 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5698 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5699 function to create the function decl for complex long double
5700 multiply and divide for -mabi=ieeelongdouble.
5701 (init_float128_ieee): Call it.
5703 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5706 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5707 -fcf-protection=branch -mibt.
5708 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5710 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5713 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5715 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5718 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5719 extern array declarations.
5721 2018-03-27 Richard Biener <rguenther@suse.de>
5724 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5725 explicit single_use checks.
5727 2018-03-27 Richard Biener <rguenther@suse.de>
5729 PR tree-optimization/85082
5730 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5733 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5735 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5736 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5737 Turn on fasynchronous-unwind-tables and funwind-tables.
5739 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5742 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5743 (*bmi_blsr_<mode>_ccz): Ditto.
5745 2018-03-26 Tom de Vries <tom@codesourcery.com>
5747 PR tree-optimization/85063
5748 * omp-general.c (offloading_function_p): New function. Factor out
5750 * omp-offload.c (pass_omp_target_link::gate): ... here.
5751 * omp-general.h (offloading_function_p): Declare.
5752 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5753 with attribute omp declare target for offloading functions.
5755 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5757 PR tree-optimization/84005
5758 * tree-data-ref.h (get_base_for_alignment): Declare.
5759 * tree-data-ref.c (get_base_for_alignment_1): New function.
5760 (get_base_for_alignment): Likewise.
5761 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5762 get_base_for_alignment to find a suitable base object, instead
5763 of always using drb->base_address.
5765 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5768 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5769 known size by default.
5771 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5774 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5775 and non BLKmode operands.
5777 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5780 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5781 Clean up attributes.
5783 2018-03-23 Richard Biener <rguenther@suse.de>
5786 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5787 we are going to emit early debug for LTO.
5789 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5792 * function.c (match_asm_constraints_1): Don't optimize if input
5793 doesn't satisfy general_operand predicate for output's mode.
5796 * alias.c (write_dependence_p): Don't require for x_canonicalized
5797 non-VOIDmode if x has VOIDmode.
5800 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5801 just don't try to optimize it rather than assert it never happens.
5803 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5805 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5806 macro expansions for definition of ST_INTERNAL_<mode> and
5807 LD_INTERNAL_<mode> builtins.
5808 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5810 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5812 (altivec_expand_st_builtin): Likewise.
5813 (altivec_expand_builtin): Remove calls to deleted functions.
5814 (rs6000_address_for_altivec): Delete this function.
5815 * config/rs6000/vector.md: Remove expands for
5816 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5818 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5821 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5822 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5823 re-computing once computed.
5824 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5825 (arm_init_machine_status): Initialize
5826 machine->static_chain_stack_bytes.
5828 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5831 * doc/extend.texi: Add four new prototypes for vec_ld.
5832 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5833 definitions for more logical presentation.
5834 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5835 entries for V1TI variants of __builtin_altivec_ld builtin.
5836 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5837 handling of V1TI variant of LVX icode pattern.
5838 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5839 (rs6000_gimple_fold_builtin): Likewise.
5840 (altivec_init_builtins): Add code to define
5841 __builtin_altivec_lvx_v1ti function.
5843 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5846 * function.c (match_asm_constraints_1): Don't do the optimization
5847 if input isn't a REG, SUBREG, MEM or constant.
5849 2018-03-22 Tom de Vries <tom@codesourcery.com>
5851 PR tree-optimization/84956
5852 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5853 bb_has_abnormal_pred.
5855 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5858 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5859 DECL_INITIAL (decl) to decl at the end.
5860 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5863 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5865 * doc/extend.texi (__builtin_tgmath): Document when complex
5866 integer types are treated as _Complex _Float64.
5868 2018-03-21 Tom de Vries <tom@codesourcery.com>
5870 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5872 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5874 PR tree-optimization/84960
5875 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5876 if it is ENTRY block, move them into single succ of ENTRY in that case.
5878 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5880 PR tree-optimization/84811
5881 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5882 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5883 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5884 (known_subrange_p): Update accordingly. Cast each value involved
5885 in the size comparison, rather than casting the result of the
5888 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5890 PR tree-optimization/84982
5891 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5892 by flipping the least significant bit rather than all bits from
5893 bitpos to bitpos + bitsize - 1.
5895 2018-03-21 Nathan Sidwell <nathan@acm.org>
5897 * doc/extend.texi (Deprecated Features): Remove mention of
5898 long-deleted deprecations.
5900 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5903 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5904 * configure: Regenerate.
5906 2018-03-21 Tom de Vries <tom@codesourcery.com>
5908 PR tree-optimization/83126
5909 * tree-parloops.c (num_phis): New function.
5910 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5912 2018-03-21 Nathan Sidwell <nathan@acm.org>
5914 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5915 mention anon-struct/union members and trailing attributes.
5917 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5919 PR tree-optimization/84969
5920 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5921 builtin memset partitions if they set different rhs values.
5923 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5925 PR rtl-optimization/84989
5926 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5927 VEC_DUPLICATE with scalar result mode.
5929 2018-03-21 Martin Liska <mliska@suse.cz>
5932 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5933 not intended return statement.
5935 2018-03-21 Martin Liska <mliska@suse.cz>
5938 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5939 (chkp_find_bound_slots_1): Limit number of iterations.
5941 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5944 * Minor grammar fixes for x86 options.
5946 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5949 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5950 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5952 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5955 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5956 (altivec_lvx_<mode>_1op): Likewise.
5957 (altivec_stvx_<mode>_2op): Likewise.
5958 (altivec_stvx_<mode>_1op): Likewise.
5959 (altivec_lvx_<VM2:mode>): New define_expand.
5960 (altivec_stvx_<VM2:mode>): Likewise.
5961 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5962 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5963 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5964 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5965 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5966 (rs6000_gen_lvx): Likewise.
5967 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5968 (altivec_expand_stv_builtin): Likewise.
5969 (altivec_expand_builtin): Likewise.
5970 * config/rs6000/vector.md: Likewise.
5972 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5975 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5978 2018-03-20 Richard Biener <rguenther@suse.de>
5981 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5982 sign-conversions as zero, fall back to standard scalar_stmt
5985 2018-03-20 Martin Liska <mliska@suse.cz>
5988 * predict.c (rebuild_frequencies): Handle case when we have
5989 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5991 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5994 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5995 flag_section_anchors.
5996 * varasm.c (use_blocks_for_decl_p): Remove hack for
5997 dw2_force_const_mem.
6000 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6002 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6003 be created, use lowpart_subreg of operands[0] rather than operands[0]
6005 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6006 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6007 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6008 and n constraint instead of aarch64_shift_imm_di and Usd.
6009 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6010 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6012 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6015 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6016 to favor GPR over NEON registers.
6017 (<shift>di3_neon): Likewise.
6019 2018-03-20 Tom de Vries <tom@codesourcery.com>
6022 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6023 (nvptx_process_pars): Emit bar.sync asap and alap.
6025 2018-03-20 Tom de Vries <tom@codesourcery.com>
6028 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6029 seen_label if seen_label is already set.
6031 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6034 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6035 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6036 Use 1U instead of 1. Formatting fixes.
6039 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6040 instead of TREE_TYPE (s1) for the return value.
6042 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6044 PR tree-optimization/84946
6045 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6046 bitsize + bitsize in poly_uint64 rather than poly_int64.
6049 * dwarf2asm.c: Include fold-const.c.
6050 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6051 of decl rather than decl itself.
6053 PR rtl-optimization/84643
6054 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6056 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6059 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6060 calling assemble_variable.
6062 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6065 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6066 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6068 2018-03-19 Jim Wilson <jimw@sifive.com>
6071 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6072 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6073 (riscv_first_stack_step): Likewise.
6074 (riscv_option_override): Use STACK_BOUNDARY instead of
6076 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6078 (BIGGEST_ALIGNMENT): Set to 128.
6079 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6080 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6083 2018-03-19 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/84933
6086 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6087 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6089 2018-03-19 Richard Biener <rguenther@suse.de>
6091 PR tree-optimization/84859
6092 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6093 (cond_if_else_store_replacement): Perform sinking operation on
6094 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6095 Generalize what a BB with a single eligible store is.
6097 2018-03-19 Richard Biener <rguenther@suse.de>
6099 PR tree-optimization/84929
6100 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6101 chrec_is_positive against non-chrec arg.
6103 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6106 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6108 2018-03-18 Martin Liska <mliska@suse.cz>
6110 PR rtl-optimization/84635
6111 * regrename.c (build_def_use): Use matches_mode only when
6114 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6116 PR tree-optimization/84913
6117 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6118 vectorize chains of COND_EXPRs.
6120 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6122 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6124 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6126 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6128 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6130 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6132 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6133 Kito Cheng <kito.cheng@gmail.com>
6135 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6136 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6137 (nds32_adjust_reg_alloc_order): New function.
6138 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6140 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6142 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6143 nds32_print_operand, nds32_print_operand_address): Use
6144 HOST_WIDE_INT_PRINT_DEC instead.
6146 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6148 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6150 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6153 * config/i386/i386.c (initial_ix86_tune_features,
6154 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6156 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6157 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6158 rather than 1u << ix86_tune. Formatting fix.
6159 (ix86_option_override_internal): Change ix86_arch_mask from
6160 unsigned int to unsigned HOST_WIDE_INT, initialize to
6161 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6162 (ix86_function_specific_restore): Likewise.
6164 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6167 * postreload.c (reload_combine_recognize_pattern): Perform
6168 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6169 truncate_int_for_mode the result for the destination's mode.
6172 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6173 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6175 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6178 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6181 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6182 regno_allocno_class_array and sorted_pseudos.
6183 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6184 insns where regno is used.
6186 2018-03-16 Martin Liska <mliska@suse.cz>
6189 * multiple_target.c (create_dispatcher_calls): Redirect
6190 reference in the symbol table.
6192 2018-03-16 Martin Liska <mliska@suse.cz>
6195 * multiple_target.c (create_dispatcher_calls): Redirect also
6198 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6202 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6203 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6204 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6206 2018-03-16 Julia Koval <julia.koval@intel.com>
6208 * doc/invoke.texi (Skylake Server): Add CLWB.
6209 Cannonlake): Remove CLWB.
6211 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6213 PR tree-optimization/84841
6214 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6216 (FLOAT_ONE_CONST_TYPE): Define.
6217 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6218 (sort_by_operand_rank): Put entries with higher constant_type last
6219 rather than first to match comments.
6221 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6223 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6226 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6229 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6230 check and use address of target always.
6232 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6235 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6236 (indirect_thunk_bnd_needed): Likewise.
6237 (indirect_thunks_used): Likewise.
6238 (indirect_thunks_bnd_used): Likewise.
6239 (indirect_return_needed): New.
6240 (indirect_return_bnd_needed): Likewise.
6241 (output_indirect_thunk_function): Add a bool argument for
6243 (output_indirect_thunk_function): Don't generate alias for
6244 function return thunk.
6245 (ix86_code_end): Call output_indirect_thunk_function to generate
6246 function return thunks.
6247 (ix86_output_function_return): Set indirect_return_bnd_needed
6248 and indirect_return_needed instead of indirect_thunk_bnd_needed
6249 and indirect_thunk_needed.
6251 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6253 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6254 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6255 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6257 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6258 Paul Hua <paul.hua.gm@gmail.com>
6261 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6263 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6265 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6266 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6269 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6272 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6273 instead of GET_MODE_SIZE when comparing Units.
6275 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6278 * varasm.c (hash_section): Return an unchangeble hash value
6279 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6280 Return !aarch64_can_use_per_function_literal_pools_p ().
6282 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6285 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6286 rather than address of cmode as last argument to prepare_cmp_insn.
6288 2018-03-15 Julia Koval <julia.koval@intel.com>
6290 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6291 F_AVX512VNNI, F_AVX512BITALG): New.
6293 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6296 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6297 insn for floating-point loads and stores.
6299 2018-03-14 Carl Love <cel@us.ibm.com>
6301 * config/rs6000/rs6000-c.c: Add macro definitions for
6302 ALTIVEC_BUILTIN_VEC_PERMXOR.
6303 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6304 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6305 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6306 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6308 * config/doc/extend.texi: Add prototypes for vec_permxor.
6310 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6313 * diagnostic-show-locus.c (class layout_point): Convert m_line
6314 from int to linenum_type.
6315 (line_span::comparator): Use linenum "compare" function when
6316 comparing line numbers.
6317 (test_line_span): New function.
6318 (layout_range::contains_point): Convert param "row" from int to
6320 (layout_range::intersects_line_p): Likewise.
6321 (layout::will_show_line_p): Likewise.
6322 (layout::print_source_line): Likewise.
6323 (layout::should_print_annotation_line_p): Likewise.
6324 (layout::print_annotation_line): Likewise.
6325 (layout::print_leading_fixits): Likewise.
6326 (layout::annotation_line_showed_range_p): Likewise.
6327 (struct line_corrections): Likewise for field m_row.
6328 (line_corrections::line_corrections): Likewise for param "row".
6329 (layout::print_trailing_fixits): Likewise.
6330 (layout::get_state_at_point): Likewise.
6331 (layout::get_x_bound_for_row): Likewise.
6332 (layout::print_line): Likewise.
6333 (diagnostic_show_locus): Likewise for locals "last_line" and
6335 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6336 * input.c (selftest::test_linenum_comparisons): New function.
6337 (selftest::input_c_tests): Call it.
6338 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6339 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6340 * selftest.h (ASSERT_GT): New macro.
6341 (ASSERT_GT_AT): New macro.
6342 (ASSERT_LT): New macro.
6343 (ASSERT_LT_AT): New macro.
6345 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6347 PR rtl-optimization/84780
6348 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6350 2018-03-14 Martin Liska <mliska@suse.cz>
6352 * tree.c (record_node_allocation_statistics): Use
6353 get_stats_node_kind.
6354 (get_stats_node_kind): New function extracted from
6355 record_node_allocation_statistics.
6356 (free_node): Use get_stats_node_kind.
6358 2018-03-14 Richard Biener <rguenther@suse.de>
6360 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6361 that the value-set of ANTIC_IN doesn't grow.
6364 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6366 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6367 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6369 2018-03-14 Julia Koval <julia.koval@intel.com>
6371 * config.gcc (icelake-client, icelake-server): New.
6373 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6374 (initial_ix86_arch_features): Ditto.
6375 (PTA_SKYLAKE): Add SGX.
6376 (PTA_ICELAKE): Remove.
6377 (PTA_ICELAKE_CLIENT): New.
6378 (PTA_ICELAKE_SERVER): New.
6379 (ix86_option_override_internal): Split up icelake on icelake client and
6381 (get_builtin_code_for_version): Ditto.
6382 (fold_builtin_cpu): Ditto.
6383 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6384 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6385 * config/i386/i386.h (processor_type): Ditto.
6386 * doc/invoke.texi: Ditto.
6388 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6391 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6392 INTEGER_CST offset, add it together with bitpos / 8 and
6393 sign extend based on POINTER_SIZE.
6397 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6400 * config/i386/constraints.md (Yc): New register constraint.
6401 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6402 Use Yc constraint for alternative 2 of operand 0. Remove
6403 preferred_for_speed attribute.
6405 2018-03-14 Richard Biener <rguenther@suse.de>
6407 PR tree-optimization/84830
6408 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6409 with the old one to avoid oscillations.
6411 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6414 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6416 (assign_by_spills): Return a flag of reload assignment failure.
6417 Do not process the reload assignment failures. Do not spill other
6418 reload pseudos if they has the same reg class. Update n if
6420 (lra_assign): Add a return arg. Set up from the result of
6421 assign_by_spills call.
6422 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6423 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6424 usage_insns if it is not NULL.
6425 (spill_hard_reg_in_range): New function.
6426 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6427 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6428 function prototypes.
6429 (lra_assign): Change prototype.
6430 * lra.c (lra): Add code to deal with fails by splitting hard reg
6433 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6435 * config/riscv/riscv.opt (mrelax): New option.
6436 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6437 "norelax" when riscv_mrelax is disabled.
6438 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6440 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6443 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6444 reassociation for int modes.
6446 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6448 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6449 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6451 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6452 * config/aarch64/aarch64-sve.md
6453 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6454 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6455 (*extend<mode><Vwide>2): Rename to...
6456 (aarch64_sve_extend<mode><Vwide>2): ...this.
6457 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6458 renaming the old pattern to...
6459 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6461 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6462 define_expand, renaming the old pattern to...
6463 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6464 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6465 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6466 account when deciding which SVE instruction the optab should use.
6467 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6469 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6471 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6472 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6473 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6474 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6475 (tlsdesc_small_<mode>): Turn a define_expand and use
6476 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6477 (tlsdesc_small_advsimd_<mode>): ...this.
6478 (tlsdesc_small_sve_<mode>): New pattern.
6480 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6482 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6483 (UNSPEC_UMUL_HIGHPART): New constants.
6484 (MUL_HIGHPART): New int iteraor.
6485 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6486 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6488 (*<su>mul<mode>3_highpart): New define_insn.
6490 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6493 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6496 2018-03-13 Martin Liska <mliska@suse.cz>
6499 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6501 (sem_item_optimizer::~sem_item_optimizer): Release it.
6502 (sem_item_optimizer::merge_classes): Register variable aliases.
6503 (sem_item_optimizer::fixup_pt_set): New function.
6504 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6505 * ipa-icf.h: Declare new variables and functions.
6507 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6510 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6511 integer_pow2p@2 and test integer_pow2p in condition.
6512 (A < 0 ? C : 0): Similarly for @1.
6515 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6516 characters starting at p contain '\0' character, don't look beyond
6520 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6521 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6524 * reg-stack.c (change_stack): Change update_end var from int to
6525 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6526 also call set_block_for_insn on the newly added insns and rescan.
6529 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6530 on the last operand.
6533 * tree.c (stabilize_reference_1): Return save_expr (e) for
6534 STATEMENT_LIST even if it doesn't have side-effects.
6536 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6538 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6540 2018-03-12 Renlin Li <renlin.li@arm.com>
6542 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6543 aarch64_output_scalar_simd_mov_immediate.
6545 2018-03-12 Martin Sebor <msebor@redhat.com>
6547 PR tree-optimization/83456
6548 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6549 for perfectly overlapping calls to memcpy.
6550 (gimple_fold_builtin_memory_chk): Same.
6551 (gimple_fold_builtin_strcpy): Handle no-warning.
6552 (gimple_fold_builtin_stxcpy_chk): Same.
6553 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6555 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6557 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6558 parameter. Use it for SFmode.
6559 (rs6000_function_arg_advance_1): Adjust.
6560 (rs6000_function_arg): Adjust.
6561 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6563 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6565 PR rtl-optimization/84169
6566 PR rtl-optimization/84780
6567 * combine.c (can_combine_p): Check for a 2-insn combination whether
6568 the destination register is used between the two insns, too.
6570 2018-03-12 Richard Biener <rguenther@suse.de>
6572 PR tree-optimization/84803
6573 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6574 for refs DR analysis didn't process.
6576 2018-03-12 Richard Biener <rguenther@suse.de>
6578 PR tree-optimization/84777
6579 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6580 force-vectorize loops ignore whether we are optimizing for size.
6582 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6584 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6585 (TARGET_MD_ASM_ADJUST): Define.
6587 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6588 Kito Cheng <kito.cheng@gmail.com>
6589 Chung-Ju Wu <jasonwucj@gmail.com>
6591 * config/nds32/nds32.c (nds32_compute_stack_frame,
6592 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6593 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6594 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6595 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6596 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6597 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6598 * config/nds32/nds32.md (prologue, epilogue): Use macro
6599 NDS32_V3PUSH_AVAILABLE_P to do checking.
6601 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6604 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6605 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6606 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6607 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6608 addition of most attributes on !orig_type_die or the attribute not
6609 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6611 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6612 Chung-Ju Wu <jasonwucj@gmail.com>
6614 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6616 * config/nds32/nds32.opt (mvh): New option.
6618 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6619 Chung-Ju Wu <jasonwucj@gmail.com>
6621 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6623 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6624 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6627 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6628 Chung-Ju Wu <jasonwucj@gmail.com>
6630 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6632 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6633 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6635 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6636 Kito Cheng <kito.cheng@gmail.com>
6637 Chung-Ju Wu <jasonwucj@gmail.com>
6639 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6640 UNSPEC_FFMISM and UNSPEC_FLMISM.
6641 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6642 for ffb, ffmism and flmism.
6643 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6644 (unspec_ffmism): Ditto.
6645 (unspec_flmism): Ditto.
6646 (nds32_expand_builtin_impl): Check if string extension is available.
6647 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6648 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6650 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6653 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6656 * lra-assigns.c (assign_by_spills): Return a flag of reload
6657 assignment failure. Do not process the reload assignment
6658 failures. Do not spill other reload pseudos if they has the same
6660 (lra_assign): Add a return arg. Set up from the result of
6661 assign_by_spills call.
6662 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6663 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6664 usage_insns if it is not NULL.
6665 (spill_hard_reg_in_range): New function.
6666 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6667 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6668 function prototypes.
6669 (lra_assign): Change prototype.
6670 * lra.c (lra): Add code to deal with fails by splitting hard reg
6673 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6676 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6678 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6681 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6682 (dw_val_node): Add val_symbolic_view.
6683 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6684 (symview_upper_bound): New.
6685 (new_line_info_table): Initialize symviews_since_reset.
6686 (dwarf2out_source_line): Count symviews_since_reset and set
6687 symview_upper_bound.
6688 (dw_val_equal_p): Handle symview.
6689 (add_AT_symview): New.
6690 (print_dw_val): Handle symview.
6691 (attr_checksum, attr_checksum_ordered): Likewise.
6692 (same_dw_val_p, size_of_die): Likewise.
6693 (value_format, output_die): Likewise.
6694 (add_high_low_attributes): Use add_AT_symview for entry_view.
6695 (dwarf2out_finish): Reset symview_upper_bound, clear
6698 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6701 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6702 Add strict argument and use it.
6703 (rs6000_split_multireg_move): Update for new strict argument.
6704 (mem_operand_gpr): Disallow all non-offsettable addresses.
6705 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6707 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6710 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6711 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6712 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6715 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6716 decl, use remap_type if we want to use the type.
6718 2018-03-09 Martin Sebor <msebor@redhat.com>
6720 PR tree-optimization/84526
6721 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6723 (builtin_access::generic_overlap): Be prepared to handle non-array
6726 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6728 PR rtl-optimization/84682
6729 * lra-constraints.c (process_address_1): Check is_address flag
6730 for address constraints.
6731 (process_alt_operands): Likewise.
6732 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6733 preprocess_constraints.
6734 * recog.h (preprocess_constraints): Add oploc parameter.
6736 * recog.c (preprocess_constraints): Test address_operand for
6737 CT_ADDRESS constraints.
6739 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6742 * lra-assigns.c (assign_by_spills): Return a flag of reload
6743 assignment failure. Do not process the reload assignment
6744 failures. Do not spill other reload pseudos if they has the same
6746 (lra_assign): Add a return arg. Set up from the result of
6747 assign_by_spills call.
6748 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6749 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6750 usage_insns if it is not NULL.
6751 (spill_hard_reg_in_range): New function.
6752 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6753 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6754 function prototypes.
6755 (lra_assign): Change prototype.
6756 * lra.c (lra): Add code to deal with fails by splitting hard reg
6759 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6762 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6763 Accept complain bool parameter. Only emit errors if it is true.
6764 (arm_parse_cpu_option_name): Likewise.
6765 (arm_target_thumb_only): Adjust callers of the above.
6766 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6767 prototype to take a default true bool parameter.
6768 (arm_parse_arch_option_name): Likewise.
6770 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6771 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6775 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6776 * configure: Regenerate.
6777 * configure.ac ("linker --version-script option"): New.
6778 ("linker soname option"): New.
6780 2018-03-09 Richard Biener <rguenther@suse.de>
6782 PR tree-optimization/84775
6783 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6784 immediate uses of predicate stmts and mark them modified.
6787 PR tree-optimization/84178
6788 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6790 (version_loop_for_if_conversion): Delay update_ssa call.
6791 (tree_if_conversion): Delay update_ssa until after predicate
6794 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6797 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6798 when the function accesses prior frames.
6800 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6803 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6804 gen_llsym, otherwise call maybe_gen_llsym.
6807 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6808 has ',' character inside of it.
6810 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6813 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6814 as clobbering CC_REGNUM.
6816 2018-03-08 Richard Biener <rguenther@suse.de>
6819 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6820 (follow_copies_to_constant): Do not follow SSA names registered
6823 2018-03-08 Richard Biener <rguenther@suse.de>
6825 PR tree-optimization/84178
6826 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6828 (version_loop_for_if_conversion): Delay update_ssa call.
6829 (tree_if_conversion): Delay update_ssa until after predicate
6832 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6834 PR tree-optimization/84178
6835 * tree-if-conv.c (release_bb_predicate): Remove the
6836 the assertion that the stmts have NULL use_ops.
6837 Discard the statements, asserting that they haven't
6838 yet been added to a BB.
6840 2018-03-08 Richard Biener <rguenther@suse.de>
6842 PR tree-optimization/84746
6843 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6844 (phi_translate): Pass in destination ANTIC_OUT set.
6845 (phi_translate_1): Likewise. For a simplified result lookup
6846 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6847 (phi_translate_set): Adjust.
6848 (do_pre_regular_insertion): Likewise.
6849 (do_pre_partial_partial_insertion): Likewise.
6851 2018-03-08 Martin Liska <mliska@suse.cz>
6853 PR gcov-profile/84735
6854 * doc/gcov.texi: Document usage of profile files.
6855 * gcov-io.h: Document changes in the format.
6857 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6861 * dwarf2out.c (struct dw_line_info_table): Update comments for
6863 (FORCE_RESET_NEXT_VIEW): New.
6864 (FORCE_RESETTING_VIEW_P): New.
6865 (RESETTING_VIEW_P): Check for -1 too.
6866 (ZERO_VIEW_P): Likewise.
6867 (new_line_info_table): Force-reset next view.
6868 (dwarf2out_begin_function): Likewise.
6869 (dwarf2out_source_line): Simplify zero_view_p initialization.
6870 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6871 view directly. Omit view when omitting .loc at line 0.
6873 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6875 PR tree-optimization/84740
6876 * tree-switch-conversion.c (process_switch): Call build_constructors
6877 only if info.phi_count is non-zero.
6879 PR tree-optimization/84739
6880 * tree-tailcall.c (find_tail_calls): Check call arguments against
6881 DECL_ARGUMENTS (current_function_decl) rather than
6882 DECL_ARGUMENTS (func) when checking for tail recursion.
6884 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6886 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6887 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6888 Volker Reichelt's entry and add entries for people that perform
6889 GCC fuzzy testing and report numerous bugs.
6891 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6894 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6895 readonly data in sdata, if that is disabled.
6896 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6897 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6898 -mreadonly-in-sdata option.
6900 2018-03-07 Martin Sebor <msebor@redhat.com>
6902 PR tree-optimization/84468
6903 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6904 basic block when looking for nul assignment.
6906 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6909 * except.h (output_function_exception_table): Adjust prototype.
6910 * except.c (output_function_exception_table): Remove FNNAME parameter
6911 and add SECTION parameter. Ouput one part of the table at a time.
6912 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6913 the first part of the exception table and emit unwind directives.
6914 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6915 (i386_pe_seh_cold_init): Likewise.
6916 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6917 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6918 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6919 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6920 * config/i386/winnt.c: Include except.h.
6921 (struct seh_frame_state): Add reg_offset, after_prologue and
6922 in_cold_section fields.
6923 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6924 (i386_pe_seh_cold_init): New function.
6925 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6926 to seh->in_cold_section.
6927 (seh_emit_push): Record the offset of the push.
6928 (seh_emit_save): Record the offet of the save.
6929 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6930 Test seh->after_prologue to disregard the epilogue.
6931 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6932 (i386_pe_end_cold_function): New function.
6934 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6937 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6938 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6941 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6943 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6944 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6947 * multiple_target.c: Include tree-inline.h and intl.h.
6948 (expand_target_clones): Diagnose and fail if node->definition and
6949 !tree_versionable_function_p (node->decl).
6951 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6953 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6955 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6956 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6957 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6959 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6962 * combine.c (try_combine): Use reg_or_subregno instead of handling
6963 just paradoxical SUBREGs and REGs.
6965 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6967 * config/arc/arc.c (arc_finalize_pic): Remove function.
6968 (arc_must_save_register): We use single base PIC register, remove
6969 checks to save/restore the PIC register.
6970 (arc_expand_prologue): Likewise.
6971 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6973 (arc_verify_short): Likewise.
6974 (arc_attr_type): Likewise.
6975 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6976 (walk_stores): Likewise.
6977 (arc_address_cost): Make it static.
6978 (arc_verify_short): Likewise.
6979 (branch_dest): Likewise.
6980 (arc_attr_type): Likewise.
6981 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6982 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6983 (arc_final_prescan_insn): Remove inserting the nops due to
6984 hardware hazards. It is done in reorg step.
6985 (insn_length_variant_t): Remove.
6986 (insn_length_parameters_t): Likewise.
6987 (arc_insn_length_parameters): Likewise.
6988 (arc_get_insn_variants): Likewise.
6989 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6991 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6994 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6997 PR tree-optimization/84687
6998 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7000 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7002 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7004 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7005 Rename to ppc_speculation_barrier.
7006 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7007 __builtin_ppc_speculation_barrier.
7009 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7012 * combine.c (combine_simplify_rtx): Don't try to simplify if
7013 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7016 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7018 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7019 to 32 bytes when compiling for POWER9.
7021 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7024 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7025 regparm >= 3 with no arg reg available also for calls with
7026 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7029 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7031 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7033 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7036 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7038 2018-03-05 Richard Biener <rguenther@suse.de>
7040 PR tree-optimization/84486
7041 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7042 When inserting a __builtin_assume_aligned call set the LHS
7043 SSA name alignment info accordingly.
7045 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7047 PR tree-optimization/84114
7048 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7049 Avoid reassociation of FLOAT_MODE addition.
7051 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7053 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7054 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7055 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7056 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7057 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7058 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7059 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7061 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7062 __builtin_ia32_wbinvd): New builtins.
7063 (SPECIAL_ARGS2): New.
7064 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7065 (SPECIAL_ARGS2): New.
7066 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7067 (ix86_valid_target_attribute_inner_p): Ditto.
7068 (ix86_init_mmx_sse_builtins): Add special_args2.
7069 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7070 TARGET_WBNOINVD_P): New.
7071 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7072 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7073 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7074 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7075 * config/i386/pconfigintrin.h: New file.
7076 * config/i386/wbnoinvdintrin.h: Ditto.
7077 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7078 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7080 2018-03-05 Richard Biener <rguenther@suse.de>
7082 PR tree-optimization/84670
7083 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7085 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7086 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7087 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7088 doesn't grow if all successors (recursively) were visited at least
7091 2018-03-05 Richard Biener <rguenther@suse.de>
7093 PR tree-optimization/84650
7094 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7095 if executed in the loop pipeline.
7097 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7099 * doc/configfiles.texi (Configuration Files): Move info about
7100 conditionalizing $target-protos.h to...
7101 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7102 differs from $target-protos.h.
7104 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7105 Chung-Ju Wu <jasonwucj@gmail.com>
7107 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7108 * config/nds32/nds32-multiple.md (setmemsi): Define.
7109 * config/nds32/nds32-memory-manipulation.c
7110 (nds32_gen_dup_4_byte_to_word_value): New.
7111 (emit_setmem_word_loop): New.
7112 (emit_setmem_byte_loop): New.
7113 (nds32_expand_setmem_loop): New.
7114 (nds32_expand_setmem_loop_v3m): New.
7115 (nds32_expand_setmem_unroll): New.
7116 (nds32_expand_setmem): New.
7118 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7119 Chung-Ju Wu <jasonwucj@gmail.com>
7121 * config/nds32/nds32-memory-manipulation.c
7122 (nds32_emit_load_store): New.
7123 (nds32_emit_post_inc_load_store): New.
7124 (nds32_emit_mem_move): New.
7125 (nds32_emit_mem_move_block): New.
7126 (nds32_expand_movmemsi_loop_unknown_size): New.
7127 (nds32_expand_movmemsi_loop_known_size): New.
7128 (nds32_expand_movmemsi_loop): New.
7129 (nds32_expand_movmemsi_unroll): New.
7130 (nds32_expand_movmemqi): Rename ...
7131 (nds32_expand_movmemsi): ... to this.
7132 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7133 (movmemsi): ... to this.
7134 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7135 (nds32_expand_movmemsi): ... to this.
7137 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7138 Monk Chiang <sh.chiang04@gmail.com>
7139 Chung-Ju Wu <jasonwucj@gmail.com>
7141 * config/nds32/nds32-protos.h
7142 (nds32_expand_load_multiple): New arguments.
7143 (nds32_expand_store_multiple): Ditto.
7144 (nds32_valid_multiple_load_store): Rename ...
7145 (nds32_valid_multiple_load_store_p): ... to this.
7146 * config/nds32/nds32-memory-manipulation.c
7147 (nds32_expand_load_multiple): Refine implementation.
7148 (nds32_expand_store_multiple): Ditto.
7149 * config/nds32/nds32-multiple.md
7150 (load_multiple): Update nds32_expand_load_multiple interface.
7151 (store_multiple): Update nds32_expand_store_multiple interface.
7152 * config/nds32/nds32-predicates.c
7153 (nds32_valid_multiple_load_store): Rename ...
7154 (nds32_valid_multiple_load_store_p): ... to this and refine
7156 * config/nds32/predicates.md
7157 (nds32_load_multiple_and_update_address_operation): New predicate.
7158 (nds32_store_multiple_and_update_address_operation): New predicate.
7160 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7161 Chung-Ju Wu <jasonwucj@gmail.com>
7163 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7164 (combo): New attribute.
7165 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7167 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7169 * config/nds32/nds32.opt: Change -mcmodel= default value.
7171 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7172 Monk Chiang <sh.chiang04@gmail.com>
7173 Chung-Ju Wu <jasonwucj@gmail.com>
7175 * config/nds32/constants.md (unspec_element): New enum.
7176 * config/nds32/constraints.md (Umw): New constraint.
7177 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7178 * config/nds32/nds32-intrinsic.md: Likewise.
7179 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7180 (nds32_valid_smw_lwm_base_p): New.
7181 (nds32_output_smw_single_word): New.
7182 (nds32_output_lmw_single_word): New.
7183 (nds32_expand_unaligned_load): New.
7184 (nds32_expand_unaligned_store): New.
7185 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7186 (nds32_output_smw_single_word): Declare.
7187 (nds32_output_lmw_single_word): Declare.
7188 (nds32_expand_unaligned_load): Declare.
7189 (nds32_expand_unaligned_store): Declare.
7190 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7191 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7192 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7193 NDS32_BUILTIN_UASTORE_DW.
7194 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7197 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7198 Kito Cheng <kito.cheng@gmail.com>
7199 Chung-Ju Wu <jasonwucj@gmail.com>
7201 * config/nds32/nds32-intrinsic.c
7202 (nds32_expand_builtin_null_ftype_reg): Delete.
7203 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7204 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7205 (nds32_read_argument): New.
7206 (nds32_legitimize_target): Ditto.
7207 (nds32_legitimize_argument): Ditto.
7208 (nds32_check_constant_argument): Ditto.
7209 (nds32_expand_unop_builtin): Ditto.
7210 (nds32_expand_unopimm_builtin): Ditto.
7211 (nds32_expand_binop_builtin): Ditto.
7212 (nds32_builtin_decl_impl): Ditto.
7213 (builtin_description): Ditto.
7214 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7215 (nds32_init_builtins_impl): Ditto.
7216 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7217 (nds32_builtin_decl): New.
7218 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7219 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7221 2018-03-02 Jeff Law <law@redhat.com>
7223 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7224 (redundant_insn, fill_simple_delay_slots): Likewise.
7225 (fill_slots_from_thread): Likewise.
7226 * resource.c (mark_referenced_resources): Likewise.
7227 (mark_set_resources, find_dead_or_set_registers): Likewise.
7229 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7231 * substring-locations.h (format_warning_va): Formatting fix for
7233 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7235 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7236 * substring-locations.c: Include intl.h.
7237 (format_warning_va): Turned into small wrapper around
7238 format_warning_n_va, renamed to ...
7239 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7240 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7242 (format_warning_at_substring_n): New function.
7243 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7244 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7245 format_warning_at_substring with just a shorter name instead of
7246 const function pointer.
7247 (fmtwarn_n): New function.
7248 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7249 appropriate, get rid of all the fmtstr temporaries, move conditionals
7250 with G_() wrapped string literals directly into fmtwarn arguments,
7251 cast dir.len to (int), formatting fixes.
7253 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7255 * doc/invoke.texi: Remove "Cilk Plus" references.
7257 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7258 Richard Biener <rguenther@suse.de>
7261 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7262 for error or warning attributes if CALL_FROM_THUNK_P is set.
7265 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7268 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7269 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7272 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7273 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7275 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7276 instead of -1U in last predictors element's probability member.
7278 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7281 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7282 arguments if they are comparable.
7284 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7286 PR tree-optimization/84634
7287 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7288 masks and masked_loop_p with a single loop_masks, making sure it's
7289 null for bb vectorization.
7291 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7293 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7294 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7295 loop->force_vectorize to check whether there is no alias.
7297 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7300 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7302 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7304 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7305 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7306 instead of a loop around prev_real_insn.
7307 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7311 * config/i386/i386.c (ix86_print_operand): Use conditional
7312 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7315 2018-03-02 Richard Biener <rguenther@suse.de>
7317 PR tree-optimization/84427
7318 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7319 (bitmap_set_subtract_values): Rewrite to handle multiple
7322 (prune_clobbered_mems): Likewise.
7323 (phi_translate): Take edge instead of pred/phiblock.
7324 (phi_translate_1): Likewise.
7325 (phi_translate_set): Likewise. Insert all translated
7326 exprs for a value into the set, keeping possibly multiple
7327 expressions per value.
7328 (compute_antic_aux): Adjust for phi_translate changes.
7329 When intersecting union the expressions and prune those
7330 not in the final value set, keeping possibly multiple
7331 expressions per value. Do not use value-insertion
7332 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7333 all expressions. Add verification that the value-sets
7334 only shrink during iteration.
7335 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7336 (do_pre_regular_insertion): Likewise.
7337 (do_pre_partial_partial_insertion): Likewise.
7339 2018-03-02 Richard Biener <rguenther@suse.de>
7342 * config/darwin.c (saved_debug_info_level): New static global.
7343 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7344 (darwin_asm_lto_end): Restore debug info generation settings.
7346 2018-03-01 Martin Liska <mliska@suse.cz>
7349 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7352 2018-03-01 Richard Biener <rguenther@suse.de>
7355 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7356 type annotation with LTO when debug was disabled at compile-time.
7358 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7360 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7362 (mips_final_postscan_insn): Likewise.
7364 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7366 PR rtl-optimization/84528
7367 * alias.c (init_alias_target): Add commentary.
7368 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7369 a unique base value if the frame pointer is not eliminated
7370 to the stack pointer.
7372 2018-03-01 Tom de Vries <tom@codesourcery.com>
7374 PR rtl-optimization/83327
7375 * lra-int.h (hard_regs_spilled_into): Declare.
7376 * lra.c (hard_regs_spilled_into): Define.
7377 (init_reg_info): Init hard_regs_spilled_into.
7378 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7379 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7380 (process_bb_lives): Handle hard_regs_spilled_into.
7381 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7382 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7384 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7386 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7387 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7388 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7389 * config/rs6000/aix72.h: New file.
7391 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7393 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7394 instead of warning_at with conditional singular and plural messages
7398 * stor-layout.c (update_alignment_for_field): For
7399 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7400 && !DECL_PACKED (field), do the alignment update, just use
7401 only desired_align instead of MAX (type_align, desired_align)
7403 (place_field): Don't do known_align < desired_align handling
7404 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7405 is non-NULL, instead do it after rli->prev_field handling and
7406 only if not within a bitfield word. For DECL_PACKED (field)
7407 use type_align of BITS_PER_UNIT.
7409 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7411 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7412 superfluous parentheses and trailing spaces.
7414 2018-02-28 Richard Biener <rguenther@suse.de>
7416 PR tree-optimization/84584
7417 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7418 SCoPs with fake exit edge.
7420 2018-02-28 Martin Liska <mliska@suse.cz>
7423 * timevar.c (timer::print): Fix format to properly print 100%
7426 2018-02-28 Richard Biener <rguenther@suse.de>
7429 * genmatch.c (capture_info::walk_match): Do not mark
7430 captured expressions without operands as expr_p given
7431 they act more like predicates and should be subject to
7432 "lost tail" side-effect preserving.
7434 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7436 PR rtl-optimization/81611
7437 * auto-inc-dec.c (attempt_change): Move dead note from
7438 mem_insn if it's the next use of regno
7439 (find_address): Take address use of reg holding
7440 non-incremented value. Add parm to limit search to the named
7442 (merge_in_block): Attempt to use a mem insn that is the next
7443 use of the original regno.
7445 2018-02-27 Martin Sebor <msebor@redhat.com>
7448 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7449 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7451 2018-02-27 Martin Sebor <msebor@redhat.com>
7453 PR translation/84207
7454 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7455 n argument to unsigned HOST_WIDE_INT.
7456 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7457 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7458 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7459 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7461 2018-02-27 Richard Biener <rguenther@suse.de>
7463 PR tree-optimization/84512
7464 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7465 Do not use the estimate returned from record_stmt_cost for
7466 the scalar iteration cost but sum properly using add_stmt_cost.
7468 2018-02-27 Richard Biener <rguenther@suse.de>
7470 PR tree-optimization/84466
7471 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7472 Adjust last change to less strictly validate use operands.
7474 2018-02-27 Martin Liska <mliska@suse.cz>
7476 PR gcov-profile/84548
7477 * gcov.c (process_file): Allow partial overlap and consider it
7478 also as group functions.
7479 (output_lines): Properly calculate range of lines for a group.
7481 2018-02-27 Martin Liska <mliska@suse.cz>
7483 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7484 'ggc' suffixes. Change first column width.
7485 (timer::print): Fix formatting of the column.
7487 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7489 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7490 preserve inline entry blocks for the sake of debug inline
7491 entry point markers alone.
7492 (remove_unused_locals): Suggest in comments a better place to
7493 force the preservation of inline entry blocks that are
7494 otherwise unused, but do not preserve them.
7496 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7498 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7500 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7503 * config/i386/constraints.md (Bs): Replace
7504 ix86_indirect_branch_register with
7505 TARGET_INDIRECT_BRANCH_REGISTER.
7507 * config/i386/i386.md (indirect_jump): Likewise.
7508 (tablejump): Likewise.
7509 (*sibcall_memory): Likewise.
7510 (*sibcall_value_memory): Likewise.
7511 Peepholes of indirect call and jump via memory: Likewise.
7512 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7513 (*sibcall_value_GOT_32): Likewise.
7514 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7515 (GOT_memory_operand): Likewise.
7516 (call_insn_operand): Likewise.
7517 (sibcall_insn_operand): Likewise.
7518 (GOT32_symbol_operand): Likewise.
7519 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7521 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7523 PR rtl-optimization/83496
7524 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7525 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7526 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7527 redundant insn, if any.
7528 (relax_delay_slots): Likewise.
7529 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7531 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7533 PR tree-optimization/83965
7534 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7535 that grouped statements are part of a reduction chain. Return
7536 true if the statement is not marked as a reduction itself but
7538 (vect_recog_dot_prod_pattern): Don't check whether the statement
7539 is part of a group here.
7540 (vect_recog_sad_pattern): Likewise.
7541 (vect_recog_widen_sum_pattern): Likewise.
7543 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7546 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7548 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7551 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7553 (ix86_output_indirect_function_return): New prototype.
7554 (ix86_split_simple_return_pop_internal): Likewise.
7555 * config/i386/i386.c (indirect_return_via_cx): New.
7556 (indirect_return_via_cx_bnd): Likewise.
7557 (indirect_thunk_name): Handle return va CX_REG.
7558 (output_indirect_thunk_function): Create alias for
7559 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7560 (ix86_output_indirect_jmp): Remove the bool argument.
7561 (ix86_output_indirect_function_return): New function.
7562 (ix86_split_simple_return_pop_internal): Likewise.
7563 * config/i386/i386.md (*indirect_jump): Don't pass false
7564 to ix86_output_indirect_jmp.
7565 (*tablejump_1): Likewise.
7566 (simple_return_pop_internal): Change it to define_insn_and_split.
7567 Call ix86_split_simple_return_pop_internal to split it for
7569 (simple_return_indirect_internal): Call
7570 ix86_output_indirect_function_return instead of
7571 ix86_output_indirect_jmp.
7573 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7576 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7577 memset and value initialization afterwards.
7579 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7581 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7583 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7586 * common/config/aarch64/aarch64-common.c
7587 (aarch_option_optimization_table[]): Switch
7588 off fomit-frame-pointer
7590 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7591 Chung-Ju Wu <jasonwucj@gmail.com>
7593 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7595 (store_multiple): Ditto.
7597 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7599 * config.gcc: Add --with-cpu support for nds32 target.
7600 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7601 * config/nds32/nds32.opt: Add -mcpu= option.
7603 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7605 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7606 isel=yes): Warn for these deprecated options.
7608 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7610 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7611 ISA_2_5_MASKS_EMBEDDED.
7613 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7615 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7616 p->max as pointers rather than using iterative_hash_expr.
7618 2018-02-23 Carl Love <cel@us.ibm.com>
7620 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7621 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7623 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7624 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7625 P8V_BUILTIN_VEC_VUNSIGNED2.
7627 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7630 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7631 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7633 * lra-constraints.c (curr_insn_transform): Set up
7634 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7636 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7637 * lra-eliminations.c (spill_pseudos): Ditto.
7638 (process_insn_for_elimination): Ditto.
7639 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7640 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7641 LRA_NON_CLOBBERED_ALT.
7643 2018-02-22 Martin Sebor <msebor@redhat.com>
7645 PR tree-optimization/84480
7646 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7647 to maybe_diag_stxncpy_trunc. Call it.
7648 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7649 from gimple_fold_builtin_strcpy. Print inlining stack.
7650 (handle_builtin_stxncpy): Print inlining stack.
7651 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7653 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7656 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7657 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7658 and -fcheck-pointer-bounds are used together.
7659 (indirect_thunk_prefix): New enum.
7660 (indirect_thunk_need_prefix): New function.
7661 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7662 "_nt" instead of "_bnd" for NOTRACK prefix.
7663 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7664 (output_indirect_thunk_function): Likewise.
7666 (ix86_code_end): Update output_indirect_thunk_function calls.
7667 (ix86_output_indirect_branch_via_reg): Replace
7668 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7669 (ix86_output_indirect_branch_via_push): Likewise.
7670 (ix86_output_function_return): Likewise.
7671 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7672 incompatible with -fcf-protection=branch and
7673 -fcheck-pointer-bounds.
7675 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7678 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7679 Change gcc_assert call to output_operand_lossage.
7681 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7683 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7685 2018-02-22 DJ Delorie <dj@redhat.com>
7686 Sebastian Perta <sebastian.perta@renesas.com>
7687 Oleg Endo <olegendo@gcc.gnu.org>
7689 * config/rx/rx.c (rx_rtx_costs): New function.
7690 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7692 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7694 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7696 2018-02-22 Martin Liska <mliska@suse.cz>
7699 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7700 Add "native" as a possible value.
7702 2018-02-22 Martin Liska <mliska@suse.cz>
7705 * config/i386/i386.c (ix86_option_override_internal):
7706 Add "native" as a possible value for -march and -mtune.
7708 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7711 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7712 to all type variants.
7714 PR tree-optimization/84503
7715 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7716 width as info->bitpos + info->bitsize - start.
7717 (merged_store_group::merge_overlapping): Simplify width computation.
7718 (check_no_overlap): New function.
7719 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7720 start + width and last_order of the group, fail if check_no_overlap
7722 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7723 to group if check_no_overlap fails.
7725 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7727 * config/rs6000/altivec.md: Delete contraint arguments to
7728 define_expand, define_split, and define_peephole2, and in
7729 define_insn_and_split if always unused.
7730 * config/rs6000/darwin.md: Ditto.
7731 * config/rs6000/dfp.md: Ditto.
7732 * config/rs6000/rs6000.md: Ditto.
7733 * config/rs6000/sync.md: Ditto.
7734 * config/rs6000/vector.md: Ditto.
7735 * config/rs6000/vsx.md: Ditto.
7737 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7739 * config/rs6000/altivec.md: Write output control strings as braced
7740 blocks instead of double-quoted strings.
7741 * config/rs6000/darwin.md: Ditto.
7742 * config/rs6000/rs6000.md: Ditto.
7743 * config/rs6000/vector.md: Ditto.
7744 * config/rs6000/vsx.md: Ditto.
7746 2018-02-21 Jason Merrill <jason@redhat.com>
7748 PR c++/84314 - ICE with templates and fastcall attribute.
7749 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7751 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7753 * ipa-cp.c (determine_versionability): Fix comment typos.
7755 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7758 * ipa-cp.c (determine_versionability): Do not version functions caling
7761 2018-02-21 Martin Liska <mliska@suse.cz>
7764 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7765 Add "native" as a possible value.
7766 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7767 the macro when native cpu detection is available.
7769 2018-02-21 Martin Liska <mliska@suse.cz>
7772 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7773 Add "native" as a possible value.
7774 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7775 when native cpu detection is available.
7777 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7778 Martin Sebor <msebor@redhat.com>
7780 PR tree-optimization/84478
7781 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7783 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7784 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7785 support which is conservatively correct, for 2 only stay conservative
7786 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7787 argument to the 2 argument get_range_strlen, adjust 6 arg
7788 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7790 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7791 (gimple_fold_builtin_strlen): Pass true as last argument to
7794 2018-02-20 Martin Sebor <msebor@redhat.com>
7797 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7798 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7799 (builtin_memref::builtin_memref): Factor out parts into
7800 set_base_and_offset and call it.
7802 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7805 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7806 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7807 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7808 search at the associated MODE_INT.
7810 2018-02-20 Jeff Law <law@redhat.com>
7813 PR tree-optimization/81592
7815 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7816 for range data rather than using global data.
7817 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7818 range data rather than using global data.
7819 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7820 pass it to children as needed.
7821 (struct directive::fmtresult): Similarly.
7822 (struct directive::set_width): Similarly.
7823 (struct directive::set_precision): Similarly.
7824 (format_integer, format_directive, parse_directive): Similarly.
7825 (format_none): Accept unnamed vr_values parameter.
7826 (format_percent, format_floating, format_character): Similarly.
7827 (format_string, format_plain): Similarly.
7828 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7829 the EVRP range analyzer for range data rather than using global data.
7830 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7831 gimple-ssa-evrp-analyze.h
7832 (class sprintf_dom_walker): Add after_dom_children member function.
7833 Add evrp_range_analyzer member.
7834 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7835 range analyzer as needed.
7836 (sprintf_dom_walker::after_dom_children): New member function.
7837 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7839 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7840 (evrp_range_analyzer::pop_to_marker): Likewise.
7842 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7844 PR tree-optimization/84419
7845 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7846 with the required type if its current type is compatible but
7849 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7852 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7853 after vectorization.
7855 2018-02-20 Martin Liska <mliska@suse.cz>
7858 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7859 possible values if we don't have a hint.
7861 2018-02-20 Martin Liska <mliska@suse.cz>
7865 * final.c (shorten_branches): Build align_tab array with one
7867 * opts.c (finish_options): Add alignment option limit check.
7868 (MAX_CODE_ALIGN): Likewise.
7869 (MAX_CODE_ALIGN_VALUE): Likewise.
7870 * doc/invoke.texi: Document maximum allowed option value for
7871 all -falign-* options.
7873 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7876 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7877 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7878 * var-tracking.c (emit_note_insn_var_location): Remove all references
7879 to NOTE_INSN_CALL_ARG_LOCATION.
7880 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7881 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7882 Use copy_rtx_if_shared.
7883 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7884 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7885 (dwarf2out_var_location): Remove handling of
7886 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7888 * final.c (final_scan_insn): Remove all references to
7889 NOTE_INSN_CALL_ARG_LOCATION.
7890 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7891 before dumping final insns.
7892 * except.c (emit_note_eh_region_end): Remove all references to
7893 NOTE_INSN_CALL_ARG_LOCATION.
7894 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7895 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7896 * config/arc/arc.c (hwloop_optimize): Likewise.
7897 * config/arm/arm.c (create_fix_barrier): Likewise.
7898 * config/s390/s390.c (s390_chunkify_start): Likewise.
7899 * config/sh/sh.c (find_barrier): Likewise.
7900 * config/i386/i386.c (rest_of_insert_endbranch,
7901 ix86_seh_fixup_eh_fallthru): Likewise.
7902 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7903 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7904 * config/frv/frv.c (frv_function_prologue): Likewise.
7905 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7907 (note_outside_basic_block_p): Remove all references to
7908 NOTE_INSN_CALL_ARG_LOCATION.
7909 * gengtype.c (adjust_field_rtx_def): Likewise.
7910 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7912 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7913 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7916 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7919 PR tree-optimization/84452
7920 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7921 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7924 2018-02-19 Martin Liska <mliska@suse.cz>
7927 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7929 2018-02-19 Martin Liska <mliska@suse.cz>
7930 Richard Sandiford <richard.sandiford@linaro.org>
7932 PR tree-optimization/82491
7933 * gimple-fold.c (get_base_constructor): Make earlier bail out
7936 2018-02-19 Carl Love <cel@us.ibm.com>
7938 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7939 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7941 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7942 P8V_BUILTIN_VEC_NEG.
7944 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7946 * config/rl78/rl78.md (movdf): New define expand.
7948 2018-02-19 Martin Liska <mliska@suse.cz>
7951 * doc/invoke.texi: Fix typo.
7952 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7954 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7956 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7957 handle rs6000_single_float and rs6000_double_float specially for
7960 2018-02-16 Jeff Law <law@redhat.com>
7962 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7965 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7968 * ipa-inline.c (inline_small_functions): Fix a typo.
7970 2018-02-16 Nathan Sidwell <nathan@acm.org>
7972 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7974 2018-02-16 Carl Love <cel@us.ibm.com>
7976 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7977 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7978 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7979 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7980 expansion to P8V_BUILTIN_VEC_FLOAT2.
7982 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7984 PR rtl-optimization/70023
7985 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7986 src_regno into account.
7988 2018-02-16 Carl Love <cel@us.ibm.com>
7990 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7991 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7992 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7993 * config/rs6000/rs6000.c: Remove case statements for
7994 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7995 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7996 and P9V_BUILTIN_VEC_VINSERT4B.
7997 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7998 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7999 * config/rs6000/vsx.md:
8000 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8003 2018-02-16 Carl Love <cel@us.ibm.com>
8005 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8007 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8009 * config/rs6000/rs6000-c.c: Add the definitions for
8010 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8011 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8012 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8013 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8014 definition for insert4b and define insn *insert3b_internal.
8015 * doc/extend.texi: Add documentation for vec_extract4b.
8017 2018-02-16 Nathan Sidwell <nathan@acm.org>
8019 * doc/extend.texi (Backwards Compatibility): Mention friend
8020 injection. Note for-scope is deprecated.
8021 * doc/invoke.texi (-ffriend-injection): Deprecate.
8023 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8025 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8026 that moved to I2, also allow destinations that are a paradoxical
8027 subreg (instead of a normal reg).
8029 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8032 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8035 2018-02-16 Richard Biener <rguenther@suse.de>
8037 PR tree-optimization/84037
8038 PR tree-optimization/84016
8040 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8041 Adjust vec_construct for the fact we need additional higher latency
8042 128bit inserts for AVX256 and AVX512 vector builds.
8043 (ix86_add_stmt_cost): Scale vector construction cost for
8046 2018-02-16 Richard Biener <rguenther@suse.de>
8048 PR tree-optimization/84417
8049 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8050 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8051 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8053 2018-02-16 Martin Liska <mliska@suse.cz>
8056 * internal-fn.def (ASAN_CHECK): Set proper flags.
8057 (ASAN_MARK): Likewise.
8059 2018-02-16 Julia Koval <julia.koval@intel.com>
8061 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8062 from PTA_CANNONLAKE.
8064 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8067 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8068 Use ++iter rather than iter++ for std::list iterators.
8069 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8070 defer deleting them until all nodes in the forest are processed. Do
8071 free even leaf nodes. Change to_process into auto_vec.
8074 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8075 * vec.h (vec_default_construct): Use memset instead of placement new
8076 if BROKEN_VALUE_INITIALIZATION is defined.
8077 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8078 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8081 PR rtl-optimization/83723
8082 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8083 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8084 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8086 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8088 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8090 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8092 PR rtl-optimization/81443
8093 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8094 from inner REGs to paradoxical SUBREGs.
8096 2018-02-16 Richard Biener <rguenther@suse.de>
8098 PR tree-optimization/84399
8099 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8100 For operands we can analyze at their definition make sure we can
8101 analyze them at each use as well.
8103 2018-02-16 Richard Biener <rguenther@suse.de>
8105 PR tree-optimization/84190
8106 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8107 volatile accesses if the decl isn't volatile.
8109 2018-02-15 Jason Merrill <jason@redhat.com>
8111 PR c++/84314 - ICE with templates and fastcall attribute.
8112 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8113 TYPE_CANONICAL on an existing type.
8115 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8117 PR tree-optimization/84383
8118 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8119 dstoff nor call operand_equal_p if dstbase is NULL.
8121 PR tree-optimization/84334
8122 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8123 also a CONSTANT_CLASS_P, punt.
8125 2018-02-14 Jim Wilson <jimw@sifive.com>
8127 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8128 first SMALL_OPERAND check. New local min_second_step. Move assert
8129 to where locals are set. Add TARGET_RVC support.
8130 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8132 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8134 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8136 2018-02-14 Martin Sebor <msebor@redhat.com>
8138 PR tree-optimization/83698
8139 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8140 arrays constrain the offset range to their bounds.
8141 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8142 (builtin_access::overlap): Avoid setting the size of overlap if it's
8144 (maybe_diag_overlap): Also consider arrays when deciding what values
8145 of offsets to include in diagnostics.
8147 2018-02-14 Martin Sebor <msebor@redhat.com>
8150 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8151 that correspond to the kind of a declaration.
8153 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8156 * config/pa/pa.md: Load address of PIC label using the linkage table
8157 if the label is nonlocal.
8159 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8161 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8162 warning message if user requests -maltivec=be.
8163 * doc/invoke.texi: Document deprecation of -maltivec=be.
8165 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8168 * config/rs6000/rs6000-c.c: Update definitions for
8169 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8170 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8172 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8175 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8176 add _get_ssp intrinsics. Remove argument from
8177 __builtin_ia32_rdssp[d|q].
8178 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8179 * config/i386/i386-builtin.def: Remove argument from
8180 __builtin_ia32_rdssp[d|q].
8181 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8182 ix86_expand_special_args_builtin for _rdssp[d|q].
8183 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8184 Clear register before usage.
8185 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8186 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8188 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8190 PR tree-optimization/84357
8191 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8192 operand 1 of an ARRAY_REF too.
8194 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8197 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8198 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8200 (set_of_reg): New struct.
8201 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8202 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8203 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8205 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8206 Split into bitclr, bitset, bitinvert patterns if appropriate.
8207 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8208 use rx_fuse_in_memory_bitop.
8209 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8210 to named insn, correct maximum insn length.
8212 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8215 * machmode.def: Define a complex mode for PARTIAL_INT.
8216 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8218 * doc/rtl.texi: Document CSPImode.
8219 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8221 (msp430_hard_regno_nregs_with_padding): Likewise.
8223 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8226 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8228 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8230 PR rtl-optimization/84169
8231 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8232 we generated a parallel as new i3 and we split that to new i2 and i3
8233 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8234 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8235 those to i2, not i1. Partially rewrite this scan code.
8237 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8240 * stor-layout.c (place_field): For variable length fields, adjust
8241 offset_align afterwards not just based on the field's alignment,
8242 but also on the size.
8245 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8246 of exps and logs in the use_exp2 case.
8248 2018-02-13 Jeff Law <law@redhat.com>
8250 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8253 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8256 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8260 * common.opt (gas-loc-support, gas-locview-support): New.
8261 (ginline-points, ginternal-reset-location-views): New.
8262 * doc/invoke.texi: Document them. Use @itemx where intended.
8263 (gvariable-location-views): Adjust.
8264 * target.def (reset_location_view): New.
8265 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8266 (TARGET_RESET_LOCATION_VIEW): New.
8267 * doc/tm.texi: Rebuilt.
8268 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8269 (dwarf2out_default_as_locview_support): New.
8270 (output_asm_line_debug_info): Use option variables.
8271 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8272 (output_loc_list): Likewise.
8273 (add_high_low_attributes): Check option variables.
8274 Don't output entry view attribute in strict mode.
8275 (gen_inlined_subroutine_die): Check option variables.
8276 (dwarf2out_inline_entry): Likewise.
8277 (init_sections_and_labels): Likewise.
8278 (dwarf2out_early_finish): Likewise.
8279 (maybe_reset_location_view): New, from...
8280 (dwarf2out_var_location): ... here. Call it.
8281 * debug.h (dwarf2out_default_as_loc_support): Declare.
8282 (dwarf2out_default_as_locview_support): Declare.
8283 * hooks.c (hook_int_rtx_insn_0): New.
8284 * hooks.h (hook_int_rtx_insn_0): Declare.
8285 * toplev.c (process_options): Take -gas-loc-support and
8286 -gas-locview-support from dwarf2out. Enable
8287 -gvariable-location-views by default only with locview
8288 assembler support. Enable -ginternal-reset-location-views by
8289 default only if the target defines the corresponding hook.
8290 Enable -ginline-points by default if location views are
8291 enabled; force it disabled if statement frontiers are
8293 * tree-inline.c (expand_call_inline): Check option variables.
8294 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8296 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8298 PR tree-optimization/84321
8299 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8300 handling. Also check whether the anti-range contains any values
8301 that satisfy the mask; switch to a VR_RANGE if not.
8303 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8306 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8308 2018-02-13 Martin Jambor <mjambor@suse.cz>
8311 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8312 of call statements, also set location of a load to a temporary.
8314 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8316 * config/rl78/rl78.c (add_vector_labels): New function.
8317 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8318 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8319 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8320 which checks that no arguments are passed.
8321 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8322 * doc/extend.texi: Documentation for the new attribute.
8324 2018-02-13 Andreas Schwab <schwab@suse.de>
8326 * config/riscv/linux.h (CPP_SPEC): Define.
8328 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8331 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8332 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8333 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8334 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8335 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8336 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8337 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8338 temporarily for AES and PCLMUL builtins.
8340 PR tree-optimization/84339
8341 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8342 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8346 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8347 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8348 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8350 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8352 * omp-simd-clone.h: New file.
8353 * omp-simd-clone.c: Include omp-simd-clone.h.
8354 (expand_simd_clones): No longer static.
8355 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8356 cgraph.h and omp-simd-clone.h.
8357 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8358 (vect_recog_widen_shift_pattern): Formatting fix.
8359 (vect_pattern_recog_1): Don't check optab for calls.
8362 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8363 operands[2] into a REG before using gen_lowpart on it.
8365 2018-02-12 Jeff Law <law@redhat.com>
8368 * config/sh/sh.c (find_barrier): Consider a sibling call
8371 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8372 successfully back substituting a reg.
8374 2018-02-12 Richard Biener <rguenther@suse.de>
8376 PR tree-optimization/84037
8377 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8378 parameter, move visited init to caller.
8379 (vect_slp_analyze_operations): Separate cost from validity
8380 check, initialize visited once for all instances.
8381 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8383 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8385 (vect_model_promotion_demotion_cost): Likewise.
8386 (vectorizable_bswap): Guard cost modeling with !slp_node
8387 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8389 (vectorizable_call): Likewise.
8390 (vectorizable_conversion): Likewise.
8391 (vectorizable_assignment): Likewise.
8392 (vectorizable_shift): Likewise.
8393 (vectorizable_operation): Likewise.
8394 (vectorizable_store): Likewise.
8395 (vectorizable_load): Likewise.
8396 (vectorizable_condition): Likewise.
8397 (vectorizable_comparison): Likewise.
8399 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8402 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8403 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8404 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8406 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8409 * params.def (inline-min-speedup): Increase from 8 to 15.
8410 (max-inline-insns-auto): Decrease from 40 to 30.
8411 * ipa-split.c (consider_split): Add some buffer for function to
8412 be considered inlining candidate.
8413 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8416 2018-02-12 Richard Biener <rguenther@suse.de>
8418 PR tree-optimization/84037
8419 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8420 matched stmts if we cannot swap the non-matched ones.
8422 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8424 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8425 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8426 _mm_maskz_scalef_round_ss): New intrinsics.
8427 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8428 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8429 __builtin_ia32_scalefss_round): Remove.
8430 (__builtin_ia32_scalefsd_mask_round,
8431 __builtin_ia32_scalefss_mask_round): New intrinsics.
8432 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8433 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8434 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8435 "<round_constraint>")): Changed to ...
8436 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8437 "<round_scalar_constraint>")): ... this.
8438 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8439 %0, %1, %2<round_op3>}"): Changed to ...
8440 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8441 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8442 %2<round_scalar_mask_op3>}"): ... this.
8443 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8445 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8447 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8448 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8449 (_mm_maskz_sqrt_round_ss): New intrinsics.
8450 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8451 (__builtin_ia32_sqrtsd_mask_round)
8452 (__builtin_ia32_sqrtss_mask_round): New builtins.
8453 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8454 (__builtin_ia32_sqrtss_round): Remove.
8455 (__builtin_ia32_sqrtsd_mask_round)
8456 (__builtin_ia32_sqrtss_mask_round): New builtins.
8457 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8458 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8459 ((match_operand:VF_128 1 "vector_operand"
8460 "xBm,<round_constraint>")): Changed to ...
8461 ((match_operand:VF_128 1 "vector_operand"
8462 "xBm,<round_scalar_constraint>")): ... this.
8463 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8464 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8465 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8466 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8467 %<iptr>1<round_scalar_mask_op3>}): ... this.
8468 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8469 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8471 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8474 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8475 Cast vec_cmpeq result to correct type.
8476 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8477 Cast vec_cmpgt result to correct type.
8479 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8481 * final.c (final_scan_insn_1): Renamed from...
8482 (final_scan_insn): ... this. New wrapper, to recover
8483 seen from the outermost call in recursive ones.
8484 * config/sparc/sparc.c (output_return): Drop seen from call.
8485 (output_sibcall): Likewise.
8486 * config/visium/visium.c (output_branch): Likewise.
8488 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8490 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8493 2018-02-10 Alan Modra <amodra@gmail.com>
8496 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8497 Specify LR as an input.
8499 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8502 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8503 remove_member_access_dummy_vars): New functions.
8504 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8505 lower_omp_1, execute_lower_omp): Use them.
8507 PR rtl-optimization/84308
8508 * shrink-wrap.c (spread_components): Release todo vector.
8510 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8512 PR rtl-optimization/57193
8513 * ira-color.c (struct allocno_color_data): Add member
8514 conflict_allocno_hard_prefs.
8515 (update_conflict_allocno_hard_prefs): New.
8516 (bucket_allocno_compare_func): Add a preference based on
8517 conflict_allocno_hard_prefs.
8518 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8519 (color_allocnos): Remove a dead code. Initiate
8520 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8522 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8525 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8526 constraint from =wa to wa. Avoid a subreg on the output operand,
8527 instead use a pseudo and subreg it in a move.
8528 (p9_xxbrd_<mode>): Changed to ...
8529 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8530 (p9_xxbrd_v2df): New expander.
8531 (p9_xxbrw_<mode>): Changed to ...
8532 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8533 (p9_xxbrw_v4sf): New expander.
8535 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8537 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8539 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8542 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8543 multiply in 32-bit mode.
8544 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8545 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8548 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8550 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8551 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8552 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8553 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8555 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8558 * dwarf2out.c (is_trivial_indirect_ref): New function.
8559 (dwarf2out_late_global_decl): Do not generate a location
8560 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8561 and that are not defined in the current unit.
8563 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8565 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8566 instead of a libcall for UNORDERED.
8568 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8571 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8572 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8574 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8577 * config/s390/s390.c (s390_set_current_function): Invoke
8578 s390_indirect_branch_settings also if fndecl didn't change.
8580 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8582 * config/rs6000/rs6000.md (blockage): Set length to zero.
8584 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8586 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8588 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8591 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8592 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8596 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8597 PARALLEL incoming that failed vt_get_decl_and_offset check.
8600 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8601 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8603 (get_variable_section): For decls in named .bss* sections pass true as
8604 second argument to bss_initializer_p.
8606 2018-02-09 Marek Polacek <polacek@redhat.com>
8607 Jakub Jelinek <jakub@redhat.com>
8610 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8611 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8612 Sync some changes from cxx_fold_indirect_ref.
8614 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8616 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8618 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8619 (BLOCK_INLINE_ENTRY_LABEL): New.
8620 (dwarf2out_var_location): Disregard inline entry markers.
8621 (inline_entry_data): New struct.
8622 (inline_entry_data_hasher): New hashtable type.
8623 (inline_entry_data_hasher::hash): New.
8624 (inline_entry_data_hasher::equal): New.
8625 (inline_entry_data_table): New variable.
8626 (add_high_low_attributes): Add DW_AT_entry_pc and
8627 DW_AT_GNU_entry_view attributes if a pending entry is found
8628 in inline_entry_data_table. Add old entry_pc attribute only
8629 if debug nonbinding markers are disabled.
8630 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8631 markers are enabled.
8632 (block_within_block_p, dwarf2out_inline_entry): New.
8633 (dwarf2out_finish): Check that no entries remained in
8634 inline_entry_data_table.
8635 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8636 (final_scan_insn, notice_source_line): Likewise.
8637 (rest_of_clean_state): Skip inline entry markers.
8638 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8640 * gimple.c (gimple_build_debug_inline_entry): New.
8641 * gimple.h (enum gimple_debug_subcode): Add
8642 GIMPLE_DEBUG_INLINE_ENTRY.
8643 (gimple_build_debug_inline_entry): Declare.
8644 (gimple_debug_inline_entry_p): New.
8645 (gimple_debug_nonbind_marker_p): Adjust.
8646 * insn-notes.def (INLINE_ENTRY): New.
8647 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8648 inline entry marker notes.
8649 (print_insn): Likewise.
8650 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8651 (INSN_DEBUG_MARKER_KIND): Likewise.
8652 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8653 * tree-inline.c (expand_call_inline): Build and insert
8654 debug_inline_entry stmt.
8655 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8656 inline entry blocks early, if nonbind markers are enabled.
8657 (dump_scope_block): Dump fragment info.
8658 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8659 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8660 (gimple_build_debug_inline_entry): New.
8661 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8662 Enable/disable inline entry points too.
8663 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8664 (DEBUG_INSN): Describe inline entry markers.
8666 * common.opt (gvariable-location-views): New.
8667 (gvariable-location-views=incompat5): New.
8668 * config.in: Rebuilt.
8669 * configure: Rebuilt.
8670 * configure.ac: Test assembler for view support.
8671 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8672 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8673 * dwarf2out.c (var_loc_view): New typedef.
8674 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8675 (dwarf2out_locviews_in_attribute): New.
8676 (dwarf2out_locviews_in_loclist): New.
8677 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8678 (enum dw_line_info_opcode): Add LI_adv_address.
8679 (struct dw_line_info_table): Add view.
8680 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8681 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8682 (zero_view_p): New variable.
8683 (ZERO_VIEW_P): New macro.
8684 (output_asm_line_debug_info): New.
8685 (struct var_loc_node): Add view.
8686 (add_AT_view_list, AT_loc_list): New.
8687 (add_var_loc_to_decl): Add view param. Test it against last.
8688 (new_loc_list): Add view params. Record them.
8689 (AT_loc_list_ptr): Handle loc and view lists.
8690 (view_list_to_loc_list_val_node): New.
8691 (print_dw_val): Handle dw_val_class_view_list.
8692 (size_of_die): Likewise.
8693 (value_format): Likewise.
8694 (loc_list_has_views): New.
8695 (gen_llsym): Set vl_symbol too.
8696 (maybe_gen_llsym, skip_loc_list_entry): New.
8697 (dwarf2out_maybe_output_loclist_view_pair): New.
8698 (output_loc_list): Output view list or entries too.
8699 (output_view_list_offset): New.
8700 (output_die): Handle dw_val_class_view_list.
8701 (output_dwarf_version): New.
8702 (output_compilation_unit_header): Use it.
8703 (output_skeleton_debug_sections): Likewise.
8704 (output_rnglists, output_line_info): Likewise.
8705 (output_pubnames, output_aranges): Update version comments.
8706 (output_one_line_info_table): Output view numbers in asm comments.
8707 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8708 Call maybe_gen_llsym.
8709 (loc_list_from_tree_1): Adjust.
8710 (add_AT_location_description): Create view list attribute if
8711 needed, check it's absent otherwise.
8712 (convert_cfa_to_fb_loc_list): Adjust.
8713 (maybe_emit_file): Call output_asm_line_debug_info for test.
8714 (dwarf2out_var_location): Reset views as needed. Precompute
8715 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8716 attribute. Set view.
8717 (new_line_info_table): Reset next view.
8718 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8719 (dwarf2out_source_line): Likewise. Output view resets and labels to
8720 the assembler, or select appropriate line info opcodes.
8721 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8722 (optimize_string_length): Catch it. Adjust.
8723 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8724 dw_val_class_view_list, and remove it if no longer needed.
8725 (hash_loc_list): Hash view numbers.
8726 (loc_list_hasher::equal): Compare them.
8727 (optimize_location_lists): Check whether a view list symbol is
8728 needed, and whether the locview attribute is present, and
8729 whether they match. Remove the locview attribute if no longer
8731 (index_location_lists): Call skip_loc_list_entry for test.
8732 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8733 Use output_dwarf_version.
8734 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8735 (struct dw_val_node): Add val_view_list.
8736 * final.c (SEEN_NEXT_VIEW): New.
8737 (set_next_view_needed): New.
8738 (clear_next_view_needed): New.
8739 (maybe_output_next_view): New.
8740 (final_start_function): Rename to...
8741 (final_start_function_1): ... this. Take pointer to FIRST,
8742 add SEEN parameter. Emit param bindings in the initial view.
8743 (final_start_function): Reintroduce SEEN-less interface.
8744 (final): Rename to...
8745 (final_1): ... this. Take SEEN parameter. Output final pending
8746 next view at the end.
8747 (final): Reintroduce seen-less interface.
8748 (final_scan_insn): Output pending next view before switching
8749 sections or ending a block. Mark the next view as needed when
8750 outputting variable locations. Notify debug backend of section
8751 changes, and of location view changes.
8752 (rest_of_handle_final): Adjust.
8753 * toplev.c (process_options): Autodetect value for debug variable
8754 location views option. Warn on incompat5 without -gdwarf-5.
8755 * doc/invoke.texi (gvariable-location-views): New.
8756 (gvariable-location-views=incompat5): New.
8757 (gno-variable-location-views): New.
8759 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8761 PR tree-optimization/84136
8762 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8763 that the result of find_edge is non-NULL.
8765 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8768 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8769 storing integer register in SImode. Fix cost of 256 and 512
8770 byte aligned SSE register store.
8772 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8774 * config/i386/i386.c (ix86_multiplication_cost): Fix
8775 multiplication cost for TARGET_AVX512DQ.
8777 2018-02-08 Marek Polacek <polacek@redhat.com>
8779 PR tree-optimization/84238
8780 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8783 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8785 PR tree-optimization/84265
8786 * tree-vect-stmts.c (vectorizable_store): Don't treat
8787 VMAT_CONTIGUOUS accesses as grouped.
8788 (vectorizable_load): Likewise.
8790 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8792 PR tree-optimization/81635
8793 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8794 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8795 (test_round_for_mask): New functions.
8796 (wide_int_cc_tests): Call test_round_for_mask.
8797 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8798 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8799 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8800 range returned by get_range_info.
8802 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8805 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8806 * symtab.c: Include builtins.h
8807 (symtab_node::output_to_lto_symbol_table_p): Move here
8808 from lto-streamer-out.c:output_symbol_p.
8809 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8810 (output_symbol_p): Move all logic to symtab.c
8811 (produce_symtab): Update.
8813 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8815 * config/s390/s390-opts.h (enum indirect_branch): Define.
8816 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8817 (s390_indirect_branch_via_thunk)
8818 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8819 (enum s390_indirect_branch_type): Define.
8820 * config/s390/s390.c (struct s390_frame_layout, struct
8821 machine_function): Remove.
8822 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8823 (indirect_branch_table_label_no, indirect_branch_table_name):
8825 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8826 (enum s390_indirect_branch_option): Define.
8827 (s390_return_addr_from_memory): New function.
8828 (s390_handle_string_attribute): New function.
8829 (s390_attribute_table): Add new attribute handler.
8830 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8831 (s390_indirect_branch_via_thunk): New function.
8832 (s390_indirect_branch_via_inline_thunk): New function.
8833 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8834 sibling call optimization for non z10 compiles.
8835 (s390_emit_call): Force indirect branch target to be a single
8836 register. Add r1 clobber for non-z10 compiles.
8837 (s390_emit_epilogue): Emit return jump via return_use expander.
8838 (s390_reorg): Handle JUMP_INSNs as execute targets.
8839 (s390_option_override_internal): Perform validity checks for the
8840 new command line options.
8841 (s390_indirect_branch_attrvalue): New function.
8842 (s390_indirect_branch_settings): New function.
8843 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8844 (s390_output_indirect_thunk_function): New function.
8845 (s390_code_end): Implement target hook.
8846 (s390_case_values_threshold): Implement target hook.
8847 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8849 * config/s390/s390.h (struct s390_frame_layout)
8850 (struct machine_function): Move here from s390.c.
8851 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8852 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8853 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8854 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8855 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8856 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8857 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8858 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8859 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8860 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8861 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8862 (mnemonic attribute): Add values which aren't recognized
8864 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8865 pattern for branch conversion. Fix mnemonic attribute.
8866 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8867 indirect branch via thunk if requested.
8868 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8869 ("*indirect_jump"): Disable for branch conversion using out of
8871 ("indirect_jump_via_thunk<mode>_z10")
8872 ("indirect_jump_via_thunk<mode>")
8873 ("indirect_jump_via_inlinethunk<mode>_z10")
8874 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8875 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8876 ("casesi_jump_via_inlinethunk<mode>_z10")
8877 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8878 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8879 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8880 ("*indirect2_jump"): Disable for branch conversion.
8881 ("casesi_jump"): Turn into expander and expand patterns for branch
8883 ("return_use"): New expander.
8884 ("*return"): Emit return via thunk and rename it to ...
8885 ("*return<mode>"): ... this one.
8886 * config/s390/s390.opt: Add new options and and enum for the
8889 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8891 * lra-constraints.c (match_reload): Unconditionally use
8892 gen_lowpart_SUBREG, rather than selecting between that
8893 and equivalent gen_rtx_SUBREG code.
8895 2018-02-08 Richard Biener <rguenther@suse.de>
8897 PR tree-optimization/84233
8898 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8899 changed flag instead of boguously re-using phi_inserted.
8901 2018-02-08 Martin Jambor <mjambor@suse.cz>
8903 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8904 static local variables.
8906 2018-02-08 Richard Biener <rguenther@suse.de>
8908 PR tree-optimization/84278
8909 * tree-vect-stmts.c (vectorizable_store): When looking for
8910 smaller vector types to perform grouped strided loads/stores
8911 make sure the mode is supported by the target.
8912 (vectorizable_load): Likewise.
8914 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8916 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8917 Increase LDP/STP opportunities by adding adjacent callee-saves.
8919 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8921 PR rtl-optimization/84068
8922 PR rtl-optimization/83459
8923 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8925 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8927 PR tree-optimization/84224
8928 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8929 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8932 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8935 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8936 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8937 count, remove one USE.
8939 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8941 * doc/install.texi (Configuration): Document the
8942 --with-long-double-format={ibm,ieee} PowerPC configuration
8946 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8947 Convert from define_expand to be define_insn_and_split. Rework
8948 float/double/_Float128 conversions to QI/HI/SImode to work with
8949 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8950 conversions to QI/HImode types did a store and then a load to
8951 truncate the value. For conversions to VSX registers, don't split
8952 the insn, instead emit the code directly. Use the code iterator
8953 any_fix to combine signed and unsigned conversions.
8954 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8955 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8956 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8957 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8958 (fix_<mode>di2_hw): Likewise.
8959 (fixuns_<mode>di2_hw): Likewise.
8960 (fix_<mode>si2_hw): Likewise.
8961 (fixuns_<mode>si2_hw): Likewise.
8962 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8963 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8964 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8965 fix<uns>_trunc<SFDF:mode>si2_p8.
8966 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8968 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8969 (fix<uns>_<mode>_mem): Likewise.
8970 (fctiw<u>z_<mode>_mem): Likewise.
8971 (fix<uns>_<mode>_mem): Likewise.
8972 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8973 the register allocator from doing a direct move to the GPRs to do
8974 a store, and instead use the ISA 3.0 store byte/half-word from
8975 vector register instruction. For IEEE 128-bit floating point,
8976 also optimize stores of 32-bit ints.
8977 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8979 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8981 * genextract.c (push_pathstr_operand): New function to support
8983 (walk_rtx): Call push_pathstr_operand.
8984 (print_path): Support [a-zA-Z].
8986 2018-02-07 Richard Biener <rguenther@suse.de>
8988 PR tree-optimization/84037
8989 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8990 (cse_and_gimplify_to_preheader): Declare.
8991 (vect_get_place_in_interleaving_chain): Likewise.
8992 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8994 (_loop_vec_info::~_loop_vec_info): Delete it.
8995 (cse_and_gimplify_to_preheader): New function.
8996 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8997 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8998 (vectorizable_load): Likewise. For grouped stores always base
8999 the IV on the first element.
9000 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9001 condition before gimplifying.
9003 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9005 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9006 *DIV_EXPR and *MOD_EXPR.
9008 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9011 * config/i386/i386.c (ix86_option_override_internal): Mask out
9012 the CF_SET bit when checking -fcf-protection.
9014 2018-02-07 Tom de Vries <tom@codesourcery.com>
9017 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9020 2018-02-07 Richard Biener <rguenther@suse.de>
9022 PR tree-optimization/84204
9023 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9026 PR tree-optimization/84205
9027 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9028 special-case isl_ast_op_zdiv_r.
9030 PR tree-optimization/84223
9031 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9032 Only add conditions from within the region.
9033 (gather_bbs::after_dom_children): Adjust.
9035 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9038 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9039 * config/avr/avr.md: Only post-reload split REG-REG moves if
9040 either register is GENERAL_REG_P.
9042 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9044 PR tree-optimization/84235
9045 * tree-ssa-scopedtables.c
9046 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9047 if the subtraction is performed in floating point type where NaNs are
9048 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9049 build 1. Formatting fix.
9051 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9054 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9055 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9056 and skip it regardless of bb boundaries. Use CALL_P macro,
9057 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9058 unnecessarily, formatting fix.
9060 2018-02-06 Michael Collison <michael.collison@arm.com>
9062 * config/arm/thumb2.md:
9063 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9064 (*thumb_mov_notscc): Ditto.
9066 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9069 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9070 unsigned_fix, not "s".
9072 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9074 * configure.ac (gcc_fn_eh_frame_ro): New function.
9075 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9076 correct .eh_frame permissions.
9077 * configure: Regenerate.
9079 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9081 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9084 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9086 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9087 Display warning message for -mno-speculate-indirect-jumps.
9089 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9091 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9093 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9095 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9097 PR tree-optimization/84225
9098 * tree-eh.c (find_trapping_overflow): Only call
9099 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9101 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9104 * config/i386/i386.c: Reimplement the check of possible options
9105 -mibt/-mshstk conbination. Change error messages.
9106 * doc/invoke.texi: Fix a typo: remove extra '='.
9108 2018-02-06 Marek Polacek <polacek@redhat.com>
9110 PR tree-optimization/84228
9111 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9113 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9116 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9117 emitted arch directives.
9118 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9119 __ARM_FEATURE_COPROC before changing architectures.
9121 2018-02-06 Richard Biener <rguenther@suse.de>
9123 * config/i386/i386.c (print_reg): Fix typo.
9124 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9126 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9128 * configure: Regenerate.
9130 2018-02-05 Martin Sebor <msebor@redhat.com>
9132 PR tree-optimization/83369
9133 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9136 2018-02-05 Martin Liska <mliska@suse.cz>
9138 * doc/invoke.texi: Cherry-pick upstream r323995.
9140 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9142 * ira.c (ira_init_register_move_cost): Adjust comment.
9144 2018-02-05 Martin Liska <mliska@suse.cz>
9146 PR gcov-profile/84137
9147 * doc/gcov.texi: Fix typo in documentation.
9149 2018-02-05 Martin Liska <mliska@suse.cz>
9151 PR gcov-profile/83879
9152 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9153 using dlopen functionality.
9155 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9157 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9158 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9159 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9160 _mm_maskz_range_round_ss): New intrinsics.
9161 (__builtin_ia32_rangesd128_round)
9162 (__builtin_ia32_rangess128_round): Remove.
9163 (__builtin_ia32_rangesd128_mask_round,
9164 __builtin_ia32_rangess128_mask_round): New builtins.
9165 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9166 __builtin_ia32_rangess128_round): Remove.
9167 (__builtin_ia32_rangesd128_mask_round,
9168 __builtin_ia32_rangess128_mask_round): New builtins.
9169 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9170 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9171 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9172 "<round_saeonly_constraint>")): Changed to ...
9173 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9174 "<round_saeonly_scalar_constraint>")): ... this.
9175 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9176 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9177 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9178 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9179 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9181 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9183 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9185 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9186 Remove all values except native, 8540 and 8548.
9188 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9190 * config/i386/i386.c (ix86_output_function_return): Pass
9191 INVALID_REGNUM, instead of -1, as invalid register number to
9192 indirect_thunk_name and output_indirect_thunk.
9194 2018-02-02 Julia Koval <julia.koval@intel.com>
9196 * config.gcc: Add -march=icelake.
9197 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9198 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9199 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9200 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9201 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9202 (processor_target_table): Add icelake.
9203 (ix86_option_override_internal): Handle new PTAs.
9204 (get_builtin_code_for_version): Handle icelake.
9205 (M_INTEL_COREI7_ICELAKE): New.
9206 (fold_builtin_cpu): Handle icelake.
9207 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9208 * doc/invoke.texi: Add -march=icelake.
9210 2018-02-02 Julia Koval <julia.koval@intel.com>
9212 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9213 to wide_int_bitmask.
9214 * wide-int-bitmask.h: New.
9216 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9219 * config/i386/i386.md: Replace Pmode with word_mode in
9220 builtin_setjmp_setup and builtin_longjmp to support x32.
9222 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9226 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9228 (rs6000_supported_cpu_names): New static variable.
9229 (linux_cpu_translation_table): Likewise.
9230 (elf_platform) <cpu>: Define new static variable and use it.
9231 Translate kernel AT_PLATFORM name to canonical name if needed.
9232 Error if platform name is unknown.
9234 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9237 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9239 2018-02-01 Jeff Law <law@redhat.com>
9242 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9243 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9244 the scratch if RELEASE_VIA_POP is false.
9245 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9246 If we have to save a temporary register, decrement SIZE appropriately.
9247 Pass new arguments to release_scratch_register_on_entry.
9248 (ix86_adjust_stack_and_probe): Likewise.
9249 (ix86_emit_probe_stack_range): Pass new arguments to
9250 release_scratch_register_on_entry.
9252 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9254 PR rtl-optimization/84157
9255 * combine.c (change_zero_ext): Use REG_P predicate in
9256 front of HARD_REGISTER_P predicate.
9258 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9260 * config/avr/avr.c (avr_option_override): Move disabling of
9261 -fdelete-null-pointer-checks to...
9262 * common/config/avr/avr-common.c (avr_option_optimization_table):
9265 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9267 PR tree-optimization/81635
9268 * tree-data-ref.c (split_constant_offset_1): For types that
9269 wrap on overflow, try to use range info to prove that wrapping
9272 2018-02-01 Renlin Li <renlin.li@arm.com>
9275 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9277 (aarch64_register_move_cost): Likewise.
9278 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9280 (REG_CLASS_NAMES): Likewise.
9281 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9282 TAILCALL_ADDR_REGS. Remove IP registers.
9283 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9285 2018-02-01 Richard Biener <rguenther@suse.de>
9287 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9288 for specifying RPO order and allow NULL for that.
9289 * domwalk.c (dom_walker::dom_walker): Likewise.
9290 (dom_walker::walk): Handle NULL RPO order.
9291 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9293 (rewrite_update_dom_walker): Likewise.
9294 (mark_def_dom_walker): Likewise.
9296 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9298 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9299 (aarch64_maybe_expand_sve_subreg_move): Declare.
9300 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9301 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9303 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9304 that are semantically a reverse operation.
9305 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9306 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9307 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9309 (aarch64_can_change_mode_class): For big-endian, forbid changes
9310 between two SVE modes if they have different element sizes.
9312 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9314 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9315 the TImode handling for big-endian targets.
9317 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9319 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9320 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9322 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9323 Remove BSWAP handing for big-endian targets and use the form of
9324 LD1RQ appropariate for the mode.
9326 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9328 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9329 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9332 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9335 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9336 check for operands that need to go through aarch64_sve_reload_be.
9338 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9340 PR tree-optimization/81661
9341 PR tree-optimization/84117
9342 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9343 * tree-eh.c: Include gimplify.h.
9344 (find_trapping_overflow, replace_trapping_overflow,
9345 rewrite_to_non_trapping_overflow): New functions.
9346 * tree-vect-loop.c: Include tree-eh.h.
9347 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9348 * tree-data-ref.c: Include tree-eh.h.
9349 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9351 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9353 PR rtl-optimization/84123
9354 * combine.c (change_zero_ext): Check if hard register satisfies
9355 can_change_dest_mode before calling gen_lowpart_SUBREG.
9357 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9360 * ira.c (ira_init_register_move_cost): Remove assert.
9362 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9364 PR rtl-optimization/84071
9365 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9366 * doc/tm.texi: Regenerate.
9368 2018-01-31 Richard Biener <rguenther@suse.de>
9370 PR tree-optimization/84132
9371 * tree-data-ref.c (analyze_miv_subscript): Properly
9372 check whether evolution_function_is_affine_multivariate_p
9373 before calling gcd_of_steps_may_divide_p.
9375 2018-01-31 Julia Koval <julia.koval@intel.com>
9378 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9379 * config/i386/i386.md (rdpid_rex64) New.
9380 (rdpid): Make 32bit only.
9382 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9385 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9386 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9388 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9391 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9393 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9395 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9397 PR rtl-optimization/84071
9398 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9399 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9401 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9403 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9404 (arc_attribute_table): Add 'aux' attribute.
9405 (arc_in_small_data_p): Consider aux like variables.
9406 (arc_is_aux_reg_p): New function.
9407 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9408 (arc_get_aux_arg): New function.
9409 (prepare_move_operands): Handle aux-register access.
9410 (arc_handle_aux_attribute): New function.
9411 * doc/extend.texi (ARC Variable attributes): Add subsection.
9413 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9415 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9416 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9417 (arc_attribute_table): Add 'uncached' attribute.
9418 (arc_print_operand): Print '.di' flag for uncached memory
9420 (arc_in_small_data_p): Do not consider for small data the uncached
9422 (arc_is_uncached_mem_p): New function.
9423 * config/arc/predicates.md (compact_store_memory_operand): Check
9424 for uncached memory accesses.
9425 (nonvol_nonimm_operand): Likewise.
9426 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9428 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9431 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9432 falign-loops=): Add Optimization flag.
9434 2018-01-30 Jeff Law <law@redhat.com>
9437 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9438 INT_REGISTERS_SAVED. Check it prior to calling
9439 get_scratch_register_on_entry.
9440 (ix86_adjust_stack_and_probe): Similarly.
9441 (ix86_emit_probe_stack_range): Similarly.
9442 (ix86_expand_prologue): Corresponding changes.
9444 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9447 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9448 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9450 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9453 * lra-constraints.c (curr_insn_transform): Process AND in the
9456 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9458 PR rtl-optimization/83986
9459 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9460 dependence against last_pending_memory_flush in addition to
9463 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9465 PR tree-optimization/81611
9466 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9469 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9472 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9475 2018-01-30 Richard Biener <rguenther@suse.de>
9476 Jakub Jelinek <jakub@redhat.com>
9478 PR tree-optimization/84111
9479 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9480 inner loops added during recursion, as they don't have up-to-date
9483 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9486 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9487 (can_inline_edge_by_limits_p): ... here.
9488 (can_early_inline_edge_p, check_callers,
9489 update_caller_keys, update_callee_keys, recursive_inlining,
9490 add_new_edges_to_heap, speculation_useful_p,
9491 inline_small_functions,
9492 inline_small_functions, flatten_function,
9493 inline_to_all_callers_1): Update.
9495 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9497 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9500 2018-01-30 Richard Biener <rguenther@suse.de>
9502 PR tree-optimization/83008
9503 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9504 invariant and constant vector uses in stmts when they need
9507 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9510 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9511 * configure: Regenerate.
9513 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9515 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9517 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9518 Use gen_rtx_REG rather than gen_lowpart.
9520 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9522 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9523 rather than 0 when creating partial subregs.
9525 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9527 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9530 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9533 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9534 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9535 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9536 flags. This restores the settings used before the 2017-07-24.
9537 Turning off pre increment/decrement/modify allows IVOPTS to
9538 optimize DF/SF loops where the index is an int.
9540 2018-01-29 Richard Biener <rguenther@suse.de>
9541 Kelvin Nilsen <kelvin@gcc.gnu.org>
9544 * tree-vect-stmts.c (vectorizable_call): Don't call
9545 targetm.vectorize_builtin_md_vectorized_function if callee is
9548 2018-01-22 Carl Love <cel@us.ibm.com>
9550 * doc/extend.tex: Fix typo in second arg in
9551 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9553 2018-01-29 Richard Biener <rguenther@suse.de>
9555 PR tree-optimization/84086
9556 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9557 (flush_ssaname_freelist): When SSA names were released reset
9558 the SCEV hash table.
9560 2018-01-29 Richard Biener <rguenther@suse.de>
9562 PR tree-optimization/84057
9563 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9564 removed paths when removing edges.
9566 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9568 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9569 -mfunction-return=@var{choice}.
9571 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9574 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9576 (layout::print_source_line): Likewise.
9577 (test_get_line_width_without_trailing_whitespace): Add test cases.
9579 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9582 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9585 2018-01-26 Jim Wilson <jimw@sifive.com>
9587 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9589 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9592 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9594 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9595 and CMP + SUB-immediate -> SUBS.
9597 2018-01-26 Martin Sebor <msebor@redhat.com>
9599 PR tree-optimization/83896
9600 * tree-ssa-strlen.c (get_string_len): Rename...
9601 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9602 Avoid assuming length is constant.
9603 (handle_char_store): Use HOST_WIDE_INT for string length.
9605 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9608 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9609 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9611 2018-01-26 Richard Biener <rguenther@suse.de>
9613 PR rtl-optimization/84003
9614 * dse.c (record_store): Only record redundant stores when
9615 the earlier store aliases at least all accesses the later one does.
9617 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9619 PR rtl-optimization/83985
9620 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9621 REG_CFA_RESTORE insns.
9622 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9623 REG_CFA_RESTORE insns here.
9626 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9627 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9629 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9631 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9632 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9633 (arc_init): Likewise.
9634 (arc_override_options): Likewise.
9635 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9637 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9639 * config/arc/arc.h (TARGET_DBNZ): Define.
9640 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9641 properly set the tune attribute.
9642 (dbnz): Use TARGET_DBNZ guard.
9643 * config/arc/arc.opt (mtune): Add core3 option.
9645 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9647 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9648 recognize new pic like addresses.
9649 (arc_delegitimize_address): Clean up.
9651 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9653 * config/arc/arc-arches.def: Option mrf16 valid for all
9655 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9656 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9657 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9658 * config/arc/arc-tables.opt: Regenerate.
9659 * config/arc/arc.c (arc_conditional_register_usage): Handle
9660 reduced register file case.
9661 (arc_file_start): Set must have build attributes.
9662 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9664 * config/arc/arc.opt (mrf16): Add new option.
9665 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9666 * config/arc/genmultilib.awk: Handle new mrf16 option.
9667 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9668 * config/arc/t-multilib: Regenerate.
9669 * doc/invoke.texi (ARC Options): Document mrf16 option.
9671 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9673 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9674 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9675 (arc_attribute_table): Add 'secure_call' attribute.
9676 (arc_print_operand): Print secure call operand.
9677 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9679 (arc_is_secure_call_p): New function. * config/arc/arc.md
9680 (call_i): Add support for sjli instruction.
9681 (call_value_i): Likewise.
9682 * config/arc/constraints.md (Csc): New constraint.
9684 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9685 John Eric Martin <John.Martin@emmicro-us.com>
9687 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9688 * config/arc/arc.c (_arc_jli_section): New struct.
9689 (arc_jli_section): New type.
9690 (rc_jli_sections): New static variable.
9691 (arc_handle_jli_attribute): New function.
9692 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9693 (arc_file_end): New function.
9694 (TARGET_ASM_FILE_END): Define.
9695 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9696 (arc_add_jli_section): New function.
9697 (jli_call_scan): Likewise.
9698 (arc_reorg): Call jli_call_scan.
9699 (arc_output_addsi): Remove 'S' from printing asm operand.
9700 (arc_is_jli_call_p): New function.
9701 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9703 (movhi_insn): Likewise.
9704 (movsi_insn): Likewise.
9705 (movsi_set_cc_insn): Likewise.
9706 (loadqi_update): Likewise.
9707 (load_zeroextendqisi_update): Likewise.
9708 (load_signextendqisi_update): Likewise.
9709 (loadhi_update): Likewise.
9710 (load_zeroextendhisi_update): Likewise.
9711 (load_signextendhisi_update): Likewise.
9712 (loadsi_update): Likewise.
9713 (loadsf_update): Likewise.
9714 (movsicc_insn): Likewise.
9715 (bset_insn): Likewise.
9716 (bxor_insn): Likewise.
9717 (bclr_insn): Likewise.
9718 (bmsk_insn): Likewise.
9719 (bicsi3_insn): Likewise.
9720 (cmpsi_cc_c_insn): Likewise.
9721 (movsi_ne): Likewise.
9722 (movsi_cond_exec): Likewise.
9723 (clrsbsi2): Likewise.
9734 (call_i): Remove 'S' asm letter, add jli instruction.
9735 (call_value_i): Likewise.
9736 * config/arc/arc.op (mjli-always): New option.
9737 * config/arc/constraints.md (Cji): New constraint.
9738 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9740 (subsf3_fpx): Likewise.
9741 (mulsf3_fpx): Likewise.
9742 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9744 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9746 * doc/invoke.texi (ARC): Document mjli-always option.
9748 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9750 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9751 avoid addition with 0 and use incw and decw where possible.
9753 2018-01-26 Richard Biener <rguenther@suse.de>
9755 PR tree-optimization/81082
9756 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9757 association if it requires casting to unsigned.
9758 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9759 from fold_plusminus_mult_expr to catch important cases late when
9760 range info is available.
9762 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9764 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9765 * configure.ac (hidden_linkonce): New test.
9766 * configure: Regenerate.
9767 * config.in: Regenerate.
9769 2018-01-26 Julia Koval <julia.koval@intel.com>
9771 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9772 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9773 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9774 _mm_mask_bitshuffle_epi64_mask): Fix type.
9775 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9776 USI_FTYPE_V4DI_V4DI_USI): Remove.
9777 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9778 __builtin_ia32_vpshufbitqmb256_mask,
9779 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9780 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9781 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9783 2018-01-26 Alan Modra <amodra@gmail.com>
9786 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9787 UNSPEC_VBPERMQ. Sort other unspecs.
9789 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9791 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9793 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9796 * predict.c (drop_profile): Do not push/pop cfun; update also
9798 (handle_missing_profiles): Fix logic looking for zero profiles.
9800 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9803 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9804 on functions with #pragma omp declare simd or functions with simd
9806 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9807 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9808 Remove trailing \n from warning_at calls.
9810 2018-01-25 Tom de Vries <tom@codesourcery.com>
9813 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9814 for neutered workers.
9816 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9819 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9820 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9822 2018-01-24 Jeff Law <law@redhat.com>
9825 * i386.c (get_probe_interval): Move to earlier point.
9826 (ix86_compute_frame_layout): If -fstack-clash-protection and
9827 the frame is larger than the probe interval, then use pushes
9828 to save registers rather than reg->mem moves.
9829 (ix86_expand_prologue): Remove conditional for int_registers_saved
9832 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9835 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9836 min/max for never referenced object.
9838 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9841 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9843 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9844 attributes from DECL_ATTRIBUTES (decl) without affecting
9845 DECL_ATTRIBUTES (current_function_decl).
9846 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9847 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9849 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9851 PR tree-optimization/83979
9852 * fold-const.c (fold_comparison): Use constant_boolean_node
9853 instead of boolean_{true,false}_node.
9855 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9857 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9860 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9862 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9863 Simplify the clause that sets the length attribute.
9864 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9865 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9866 clause that sets the length attribute.
9867 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9869 2018-01-24 Tom de Vries <tom@codesourcery.com>
9872 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9873 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9874 Add strict parameter.
9875 (prevent_branch_around_nothing): Insert dummy insn between branch to
9876 label and label with no ptx insn inbetween.
9877 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9879 2018-01-24 Tom de Vries <tom@codesourcery.com>
9882 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9883 for neutered threads in warp.
9884 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9886 2018-01-24 Richard Biener <rguenther@suse.de>
9888 PR tree-optimization/83176
9889 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9892 2018-01-24 Richard Biener <rguenther@suse.de>
9894 PR tree-optimization/82819
9895 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9896 code generating pluses that are no-ops in the target precision.
9898 2018-01-24 Richard Biener <rguenther@suse.de>
9901 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9903 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9905 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9906 to merge probabilities.
9907 * predict.c (probably_never_executed): Also mark as cold functions
9908 with global 0 profile and guessed local profile.
9909 * profile-count.c (profile_probability::combine_with_count): New
9911 * profile-count.h (profile_probability::operator*,
9912 profile_probability::operator*=, profile_probability::operator/,
9913 profile_probability::operator/=): Reduce precision to adjusted
9914 and set value to guessed on contradictory divisions.
9915 (profile_probability::combine_with_freq): Remove.
9916 (profile_probability::combine_wiht_count): Declare.
9917 (profile_count::force_nonzero):: Set to adjusted.
9918 (profile_count::probability_in):: Set quality to adjusted.
9919 * tree-ssa-tail-merge.c (replace_block_by): Use
9922 2018-01-23 Andrew Waterman <andrew@sifive.com>
9923 Jim Wilson <jimw@sifive.com>
9925 * config/riscv/riscv.c (riscv_stack_boundary): New.
9926 (riscv_option_override): Set riscv_stack_boundary. Handle
9927 riscv_preferred_stack_boundary_arg.
9928 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9929 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9930 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9931 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9932 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9933 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9935 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9938 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9939 of struct ix86_frame.
9940 (ix86_expand_epilogue): Likewise. Add a local variable for
9941 the reg_save_offset field in struct ix86_frame.
9943 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9945 PR tree-optimization/82604
9946 * tree-loop-distribution.c (enum partition_kind): New enum item
9947 PKIND_PARTIAL_MEMSET.
9948 (partition_builtin_p): Support above new enum item.
9949 (generate_code_for_partition): Ditto.
9950 (compute_access_range): Differentiate cases that equality can be
9951 proven at all loops, the innermost loops or no loops.
9952 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9953 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9954 (finalize_partitions, distribute_loop): Don't fuse partition of
9955 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9956 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9959 2018-01-23 Martin Liska <mliska@suse.cz>
9961 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9962 order to ignore the predictor.
9963 (PRED_POLYMORPHIC_CALL): Likewise.
9964 (PRED_RECURSIVE_CALL): Likewise.
9966 2018-01-23 Martin Liska <mliska@suse.cz>
9968 * tree-profile.c (tree_profiling): Print function header to
9969 aware reader which function we are working on.
9970 * value-prof.c (gimple_find_values_to_profile): Do not print
9971 not interesting value histograms.
9973 2018-01-23 Martin Liska <mliska@suse.cz>
9975 * profile-count.h (enum profile_quality): Add
9976 profile_uninitialized as the first value. Do not number values
9977 as they are zero based.
9978 (profile_count::verify): Update sanity check.
9979 (profile_probability::verify): Likewise.
9981 2018-01-23 Nathan Sidwell <nathan@acm.org>
9983 * doc/invoke.texi (ffor-scope): Deprecate.
9985 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9987 PR tree-optimization/83510
9988 * domwalk.c (set_all_edges_as_executable): New function.
9989 (dom_walker::dom_walker): Convert bool param
9990 "skip_unreachable_blocks" to enum reachability. Move setup of
9991 edge flags to set_all_edges_as_executable and only do it when
9992 reachability is REACHABLE_BLOCKS.
9993 * domwalk.h (enum dom_walker::reachability): New enum.
9994 (dom_walker::dom_walker): Convert bool param
9995 "skip_unreachable_blocks" to enum reachability.
9996 (set_all_edges_as_executable): New decl.
9997 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9998 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10000 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10001 but converting true to REACHABLE_BLOCKS.
10002 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10004 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10005 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10006 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10008 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10009 if check_all_array_refs will be called.
10011 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10013 * tree.c (selftest::test_location_wrappers): Add more test
10016 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10018 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10019 (selftest::test_bit_in_range): Likewise.
10021 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10024 * doc/sourcebuild.texi (vect_float): Say that the selector
10025 only describes the situation when -funsafe-math-optimizations is on.
10026 (vect_float_strict): Document.
10028 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10030 PR tree-optimization/83965
10031 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10032 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10033 instead of checking only for a reduction.
10034 (vect_recog_widen_sum_pattern): Likewise.
10036 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10038 * predict.c (probably_never_executed): Only use precise profile info.
10039 (compute_function_frequency): Skip after inlining hack since we now
10040 have quality checking.
10042 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10044 * profile-count.h (profile_probability::very_unlikely,
10045 profile_probability::unlikely, profile_probability::even): Set
10046 precision to guessed.
10048 2018-01-23 Richard Biener <rguenther@suse.de>
10050 PR tree-optimization/83963
10051 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10052 Properly terminate dominator walk when crossing the exit edge not
10053 when visiting its source block.
10055 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10058 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10059 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10061 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10063 PR tree-optimization/83957
10064 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10065 semicolon after for body surrounded by braces.
10067 PR tree-optimization/83081
10068 * profile-count.h (profile_probability::split): New method.
10069 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10070 Use profile_probability::split.
10071 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10072 when splitting a single conditional jump into 2.
10074 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10076 PR tree-optimization/69452
10077 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10080 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10082 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10083 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10084 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10086 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10088 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10089 * config/rl78/rl78.md (movdi): New define_expand.
10090 * config/rl78/rl78.c (rl78_split_movdi): New function.
10092 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10095 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10097 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10098 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10099 128-bit to produce an UNSPEC move to get the double word with the
10100 signbit and then a shift directly to do signbit.
10101 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10102 implementation with a new version that just does either a direct
10103 move or a regular move. Move memory interface to separate insns.
10104 Move insns so they are next to the expander.
10105 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10106 with signbit move. Split big and little endian case.
10107 (signbit<mode>2_dm_mem_le): Likewise.
10108 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10109 (signbit<mode>2_dm2): Likewise.
10111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10113 * config/rl78/rl78.md (anddi3): New define_expand.
10115 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10117 * config/rl78/rl78.md (umindi3): New define_expand.
10119 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10121 * config/rl78/rl78.md (smindi3): New define_expand.
10123 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10125 * config/rl78/rl78.md (smaxdi3): New define_expand.
10127 2018-01-22 Carl Love <cel@us.ibm.com>
10129 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10130 LVX_V1TI): Add macro expansion.
10131 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10132 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10133 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10134 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10135 Change check to determine if the instruction is a byte reversing
10136 entry. Fix typo in comment.
10137 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10138 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10139 Add def_builtin calls for new builtins.
10140 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10141 Add define_insn expansion.
10143 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10145 * config/rl78/rl78.md (umaxdi3): New define_expand.
10147 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10149 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10150 for non-QImode registers.
10152 2018-01-22 Richard Biener <rguenther@suse.de>
10154 PR tree-optimization/83963
10155 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10156 including the loop exit block.
10157 (scop_detection::merge_sese): Likewise.
10158 (scop_detection::add_scop): Do it here instead.
10160 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10162 * doc/sourcebuild.texi (arm_softfloat): Document.
10164 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10167 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10168 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10169 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10171 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10172 David Edelsohn <dje.gcc@gmail.com>
10175 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10176 Change "crset eq" to "crset 2".
10177 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10178 (*call_indirect_aix<mode>_nospec): Likewise.
10179 (*call_value_indirect_aix<mode>_nospec): Likewise.
10180 (*call_indirect_elfv2<mode>_nospec): Likewise.
10181 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10182 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10183 change assembly output from . to $.
10184 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10185 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10186 (*tablejump<mode>_internal1_nospec): Likewise.
10188 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10191 * config/sh/sh_optimize_sett_clrt.cc:
10192 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10194 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10196 PR tree-optimization/83940
10197 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10198 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10199 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10200 use it to pass back the definition type.
10201 (vect_check_store_rhs): Likewise rhs_dt_out.
10202 (vect_build_gather_load_calls): Add a mask_dt argument and use
10203 it instead of a call to vect_is_simple_use.
10204 (vectorizable_store): Update calls to vect_check_load_store_mask
10205 and vect_check_store_rhs. Use the dt returned by the latter instead
10206 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10207 instead of calls to vect_is_simple_use. Pass the scalar rather
10208 than the vector operand to vect_is_simple_use when handling
10209 second and subsequent copies of an rhs value.
10210 (vectorizable_load): Update calls to vect_check_load_store_mask
10211 and vect_build_gather_load_calls. Use the cached mask_dt and
10212 gs_info.offset_dt instead of calls to vect_is_simple_use.
10214 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10216 PR middle-end/83945
10217 * tree-emutls.c: Include gimplify.h.
10218 (lower_emutls_2): New function.
10219 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10220 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10221 it before further processing.
10224 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10225 UINTVAL (trueop1) instead of INTVAL (op1).
10227 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10231 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10232 INCOMING_FRAME_SP_OFFSET if not defined.
10233 (scan_trace): Add ENTRY argument. If true and
10234 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10235 emit a note to adjust the CFA offset.
10236 (create_cfi_notes): Adjust scan_trace callers.
10237 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10238 INCOMING_FRAME_SP_OFFSET in the CIE.
10239 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10240 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10242 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10243 * doc/tm.texi: Regenerated.
10245 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10247 PR rtl-optimization/83147
10248 * lra-constraints.c (remove_inheritance_pseudos): Use
10249 lra_substitute_pseudo_within_insn.
10251 2018-01-19 Tom de Vries <tom@codesourcery.com>
10252 Cesar Philippidis <cesar@codesourcery.com>
10255 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10257 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10260 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10261 spaces for function labels.
10263 2018-01-19 Martin Liska <mliska@suse.cz>
10265 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10266 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10267 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10268 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10269 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10270 (PRED_CONST_RETURN): Change from 69 to 65.
10271 (PRED_NULL_RETURN): Change from 91 to 71.
10272 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10273 (PRED_LOOP_GUARD): Change from 66 to 73.
10275 2018-01-19 Martin Liska <mliska@suse.cz>
10277 * predict.c (predict_insn_def): Add new assert.
10278 (struct branch_predictor): Change type to signed integer.
10279 (test_prediction_value_range): Amend test to cover
10280 PROB_UNINITIALIZED.
10281 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10282 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10283 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10284 (PRED_LOOP_IV_COMPARE): Likewise.
10285 * predict.h (PROB_UNINITIALIZED): Define new constant.
10287 2018-01-19 Martin Liska <mliska@suse.cz>
10289 * predict.c (dump_prediction): Add new format for
10290 analyze_brprob.py script which is enabled with -details
10292 * profile-count.h (precise_p): New function.
10294 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10296 PR tree-optimization/83922
10297 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10298 there are no statements that need masking.
10299 (vect_active_double_reduction_p): New function.
10300 (vect_analyze_loop_operations): Use it when handling phis that
10301 are not in the loop header.
10303 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10305 PR tree-optimization/83914
10306 * tree-vect-loop.c (vectorizable_induction): Don't convert
10307 init_expr or apply the peeling adjustment for inductions
10308 that are nested within the vectorized loop.
10310 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10312 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10315 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10319 * function.h (gimplify_parameters): Add gimple_seq * argument.
10320 * function.c: Include gimple.h and options.h.
10321 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10322 for the added local temporaries if needed.
10323 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10324 if there are any parameter cleanups, wrap whole body into a
10325 try/finally with the cleanups.
10327 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10330 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10331 Use GET_MODE_CLASS for scalar floating point.
10333 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10337 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10338 Fix call of call_cgraph_insertion_hooks.
10340 2018-01-18 Martin Sebor <msebor@redhat.com>
10342 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10344 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10347 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10350 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10353 * common.opt: (-ffile-prefix-map): New option.
10354 * opts.c (common_handle_option): Defer it.
10355 * opts-global.c (handle_common_deferred_options): Handle it.
10356 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10357 * file-prefix-map.h: New file.
10358 (remap_debug_filename, add_debug_prefix_map): ...here.
10359 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10360 * final.c (debug_prefix_map, add_debug_prefix_map
10361 remap_debug_filename): Move to...
10362 * file-prefix-map.c: New file.
10363 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10364 generalize, get rid of alloca(), use strrchr() instead of strchr().
10365 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10366 Implement in terms of add_prefix_map().
10367 (remap_macro_filename, remap_debug_filename): Implement in term of
10369 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10370 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10371 * dbxout.c: Include file-prefix-map.h.
10372 * varasm.c: Likewise.
10373 * vmsdbgout.c: Likewise.
10374 * xcoffout.c: Likewise.
10375 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10376 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10377 * doc/invoke.texi (-ffile-prefix-map): Document.
10378 (-fdebug-prefix-map): Update description.
10380 2018-01-18 Martin Liska <mliska@suse.cz>
10382 * config/i386/i386.c (indirect_thunk_name): Document that also
10384 (output_indirect_thunk): Document why both instructions
10385 (pause and lfence) are generated.
10387 2018-01-18 Richard Biener <rguenther@suse.de>
10389 PR tree-optimization/83887
10390 * graphite-scop-detection.c
10391 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10392 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10393 (scop_detection::merge_sese): Re-implement with a flood-fill
10394 algorithm that properly finds a SESE region if it exists.
10396 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10399 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10400 pointer_diff optimizations use view_convert instead of convert.
10402 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10404 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10405 Generate different code for -mno-speculate-indirect-jumps.
10406 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10407 (*call_indirect_aix<mode>): Disable for
10408 -mno-speculate-indirect-jumps.
10409 (*call_indirect_aix<mode>_nospec): New define_insn.
10410 (*call_value_indirect_aix<mode>): Disable for
10411 -mno-speculate-indirect-jumps.
10412 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10413 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10414 -mno-speculate-indirect-jumps.
10415 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10417 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10419 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10420 long double type, set the flags for noting the default long double
10421 type, even if we don't pass or return a long double type.
10423 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10426 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10429 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10431 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10432 support for merge[hl].
10433 (fold_mergehl_helper): New helper function.
10434 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10435 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10436 (altivec_vmrglw_direct): Add xxmrglw insn.
10438 2018-01-17 Andrew Waterman <andrew@sifive.com>
10440 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10441 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10443 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10446 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10447 call the lto_location_cache before reading the
10448 DECL_SOURCE_LOCATION of the types.
10450 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10451 Richard Sandiford <richard.sandiford@linaro.org>
10453 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10454 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10455 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10456 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10457 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10459 * config/aarch64/constraints.md (aarch64_movti_operand):
10461 * config/aarch64/predicates.md (Uti): Add new constraint.
10463 2018-01-17 Carl Love <cel@us.ibm.com>
10465 * config/rs6000/vsx.md (define_expand xl_len_r,
10466 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10467 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10469 (define_expand, define_insn): Move the shift left from the
10470 define_insn to the define_expand for lxvl and stxvl instructions.
10471 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10472 and XL_LEN_R definitions to PURE.
10474 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10476 * config/i386/i386.c (indirect_thunk_name): Declare regno
10477 as unsigned int. Compare regno with INVALID_REGNUM.
10478 (output_indirect_thunk): Ditto.
10479 (output_indirect_thunk_function): Ditto.
10480 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10481 in the call to output_indirect_thunk_function.
10483 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10485 PR middle-end/83884
10486 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10487 rather than the size of inner_type to determine the stack slot size
10488 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10490 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10493 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10496 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10498 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10499 endian Linux systems to optionally enable multilibs for selecting
10500 the long double type if the user configured an explicit type.
10501 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10502 have no long double multilibs if not defined.
10503 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10504 warn if the user used -mabi={ieee,ibm}longdouble and we built
10505 multilibs for long double.
10506 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10507 appropriate multilib option.
10508 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10510 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10511 for building long double multilibs.
10512 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10514 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10516 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10519 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10521 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10524 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10527 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10530 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10532 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10533 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10535 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10537 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10538 different rtl trees depending on TARGET_64BIT.
10539 (rs6000_gen_lvx): Likewise.
10541 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10543 * config/visium/visium.md (nop): Tweak comment.
10544 (hazard_nop): Likewise.
10546 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10548 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10549 -mspeculate-indirect-jumps.
10550 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10551 for -mno-speculate-indirect-jumps.
10552 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10553 (*call_value_indirect_elfv2<mode>): Disable for
10554 -mno-speculate-indirect-jumps.
10555 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10556 (indirect_jump): Emit different RTL for
10557 -mno-speculate-indirect-jumps.
10558 (*indirect_jump<mode>): Disable for
10559 -mno-speculate-indirect-jumps.
10560 (*indirect_jump<mode>_nospec): New define_insn.
10561 (tablejump): Emit different RTL for
10562 -mno-speculate-indirect-jumps.
10563 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10564 (tablejumpsi_nospec): New define_expand.
10565 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10566 (tablejumpdi_nospec): New define_expand.
10567 (*tablejump<mode>_internal1): Disable for
10568 -mno-speculate-indirect-jumps.
10569 (*tablejump<mode>_internal1_nospec): New define_insn.
10570 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10573 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10575 * caller-save.c (insert_save): Drop unnecessary parameter. All
10578 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10579 Richard Biener <rguenth@suse.de>
10582 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10583 return early, inline manually is_gimple_sizepos. Make sure if we
10584 call gimplify_expr we don't end up with a gimple constant.
10585 * tree.c (variably_modified_type_p): Don't return true for
10586 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10587 * gimplify.h (is_gimple_sizepos): Remove.
10589 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10591 PR tree-optimization/83857
10592 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10593 vectorizable_live_operation for pure SLP statements.
10594 (vectorizable_live_operation): Handle PHIs.
10596 2018-01-16 Richard Biener <rguenther@suse.de>
10598 PR tree-optimization/83867
10599 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10600 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10602 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10605 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10606 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10607 If off is not INTEGER_CST, issue a may not be aligned warning
10608 rather than isn't aligned. Use isn%'t rather than isn't.
10609 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10611 <case MULT_EXPR>: Improve the case when bottom and one of the
10612 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10613 operand, in that case check if the other operand is multiple of
10614 bottom divided by the INTEGER_CST operand.
10616 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10619 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10620 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10621 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10622 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10623 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10624 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10625 (pa_function_arg_size): New function.
10627 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10629 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10630 in a separate statement.
10632 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10634 PR tree-optimization/83847
10635 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10636 group gathers and scatters.
10638 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10640 PR rtl-optimization/86620
10641 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10643 PR rtl-optimization/83213
10644 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10645 to last if both are JUMP_INSNs.
10647 PR tree-optimization/83843
10648 * gimple-ssa-store-merging.c
10649 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10650 store_immediate_info for bswap/nop orig_stores.
10652 2018-01-15 Andrew Waterman <andrew@sifive.com>
10654 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10656 <UDIV>: Increase cost if !TARGET_DIV.
10658 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10660 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10661 (define_attr "cr_logical_3op"): New.
10662 (cceq_ior_compare): Adjust.
10663 (cceq_ior_compare_complement): Adjust.
10664 (*cceq_rev_compare): Adjust.
10665 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10666 (is_cracked_insn): Adjust.
10667 (insn_must_be_first_in_group): Adjust.
10668 * config/rs6000/40x.md: Adjust.
10669 * config/rs6000/440.md: Adjust.
10670 * config/rs6000/476.md: Adjust.
10671 * config/rs6000/601.md: Adjust.
10672 * config/rs6000/603.md: Adjust.
10673 * config/rs6000/6xx.md: Adjust.
10674 * config/rs6000/7450.md: Adjust.
10675 * config/rs6000/7xx.md: Adjust.
10676 * config/rs6000/8540.md: Adjust.
10677 * config/rs6000/cell.md: Adjust.
10678 * config/rs6000/e300c2c3.md: Adjust.
10679 * config/rs6000/e500mc.md: Adjust.
10680 * config/rs6000/e500mc64.md: Adjust.
10681 * config/rs6000/e5500.md: Adjust.
10682 * config/rs6000/e6500.md: Adjust.
10683 * config/rs6000/mpc.md: Adjust.
10684 * config/rs6000/power4.md: Adjust.
10685 * config/rs6000/power5.md: Adjust.
10686 * config/rs6000/power6.md: Adjust.
10687 * config/rs6000/power7.md: Adjust.
10688 * config/rs6000/power8.md: Adjust.
10689 * config/rs6000/power9.md: Adjust.
10690 * config/rs6000/rs64.md: Adjust.
10691 * config/rs6000/titan.md: Adjust.
10693 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10695 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10696 ix86_indirect_branch_register logic.
10698 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10700 * config/i386/constraints.md (Bs): Update
10701 ix86_indirect_branch_register check. Don't check
10702 ix86_indirect_branch_register with GOT_memory_operand.
10704 * config/i386/predicates.md (GOT_memory_operand): Don't check
10705 ix86_indirect_branch_register here.
10706 (GOT32_symbol_operand): Likewise.
10708 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10710 * config/i386/predicates.md (constant_call_address_operand):
10711 Rewrite ix86_indirect_branch_register logic.
10712 (sibcall_insn_operand): Likewise.
10714 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10716 * config/i386/constraints.md (Bs): Replace
10717 ix86_indirect_branch_thunk_register with
10718 ix86_indirect_branch_register.
10720 * config/i386/i386.md (indirect_jump): Likewise.
10721 (tablejump): Likewise.
10722 (*sibcall_memory): Likewise.
10723 (*sibcall_value_memory): Likewise.
10724 Peepholes of indirect call and jump via memory: Likewise.
10725 * config/i386/i386.opt: Likewise.
10726 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10727 (GOT_memory_operand): Likewise.
10728 (call_insn_operand): Likewise.
10729 (sibcall_insn_operand): Likewise.
10730 (GOT32_symbol_operand): Likewise.
10732 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10734 PR middle-end/83837
10735 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10736 type rather than type addr's type points to.
10737 (expand_omp_atomic_mutex): Likewise.
10738 (expand_omp_atomic): Likewise.
10740 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10743 * config/i386/i386.c (output_indirect_thunk_function): Use
10744 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10745 for __x86_return_thunk.
10747 2018-01-15 Richard Biener <rguenther@suse.de>
10749 PR middle-end/83850
10750 * expmed.c (extract_bit_field_1): Fix typo.
10752 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10755 * config/arm/iterators.md (VF): New mode iterator.
10756 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10757 Remove integer-related logic from pattern.
10758 (neon_vabd<mode>_3): Likewise.
10760 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10762 PR middle-end/82694
10763 * common.opt (fstrict-overflow): No longer an alias.
10764 (fwrapv-pointer): New option.
10765 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10766 also for pointer types based on flag_wrapv_pointer.
10767 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10768 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10769 opts->x_flag_wrapv got set.
10770 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10771 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10772 POINTER_TYPE_OVERFLOW_UNDEFINED.
10773 * match.pd: Likewise in address comparison pattern.
10774 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10776 2018-01-15 Richard Biener <rguenther@suse.de>
10779 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10780 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10781 Reset type names to their identifier if their TYPE_DECL doesn't
10782 have linkage (and thus is used for ODR and devirt).
10783 (save_debug_info_for_decl): Remove.
10784 (save_debug_info_for_type): Likewise.
10785 (add_tree_to_fld_list): Adjust.
10786 * tree-pretty-print.c (dump_generic_node): Make dumping of
10787 type names more robust.
10789 2018-01-15 Richard Biener <rguenther@suse.de>
10791 * BASE-VER: Bump to 8.0.1.
10793 2018-01-14 Martin Sebor <msebor@redhat.com>
10796 * builtins.c (check_access): Avoid warning when the no-warning bit
10799 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10801 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10802 * ira-color (allocno_hard_regs_compare): Likewise.
10804 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10807 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10808 Use .pushsection/.popsection.
10810 2018-01-14 Martin Sebor <msebor@redhat.com>
10813 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10815 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10817 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10818 entry from extra_headers.
10819 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10820 extra_headers, make the list bitwise identical to the i?86-*-* one.
10822 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10824 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10825 -mcmodel=large with -mindirect-branch=thunk,
10826 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10827 -mfunction-return=thunk-extern.
10828 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10829 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10830 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10832 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10834 * config/i386/i386.c (print_reg): Print the name of the full
10835 integer register without '%'.
10836 (ix86_print_operand): Handle 'V'.
10837 * doc/extend.texi: Document 'V' modifier.
10839 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10841 * config/i386/constraints.md (Bs): Disallow memory operand for
10842 -mindirect-branch-register.
10844 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10845 (GOT_memory_operand): Likewise.
10846 (call_insn_operand): Likewise.
10847 (sibcall_insn_operand): Likewise.
10848 (GOT32_symbol_operand): Likewise.
10849 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10850 for -mindirect-branch-register.
10851 (tablejump): Likewise.
10852 (*sibcall_memory): Likewise.
10853 (*sibcall_value_memory): Likewise.
10854 Disallow peepholes of indirect call and jump via memory for
10855 -mindirect-branch-register.
10856 (*call_pop): Replace m with Bw.
10857 (*call_value_pop): Likewise.
10858 (*sibcall_pop_memory): Replace m with Bs.
10859 * config/i386/i386.opt (mindirect-branch-register): New option.
10860 * doc/invoke.texi: Document -mindirect-branch-register option.
10862 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10864 * config/i386/i386-protos.h (ix86_output_function_return): New.
10865 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10866 set function_return_type.
10867 (indirect_thunk_name): Add ret_p to indicate thunk for function
10869 (output_indirect_thunk_function): Pass false to
10870 indirect_thunk_name.
10871 (ix86_output_indirect_branch_via_reg): Likewise.
10872 (ix86_output_indirect_branch_via_push): Likewise.
10873 (output_indirect_thunk_function): Create alias for function
10874 return thunk if regno < 0.
10875 (ix86_output_function_return): New function.
10876 (ix86_handle_fndecl_attribute): Handle function_return.
10877 (ix86_attribute_table): Add function_return.
10878 * config/i386/i386.h (machine_function): Add
10879 function_return_type.
10880 * config/i386/i386.md (simple_return_internal): Use
10881 ix86_output_function_return.
10882 (simple_return_internal_long): Likewise.
10883 * config/i386/i386.opt (mfunction-return=): New option.
10884 (indirect_branch): Mention -mfunction-return=.
10885 * doc/extend.texi: Document function_return function attribute.
10886 * doc/invoke.texi: Document -mfunction-return= option.
10888 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10890 * config/i386/i386-opts.h (indirect_branch): New.
10891 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10892 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10893 with local indirect jump when converting indirect call and jump.
10894 (ix86_set_indirect_branch_type): New.
10895 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10896 (indirectlabelno): New.
10897 (indirect_thunk_needed): Likewise.
10898 (indirect_thunk_bnd_needed): Likewise.
10899 (indirect_thunks_used): Likewise.
10900 (indirect_thunks_bnd_used): Likewise.
10901 (INDIRECT_LABEL): Likewise.
10902 (indirect_thunk_name): Likewise.
10903 (output_indirect_thunk): Likewise.
10904 (output_indirect_thunk_function): Likewise.
10905 (ix86_output_indirect_branch_via_reg): Likewise.
10906 (ix86_output_indirect_branch_via_push): Likewise.
10907 (ix86_output_indirect_branch): Likewise.
10908 (ix86_output_indirect_jmp): Likewise.
10909 (ix86_code_end): Call output_indirect_thunk_function if needed.
10910 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10912 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10913 (ix86_attribute_table): Add indirect_branch.
10914 * config/i386/i386.h (machine_function): Add indirect_branch_type
10915 and has_local_indirect_jump.
10916 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10918 (tablejump): Likewise.
10919 (*indirect_jump): Use ix86_output_indirect_jmp.
10920 (*tablejump_1): Likewise.
10921 (simple_return_indirect_internal): Likewise.
10922 * config/i386/i386.opt (mindirect-branch=): New option.
10923 (indirect_branch): New.
10926 (thunk-inline): Likewise.
10927 (thunk-extern): Likewise.
10928 * doc/extend.texi: Document indirect_branch function attribute.
10929 * doc/invoke.texi: Document -mindirect-branch= option.
10931 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10934 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10936 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10938 * ipa-inline.c (want_inline_small_function_p): Return false if
10939 inlining has already failed with CIF_FINAL_ERROR.
10940 (update_caller_keys): Call want_inline_small_function_p before
10942 (update_callee_keys): Likewise.
10944 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10946 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10948 (rs6000_quadword_masked_address_p): Likewise.
10949 (quad_aligned_load_p): Likewise.
10950 (quad_aligned_store_p): Likewise.
10951 (const_load_sequence_p): Add comment to describe the outer-most loop.
10952 (mimic_memory_attributes_and_flags): New function.
10953 (rs6000_gen_stvx): Likewise.
10954 (replace_swapped_aligned_store): Likewise.
10955 (rs6000_gen_lvx): Likewise.
10956 (replace_swapped_aligned_load): Likewise.
10957 (replace_swapped_load_constant): Capitalize argument name in
10958 comment describing this function.
10959 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10960 and stores that access quad-word aligned addresses and replace
10961 with stvx or lvx instructions when appropriate.
10962 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10963 New function prototype.
10964 (rs6000_quadword_masked_address_p): Likewise.
10965 (rs6000_gen_lvx): Likewise.
10966 (rs6000_gen_stvx): Likewise.
10967 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10968 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10969 when memory address is aligned.
10970 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10971 this split to select lvx instruction when memory address is aligned.
10972 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10973 instruction when memory address is aligned.
10974 (*vsx_le_perm_load_v16qi): Likewise.
10975 (four unnamed splitters): Modify to select the stvx instruction
10976 when memory is aligned.
10978 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10980 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10981 which appears in the queue multiple times.
10983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10984 Alan Hayward <alan.hayward@arm.com>
10985 David Sherwood <david.sherwood@arm.com>
10987 * tree-vectorizer.h (vec_lower_bound): New structure.
10988 (_loop_vec_info): Add check_nonzero and lower_bounds.
10989 (LOOP_VINFO_CHECK_NONZERO): New macro.
10990 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10991 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10992 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10993 fields. Make seg_len the distance travelled, not including the
10995 (dr_direction_indicator): Declare.
10996 (dr_zero_step_indicator): Likewise.
10997 (dr_known_forward_stride_p): Likewise.
10998 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11000 (runtime_alias_check_p): Allow runtime alias checks with
11002 (operator ==): Compare access_size and align.
11003 (prune_runtime_alias_test_list): Rework for new distinction between
11004 the access_size and seg_len.
11005 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11007 (get_segment_min_max): New function.
11008 (create_intersect_range_checks): Use it.
11009 (dr_step_indicator): New function.
11010 (dr_direction_indicator): Likewise.
11011 (dr_zero_step_indicator): Likewise.
11012 (dr_known_forward_stride_p): Likewise.
11013 * tree-loop-distribution.c (data_ref_segment_size): Return
11014 DR_STEP * (niters - 1).
11015 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11017 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11018 (vect_preserves_scalar_order_p): New function, split out from...
11019 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11020 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11021 (vect_vfa_access_size): New function.
11022 (vect_vfa_align): Likewise.
11023 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11024 (dump_lower_bound): New function.
11025 (vect_check_lower_bound): Likewise.
11026 (vect_small_gap_p): Likewise.
11027 (vectorizable_with_step_bound_p): Likewise.
11028 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11029 depencies if the vectorization factor is 1. Convert the checks
11030 for nonzero steps into checks on the bounds of DR_STEP. Try using
11031 a bunds check for variable steps if the minimum required step is
11032 relatively small. Update calls to the dr_with_seg_len
11033 constructor and to vect_compile_time_alias.
11034 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11036 (vect_loop_versioning): Call it.
11037 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11039 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * doc/sourcebuild.texi (vect_scatter_store): Document.
11046 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11048 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11050 * genopinit.c (main): Add supports_vec_scatter_store and
11051 supports_vec_scatter_store_cached to target_optabs.
11052 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11053 IFN_MASK_SCATTER_STORE.
11054 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11056 * internal-fn.h (internal_store_fn_p): Declare.
11057 (internal_fn_stored_value_index): Likewise.
11058 * internal-fn.c (scatter_store_direct): New macro.
11059 (expand_scatter_store_optab_fn): New function.
11060 (direct_scatter_store_optab_supported_p): New macro.
11061 (internal_store_fn_p): New function.
11062 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11063 IFN_MASK_SCATTER_STORE.
11064 (internal_fn_mask_index): Likewise.
11065 (internal_fn_stored_value_index): New function.
11066 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11067 for scatter stores.
11068 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11069 * optabs-query.c (supports_vec_scatter_store_p): New function.
11070 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11071 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11072 true for scatter stores.
11073 (vect_gather_scatter_fn_p): Handle scatter stores too.
11074 (vect_check_gather_scatter): Consider using scatter stores if
11075 supports_vec_scatter_store_p.
11076 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11077 scatter stores too.
11078 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11079 internal_fn_stored_value_index.
11080 (check_load_store_masking): Handle scatter stores too.
11081 (vect_get_store_rhs): Make public.
11082 (vectorizable_call): Use internal_store_fn_p.
11083 (vectorizable_store): Handle scatter store internal functions.
11084 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11085 when deciding whether the end of the group has been reached.
11086 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11087 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11088 (mask_scatter_store<mode>): New insns.
11090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11091 Alan Hayward <alan.hayward@arm.com>
11092 David Sherwood <david.sherwood@arm.com>
11094 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11095 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11096 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11098 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11099 Use vect_truncate_gather_scatter_offset if we can't treat the
11100 operation as a normal gather load or scatter store.
11101 (get_group_load_store_type): Take the gather_scatter_info
11102 as argument. Try using a gather load or scatter store for
11103 single-element groups.
11104 (get_load_store_type): Update calls to get_group_load_store_type
11105 and vect_use_strided_gather_scatters_p.
11107 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11108 Alan Hayward <alan.hayward@arm.com>
11109 David Sherwood <david.sherwood@arm.com>
11111 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11112 optional tree argument.
11113 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11115 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11116 but continue to use the current value as a fallback.
11117 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11118 to compare the updates.
11119 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11120 (get_load_store_type): Use it when handling a strided access.
11121 (vect_get_strided_load_store_ops): New function.
11122 (vect_get_data_ptr_increment): Likewise.
11123 (vectorizable_load): Handle strided gather loads. Always pass
11124 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11126 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11127 Alan Hayward <alan.hayward@arm.com>
11128 David Sherwood <david.sherwood@arm.com>
11130 * doc/md.texi (gather_load@var{m}): Document.
11131 (mask_gather_load@var{m}): Likewise.
11132 * genopinit.c (main): Add supports_vec_gather_load and
11133 supports_vec_gather_load_cached to target_optabs.
11134 * optabs-tree.c (init_tree_optimization_optabs): Use
11135 ggc_cleared_alloc to allocate target_optabs.
11136 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11137 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11139 * internal-fn.h (internal_load_fn_p): Declare.
11140 (internal_gather_scatter_fn_p): Likewise.
11141 (internal_fn_mask_index): Likewise.
11142 (internal_gather_scatter_fn_supported_p): Likewise.
11143 * internal-fn.c (gather_load_direct): New macro.
11144 (expand_gather_load_optab_fn): New function.
11145 (direct_gather_load_optab_supported_p): New macro.
11146 (direct_internal_fn_optab): New function.
11147 (internal_load_fn_p): Likewise.
11148 (internal_gather_scatter_fn_p): Likewise.
11149 (internal_fn_mask_index): Likewise.
11150 (internal_gather_scatter_fn_supported_p): Likewise.
11151 * optabs-query.c (supports_at_least_one_mode_p): New function.
11152 (supports_vec_gather_load_p): Likewise.
11153 * optabs-query.h (supports_vec_gather_load_p): Declare.
11154 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11155 and memory_type field.
11156 (NUM_PATTERNS): Bump to 15.
11157 * tree-vect-data-refs.c: Include internal-fn.h.
11158 (vect_gather_scatter_fn_p): New function.
11159 (vect_describe_gather_scatter_call): Likewise.
11160 (vect_check_gather_scatter): Try using internal functions for
11161 gather loads. Recognize existing calls to a gather load function.
11162 (vect_analyze_data_refs): Consider using gather loads if
11163 supports_vec_gather_load_p.
11164 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11165 (vect_get_gather_scatter_offset_type): Likewise.
11166 (vect_convert_mask_for_vectype): Likewise.
11167 (vect_add_conversion_to_patterm): Likewise.
11168 (vect_try_gather_scatter_pattern): Likewise.
11169 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11170 (vect_vect_recog_func_ptrs): Add it.
11171 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11172 internal_fn_mask_index and internal_gather_scatter_fn_p.
11173 (check_load_store_masking): Take the gather_scatter_info as an
11174 argument and handle gather loads.
11175 (vect_get_gather_scatter_ops): New function.
11176 (vectorizable_call): Check internal_load_fn_p.
11177 (vectorizable_load): Likewise. Handle gather load internal
11179 (vectorizable_store): Update call to check_load_store_masking.
11180 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11181 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11182 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11183 (aarch64_gather_scale_operand_d): New predicates.
11184 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11185 (mask_gather_load<mode>): New insns.
11187 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11188 Alan Hayward <alan.hayward@arm.com>
11189 David Sherwood <david.sherwood@arm.com>
11191 * optabs.def (fold_left_plus_optab): New optab.
11192 * doc/md.texi (fold_left_plus_@var{m}): Document.
11193 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11194 * internal-fn.c (fold_left_direct): Define.
11195 (expand_fold_left_optab_fn): Likewise.
11196 (direct_fold_left_optab_supported_p): Likewise.
11197 * fold-const-call.c (fold_const_fold_left): New function.
11198 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11199 * tree-parloops.c (valid_reduction_p): New function.
11200 (gather_scalar_reductions): Use it.
11201 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11202 (vect_finish_replace_stmt): Declare.
11203 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11204 (needs_fold_left_reduction_p): New function, split out from...
11205 (vect_is_simple_reduction): ...here. Accept reductions that
11206 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11207 (vect_force_simple_reduction): Also store the reduction type in
11208 the assignment's STMT_VINFO_REDUC_TYPE.
11209 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11210 (merge_with_identity): New function.
11211 (vect_expand_fold_left): Likewise.
11212 (vectorize_fold_left_reduction): Likewise.
11213 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11214 scalar phi in place for it. Check for target support and reject
11215 cases that would reassociate the operation. Defer the transform
11216 phase to vectorize_fold_left_reduction.
11217 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11218 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11219 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11223 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11224 call to ifc_temp_var.
11226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11227 Alan Hayward <alan.hayward@arm.com>
11228 David Sherwood <david.sherwood@arm.com>
11230 * target.def (legitimize_address_displacement): Take the original
11231 offset as a poly_int.
11232 * targhooks.h (default_legitimize_address_displacement): Update
11234 * targhooks.c (default_legitimize_address_displacement): Likewise.
11235 * doc/tm.texi: Regenerate.
11236 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11237 as an argument, moving assert of ad->disp == ad->disp_term to...
11238 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11239 Try calling targetm.legitimize_address_displacement before expanding
11240 the address rather than afterwards, and adjust for the new interface.
11241 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11242 Match the new hook interface. Handle SVE addresses.
11243 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11244 new hook interface.
11246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11248 * Makefile.in (OBJS): Add early-remat.o.
11249 * target.def (select_early_remat_modes): New hook.
11250 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11251 * doc/tm.texi: Regenerate.
11252 * targhooks.h (default_select_early_remat_modes): Declare.
11253 * targhooks.c (default_select_early_remat_modes): New function.
11254 * timevar.def (TV_EARLY_REMAT): New timevar.
11255 * passes.def (pass_early_remat): New pass.
11256 * tree-pass.h (make_pass_early_remat): Declare.
11257 * early-remat.c: New file.
11258 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11260 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11262 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11263 Alan Hayward <alan.hayward@arm.com>
11264 David Sherwood <david.sherwood@arm.com>
11266 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11267 vfm1 with a bound_epilog parameter.
11268 (vect_do_peeling): Update calls accordingly, and move the prologue
11269 call earlier in the function. Treat the base bound_epilog as 0 for
11270 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11271 this base when peeling for gaps.
11272 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11273 with fully-masked loops.
11274 (vect_estimate_min_profitable_iters): Handle the single peeled
11275 iteration in that case.
11277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11278 Alan Hayward <alan.hayward@arm.com>
11279 David Sherwood <david.sherwood@arm.com>
11281 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11282 single-element interleaving even if the size is not a power of 2.
11283 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11284 accesses for single-element interleaving if the group size is
11287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11288 Alan Hayward <alan.hayward@arm.com>
11289 David Sherwood <david.sherwood@arm.com>
11291 * doc/md.texi (fold_extract_last_@var{m}): Document.
11292 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11293 * optabs.def (fold_extract_last_optab): New optab.
11294 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11295 * internal-fn.c (fold_extract_direct): New macro.
11296 (expand_fold_extract_optab_fn): Likewise.
11297 (direct_fold_extract_optab_supported_p): Likewise.
11298 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11299 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11300 EXTRACT_LAST_REDUCTION.
11301 (get_initial_def_for_reduction): Do not create an initial vector
11302 for EXTRACT_LAST_REDUCTION reductions.
11303 (vectorizable_reduction): Leave the scalar phi in place for
11304 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11305 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11306 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11307 transform phase to vectorizable_condition.
11308 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11310 (vect_finish_stmt_generation): ...here.
11311 (vect_finish_replace_stmt): New function.
11312 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11313 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11315 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11317 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * doc/md.texi (extract_last_@var{m}): Document.
11322 * optabs.def (extract_last_optab): New optab.
11323 * internal-fn.def (EXTRACT_LAST): New internal function.
11324 * internal-fn.c (cond_unary_direct): New macro.
11325 (expand_cond_unary_optab_fn): Likewise.
11326 (direct_cond_unary_optab_supported_p): Likewise.
11327 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11328 loops using EXTRACT_LAST.
11329 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11330 (extract_last_<mode>): ...this optab.
11331 (vec_extract<mode><Vel>): Update accordingly.
11333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11334 Alan Hayward <alan.hayward@arm.com>
11335 David Sherwood <david.sherwood@arm.com>
11337 * target.def (empty_mask_is_expensive): New hook.
11338 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11339 * doc/tm.texi: Regenerate.
11340 * targhooks.h (default_empty_mask_is_expensive): Declare.
11341 * targhooks.c (default_empty_mask_is_expensive): New function.
11342 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11343 if the target says that empty masks are expensive.
11344 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11346 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11349 Alan Hayward <alan.hayward@arm.com>
11350 David Sherwood <david.sherwood@arm.com>
11352 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11353 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11354 (vect_use_loop_mask_for_alignment_p): New function.
11355 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11356 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11357 niters_skip argument. Make sure that the first niters_skip elements
11358 of the first iteration are inactive.
11359 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11360 Update call to vect_set_loop_masks_directly.
11361 (get_misalign_in_elems): New function, split out from...
11362 (vect_gen_prolog_loop_niters): ...here.
11363 (vect_update_init_of_dr): Take a code argument that specifies whether
11364 the adjustment should be added or subtracted.
11365 (vect_update_init_of_drs): Likewise.
11366 (vect_prepare_for_masked_peels): New function.
11367 (vect_do_peeling): Skip prologue peeling if we're using a mask
11368 instead. Update call to vect_update_inits_of_drs.
11369 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11371 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11372 alignment. Do not include the number of peeled iterations in
11373 the minimum threshold in that case.
11374 (vectorizable_induction): Adjust the start value down by
11375 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11376 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11377 Take the number of skipped iterations into account when calculating
11379 * tree-vect-stmts.c (vect_gen_while_not): New function.
11381 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11382 Alan Hayward <alan.hayward@arm.com>
11383 David Sherwood <david.sherwood@arm.com>
11385 * doc/sourcebuild.texi (vect_fully_masked): Document.
11386 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11387 default value to 0.
11388 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11390 (vect_analyze_loop_2): ...here. Don't check the vectorization
11391 factor against the number of loop iterations if the loop is
11394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11395 Alan Hayward <alan.hayward@arm.com>
11396 David Sherwood <david.sherwood@arm.com>
11398 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11399 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11400 (dump_groups): Update accordingly.
11401 (iv_use::mem_type): New member variable.
11402 (address_p): New function.
11403 (record_use): Add a mem_type argument and initialize the new
11405 (record_group_use): Add a mem_type argument. Use address_p.
11406 Remove obsolete null checks of base_object. Update call to record_use.
11407 (find_interesting_uses_op): Update call to record_group_use.
11408 (find_interesting_uses_cond): Likewise.
11409 (find_interesting_uses_address): Likewise.
11410 (get_mem_type_for_internal_fn): New function.
11411 (find_address_like_use): Likewise.
11412 (find_interesting_uses_stmt): Try find_address_like_use before
11413 calling find_interesting_uses_op.
11414 (addr_offset_valid_p): Use the iv mem_type field as the type
11415 of the addressed memory.
11416 (add_autoinc_candidates): Likewise.
11417 (get_address_cost): Likewise.
11418 (split_small_address_groups_p): Use address_p.
11419 (split_address_groups): Likewise.
11420 (add_iv_candidate_for_use): Likewise.
11421 (autoinc_possible_for_pair): Likewise.
11422 (rewrite_groups): Likewise.
11423 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11424 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11425 (get_alias_ptr_type_for_ptr_address): New function.
11426 (rewrite_use_address): Rewrite address uses in calls that were
11427 identified by find_address_like_use.
11429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11430 Alan Hayward <alan.hayward@arm.com>
11431 David Sherwood <david.sherwood@arm.com>
11433 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11435 * gimple-expr.h (is_gimple_addressable: Likewise.
11436 * gimple-expr.c (is_gimple_address): Likewise.
11437 * internal-fn.c (expand_call_mem_ref): New function.
11438 (expand_mask_load_optab_fn): Use it.
11439 (expand_mask_store_optab_fn): Likewise.
11441 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11442 Alan Hayward <alan.hayward@arm.com>
11443 David Sherwood <david.sherwood@arm.com>
11445 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11446 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11447 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11448 (cond_umax@var{mode}): Document.
11449 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11450 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11451 (cond_umin_optab, cond_umax_optab): New optabs.
11452 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11453 (COND_IOR, COND_XOR): New internal functions.
11454 * internal-fn.h (get_conditional_internal_fn): Declare.
11455 * internal-fn.c (cond_binary_direct): New macro.
11456 (expand_cond_binary_optab_fn): Likewise.
11457 (direct_cond_binary_optab_supported_p): Likewise.
11458 (get_conditional_internal_fn): New function.
11459 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11460 Cope with reduction statements that are vectorized as calls rather
11462 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11463 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11464 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11465 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11466 (UNSPEC_COND_EOR): New unspecs.
11467 (optab): Add mappings for them.
11468 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11469 (sve_int_op, sve_fp_op): New int attributes.
11471 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11472 Alan Hayward <alan.hayward@arm.com>
11473 David Sherwood <david.sherwood@arm.com>
11475 * optabs.def (while_ult_optab): New optab.
11476 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11477 * internal-fn.def (WHILE_ULT): New internal function.
11478 * internal-fn.h (direct_internal_fn_supported_p): New override
11479 that takes two types as argument.
11480 * internal-fn.c (while_direct): New macro.
11481 (expand_while_optab_fn): New function.
11482 (convert_optab_supported_p): Likewise.
11483 (direct_while_optab_supported_p): New macro.
11484 * wide-int.h (wi::udiv_ceil): New function.
11485 * tree-vectorizer.h (rgroup_masks): New structure.
11486 (vec_loop_masks): New typedef.
11487 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11488 and fully_masked_p.
11489 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11490 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11491 (vect_max_vf): New function.
11492 (slpeel_make_loop_iterate_ntimes): Delete.
11493 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11494 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11495 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11496 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11497 internal-fn.h, stor-layout.h and optabs-query.h.
11498 (vect_set_loop_mask): New function.
11499 (add_preheader_seq): Likewise.
11500 (add_header_seq): Likewise.
11501 (interleave_supported_p): Likewise.
11502 (vect_maybe_permute_loop_masks): Likewise.
11503 (vect_set_loop_masks_directly): Likewise.
11504 (vect_set_loop_condition_masked): Likewise.
11505 (vect_set_loop_condition_unmasked): New function, split out from
11506 slpeel_make_loop_iterate_ntimes.
11507 (slpeel_make_loop_iterate_ntimes): Rename to..
11508 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11509 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11510 (vect_do_peeling): Update call accordingly.
11511 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11513 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11514 mask_compare_type, can_fully_mask_p and fully_masked_p.
11515 (release_vec_loop_masks): New function.
11516 (_loop_vec_info): Use it to free the loop masks.
11517 (can_produce_all_loop_masks_p): New function.
11518 (vect_get_max_nscalars_per_iter): Likewise.
11519 (vect_verify_full_masking): Likewise.
11520 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11521 retries, and free the mask rgroups before retrying. Check loop-wide
11522 reasons for disallowing fully-masked loops. Make the final decision
11523 about whether use a fully-masked loop or not.
11524 (vect_estimate_min_profitable_iters): Do not assume that peeling
11525 for the number of iterations will be needed for fully-masked loops.
11526 (vectorizable_reduction): Disable fully-masked loops.
11527 (vectorizable_live_operation): Likewise.
11528 (vect_halve_mask_nunits): New function.
11529 (vect_double_mask_nunits): Likewise.
11530 (vect_record_loop_mask): Likewise.
11531 (vect_get_loop_mask): Likewise.
11532 (vect_transform_loop): Handle the case in which the final loop
11533 iteration might handle a partial vector. Call vect_set_loop_condition
11534 instead of slpeel_make_loop_iterate_ntimes.
11535 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11536 (check_load_store_masking): New function.
11537 (prepare_load_store_mask): Likewise.
11538 (vectorizable_store): Handle fully-masked loops.
11539 (vectorizable_load): Likewise.
11540 (supportable_widening_operation): Use vect_halve_mask_nunits for
11542 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11543 (vect_gen_while): New function.
11544 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11545 (aarch64_uqdec<mode>): New insn.
11547 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11548 Alan Hayward <alan.hayward@arm.com>
11549 David Sherwood <david.sherwood@arm.com>
11551 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11552 (reduc_xor_scal_optab): New optabs.
11553 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11554 (reduc_xor_scal_@var{m}): Document.
11555 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11556 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11557 internal functions.
11558 * fold-const-call.c (fold_const_call): Handle them.
11559 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11560 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11561 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11562 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11563 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11564 (UNSPEC_XORV): New unspecs.
11565 (optab): Add entries for them.
11566 (BITWISEV): New int iterator.
11567 (bit_reduc_op): New int attributes.
11569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11570 Alan Hayward <alan.hayward@arm.com>
11571 David Sherwood <david.sherwood@arm.com>
11573 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11574 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11575 * optabs.def (vec_shl_insert_optab): New optab.
11576 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11577 (duplicate_and_interleave): Likewise.
11578 * tree-vect-loop.c: Include internal-fn.h.
11579 (neutral_op_for_slp_reduction): New function, split out from
11580 get_initial_defs_for_reduction.
11581 (get_initial_def_for_reduction): Handle option 2 for variable-length
11582 vectors by loading the neutral value into a vector and then shifting
11583 the initial value into element 0.
11584 (get_initial_defs_for_reduction): Replace the code argument with
11585 the neutral value calculated by neutral_op_for_slp_reduction.
11586 Use gimple_build_vector for constant-length vectors.
11587 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11588 but the first group_size elements have a neutral value.
11589 Use duplicate_and_interleave otherwise.
11590 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11591 Update call to get_initial_defs_for_reduction. Handle SLP
11592 reductions for variable-length vectors by creating one vector
11593 result for each scalar result, with the elements associated
11594 with other scalar results stubbed out with the neutral value.
11595 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11596 Require IFN_VEC_SHL_INSERT for double reductions on
11597 variable-length vectors, or SLP reductions that have
11598 a neutral value. Require can_duplicate_and_interleave_p
11599 support for variable-length unchained SLP reductions if there
11600 is no neutral value, such as for MIN/MAX reductions. Also require
11601 the number of vector elements to be a multiple of the number of
11602 SLP statements when doing variable-length unchained SLP reductions.
11603 Update call to vect_create_epilog_for_reduction.
11604 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11605 and remove initial values.
11606 (duplicate_and_interleave): Make public.
11607 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11608 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11611 Alan Hayward <alan.hayward@arm.com>
11612 David Sherwood <david.sherwood@arm.com>
11614 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11615 (can_duplicate_and_interleave_p): New function.
11616 (vect_get_and_check_slp_defs): Take the vector of statements
11617 rather than just the current one. Remove excess parentheses.
11618 Restriction rejectinon of vect_constant_def and vect_external_def
11619 for variable-length vectors to boolean types, or types for which
11620 can_duplicate_and_interleave_p is false.
11621 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11622 (duplicate_and_interleave): New function.
11623 (vect_get_constant_vectors): Use gimple_build_vector for
11624 constant-length vectors and suitable variable-length constant
11625 vectors. Use duplicate_and_interleave for other variable-length
11626 vectors. Don't defer the update when inserting new statements.
11628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11629 Alan Hayward <alan.hayward@arm.com>
11630 David Sherwood <david.sherwood@arm.com>
11632 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11633 min_profitable_iters doesn't go negative.
11635 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11636 Alan Hayward <alan.hayward@arm.com>
11637 David Sherwood <david.sherwood@arm.com>
11639 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11640 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11641 * optabs.def (vec_mask_load_lanes_optab): New optab.
11642 (vec_mask_store_lanes_optab): Likewise.
11643 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11644 (MASK_STORE_LANES): Likewise.
11645 * internal-fn.c (mask_load_lanes_direct): New macro.
11646 (mask_store_lanes_direct): Likewise.
11647 (expand_mask_load_optab_fn): Handle masked operations.
11648 (expand_mask_load_lanes_optab_fn): New macro.
11649 (expand_mask_store_optab_fn): Handle masked operations.
11650 (expand_mask_store_lanes_optab_fn): New macro.
11651 (direct_mask_load_lanes_optab_supported_p): Likewise.
11652 (direct_mask_store_lanes_optab_supported_p): Likewise.
11653 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11655 (vect_load_lanes_supported): Likewise.
11656 * tree-vect-data-refs.c (strip_conversion): New function.
11657 (can_group_stmts_p): Likewise.
11658 (vect_analyze_data_ref_accesses): Use it instead of checking
11659 for a pair of assignments.
11660 (vect_store_lanes_supported): Take a masked_p parameter.
11661 (vect_load_lanes_supported): Likewise.
11662 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11663 vect_store_lanes_supported and vect_load_lanes_supported.
11664 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11665 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11666 parameter. Don't allow gaps for masked accesses.
11667 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11668 and vect_load_lanes_supported.
11669 (get_load_store_type): Take a masked_p parameter and update
11670 call to get_group_load_store_type.
11671 (vectorizable_store): Update call to get_load_store_type.
11672 Handle IFN_MASK_STORE_LANES.
11673 (vectorizable_load): Update call to get_load_store_type.
11674 Handle IFN_MASK_LOAD_LANES.
11676 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11677 Alan Hayward <alan.hayward@arm.com>
11678 David Sherwood <david.sherwood@arm.com>
11680 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11682 * config/aarch64/aarch64-protos.h
11683 (aarch64_sve_struct_memory_operand_p): Declare.
11684 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11685 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11686 (VPRED, vpred): Handle SVE structure modes.
11687 * config/aarch64/constraints.md (Utx): New constraint.
11688 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11689 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11690 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11691 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11692 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11693 structure modes. Split into pieces after RA.
11694 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11695 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11697 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11698 SVE structure modes.
11699 (aarch64_classify_address): Likewise.
11700 (sizetochar): Move earlier in file.
11701 (aarch64_print_operand): Handle SVE register lists.
11702 (aarch64_array_mode): New function.
11703 (aarch64_sve_struct_memory_operand_p): Likewise.
11704 (TARGET_ARRAY_MODE): Redefine.
11706 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11707 Alan Hayward <alan.hayward@arm.com>
11708 David Sherwood <david.sherwood@arm.com>
11710 * target.def (array_mode): New target hook.
11711 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11712 * doc/tm.texi: Regenerate.
11713 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11714 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11715 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11716 targetm.array_mode.
11717 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11721 Alan Hayward <alan.hayward@arm.com>
11722 David Sherwood <david.sherwood@arm.com>
11724 * fold-const.c (fold_binary_loc): Check the argument types
11725 rather than the result type when testing for a vector operation.
11727 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11729 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11730 * doc/tm.texi: Regenerate.
11732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11733 Alan Hayward <alan.hayward@arm.com>
11734 David Sherwood <david.sherwood@arm.com>
11736 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11737 (sve): Document new AArch64 extension.
11738 * doc/md.texi (w): Extend the description of the AArch64
11739 constraint to include SVE vectors.
11740 (Upl, Upa): Document new AArch64 predicate constraints.
11741 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11743 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11744 (msve-vector-bits=): New option.
11745 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11746 SVE when these are disabled.
11747 (sve): New extension.
11748 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11749 modes. Adjust their number of units based on aarch64_sve_vg.
11750 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11751 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11752 aarch64_addr_query_type.
11753 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11754 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11755 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11756 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11757 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11758 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11759 (aarch64_simd_imm_zero_p): Delete.
11760 (aarch64_check_zero_based_sve_index_immediate): Declare.
11761 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11762 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11763 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11764 (aarch64_sve_float_mul_immediate_p): Likewise.
11765 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11766 rather than an rtx.
11767 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11768 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11769 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11770 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11771 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11772 (aarch64_regmode_natural_size): Likewise.
11773 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11774 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11776 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11777 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11778 for VG and the SVE predicate registers.
11779 (V_ALIASES): Add a "z"-prefixed alias.
11780 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11781 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11782 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11783 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11784 (REG_CLASS_NAMES): Add entries for them.
11785 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11786 and the predicate registers.
11787 (aarch64_sve_vg): Declare.
11788 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11789 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11790 (REGMODE_NATURAL_SIZE): Define.
11791 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11793 * config/aarch64/aarch64.c: Include cfgrtl.h.
11794 (simd_immediate_info): Add a constructor for series vectors,
11795 and an associated step field.
11796 (aarch64_sve_vg): New variable.
11797 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11798 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11799 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11800 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11801 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11802 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11803 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11804 (aarch64_get_mask_mode): New functions.
11805 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11806 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11807 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11808 predicate modes and predicate registers. Explicitly restrict
11809 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11810 to store a vector mode if it is recognized by
11811 aarch64_classify_vector_mode.
11812 (aarch64_regmode_natural_size): New function.
11813 (aarch64_hard_regno_caller_save_mode): Return the original mode
11815 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11816 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11817 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11818 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11820 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11821 does not overlap dest if the function is frame-related. Handle
11823 (aarch64_split_add_offset): New function.
11824 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11825 them aarch64_add_offset.
11826 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11827 and update call to aarch64_sub_sp.
11828 (aarch64_add_cfa_expression): New function.
11829 (aarch64_expand_prologue): Pass extra temporary registers to the
11830 functions above. Handle the case in which we need to emit new
11831 DW_CFA_expressions for registers that were originally saved
11832 relative to the stack pointer, but now have to be expressed
11833 relative to the frame pointer.
11834 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11836 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11837 IP0 and IP1 values for SVE frames.
11838 (aarch64_expand_vec_series): New function.
11839 (aarch64_expand_sve_widened_duplicate): Likewise.
11840 (aarch64_expand_sve_const_vector): Likewise.
11841 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11842 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11843 into the register, rather than emitting a SET directly.
11844 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11845 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11846 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11847 (offset_9bit_signed_scaled_p): New functions.
11848 (aarch64_replicate_bitmask_imm): New function.
11849 (aarch64_bitmask_imm): Use it.
11850 (aarch64_cannot_force_const_mem): Reject expressions involving
11851 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11852 (aarch64_classify_index): Handle SVE indices, by requiring
11853 a plain register index with a scale that matches the element size.
11854 (aarch64_classify_address): Handle SVE addresses. Assert that
11855 the mode of the address is VOIDmode or an integer mode.
11856 Update call to aarch64_classify_symbol.
11857 (aarch64_classify_symbolic_expression): Update call to
11858 aarch64_classify_symbol.
11859 (aarch64_const_vec_all_in_range_p): New function.
11860 (aarch64_print_vector_float_operand): Likewise.
11861 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11862 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11863 and the FP immediates 1.0 and 0.5.
11864 (aarch64_print_address_internal): Handle SVE addresses.
11865 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11866 (aarch64_regno_regclass): Handle predicate registers.
11867 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11869 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11870 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11871 (aarch64_convert_sve_vector_bits): New function.
11872 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11873 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11874 rather than an rtx.
11875 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11876 Handle SVE vector and predicate modes. Accept VL-based constants
11877 that need only one temporary register, and VL offsets that require
11878 no temporary registers.
11879 (aarch64_conditional_register_usage): Mark the predicate registers
11880 as fixed if SVE isn't available.
11881 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11882 Return true for SVE vector and predicate modes.
11883 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11884 rather than an unsigned int. Handle SVE modes.
11885 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11887 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11889 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11890 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11891 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11892 (aarch64_sve_float_mul_immediate_p): New functions.
11893 (aarch64_sve_valid_immediate): New function.
11894 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11895 Explicitly reject structure modes. Check for INDEX constants.
11896 Handle PTRUE and PFALSE constants.
11897 (aarch64_check_zero_based_sve_index_immediate): New function.
11898 (aarch64_simd_imm_zero_p): Delete.
11899 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11900 vector modes. Accept constants in the range of CNT[BHWD].
11901 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11902 ask for an Advanced SIMD mode.
11903 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11904 (aarch64_simd_vector_alignment): Handle SVE predicates.
11905 (aarch64_vectorize_preferred_vector_alignment): New function.
11906 (aarch64_simd_vector_alignment_reachable): Use it instead of
11908 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11909 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11911 (MAX_VECT_LEN): Delete.
11912 (expand_vec_perm_d): Add a vec_flags field.
11913 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11914 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11915 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11917 (aarch64_evpc_rev): Rename to...
11918 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11919 (aarch64_evpc_rev_global): New function.
11920 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11921 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11923 (aarch64_evpc_sve_tbl): New function.
11924 (aarch64_expand_vec_perm_const_1): Update after rename of
11925 aarch64_evpc_rev. Handle SVE permutes too, trying
11926 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11927 than aarch64_evpc_tbl.
11928 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11929 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11930 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11931 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11932 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11933 (aarch64_expand_sve_vcond): New functions.
11934 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11935 of aarch64_vector_mode_p.
11936 (aarch64_dwarf_poly_indeterminate_value): New function.
11937 (aarch64_compute_pressure_classes): Likewise.
11938 (aarch64_can_change_mode_class): Likewise.
11939 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11940 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11941 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11942 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11943 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11944 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11945 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11946 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11948 (Dn, Dl, Dr): Accept const as well as const_vector.
11949 (Dz): Likewise. Compare against CONST0_RTX.
11950 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11951 of "vector" where appropriate.
11952 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11953 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11954 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11955 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11956 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11957 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11958 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11959 (v_int_equiv): Extend to SVE modes.
11960 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11962 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11963 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11964 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11965 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11966 (SVE_COND_FP_CMP): New int iterators.
11967 (perm_hilo): Handle the new unpack unspecs.
11968 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11970 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11971 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11972 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11973 (aarch64_equality_operator, aarch64_constant_vector_operand)
11974 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11975 (aarch64_sve_nonimmediate_operand): Likewise.
11976 (aarch64_sve_general_operand): Likewise.
11977 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11978 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11979 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11980 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11981 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11982 (aarch64_sve_float_arith_immediate): Likewise.
11983 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11984 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11985 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11986 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11987 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11988 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11989 (aarch64_sve_float_arith_operand): Likewise.
11990 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11991 (aarch64_sve_float_mul_operand): Likewise.
11992 (aarch64_sve_vec_perm_operand): Likewise.
11993 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11994 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11995 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11996 as well as const_vector.
11997 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11998 in file. Use CONST0_RTX and CONSTM1_RTX.
11999 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12000 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12001 Use aarch64_simd_imm_zero.
12002 * config/aarch64/aarch64-sve.md: New file.
12003 * config/aarch64/aarch64.md: Include it.
12004 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12005 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12006 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12007 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12008 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12009 (sve): New attribute.
12010 (enabled): Disable instructions with the sve attribute unless
12012 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12013 aarch64_expand_mov_immediate.
12014 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12015 CNT[BHSD] immediates.
12016 (movti): Split CONST_POLY_INT moves into two halves.
12017 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12018 Split additions that need a temporary here if the destination
12019 is the stack pointer.
12020 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12021 (*add<mode>3_poly_1): New instruction.
12022 (set_clobber_cc): New expander.
12024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12026 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12027 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12028 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12029 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12030 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12031 Change innermode from fixed_mode_size to machine_mode.
12032 (simplify_subreg): Update call accordingly. Handle a constant-sized
12033 subreg of a variable-length CONST_VECTOR.
12035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12036 Alan Hayward <alan.hayward@arm.com>
12037 David Sherwood <david.sherwood@arm.com>
12039 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12040 (add_offset_to_base): New function, split out from...
12041 (create_mem_ref): ...here. When handling a scale other than 1,
12042 check first whether the address is valid without the offset.
12043 Add it into the base if so, leaving the index and scale as-is.
12045 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12048 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12049 fold_for_warn before checking if arg2 is INTEGER_CST.
12051 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12053 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12054 (store_multiple_operation): Delete.
12055 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12056 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12057 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12058 guarded by TARGET_STRING.
12059 (rs6000_output_load_multiple): Delete.
12060 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12061 OPTION_MASK_STRING / TARGET_STRING handling.
12062 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12063 (const rs6000_opt_masks) <"string">: Change mask to 0.
12064 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12065 (MASK_STRING): Delete.
12066 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12068 (load_multiple): Delete.
12075 (store_multiple): Delete.
12082 (movmemsi_8reg): Delete.
12083 (corresponding unnamed define_insn): Delete.
12084 (movmemsi_6reg): Delete.
12085 (corresponding unnamed define_insn): Delete.
12086 (movmemsi_4reg): Delete.
12087 (corresponding unnamed define_insn): Delete.
12088 (movmemsi_2reg): Delete.
12089 (corresponding unnamed define_insn): Delete.
12090 (movmemsi_1reg): Delete.
12091 (corresponding unnamed define_insn): Delete.
12092 * config/rs6000/rs6000.opt (mno-string): New.
12093 (mstring): Replace by deprecation warning stub.
12094 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12096 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12098 * regrename.c (regrename_do_replace): If replacing the same
12099 reg multiple times, try to reuse last created gen_raw_REG.
12102 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12103 main to workaround a bug in GDB.
12105 2018-01-12 Tom de Vries <tom@codesourcery.com>
12108 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12110 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12112 PR rtl-optimization/80481
12113 * ira-color.c (get_cap_member): New function.
12114 (allocnos_conflict_by_live_ranges_p): Use it.
12115 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12116 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12118 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12121 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12122 (*saddl_se_1): Ditto.
12123 (*ssubsi_1): Ditto.
12124 (*ssubl_se_1): Ditto.
12126 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12128 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12129 rather than wi::to_widest for DR_INITs.
12130 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12131 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12132 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12134 (vect_analyze_group_access_1): Note that here.
12136 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12138 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12139 polynomial type sizes.
12141 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12143 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12144 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12145 (gimple_add_tmp_var): Likewise.
12147 2018-01-12 Martin Liska <mliska@suse.cz>
12149 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12150 (gimple_alloc_sizes): Likewise.
12151 (dump_gimple_statistics): Use PRIu64 in printf format.
12152 * gimple.h: Change uint64_t to int.
12154 2018-01-12 Martin Liska <mliska@suse.cz>
12156 * tree-core.h: Use uint64_t instead of int.
12157 * tree.c (tree_node_counts): Likewise.
12158 (tree_node_sizes): Likewise.
12159 (dump_tree_statistics): Use PRIu64 in printf format.
12161 2018-01-12 Martin Liska <mliska@suse.cz>
12163 * Makefile.in: As qsort_chk is implemented in vec.c, add
12164 vec.o to linkage of gencfn-macros.
12165 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12166 passing the info to record_node_allocation_statistics.
12167 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12169 * ggc-common.c (struct ggc_usage): Add operator== and use
12170 it in operator< and compare function.
12171 * mem-stats.h (struct mem_usage): Likewise.
12172 * vec.c (struct vec_usage): Remove operator< and compare
12173 function. Can be simply inherited.
12175 2018-01-12 Martin Jambor <mjambor@suse.cz>
12178 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12179 * tree-ssa-math-opts.c: Include domwalk.h.
12180 (convert_mult_to_fma_1): New function.
12181 (fma_transformation_info): New type.
12182 (fma_deferring_state): Likewise.
12183 (cancel_fma_deferring): New function.
12184 (result_of_phi): Likewise.
12185 (last_fma_candidate_feeds_initial_phi): Likewise.
12186 (convert_mult_to_fma): Added deferring logic, split actual
12187 transformation to convert_mult_to_fma_1.
12188 (math_opts_dom_walker): New type.
12189 (math_opts_dom_walker::after_dom_children): New method, body moved
12190 here from pass_optimize_widening_mul::execute, added deferring logic
12192 (pass_optimize_widening_mul::execute): Moved most of code to
12193 math_opts_dom_walker::after_dom_children.
12194 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12195 * config/i386/i386.c (ix86_option_override_internal): Added
12196 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12198 2018-01-12 Richard Biener <rguenther@suse.de>
12201 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12202 inline instance vars.
12204 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12207 * config/rx/rx.c (rx_is_restricted_memory_address):
12208 Handle SUBREG case.
12210 2018-01-12 Richard Biener <rguenther@suse.de>
12212 PR tree-optimization/80846
12213 * target.def (split_reduction): New target hook.
12214 * targhooks.c (default_split_reduction): New function.
12215 * targhooks.h (default_split_reduction): Declare.
12216 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12217 target requests first reduce vectors by combining low and high
12219 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12220 (get_vectype_for_scalar_type_and_size): Export.
12221 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12222 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12223 * doc/tm.texi: Regenerate.
12224 * config/i386/i386.c (ix86_split_reduction): Implement
12225 TARGET_VECTORIZE_SPLIT_REDUCTION.
12227 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12230 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12231 in PIC mode except for TARGET_VXWORKS_RTP.
12232 * config/sparc/sparc.c: Include cfgrtl.h.
12233 (TARGET_INIT_PIC_REG): Define.
12234 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12235 (sparc_pic_register_p): New predicate.
12236 (sparc_legitimate_address_p): Use it.
12237 (sparc_legitimize_pic_address): Likewise.
12238 (sparc_delegitimize_address): Likewise.
12239 (sparc_mode_dependent_address_p): Likewise.
12240 (gen_load_pcrel_sym): Remove 4th parameter.
12241 (load_got_register): Adjust call to above. Remove obsolete stuff.
12242 (sparc_expand_prologue): Do not call load_got_register here.
12243 (sparc_flat_expand_prologue): Likewise.
12244 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12245 (sparc_use_pseudo_pic_reg): New function.
12246 (sparc_init_pic_reg): Likewise.
12247 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12248 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12250 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12252 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12253 Add item for branch_cost.
12255 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12257 PR rtl-optimization/83565
12258 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12259 not extend the result to a larger mode for rotate operations.
12260 (num_sign_bit_copies1): Likewise.
12262 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12265 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12267 Use values-Xc.o for -pedantic.
12268 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12270 2018-01-12 Martin Liska <mliska@suse.cz>
12273 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12275 (possible_polymorphic_call_targets): Use it.
12276 (ipa_devirt): Likewise.
12278 2018-01-12 Martin Liska <mliska@suse.cz>
12280 * profile-count.h (enum profile_quality): Use 0 as invalid
12281 enum value of profile_quality.
12283 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12285 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12286 -mext-string options.
12288 2018-01-12 Richard Biener <rguenther@suse.de>
12290 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12291 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12292 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12294 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12296 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12298 * configure.ac (--with-long-double-format): Add support for the
12299 configuration option to change the default long double format on
12301 * config.gcc (powerpc*-linux*-*): Likewise.
12302 * configure: Regenerate.
12303 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12304 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12305 used without modification.
12307 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12309 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12310 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12311 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12312 MISC_BUILTIN_SPEC_BARRIER.
12313 (rs6000_init_builtins): Likewise.
12314 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12316 (speculation_barrier): New define_insn.
12317 * doc/extend.texi: Document __builtin_speculation_barrier.
12319 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12322 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12323 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12324 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12326 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12327 integral modes instead of "ss" and "sd".
12328 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12329 vectors with 32-bit and 64-bit elements.
12330 (vecdupssescalarmodesuffix): New mode attribute.
12331 (vec_dup<mode>): Use it.
12333 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12336 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12337 frame if argument is passed on stack.
12339 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12342 * ree.c (combine_reaching_defs): Optimize also
12343 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12344 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12346 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12348 PR middle-end/83189
12349 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12351 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12353 PR middle-end/83718
12354 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12355 after they are computed.
12357 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12359 PR tree-optimization/83695
12360 * gimple-loop-linterchange.cc
12361 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12362 reset cached scev information after interchange.
12363 (pass_linterchange::execute): Remove call to scev_reset_htab.
12365 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12367 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12368 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12369 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12370 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12371 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12372 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12373 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12374 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12375 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12376 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12377 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12378 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12379 (V_lane_reg): Likewise.
12380 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12382 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12383 (vfmal_lane_low<mode>_intrinsic,
12384 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12385 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12386 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12387 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12388 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12389 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12391 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12393 * config/arm/arm-cpus.in (fp16fml): New feature.
12394 (ALL_SIMD): Add fp16fml.
12395 (armv8.2-a): Add fp16fml as an option.
12396 (armv8.3-a): Likewise.
12397 (armv8.4-a): Add fp16fml as part of fp16.
12398 * config/arm/arm.h (TARGET_FP16FML): Define.
12399 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12401 * config/arm/arm-modes.def (V2HF): Define.
12402 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12403 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12404 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12405 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12406 vfmsl_low, vfmsl_high): New set of builtins.
12407 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12408 (vfml_op): New code attribute.
12409 (VFMLHALVES): New int iterator.
12410 (VFML, VFMLSEL): New mode attributes.
12411 (V_reg): Define mapping for V2HF.
12412 (V_hi, V_lo): New mode attributes.
12413 (VF_constraint): Likewise.
12414 (vfml_half, vfml_half_selector): New int attributes.
12415 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12417 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12418 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12420 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12421 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12422 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12423 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12425 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12426 Document new effective target and option set.
12428 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12430 * config/arm/arm-cpus.in (armv8_4): New feature.
12431 (ARMv8_4a): New fgroup.
12432 (armv8.4-a): New arch.
12433 * config/arm/arm-tables.opt: Regenerate.
12434 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12435 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12436 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12437 Add matching rules for -march=armv8.4-a and extensions.
12438 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12440 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12443 * config/rx/rx.md (BW): New mode attribute.
12444 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12446 2018-01-11 Richard Biener <rguenther@suse.de>
12448 PR tree-optimization/83435
12449 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12450 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12451 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12453 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12454 Alan Hayward <alan.hayward@arm.com>
12455 David Sherwood <david.sherwood@arm.com>
12457 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12459 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12460 (aarch64_print_address_internal): Use it to check for a zero offset.
12462 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12463 Alan Hayward <alan.hayward@arm.com>
12464 David Sherwood <david.sherwood@arm.com>
12466 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12467 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12468 Return a poly_int64 rather than a HOST_WIDE_INT.
12469 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12470 rather than a HOST_WIDE_INT.
12471 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12472 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12473 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12474 final_offset from HOST_WIDE_INT to poly_int64.
12475 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12476 to_constant when getting the number of units in an Advanced SIMD
12478 (aarch64_builtin_vectorized_function): Check for a constant number
12480 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12482 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12483 attribute instead of GET_MODE_NUNITS.
12484 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12485 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12486 GET_MODE_SIZE for fixed-size registers.
12487 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12488 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12489 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12490 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12491 (aarch64_print_operand, aarch64_print_address_internal)
12492 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12493 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12494 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12495 Handle polynomial GET_MODE_SIZE.
12496 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12497 wider than SImode without modification.
12498 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12499 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12500 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12501 passing and returning SVE modes.
12502 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12503 rather than GEN_INT.
12504 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12505 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12506 (aarch64_allocate_and_probe_stack_space): Likewise.
12507 (aarch64_layout_frame): Cope with polynomial offsets.
12508 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12509 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12510 polynomial offsets.
12511 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12512 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12513 poly_int64 rather than a HOST_WIDE_INT.
12514 (aarch64_get_separate_components, aarch64_process_components)
12515 (aarch64_expand_prologue, aarch64_expand_epilogue)
12516 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12517 (aarch64_anchor_offset): New function, split out from...
12518 (aarch64_legitimize_address): ...here.
12519 (aarch64_builtin_vectorization_cost): Handle polynomial
12520 TYPE_VECTOR_SUBPARTS.
12521 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12523 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12524 number of elements from the PARALLEL rather than the mode.
12525 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12526 rather than GET_MODE_BITSIZE.
12527 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12528 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12529 (aarch64_expand_vec_perm_const_1): Handle polynomial
12530 d->perm.length () and d->perm elements.
12531 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12532 Apply to_constant to d->perm elements.
12533 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12534 polynomial CONST_VECTOR_NUNITS.
12535 (aarch64_move_pointer): Take amount as a poly_int64 rather
12537 (aarch64_progress_pointer): Avoid temporary variable.
12538 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12539 the mode attribute instead of GET_MODE.
12541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12542 Alan Hayward <alan.hayward@arm.com>
12543 David Sherwood <david.sherwood@arm.com>
12545 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12546 x exists before using it.
12547 (aarch64_add_constant_internal): Rename to...
12548 (aarch64_add_offset_1): ...this. Replace regnum with separate
12549 src and dest rtxes. Handle the case in which they're different,
12550 including when the offset is zero. Replace scratchreg with an rtx.
12551 Use 2 additions if there is no spare register into which we can
12552 move a 16-bit constant.
12553 (aarch64_add_constant): Delete.
12554 (aarch64_add_offset): Replace reg with separate src and dest
12555 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12556 Use aarch64_add_offset_1.
12557 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12558 an rtx rather than an int. Take the delta as a poly_int64
12559 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12560 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12561 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12562 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12563 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12564 and aarch64_add_sp.
12565 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12566 aarch64_add_constant.
12568 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12570 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12571 Use scalar_float_mode.
12573 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12575 * config/aarch64/aarch64-simd.md
12576 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12577 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12578 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12579 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12580 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12581 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12582 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12583 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12584 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12585 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12587 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12590 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12591 targ_options->x_arm_arch_string is non NULL.
12593 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12595 * config/aarch64/aarch64.h
12596 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12598 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12601 * expmed.c (emit_store_flag_force): Swap if const op0
12602 and change VOIDmode to mode of op0.
12604 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12606 PR rtl-optimization/83761
12607 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12608 than bytes to mode_for_size.
12610 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12612 PR middle-end/83189
12613 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12614 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12617 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12619 PR middle-end/83575
12620 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12621 when in layout mode.
12622 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12623 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12626 2018-01-10 Michael Collison <michael.collison@arm.com>
12628 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12629 * config/aarch64/aarch64-option-extension.def: Add
12630 AARCH64_OPT_EXTENSION of 'fp16fml'.
12631 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12632 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12633 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12634 * config/aarch64/constraints.md (Ui7): New constraint.
12635 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12636 (VFMLA_SEL_W): Ditto.
12639 (VFMLA16_LOW): New int iterator.
12640 (VFMLA16_HIGH): Ditto.
12641 (UNSPEC_FMLAL): New unspec.
12642 (UNSPEC_FMLSL): Ditto.
12643 (UNSPEC_FMLAL2): Ditto.
12644 (UNSPEC_FMLSL2): Ditto.
12645 (f16mac): New code attribute.
12646 * config/aarch64/aarch64-simd-builtins.def
12647 (aarch64_fmlal_lowv2sf): Ditto.
12648 (aarch64_fmlsl_lowv2sf): Ditto.
12649 (aarch64_fmlalq_lowv4sf): Ditto.
12650 (aarch64_fmlslq_lowv4sf): Ditto.
12651 (aarch64_fmlal_highv2sf): Ditto.
12652 (aarch64_fmlsl_highv2sf): Ditto.
12653 (aarch64_fmlalq_highv4sf): Ditto.
12654 (aarch64_fmlslq_highv4sf): Ditto.
12655 (aarch64_fmlal_lane_lowv2sf): Ditto.
12656 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12657 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12658 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12659 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12660 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12661 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12662 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12663 (aarch64_fmlal_lane_highv2sf): Ditto.
12664 (aarch64_fmlsl_lane_highv2sf): Ditto.
12665 (aarch64_fmlal_laneq_highv2sf): Ditto.
12666 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12667 (aarch64_fmlalq_lane_highv4sf): Ditto.
12668 (aarch64_fmlsl_lane_highv4sf): Ditto.
12669 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12670 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12671 * config/aarch64/aarch64-simd.md:
12672 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12673 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12674 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12675 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12676 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12677 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12678 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12679 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12680 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12681 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12682 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12683 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12684 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12685 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12686 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12687 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12688 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12689 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12690 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12691 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12692 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12693 (vfmlsl_low_u32): Ditto.
12694 (vfmlalq_low_u32): Ditto.
12695 (vfmlslq_low_u32): Ditto.
12696 (vfmlal_high_u32): Ditto.
12697 (vfmlsl_high_u32): Ditto.
12698 (vfmlalq_high_u32): Ditto.
12699 (vfmlslq_high_u32): Ditto.
12700 (vfmlal_lane_low_u32): Ditto.
12701 (vfmlsl_lane_low_u32): Ditto.
12702 (vfmlal_laneq_low_u32): Ditto.
12703 (vfmlsl_laneq_low_u32): Ditto.
12704 (vfmlalq_lane_low_u32): Ditto.
12705 (vfmlslq_lane_low_u32): Ditto.
12706 (vfmlalq_laneq_low_u32): Ditto.
12707 (vfmlslq_laneq_low_u32): Ditto.
12708 (vfmlal_lane_high_u32): Ditto.
12709 (vfmlsl_lane_high_u32): Ditto.
12710 (vfmlal_laneq_high_u32): Ditto.
12711 (vfmlsl_laneq_high_u32): Ditto.
12712 (vfmlalq_lane_high_u32): Ditto.
12713 (vfmlslq_lane_high_u32): Ditto.
12714 (vfmlalq_laneq_high_u32): Ditto.
12715 (vfmlslq_laneq_high_u32): Ditto.
12716 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12717 (AARCH64_FL_FOR_ARCH8_4): New.
12718 (AARCH64_ISA_F16FML): New ISA flag.
12719 (TARGET_F16FML): New feature flag for fp16fml.
12720 (doc/invoke.texi): Document new fp16fml option.
12722 2018-01-10 Michael Collison <michael.collison@arm.com>
12724 * config/aarch64/aarch64-builtins.c:
12725 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12726 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12727 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12728 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12729 (AARCH64_ISA_SHA3): New ISA flag.
12730 (TARGET_SHA3): New feature flag for sha3.
12731 * config/aarch64/iterators.md (sha512_op): New int attribute.
12732 (CRYPTO_SHA512): New int iterator.
12733 (UNSPEC_SHA512H): New unspec.
12734 (UNSPEC_SHA512H2): Ditto.
12735 (UNSPEC_SHA512SU0): Ditto.
12736 (UNSPEC_SHA512SU1): Ditto.
12737 * config/aarch64/aarch64-simd-builtins.def
12738 (aarch64_crypto_sha512hqv2di): New builtin.
12739 (aarch64_crypto_sha512h2qv2di): Ditto.
12740 (aarch64_crypto_sha512su0qv2di): Ditto.
12741 (aarch64_crypto_sha512su1qv2di): Ditto.
12742 (aarch64_eor3qv8hi): Ditto.
12743 (aarch64_rax1qv2di): Ditto.
12744 (aarch64_xarqv2di): Ditto.
12745 (aarch64_bcaxqv8hi): Ditto.
12746 * config/aarch64/aarch64-simd.md:
12747 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12748 (aarch64_crypto_sha512su0qv2di): Ditto.
12749 (aarch64_crypto_sha512su1qv2di): Ditto.
12750 (aarch64_eor3qv8hi): Ditto.
12751 (aarch64_rax1qv2di): Ditto.
12752 (aarch64_xarqv2di): Ditto.
12753 (aarch64_bcaxqv8hi): Ditto.
12754 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12755 (vsha512h2q_u64): Ditto.
12756 (vsha512su0q_u64): Ditto.
12757 (vsha512su1q_u64): Ditto.
12758 (veor3q_u16): Ditto.
12759 (vrax1q_u64): Ditto.
12760 (vxarq_u64): Ditto.
12761 (vbcaxq_u16): Ditto.
12762 * config/arm/types.md (crypto_sha512): New type attribute.
12763 (crypto_sha3): Ditto.
12764 (doc/invoke.texi): Document new sha3 option.
12766 2018-01-10 Michael Collison <michael.collison@arm.com>
12768 * config/aarch64/aarch64-builtins.c:
12769 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12770 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12771 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12772 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12773 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12774 (AARCH64_ISA_SM4): New ISA flag.
12775 (TARGET_SM4): New feature flag for sm4.
12776 * config/aarch64/aarch64-simd-builtins.def
12777 (aarch64_sm3ss1qv4si): Ditto.
12778 (aarch64_sm3tt1aq4si): Ditto.
12779 (aarch64_sm3tt1bq4si): Ditto.
12780 (aarch64_sm3tt2aq4si): Ditto.
12781 (aarch64_sm3tt2bq4si): Ditto.
12782 (aarch64_sm3partw1qv4si): Ditto.
12783 (aarch64_sm3partw2qv4si): Ditto.
12784 (aarch64_sm4eqv4si): Ditto.
12785 (aarch64_sm4ekeyqv4si): Ditto.
12786 * config/aarch64/aarch64-simd.md:
12787 (aarch64_sm3ss1qv4si): Ditto.
12788 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12789 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12790 (aarch64_sm4eqv4si): Ditto.
12791 (aarch64_sm4ekeyqv4si): Ditto.
12792 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12793 (sm3part_op): Ditto.
12794 (CRYPTO_SM3TT): Ditto.
12795 (CRYPTO_SM3PART): Ditto.
12796 (UNSPEC_SM3SS1): New unspec.
12797 (UNSPEC_SM3TT1A): Ditto.
12798 (UNSPEC_SM3TT1B): Ditto.
12799 (UNSPEC_SM3TT2A): Ditto.
12800 (UNSPEC_SM3TT2B): Ditto.
12801 (UNSPEC_SM3PARTW1): Ditto.
12802 (UNSPEC_SM3PARTW2): Ditto.
12803 (UNSPEC_SM4E): Ditto.
12804 (UNSPEC_SM4EKEY): Ditto.
12805 * config/aarch64/constraints.md (Ui2): New constraint.
12806 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12807 * config/arm/types.md (crypto_sm3): New type attribute.
12808 (crypto_sm4): Ditto.
12809 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12810 (vsm3tt1aq_u32): Ditto.
12811 (vsm3tt1bq_u32): Ditto.
12812 (vsm3tt2aq_u32): Ditto.
12813 (vsm3tt2bq_u32): Ditto.
12814 (vsm3partw1q_u32): Ditto.
12815 (vsm3partw2q_u32): Ditto.
12816 (vsm4eq_u32): Ditto.
12817 (vsm4ekeyq_u32): Ditto.
12818 (doc/invoke.texi): Document new sm4 option.
12820 2018-01-10 Michael Collison <michael.collison@arm.com>
12822 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12823 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12824 (AARCH64_FL_FOR_ARCH8_4): New.
12825 (AARCH64_FL_V8_4): New flag.
12826 (doc/invoke.texi): Document new armv8.4-a option.
12828 2018-01-10 Michael Collison <michael.collison@arm.com>
12830 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12831 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12832 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12833 * config/aarch64/aarch64-option-extension.def: Add
12834 AARCH64_OPT_EXTENSION of 'sha2'.
12835 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12836 (crypto): Disable sha2 and aes if crypto disabled.
12837 (crypto): Enable aes and sha2 if enabled.
12838 (simd): Disable sha2 and aes if simd disabled.
12839 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12841 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12842 (TARGET_SHA2): New feature flag for sha2.
12843 (TARGET_AES): New feature flag for aes.
12844 * config/aarch64/aarch64-simd.md:
12845 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12846 conditional on TARGET_AES.
12847 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12848 (aarch64_crypto_sha1hsi): Make pattern conditional
12850 (aarch64_crypto_sha1hv4si): Ditto.
12851 (aarch64_be_crypto_sha1hv4si): Ditto.
12852 (aarch64_crypto_sha1su1v4si): Ditto.
12853 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12854 (aarch64_crypto_sha1su0v4si): Ditto.
12855 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12856 (aarch64_crypto_sha256su0v4si): Ditto.
12857 (aarch64_crypto_sha256su1v4si): Ditto.
12858 (doc/invoke.texi): Document new aes and sha2 options.
12860 2018-01-10 Martin Sebor <msebor@redhat.com>
12862 PR tree-optimization/83781
12863 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12866 2018-01-11 Martin Sebor <msebor@gmail.com>
12867 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12869 PR tree-optimization/83501
12870 PR tree-optimization/81703
12872 * tree-ssa-strlen.c (get_string_cst): Rename...
12873 (get_string_len): ...to this. Handle global constants.
12874 (handle_char_store): Adjust.
12876 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12877 Jim Wilson <jimw@sifive.com>
12879 * config/riscv/riscv-protos.h (riscv_output_return): New.
12880 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12881 (riscv_attribute_table, riscv_output_return),
12882 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12883 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12884 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12886 (riscv_expand_prologue): Add early return for naked function.
12887 (riscv_expand_epilogue): Likewise.
12888 (riscv_function_ok_for_sibcall): Return false for naked function.
12889 (riscv_set_current_function): New.
12890 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12891 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12892 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12893 * doc/extend.texi (RISC-V Function Attributes): New.
12895 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12897 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12898 check for 128-bit long double before checking TCmode.
12899 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12900 128-bit long doubles before checking TFmode or TCmode.
12901 (FLOAT128_IBM_P): Likewise.
12903 2018-01-10 Martin Sebor <msebor@redhat.com>
12905 PR tree-optimization/83671
12906 * builtins.c (c_strlen): Unconditionally return zero for the empty
12908 Use -Warray-bounds for warnings.
12909 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12910 for non-constant array indices with COMPONENT_REF, arrays of
12911 arrays, and pointers to arrays.
12912 (gimple_fold_builtin_strlen): Determine and set length range for
12913 non-constant character arrays.
12915 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12917 PR middle-end/81897
12918 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12921 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12923 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12925 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12928 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12929 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12930 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12931 indexed_or_indirect_operand predicate.
12932 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12933 (*vsx_le_perm_load_v8hi): Likewise.
12934 (*vsx_le_perm_load_v16qi): Likewise.
12935 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12936 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12937 (*vsx_le_perm_store_v8hi): Likewise.
12938 (*vsx_le_perm_store_v16qi): Likewise.
12939 (eight unnamed splitters): Likewise.
12941 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12943 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12944 * config/rs6000/emmintrin.h: Likewise.
12945 * config/rs6000/mmintrin.h: Likewise.
12946 * config/rs6000/xmmintrin.h: Likewise.
12948 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12951 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12953 * tree.c (tree_nop_conversion): Return true for location wrapper
12955 (maybe_wrap_with_location): New function.
12956 (selftest::check_strip_nops): New function.
12957 (selftest::test_location_wrappers): New function.
12958 (selftest::tree_c_tests): Call it.
12959 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12960 (maybe_wrap_with_location): New decl.
12961 (EXPR_LOCATION_WRAPPER_P): New macro.
12962 (location_wrapper_p): New inline function.
12963 (tree_strip_any_location_wrapper): New inline function.
12965 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12968 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12969 stack_realign_offset for the largest alignment of stack slot
12971 (ix86_find_max_used_stack_alignment): New function.
12972 (ix86_finalize_stack_frame_flags): Use it. Set
12973 max_used_stack_alignment if we don't realign stack.
12974 * config/i386/i386.h (machine_function): Add
12975 max_used_stack_alignment.
12977 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12979 * config/arm/arm.opt (-mbranch-cost): New option.
12980 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12983 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12986 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12987 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12989 2018-01-10 Richard Biener <rguenther@suse.de>
12992 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12993 early out so it also covers the case where we have a non-NULL
12996 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12998 PR tree-optimization/83753
12999 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13000 for non-strided grouped accesses if the number of elements is 1.
13002 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13005 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13006 * i386.h (TARGET_USE_GATHER): Define.
13007 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13009 2018-01-10 Martin Liska <mliska@suse.cz>
13012 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13013 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13015 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13016 CLEANUP_NO_PARTITIONING is not set.
13018 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13020 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13021 for vectors, as a partial revert of r254296.
13022 * rtl.h (const_vec_p): Delete.
13023 (const_vec_duplicate_p): Don't test for vector CONSTs.
13024 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13025 * expmed.c (make_tree): Likewise.
13028 * common.md (E, F): Use CONSTANT_P instead of checking for
13030 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13031 checking for CONST_VECTOR.
13033 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13035 PR middle-end/83575
13036 * predict.c (force_edge_cold): Handle in more sane way edges
13037 with no prediction.
13039 2018-01-09 Carl Love <cel@us.ibm.com>
13041 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13043 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13044 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13045 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13046 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13047 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13048 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13049 * config/rs6000/rs6000-protos.h: Add extern defition for
13050 rs6000_generate_float2_double_code.
13051 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13053 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13054 (float2_v2df): Add define_expand.
13056 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13059 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13060 op_mode in the force_to_mode call.
13062 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13064 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13065 instead of checking each element individually.
13066 (aarch64_evpc_uzp): Likewise.
13067 (aarch64_evpc_zip): Likewise.
13068 (aarch64_evpc_ext): Likewise.
13069 (aarch64_evpc_rev): Likewise.
13070 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13071 instead of checking each element individually. Return true without
13073 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13074 whether all selected elements come from the same input, instead of
13075 checking each element individually. Remove calls to gen_rtx_REG,
13076 start_sequence and end_sequence and instead assert that no rtl is
13079 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13081 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13082 order of HIGH and CONST checks.
13084 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13086 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13087 if the destination isn't an SSA_NAME.
13089 2018-01-09 Richard Biener <rguenther@suse.de>
13091 PR tree-optimization/83668
13092 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13094 (canonicalize_loop_form): ... here, renamed from ...
13095 (canonicalize_loop_closed_ssa_form): ... this and amended to
13096 swap successor edges for loop exit blocks to make us use
13097 the RPO order we need for initial schedule generation.
13099 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13101 PR tree-optimization/64811
13102 * match.pd: When optimizing comparisons with Inf, avoid
13103 introducing or losing exceptions from comparisons with NaN.
13105 2018-01-09 Martin Liska <mliska@suse.cz>
13108 * asan.c (shadow_mem_size): Add gcc_assert.
13110 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13112 Don't save registers in main().
13115 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13116 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13117 * config/avr/avr.c (avr_set_current_function): Don't error if
13118 naked, OS_task or OS_main are specified at the same time.
13119 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13121 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13123 * common/config/avr/avr-common.c (avr_option_optimization_table):
13124 Switch on -mmain-is-OS_task for optimizing compilations.
13126 2018-01-09 Richard Biener <rguenther@suse.de>
13128 PR tree-optimization/83572
13129 * graphite.c: Include cfganal.h.
13130 (graphite_transform_loops): Connect infinite loops to exit
13131 and remove fake edges at the end.
13133 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13135 * ipa-inline.c (edge_badness): Revert accidental checkin.
13137 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13140 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13141 symbols; not inline clones.
13143 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13146 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13147 hard registers. Formatting fixes.
13149 PR preprocessor/83722
13150 * gcc.c (try_generate_repro): Pass
13151 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13152 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13155 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13156 Kito Cheng <kito.cheng@gmail.com>
13158 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13159 (riscv_leaf_function_p): Delete.
13160 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13162 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13164 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13166 (do_ifelse): New function.
13167 (do_isel): New function.
13168 (do_sub3): New function.
13169 (do_add3): New function.
13170 (do_load_mask_compare): New function.
13171 (do_overlap_load_compare): New function.
13172 (expand_compare_loop): New function.
13173 (expand_block_compare): Call expand_compare_loop() when appropriate.
13174 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13175 option description.
13176 (-mblock-compare-inline-loop-limit): New option.
13178 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13181 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13182 Reverse order of second and third operands in first alternative.
13183 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13184 of first and second elements in UNSPEC_VPERMR vector.
13185 (altivec_expand_vec_perm_le): Likewise.
13187 2018-01-08 Jeff Law <law@redhat.com>
13189 PR rtl-optimizatin/81308
13190 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13191 (process_switch): If group_case_labels makes a change, then set
13193 (pass_convert_switch::execute): If a switch is converted, then
13194 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13196 PR rtl-optimization/81308
13197 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13200 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13202 PR target/83663 - Revert r255946
13203 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13204 generation for cases where splatting a value is not useful.
13205 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13206 across a vec_duplicate and a paradoxical subreg forming a vector
13207 mode to a vec_concat.
13209 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13211 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13212 -march=armv8.3-a variants.
13213 * config/arm/t-multilib: Likewise.
13214 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13216 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13218 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13220 (cceq_ior_compare_complement): Give it a name so I can use it, and
13221 change boolean_or_operator predicate to boolean_operator so it can
13222 be used to generate a crand.
13223 (eqne): New code iterator.
13224 (bd/bd_neg): New code_attrs.
13225 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13226 a single define_insn.
13227 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13228 decrement (bdnzt/bdnzf/bdzt/bdzf).
13229 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13230 with the new names of the branch decrement patterns, and added the
13231 names of the branch decrement conditional patterns.
13233 2018-01-08 Richard Biener <rguenther@suse.de>
13235 PR tree-optimization/83563
13236 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13239 2018-01-08 Richard Biener <rguenther@suse.de>
13241 PR middle-end/83713
13242 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13244 2018-01-08 Richard Biener <rguenther@suse.de>
13246 PR tree-optimization/83685
13247 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13248 references to abnormals.
13250 2018-01-08 Richard Biener <rguenther@suse.de>
13253 * dwarf2out.c (output_indirect_strings): Handle empty
13254 skeleton_debug_str_hash.
13255 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13257 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13259 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13260 (emit_store_direct): Likewise.
13261 (arc_trampoline_adjust_address): Likewise.
13262 (arc_asm_trampoline_template): New function.
13263 (arc_initialize_trampoline): Use asm_trampoline_template.
13264 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13265 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13266 * config/arc/arc.md (flush_icache): Delete pattern.
13268 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13270 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13271 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13274 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13277 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13278 by not USED_FOR_TARGET.
13279 (make_pass_resolve_sw_modes): Likewise.
13281 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13283 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13286 2018-01-08 Richard Biener <rguenther@suse.de>
13288 PR middle-end/83580
13289 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13291 2018-01-08 Richard Biener <rguenther@suse.de>
13293 PR middle-end/83517
13294 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13296 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13298 PR middle-end/81897
13299 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13300 basic blocks with a small number of successors.
13301 (convert_control_dep_chain_into_preds): Improve handling of
13303 (dump_predicates): Split apart into...
13304 (dump_pred_chain): ...here...
13305 (dump_pred_info): ...and here.
13306 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13307 (can_chain_union_be_invalidated_p): Improve check for invalidation
13309 (uninit_uses_cannot_happen): Avoid unnecessary if
13310 convert_control_dep_chain_into_preds yielded nothing.
13312 2018-01-06 Martin Sebor <msebor@redhat.com>
13314 PR tree-optimization/83640
13315 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13316 subtracting negative offset from size.
13317 (builtin_access::overlap): Adjust offset bounds of the access to fall
13318 within the size of the object if possible.
13320 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13322 PR rtl-optimization/83699
13323 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13324 extract_bit_field_as_subreg to cases in which the extracted
13325 value is also a vector.
13327 * lra-constraints.c (process_alt_operands): Test for the equivalence
13328 substitutions when detecting a possible reload cycle.
13330 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13333 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13334 by default if flag_selective_schedling{,2}. Formatting fixes.
13336 PR rtl-optimization/83682
13337 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13338 if it has non-VECTOR_MODE element mode.
13339 (vec_duplicate_p): Likewise.
13341 PR middle-end/83694
13342 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13343 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13345 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13348 * config/i386/i386-builtin.def
13349 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13350 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13351 Require also OPTION_MASK_ISA_AVX512F in addition to
13352 OPTION_MASK_ISA_GFNI.
13353 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13354 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13355 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13356 to OPTION_MASK_ISA_GFNI.
13357 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13358 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13359 OPTION_MASK_ISA_AVX512BW.
13360 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13361 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13362 addition to OPTION_MASK_ISA_GFNI.
13363 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13364 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13365 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13366 to OPTION_MASK_ISA_GFNI.
13367 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13368 a requirement for all ISAs rather than any of them with a few
13370 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13372 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13373 bitmasks to be enabled with 3 exceptions, instead of requiring any
13374 enabled ISA with lots of exceptions.
13375 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13376 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13377 Change avx512bw in isa attribute to avx512f.
13378 * config/i386/sgxintrin.h: Add license boilerplate.
13379 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13380 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13381 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13382 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13384 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13385 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13386 temporarily sse2 rather than sse if not enabled already.
13389 * config/i386/sse.md (VI248_VLBW): Rename to ...
13390 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13391 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13392 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13393 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13394 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13395 mode iterator instead of VI248_VLBW.
13397 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13399 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13400 (record_modified): Skip clobbers; add debug output.
13401 (param_change_prob): Use sreal frequencies.
13403 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13405 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13406 punt for user-aligned variables.
13408 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13410 * tree-chrec.c (chrec_contains_symbols): Return true for
13413 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13416 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13417 of (x|y) == x for BICS pattern.
13419 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13421 PR tree-optimization/83605
13422 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13423 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13426 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13428 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13429 * config/epiphany/rtems.h: New file.
13431 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13432 Uros Bizjak <ubizjak@gmail.com>
13435 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13436 QIreg_operand instead of register_operand predicate.
13437 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13438 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13439 comments instead of -fmitigate[-_]rop.
13441 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13444 * cgraphunit.c (symbol_table::compile): Switch to text_section
13445 before calling assembly_start debug hook.
13446 * run-rtl-passes.c (run_rtl_passes): Likewise.
13449 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13451 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13452 range_int_cst_p rather than !symbolic_range_p before calling
13453 extract_range_from_multiplicative_op_1.
13455 2018-01-04 Jeff Law <law@redhat.com>
13457 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13458 redundant test in assertion.
13460 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13462 * doc/rtl.texi: Document machine_mode wrapper classes.
13464 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13466 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13467 using tree_to_uhwi.
13469 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13471 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13472 the VEC_PERM_EXPR fold to fail.
13474 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13477 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13478 to switched_sections.
13480 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13483 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13484 test for d.testing.
13486 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13489 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13490 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13492 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13495 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13496 is BLKmode and bitpos not zero or mode change is needed.
13498 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13501 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13504 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13507 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13508 instead of MULT rtx. Update all corresponding splitters.
13509 (*saddl_se): Ditto.
13510 (*ssub<modesuffix>): Ditto.
13511 (*ssubl_se): Ditto.
13512 (*cmp_sadd_di): Update split patterns.
13513 (*cmp_sadd_si): Ditto.
13514 (*cmp_sadd_sidi): Ditto.
13515 (*cmp_ssub_di): Ditto.
13516 (*cmp_ssub_si): Ditto.
13517 (*cmp_ssub_sidi): Ditto.
13518 * config/alpha/predicates.md (const23_operand): New predicate.
13519 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13520 Look for ASHIFT, not MULT inner operand.
13521 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13523 2018-01-04 Martin Liska <mliska@suse.cz>
13525 PR gcov-profile/83669
13526 * gcov.c (output_intermediate_file): Add version to intermediate
13528 * doc/gcov.texi: Document new field 'version' in intermediate
13529 file format. Fix location of '-k' option of gcov command.
13531 2018-01-04 Martin Liska <mliska@suse.cz>
13534 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13536 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13538 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13540 2018-01-03 Martin Sebor <msebor@redhat.com>
13542 PR tree-optimization/83655
13543 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13544 checking calls with invalid arguments.
13546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13548 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13549 (vectorizable_mask_load_store): Delete.
13550 (vectorizable_call): Return false for masked loads and stores.
13551 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13552 instead of gimple_assign_rhs1.
13553 (vectorizable_load): Handle IFN_MASK_LOAD.
13554 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13558 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13560 (vectorizable_mask_load_store): ...here.
13561 (vectorizable_load): ...and here.
13563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13565 * tree-vect-stmts.c (vect_build_all_ones_mask)
13566 (vect_build_zero_merge_argument): New functions, split out from...
13567 (vectorizable_load): ...here.
13569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13571 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13573 (vectorizable_mask_load_store): ...here.
13574 (vectorizable_store): ...and here.
13576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13578 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13580 (vectorizable_mask_load_store): ...here.
13582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13584 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13585 (vect_model_store_cost): Take a vec_load_store_type instead of a
13587 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13588 (vect_model_store_cost): Take a vec_load_store_type instead of a
13590 (vectorizable_mask_load_store): Update accordingly.
13591 (vectorizable_store): Likewise.
13592 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13596 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13597 IFN_MASK_LOAD calls here rather than...
13598 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13601 Alan Hayward <alan.hayward@arm.com>
13602 David Sherwood <david.sherwood@arm.com>
13604 * expmed.c (extract_bit_field_1): For vector extracts,
13605 fall back to extract_bit_field_as_subreg if vec_extract
13608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13609 Alan Hayward <alan.hayward@arm.com>
13610 David Sherwood <david.sherwood@arm.com>
13612 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13613 they are variable or constant sized.
13614 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13615 slots for constant-sized data.
13617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13618 Alan Hayward <alan.hayward@arm.com>
13619 David Sherwood <david.sherwood@arm.com>
13621 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13622 handling COND_EXPRs with boolean comparisons, try to find a better
13623 basis for the mask type than the boolean itself.
13625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13627 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13628 is calculated and how it can be overridden.
13629 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13630 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13632 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13636 Alan Hayward <alan.hayward@arm.com>
13637 David Sherwood <david.sherwood@arm.com>
13639 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13640 Remove the mode argument.
13641 (aarch64_simd_valid_immediate): Remove the mode and inverse
13643 * config/aarch64/iterators.md (bitsize): New iterator.
13644 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13645 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13646 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13647 aarch64_simd_valid_immediate.
13648 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13649 (aarch64_reg_or_bic_imm): Likewise.
13650 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13651 with an insn_type enum and msl with a modifier_type enum.
13652 Replace element_width with a scalar_mode. Change the shift
13653 to unsigned int. Add constructors for scalar_float_mode and
13654 scalar_int_mode elements.
13655 (aarch64_vect_float_const_representable_p): Delete.
13656 (aarch64_can_const_movi_rtx_p)
13657 (aarch64_simd_scalar_immediate_valid_for_move)
13658 (aarch64_simd_make_constant): Update call to
13659 aarch64_simd_valid_immediate.
13660 (aarch64_advsimd_valid_immediate_hs): New function.
13661 (aarch64_advsimd_valid_immediate): Likewise.
13662 (aarch64_simd_valid_immediate): Remove mode and inverse
13663 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13664 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13665 and aarch64_float_const_representable_p on the result.
13666 (aarch64_output_simd_mov_immediate): Remove mode argument.
13667 Update call to aarch64_simd_valid_immediate and use of
13668 simd_immediate_info.
13669 (aarch64_output_scalar_simd_mov_immediate): Update call
13672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13673 Alan Hayward <alan.hayward@arm.com>
13674 David Sherwood <david.sherwood@arm.com>
13676 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13677 (mode_nunits): Likewise CONST_MODE_NUNITS.
13678 * machmode.def (ADJUST_NUNITS): Document.
13679 * genmodes.c (mode_data::need_nunits_adj): New field.
13680 (blank_mode): Update accordingly.
13681 (adj_nunits): New variable.
13682 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13684 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13685 listed in adj_nunits.
13686 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13687 listed in adj_nunits. Don't emit case statements for such modes.
13688 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13689 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13690 nothing if adj_nunits is nonnull.
13691 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13692 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13693 (emit_mode_fbit): Update use of print_maybe_const_decl.
13694 (emit_move_size): Likewise. Treat the array as non-const
13696 (emit_mode_adjustments): Handle adj_nunits.
13698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13700 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13701 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13702 (VECTOR_MODES): Use it.
13703 (make_vector_modes): Take the prefix as an argument.
13705 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13706 Alan Hayward <alan.hayward@arm.com>
13707 David Sherwood <david.sherwood@arm.com>
13709 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13710 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13711 for MODE_VECTOR_BOOL.
13712 * machmode.def (VECTOR_BOOL_MODE): Document.
13713 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13714 (make_vector_bool_mode): New function.
13715 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13717 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13718 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13720 * stor-layout.c (int_mode_for_mode): Likewise.
13721 * tree.c (build_vector_type_for_mode): Likewise.
13722 * varasm.c (output_constant_pool_2): Likewise.
13723 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13724 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13725 for MODE_VECTOR_BOOL.
13726 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13727 of mode class checks.
13728 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13729 instead of a list of mode class checks.
13730 (expand_vector_scalar_condition): Likewise.
13731 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13734 Alan Hayward <alan.hayward@arm.com>
13735 David Sherwood <david.sherwood@arm.com>
13737 * machmode.h (mode_size): Change from unsigned short to
13739 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13740 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13741 or if measurement_type is not polynomial.
13742 (fixed_size_mode::includes_p): Check for constant-sized modes.
13743 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13744 return a poly_uint16 rather than an unsigned short.
13745 (emit_mode_size): Change the type of mode_size from unsigned short
13746 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13747 (emit_mode_adjustments): Cope with polynomial vector sizes.
13748 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13750 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13752 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13753 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13754 * caller-save.c (setup_save_areas): Likewise.
13755 (replace_reg_with_saved_mem): Likewise.
13756 * calls.c (emit_library_call_value_1): Likewise.
13757 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13758 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13759 (gen_lowpart_for_combine): Likewise.
13760 * convert.c (convert_to_integer_1): Likewise.
13761 * cse.c (equiv_constant, cse_insn): Likewise.
13762 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13763 (cselib_subst_to_values): Likewise.
13764 * dce.c (word_dce_process_block): Likewise.
13765 * df-problems.c (df_word_lr_mark_ref): Likewise.
13766 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13767 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13768 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13769 (rtl_for_decl_location): Likewise.
13770 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13771 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13772 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13773 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13774 (expand_expr_real_1): Likewise.
13775 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13776 (pad_below): Likewise.
13777 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13778 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13779 * ira.c (get_subreg_tracking_sizes): Likewise.
13780 * ira-build.c (ira_create_allocno_objects): Likewise.
13781 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13782 (ira_sort_regnos_for_alter_reg): Likewise.
13783 * ira-costs.c (record_operand_costs): Likewise.
13784 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13785 (resolve_simple_move): Likewise.
13786 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13787 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13788 (lra_constraints): Likewise.
13789 (CONST_POOL_OK_P): Reject variable-sized modes.
13790 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13791 (add_pseudo_to_slot, lra_spill): Likewise.
13792 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13793 * optabs-query.c (get_best_extraction_insn): Likewise.
13794 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13795 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13796 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13797 * recog.c (offsettable_address_addr_space_p): Likewise.
13798 * regcprop.c (maybe_mode_change): Likewise.
13799 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13800 * regrename.c (build_def_use): Likewise.
13801 * regstat.c (dump_reg_info): Likewise.
13802 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13803 (find_reloads, find_reloads_subreg_address): Likewise.
13804 * reload1.c (eliminate_regs_1): Likewise.
13805 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13806 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13807 (simplify_binary_operation_1, simplify_subreg): Likewise.
13808 * targhooks.c (default_function_arg_padding): Likewise.
13809 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13810 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13811 (verify_gimple_assign_ternary): Likewise.
13812 * tree-inline.c (estimate_move_cost): Likewise.
13813 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13814 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13815 (get_address_cost_ainc): Likewise.
13816 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13817 (vect_supportable_dr_alignment): Likewise.
13818 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13819 (vectorizable_reduction): Likewise.
13820 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13821 (vectorizable_operation, vectorizable_load): Likewise.
13822 * tree.c (build_same_sized_truth_vector_type): Likewise.
13823 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13824 * var-tracking.c (emit_note_insn_var_location): Likewise.
13825 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13826 (ADDR_VEC_ALIGN): Likewise.
13828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13829 Alan Hayward <alan.hayward@arm.com>
13830 David Sherwood <david.sherwood@arm.com>
13832 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13834 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13835 or if measurement_type is polynomial.
13836 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13837 * combine.c (make_extraction): Likewise.
13838 * dse.c (find_shift_sequence): Likewise.
13839 * dwarf2out.c (mem_loc_descriptor): Likewise.
13840 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13841 (extract_bit_field, extract_low_bits): Likewise.
13842 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13843 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13844 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13845 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13846 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13847 * reload.c (find_reloads): Likewise.
13848 * reload1.c (alter_reg): Likewise.
13849 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13850 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13851 * tree-if-conv.c (predicate_mem_writes): Likewise.
13852 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13853 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13854 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13855 * valtrack.c (dead_debug_insert_temp): Likewise.
13856 * varasm.c (mergeable_constant_section): Likewise.
13857 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13860 Alan Hayward <alan.hayward@arm.com>
13861 David Sherwood <david.sherwood@arm.com>
13863 * expr.c (expand_assignment): Cope with polynomial mode sizes
13864 when assigning to a CONCAT.
13866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13867 Alan Hayward <alan.hayward@arm.com>
13868 David Sherwood <david.sherwood@arm.com>
13870 * machmode.h (mode_precision): Change from unsigned short to
13872 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13874 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13875 or if measurement_type is not polynomial.
13876 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13877 in which the mode is already known to be a scalar_int_mode.
13878 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13879 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13881 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13882 for GET_MODE_PRECISION.
13883 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13884 for GET_MODE_PRECISION.
13885 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13887 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13888 (expand_field_assignment, make_extraction): Likewise.
13889 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13890 (get_last_value): Likewise.
13891 * convert.c (convert_to_integer_1): Likewise.
13892 * cse.c (cse_insn): Likewise.
13893 * expr.c (expand_expr_real_1): Likewise.
13894 * lra-constraints.c (simplify_operand_subreg): Likewise.
13895 * optabs-query.c (can_atomic_load_p): Likewise.
13896 * optabs.c (expand_atomic_load): Likewise.
13897 (expand_atomic_store): Likewise.
13898 * ree.c (combine_reaching_defs): Likewise.
13899 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13900 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13901 * tree.h (type_has_mode_precision_p): Likewise.
13902 * ubsan.c (instrument_si_overflow): Likewise.
13904 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13905 Alan Hayward <alan.hayward@arm.com>
13906 David Sherwood <david.sherwood@arm.com>
13908 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13909 polynomial numbers of units.
13910 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13911 (valid_vector_subparts_p): New function.
13912 (build_vector_type): Remove temporary shim and take the number
13913 of units as a poly_uint64 rather than an int.
13914 (build_opaque_vector_type): Take the number of units as a
13915 poly_uint64 rather than an int.
13916 * tree.c (build_vector_from_ctor): Handle polynomial
13917 TYPE_VECTOR_SUBPARTS.
13918 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13919 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13920 (build_vector_from_val): If the number of units is variable,
13921 use build_vec_duplicate_cst for constant operands and
13922 VEC_DUPLICATE_EXPR otherwise.
13923 (make_vector_type): Remove temporary is_constant ().
13924 (build_vector_type, build_opaque_vector_type): Take the number of
13925 units as a poly_uint64 rather than an int.
13926 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13928 * cfgexpand.c (expand_debug_expr): Likewise.
13929 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13930 (store_constructor, expand_expr_real_1): Likewise.
13931 (const_scalar_mask_from_tree): Likewise.
13932 * fold-const-call.c (fold_const_reduction): Likewise.
13933 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13934 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13935 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13936 (fold_relational_const): Likewise.
13937 (native_interpret_vector): Likewise. Change the size from an
13938 int to an unsigned int.
13939 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13940 TYPE_VECTOR_SUBPARTS.
13941 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13942 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13943 duplicating a non-constant operand into a variable-length vector.
13944 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13945 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13946 * ipa-icf.c (sem_variable::equals): Likewise.
13947 * match.pd: Likewise.
13948 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13949 * print-tree.c (print_node): Likewise.
13950 * stor-layout.c (layout_type): Likewise.
13951 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13952 * tree-cfg.c (verify_gimple_comparison): Likewise.
13953 (verify_gimple_assign_binary): Likewise.
13954 (verify_gimple_assign_ternary): Likewise.
13955 (verify_gimple_assign_single): Likewise.
13956 * tree-pretty-print.c (dump_generic_node): Likewise.
13957 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13958 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13959 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13960 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13961 (vect_shift_permute_load_chain): Likewise.
13962 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13963 (expand_vector_condition, optimize_vector_constructor): Likewise.
13964 (lower_vec_perm, get_compute_type): Likewise.
13965 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13966 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13967 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13968 (vect_recog_mask_conversion_pattern): Likewise.
13969 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13970 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13971 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13972 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13973 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13974 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13975 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13976 (supportable_widening_operation): Likewise.
13977 (supportable_narrowing_operation): Likewise.
13978 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13980 * varasm.c (output_constant): Likewise.
13982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13983 Alan Hayward <alan.hayward@arm.com>
13984 David Sherwood <david.sherwood@arm.com>
13986 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13987 so that both the length == 3 and length != 3 cases set up their
13988 own permute vectors. Add comments explaining why we know the
13989 number of elements is constant.
13990 (vect_permute_load_chain): Likewise.
13992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13993 Alan Hayward <alan.hayward@arm.com>
13994 David Sherwood <david.sherwood@arm.com>
13996 * machmode.h (mode_nunits): Change from unsigned char to
13998 (ONLY_FIXED_SIZE_MODES): New macro.
13999 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14000 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14001 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14003 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14004 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14005 or if measurement_type is not polynomial.
14006 * genmodes.c (ZERO_COEFFS): New macro.
14007 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14009 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14010 Use ZERO_COEFFS when emitting initializers.
14011 * data-streamer.h (bp_pack_poly_value): New function.
14012 (bp_unpack_poly_value): Likewise.
14013 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14014 for GET_MODE_NUNITS.
14015 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14016 for GET_MODE_NUNITS.
14017 * tree.c (make_vector_type): Remove temporary shim and make
14018 the real function take the number of units as a poly_uint64
14019 rather than an int.
14020 (build_vector_type_for_mode): Handle polynomial nunits.
14021 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14022 * emit-rtl.c (const_vec_series_p_1): Likewise.
14023 (gen_rtx_CONST_VECTOR): Likewise.
14024 * fold-const.c (test_vec_duplicate_folding): Likewise.
14025 * genrecog.c (validate_pattern): Likewise.
14026 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14027 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14028 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14029 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14030 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14031 * rtlanal.c (subreg_get_info): Likewise.
14032 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14033 (vect_grouped_load_supported): Likewise.
14034 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14035 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14036 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14037 (simplify_const_unary_operation, simplify_binary_operation_1)
14038 (simplify_const_binary_operation, simplify_ternary_operation)
14039 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14040 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14041 instead of CONST_VECTOR_NUNITS.
14042 * varasm.c (output_constant_pool_2): Likewise.
14043 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14044 explicit-encoded elements in the XVEC for variable-length vectors.
14046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14048 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14051 Alan Hayward <alan.hayward@arm.com>
14052 David Sherwood <david.sherwood@arm.com>
14054 * coretypes.h (fixed_size_mode): Declare.
14055 (fixed_size_mode_pod): New typedef.
14056 * builtins.h (target_builtins::x_apply_args_mode)
14057 (target_builtins::x_apply_result_mode): Change type to
14058 fixed_size_mode_pod.
14059 * builtins.c (apply_args_size, apply_result_size, result_vector)
14060 (expand_builtin_apply_args_1, expand_builtin_apply)
14061 (expand_builtin_return): Update accordingly.
14063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14065 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14066 * cselib.c (cselib_hash_rtx): Likewise.
14067 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14068 CONST_VECTOR encoding.
14070 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14071 Jeff Law <law@redhat.com>
14074 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14075 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14076 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14077 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14080 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14081 explicitly probe *sp in a noreturn function if there were any callee
14082 register saves or frame pointer is needed.
14084 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14087 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14088 BLKmode for ternary, binary or unary expressions.
14091 * var-tracking.c (delete_vta_debug_insn): New inline function.
14092 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14093 insns from get_insns () to NULL instead of each bb separately.
14094 Use delete_vta_debug_insn. No longer static.
14095 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14096 delete_vta_debug_insns callers.
14097 * rtl.h (delete_vta_debug_insns): Declare.
14098 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14099 instead of variable_tracking_main.
14101 2018-01-03 Martin Sebor <msebor@redhat.com>
14103 PR tree-optimization/83603
14104 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14105 arguments past the endof the argument list in functions declared
14106 without a prototype.
14107 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14108 Avoid checking when arguments are null.
14110 2018-01-03 Martin Sebor <msebor@redhat.com>
14113 * doc/extend.texi (attribute const): Fix a typo.
14114 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14115 issuing -Wsuggest-attribute for void functions.
14117 2018-01-03 Martin Sebor <msebor@redhat.com>
14119 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14120 offset_int::from instead of wide_int::to_shwi.
14121 (maybe_diag_overlap): Remove assertion.
14122 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14123 * gimple-ssa-sprintf.c (format_directive): Same.
14124 (parse_directive): Same.
14125 (sprintf_dom_walker::compute_format_length): Same.
14126 (try_substitute_return_value): Same.
14128 2018-01-03 Jeff Law <law@redhat.com>
14130 PR middle-end/83654
14131 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14132 non-constant residual for zero at runtime and avoid probing in
14133 that case. Reorganize code for trailing problem to mirror handling
14136 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14138 PR tree-optimization/83501
14139 * tree-ssa-strlen.c (get_string_cst): New.
14140 (handle_char_store): Call get_string_cst.
14142 2018-01-03 Martin Liska <mliska@suse.cz>
14144 PR tree-optimization/83593
14145 * tree-ssa-strlen.c: Include tree-cfg.h.
14146 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14147 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14148 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14150 (strlen_dom_walker::before_dom_children): Call
14151 gimple_purge_dead_eh_edges. Dump tranformation with details
14153 (strlen_dom_walker::before_dom_children): Update call by adding
14154 new argument cleanup_eh.
14155 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14157 2018-01-03 Martin Liska <mliska@suse.cz>
14160 * cif-code.def (VARIADIC_THUNK): New enum value.
14161 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14164 2018-01-03 Jan Beulich <jbeulich@suse.com>
14166 * sse.md (mov<mode>_internal): Tighten condition for when to use
14167 vmovdqu<ssescalarsize> for TI and OI modes.
14169 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14171 Update copyright years.
14173 2018-01-03 Martin Liska <mliska@suse.cz>
14176 * ipa-visibility.c (function_and_variable_visibility): Skip
14177 functions with noipa attribure.
14179 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14181 * gcc.c (process_command): Update copyright notice dates.
14182 * gcov-dump.c (print_version): Ditto.
14183 * gcov.c (print_version): Ditto.
14184 * gcov-tool.c (print_version): Ditto.
14185 * gengtype.c (create_file): Ditto.
14186 * doc/cpp.texi: Bump @copying's copyright year.
14187 * doc/cppinternals.texi: Ditto.
14188 * doc/gcc.texi: Ditto.
14189 * doc/gccint.texi: Ditto.
14190 * doc/gcov.texi: Ditto.
14191 * doc/install.texi: Ditto.
14192 * doc/invoke.texi: Ditto.
14194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14196 * vector-builder.h (vector_builder::m_full_nelts): Change from
14197 unsigned int to poly_uint64.
14198 (vector_builder::full_nelts): Update prototype accordingly.
14199 (vector_builder::new_vector): Likewise.
14200 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14201 (vector_builder::operator ==): Likewise.
14202 (vector_builder::finalize): Likewise.
14203 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14204 Take the number of elements as a poly_uint64 rather than an
14206 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14207 from unsigned int to poly_uint64.
14208 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14209 (vec_perm_indices::new_vector): Likewise.
14210 (vec_perm_indices::length): Likewise.
14211 (vec_perm_indices::nelts_per_input): Likewise.
14212 (vec_perm_indices::input_nelts): Likewise.
14213 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14214 number of elements per input as a poly_uint64 rather than an
14215 unsigned int. Use the original encoding for variable-length
14216 vectors, rather than clamping each individual element.
14217 For the second and subsequent elements in each pattern,
14218 clamp the step and base before clamping their sum.
14219 (vec_perm_indices::series_p): Handle polynomial element counts.
14220 (vec_perm_indices::all_in_range_p): Likewise.
14221 (vec_perm_indices_to_tree): Likewise.
14222 (vec_perm_indices_to_rtx): Likewise.
14223 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14224 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14225 (tree_vector_builder::new_binary_operation): Handle polynomial
14226 element counts. Return false if we need to know the number
14227 of elements at compile time.
14228 * fold-const.c (fold_vec_perm): Punt if the number of elements
14229 isn't known at compile time.
14231 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14233 * vec-perm-indices.h (vec_perm_builder): Change element type
14234 from HOST_WIDE_INT to poly_int64.
14235 (vec_perm_indices::element_type): Update accordingly.
14236 (vec_perm_indices::clamp): Handle polynomial element_types.
14237 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14238 (vec_perm_indices::all_in_range_p): Likewise.
14239 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14241 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14242 polynomial vec_perm_indices element types.
14243 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14244 * fold-const.c (fold_vec_perm): Likewise.
14245 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14246 * tree-vect-generic.c (lower_vec_perm): Likewise.
14247 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14248 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14249 element type to HOST_WIDE_INT.
14251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14252 Alan Hayward <alan.hayward@arm.com>
14253 David Sherwood <david.sherwood@arm.com>
14255 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14256 rather than an int. Use plus_constant.
14257 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14258 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14261 Alan Hayward <alan.hayward@arm.com>
14262 David Sherwood <david.sherwood@arm.com>
14264 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14265 a HOST_WIDE_INT to a poly_int64.
14267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14268 Alan Hayward <alan.hayward@arm.com>
14269 David Sherwood <david.sherwood@arm.com>
14271 * calls.c (load_register_parameters): Cope with polynomial
14272 mode sizes. Require a constant size for BLKmode parameters
14273 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14274 forces a parameter to be padded at the lsb end in order to
14275 fill a complete number of words, require the parameter size
14276 to be ordered wrt UNITS_PER_WORD.
14278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14279 Alan Hayward <alan.hayward@arm.com>
14280 David Sherwood <david.sherwood@arm.com>
14282 * reload1.c (spill_stack_slot_width): Change element type
14283 from unsigned int to poly_uint64_pod.
14284 (alter_reg): Treat mode sizes as polynomial.
14286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14287 Alan Hayward <alan.hayward@arm.com>
14288 David Sherwood <david.sherwood@arm.com>
14290 * reload.c (complex_word_subreg_p): New function.
14291 (reload_inner_reg_of_subreg, push_reload): Use it.
14293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14294 Alan Hayward <alan.hayward@arm.com>
14295 David Sherwood <david.sherwood@arm.com>
14297 * lra-constraints.c (process_alt_operands): Reject matched
14298 operands whose sizes aren't ordered.
14299 (match_reload): Refer to this check here.
14301 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14302 Alan Hayward <alan.hayward@arm.com>
14303 David Sherwood <david.sherwood@arm.com>
14305 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14306 that the mode size is in the set {1, 2, 4, 8, 16}.
14308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14309 Alan Hayward <alan.hayward@arm.com>
14310 David Sherwood <david.sherwood@arm.com>
14312 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14313 Use plus_constant instead of gen_rtx_PLUS.
14315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14316 Alan Hayward <alan.hayward@arm.com>
14317 David Sherwood <david.sherwood@arm.com>
14319 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14320 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14321 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14322 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14323 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14324 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14325 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14326 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14327 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14328 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14330 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14331 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14332 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14333 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14334 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14335 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14336 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14337 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14338 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14339 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14341 * expr.c (emit_move_resolve_push): Treat the input and result
14342 of PUSH_ROUNDING as a poly_int64.
14343 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14344 (emit_push_insn): Likewise.
14345 * lra-eliminations.c (mark_not_eliminable): Likewise.
14346 * recog.c (push_operand): Likewise.
14347 * reload1.c (elimination_effects): Likewise.
14348 * rtlanal.c (nonzero_bits1): Likewise.
14349 * calls.c (store_one_arg): Likewise. Require the padding to be
14350 known at compile time.
14352 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14353 Alan Hayward <alan.hayward@arm.com>
14354 David Sherwood <david.sherwood@arm.com>
14356 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14357 Use plus_constant instead of gen_rtx_PLUS.
14359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14360 Alan Hayward <alan.hayward@arm.com>
14361 David Sherwood <david.sherwood@arm.com>
14363 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14364 rather than an int.
14366 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14367 Alan Hayward <alan.hayward@arm.com>
14368 David Sherwood <david.sherwood@arm.com>
14370 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14371 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14372 via stack temporaries. Treat the mode size as polynomial too.
14374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14375 Alan Hayward <alan.hayward@arm.com>
14376 David Sherwood <david.sherwood@arm.com>
14378 * expr.c (expand_expr_real_2): When handling conversions involving
14379 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14380 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14381 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14390 Alan Hayward <alan.hayward@arm.com>
14391 David Sherwood <david.sherwood@arm.com>
14393 * combine.c (can_change_dest_mode): Handle polynomial
14394 REGMODE_NATURAL_SIZE.
14395 * expmed.c (store_bit_field_1): Likewise.
14396 * expr.c (store_constructor): Likewise.
14397 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14398 and polynomial REGMODE_NATURAL_SIZE.
14399 (gen_lowpart_common): Likewise.
14400 * reginfo.c (record_subregs_of_mode): Likewise.
14401 * rtlanal.c (read_modify_subreg_p): Likewise.
14403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14404 Alan Hayward <alan.hayward@arm.com>
14405 David Sherwood <david.sherwood@arm.com>
14407 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14408 numbers of elements.
14410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14411 Alan Hayward <alan.hayward@arm.com>
14412 David Sherwood <david.sherwood@arm.com>
14414 * match.pd: Cope with polynomial numbers of vector elements.
14416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14417 Alan Hayward <alan.hayward@arm.com>
14418 David Sherwood <david.sherwood@arm.com>
14420 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14421 in a POINTER_PLUS_EXPR.
14423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14424 Alan Hayward <alan.hayward@arm.com>
14425 David Sherwood <david.sherwood@arm.com>
14427 * omp-simd-clone.c (simd_clone_subparts): New function.
14428 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14429 (ipa_simd_modify_function_body): Likewise.
14431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14432 Alan Hayward <alan.hayward@arm.com>
14433 David Sherwood <david.sherwood@arm.com>
14435 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14436 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14437 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14438 (expand_vector_condition, vector_element): Likewise.
14439 (subparts_gt): New function.
14440 (get_compute_type): Use subparts_gt.
14441 (count_type_subparts): Delete.
14442 (expand_vector_operations_1): Use subparts_gt instead of
14443 count_type_subparts.
14445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14446 Alan Hayward <alan.hayward@arm.com>
14447 David Sherwood <david.sherwood@arm.com>
14449 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14450 (vect_compile_time_alias): ...this new function. Do the calculation
14451 on poly_ints rather than trees.
14452 (vect_prune_runtime_alias_test_list): Update call accordingly.
14454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14455 Alan Hayward <alan.hayward@arm.com>
14456 David Sherwood <david.sherwood@arm.com>
14458 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14460 (vect_schedule_slp_instance): Likewise.
14462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14463 Alan Hayward <alan.hayward@arm.com>
14464 David Sherwood <david.sherwood@arm.com>
14466 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14467 constant and extern definitions for variable-length vectors.
14468 (vect_get_constant_vectors): Note that the number of units
14469 is known to be constant.
14471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14472 Alan Hayward <alan.hayward@arm.com>
14473 David Sherwood <david.sherwood@arm.com>
14475 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14476 of units as polynomial. Choose between WIDE and NARROW based
14479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14480 Alan Hayward <alan.hayward@arm.com>
14481 David Sherwood <david.sherwood@arm.com>
14483 * tree-vect-stmts.c (simd_clone_subparts): New function.
14484 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
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-stmts.c (vectorizable_call): Treat the number of
14491 vectors as polynomial. Use build_index_vector for
14492 IFN_GOMP_SIMD_LANE.
14494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14495 Alan Hayward <alan.hayward@arm.com>
14496 David Sherwood <david.sherwood@arm.com>
14498 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14499 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14500 for variable-length vectors.
14501 (vectorizable_mask_load_store): Treat the number of units as
14502 polynomial, asserting that it is constant if the condition has
14503 already been enforced.
14504 (vectorizable_store, vectorizable_load): Likewise.
14506 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14507 Alan Hayward <alan.hayward@arm.com>
14508 David Sherwood <david.sherwood@arm.com>
14510 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14511 of units as polynomial. Punt if we can't tell at compile time
14512 which vector contains the final result.
14514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14515 Alan Hayward <alan.hayward@arm.com>
14516 David Sherwood <david.sherwood@arm.com>
14518 * tree-vect-loop.c (vectorizable_induction): Treat the number
14519 of units as polynomial. Punt on SLP inductions. Use an integer
14520 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14521 cast of such a series for variable-length floating-point
14524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14525 Alan Hayward <alan.hayward@arm.com>
14526 David Sherwood <david.sherwood@arm.com>
14528 * tree.h (build_index_vector): Declare.
14529 * tree.c (build_index_vector): New function.
14530 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14531 of units as polynomial, forcibly converting it to a constant if
14532 vectorizable_reduction has already enforced the condition.
14533 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14534 to create a {1,2,3,...} vector.
14535 (vectorizable_reduction): Treat the number of units as polynomial.
14536 Choose vectype_in based on the largest scalar element size rather
14537 than the smallest number of units. Enforce the restrictions
14540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14541 Alan Hayward <alan.hayward@arm.com>
14542 David Sherwood <david.sherwood@arm.com>
14544 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14545 number of units as polynomial.
14547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14548 Alan Hayward <alan.hayward@arm.com>
14549 David Sherwood <david.sherwood@arm.com>
14551 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14552 * target.def (autovectorize_vector_sizes): Return the vector sizes
14553 by pointer, using vector_sizes rather than a bitmask.
14554 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14555 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14556 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14558 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14559 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14560 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14561 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14562 * omp-general.c (omp_max_vf): Likewise.
14563 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14564 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14565 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14566 * tree-vect-slp.c (vect_slp_bb): Likewise.
14567 * doc/tm.texi: Regenerate.
14568 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14570 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14571 the vector size as a poly_uint64 rather than an unsigned int.
14572 (current_vector_size): Change from an unsigned int to a poly_uint64.
14573 (get_vectype_for_scalar_type): Update accordingly.
14574 * tree.h (build_truth_vector_type): Take the size and number of
14575 units as a poly_uint64 rather than an unsigned int.
14576 (build_vector_type): Add a temporary overload that takes
14577 the number of units as a poly_uint64 rather than an unsigned int.
14578 * tree.c (make_vector_type): Likewise.
14579 (build_truth_vector_type): Take the number of units as a poly_uint64
14580 rather than an unsigned int.
14582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14583 Alan Hayward <alan.hayward@arm.com>
14584 David Sherwood <david.sherwood@arm.com>
14586 * target.def (get_mask_mode): Take the number of units and length
14587 as poly_uint64s rather than unsigned ints.
14588 * targhooks.h (default_get_mask_mode): Update accordingly.
14589 * targhooks.c (default_get_mask_mode): Likewise.
14590 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14591 * doc/tm.texi: Regenerate.
14593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14594 Alan Hayward <alan.hayward@arm.com>
14595 David Sherwood <david.sherwood@arm.com>
14597 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14598 * omp-general.c (omp_max_vf): Likewise.
14599 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14600 (expand_omp_simd): Handle polynomial safelen.
14601 * omp-low.c (omplow_simd_context): Add a default constructor.
14602 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14603 (lower_rec_simd_input_clauses): Update accordingly.
14604 (lower_rec_input_clauses): Likewise.
14606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14607 Alan Hayward <alan.hayward@arm.com>
14608 David Sherwood <david.sherwood@arm.com>
14610 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14611 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14612 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14613 (vect_analyze_slp_cost): Likewise.
14614 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14615 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14622 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14623 from an unsigned int * to a poly_uint64_pod *.
14624 (calculate_unrolling_factor): New function.
14625 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14628 Alan Hayward <alan.hayward@arm.com>
14629 David Sherwood <david.sherwood@arm.com>
14631 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14632 from an unsigned int to a poly_uint64.
14633 (_loop_vec_info::slp_unrolling_factor): Likewise.
14634 (_loop_vec_info::vectorization_factor): Change from an int
14636 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14637 (vect_get_num_vectors): New function.
14638 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14639 (vect_get_num_copies): Use vect_get_num_vectors.
14640 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14641 to an unsigned int *.
14642 (vect_analyze_data_refs): Change min_vf from an int * to a
14644 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14645 than an unsigned HOST_WIDE_INT.
14646 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14647 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14648 to an unsigned int *.
14649 (vect_analyze_data_ref_dependences): Likewise.
14650 (vect_compute_data_ref_alignment): Handle polynomial vf.
14651 (vect_enhance_data_refs_alignment): Likewise.
14652 (vect_prune_runtime_alias_test_list): Likewise.
14653 (vect_shift_permute_load_chain): Likewise.
14654 (vect_supportable_dr_alignment): Likewise.
14655 (dependence_distance_ge_vf): Take the vectorization factor as a
14656 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14657 (vect_analyze_data_refs): Change min_vf from an int * to a
14659 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14660 vfm1 as a poly_uint64 rather than an int. Make the same change
14661 for the returned bound_scalar.
14662 (vect_gen_vector_loop_niters): Handle polynomial vf.
14663 (vect_do_peeling): Likewise. Update call to
14664 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14665 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14667 * tree-vect-loop.c (vect_determine_vectorization_factor)
14668 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14669 (vect_get_known_peeling_cost): Likewise.
14670 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14671 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14672 (vect_transform_loop): Likewise. Use the lowest possible VF when
14673 updating the upper bounds of the loop.
14674 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14675 rather than an int.
14676 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14677 polynomial unroll factors.
14678 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14679 (vect_make_slp_decision): Likewise.
14680 (vect_supported_load_permutation_p): Likewise, and polynomial
14682 (vect_analyze_slp_cost): Handle polynomial vf.
14683 (vect_slp_analyze_node_operations): Likewise.
14684 (vect_slp_analyze_bb_1): Likewise.
14685 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14686 than an unsigned HOST_WIDE_INT.
14687 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14688 (vectorizable_load): Handle polynomial vf.
14689 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14691 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14694 Alan Hayward <alan.hayward@arm.com>
14695 David Sherwood <david.sherwood@arm.com>
14697 * match.pd: Handle bit operations involving three constants
14698 and try to fold one pair.
14700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14702 * tree-vect-loop-manip.c: Include gimple-fold.h.
14703 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14704 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14705 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14706 Add a path that uses a step of VF instead of 1, but disable it
14708 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14709 and niters_no_overflow parameters. Update calls to
14710 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14711 Create a new SSA name if the latter choses to use a ste other
14712 than zero, and return it via niters_vector_mult_vf_var.
14713 * tree-vect-loop.c (vect_transform_loop): Update calls to
14714 vect_do_peeling, vect_gen_vector_loop_niters and
14715 slpeel_make_loop_iterate_ntimes.
14716 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14717 (vect_gen_vector_loop_niters): Update declarations after above changes.
14719 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14721 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14722 128-bit round to integer instructions.
14723 (ceil<mode>2): Likewise.
14724 (btrunc<mode>2): Likewise.
14725 (round<mode>2): Likewise.
14727 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14729 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14730 unaligned VSX load/store on P8/P9.
14731 (expand_block_clear): Allow the use of unaligned VSX
14732 load/store on P8/P9.
14734 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14736 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14738 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14739 swap associated with both a load and a store.
14741 2018-01-02 Andrew Waterman <andrew@sifive.com>
14743 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14744 * config/riscv/riscv.md (clear_cache): Use it.
14746 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14748 * web.c: Remove out-of-date comment.
14750 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14752 * expr.c (fixup_args_size_notes): Check that any existing
14753 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14754 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14755 (emit_single_push_insn): ...here.
14757 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14759 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14760 (const_vector_encoded_nelts): New function.
14761 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14762 (const_vector_int_elt, const_vector_elt): Declare.
14763 * emit-rtl.c (const_vector_int_elt_1): New function.
14764 (const_vector_elt): Likewise.
14765 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14766 of CONST_VECTOR_ELT.
14768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14770 * expr.c: Include rtx-vector-builder.h.
14771 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14772 directly on the tree encoding.
14773 (const_vector_from_tree): Likewise.
14774 * optabs.c: Include rtx-vector-builder.h.
14775 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14776 sequence of "u" values.
14777 * vec-perm-indices.c: Include rtx-vector-builder.h.
14778 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14779 directly on the vec_perm_indices encoding.
14781 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14783 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14784 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14785 * rtx-vector-builder.h: New file.
14786 * rtx-vector-builder.c: Likewise.
14787 * rtl.h (rtx_def::u2): Add a const_vector field.
14788 (CONST_VECTOR_NPATTERNS): New macro.
14789 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14790 (CONST_VECTOR_DUPLICATE_P): Likewise.
14791 (CONST_VECTOR_STEPPED_P): Likewise.
14792 (CONST_VECTOR_ENCODED_ELT): Likewise.
14793 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14794 (unwrap_const_vec_duplicate): Likewise.
14795 (const_vec_series_p): Check for a non-duplicated vector encoding.
14796 Say that the function only returns true for integer vectors.
14797 * emit-rtl.c: Include rtx-vector-builder.h.
14798 (gen_const_vec_duplicate_1): Delete.
14799 (gen_const_vector): Call gen_const_vec_duplicate instead of
14800 gen_const_vec_duplicate_1.
14801 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14802 (gen_const_vec_duplicate): Use rtx_vector_builder.
14803 (gen_const_vec_series): Likewise.
14804 (gen_rtx_CONST_VECTOR): Likewise.
14805 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14806 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14807 Build a new vector rather than modifying a CONST_VECTOR in-place.
14808 (handle_special_swappables): Update call accordingly.
14809 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14810 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14811 Build a new vector rather than modifying a CONST_VECTOR in-place.
14812 (handle_special_swappables): Update call accordingly.
14814 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14816 * simplify-rtx.c (simplify_const_binary_operation): Use
14817 CONST_VECTOR_ELT instead of XVECEXP.
14819 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14821 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14822 the selector elements to be different from the data elements
14823 if the selector is a VECTOR_CST.
14824 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14825 ssizetype for the selector.
14827 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14829 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14830 before testing each element individually.
14831 * tree-vect-generic.c (lower_vec_perm): Likewise.
14833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14835 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14836 * selftest-run-tests.c (selftest::run_tests): Call it.
14837 * vector-builder.h (vector_builder::operator ==): New function.
14838 (vector_builder::operator !=): Likewise.
14839 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14840 (vec_perm_indices::all_from_input_p): New function.
14841 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14842 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14843 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14844 instead of reading the VECTOR_CST directly. Detect whether both
14845 vector inputs are the same before constructing the vec_perm_indices,
14846 and update the number of inputs argument accordingly. Use the
14847 utility functions added above. Only construct sel2 if we need to.
14849 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14851 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14852 the broadcast of the low byte.
14853 (expand_mult_highpart): Use an explicit encoding for the permutes.
14854 * optabs-query.c (can_mult_highpart_p): Likewise.
14855 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14856 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14857 (vectorizable_bswap): Likewise.
14858 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14859 explicit encoding for the power-of-2 permutes.
14860 (vect_permute_store_chain): Likewise.
14861 (vect_grouped_load_supported): Likewise.
14862 (vect_permute_load_chain): Likewise.
14864 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14866 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14867 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14868 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14869 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14870 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14871 (vect_gen_perm_mask_any): Likewise.
14873 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14875 * int-vector-builder.h: New file.
14876 * vec-perm-indices.h: Include int-vector-builder.h.
14877 (vec_perm_indices): Redefine as an int_vector_builder.
14878 (auto_vec_perm_indices): Delete.
14879 (vec_perm_builder): Redefine as a stand-alone class.
14880 (vec_perm_indices::vec_perm_indices): New function.
14881 (vec_perm_indices::clamp): Likewise.
14882 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14883 (vec_perm_indices::new_vector): New function.
14884 (vec_perm_indices::new_expanded_vector): Update for new
14885 vec_perm_indices class.
14886 (vec_perm_indices::rotate_inputs): New function.
14887 (vec_perm_indices::all_in_range_p): Operate directly on the
14888 encoded form, without computing elided elements.
14889 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14890 encoding. Update for new vec_perm_indices class.
14891 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14892 the given vec_perm_builder.
14893 (expand_vec_perm_var): Update vec_perm_builder constructor.
14894 (expand_mult_highpart): Use vec_perm_builder instead of
14895 auto_vec_perm_indices.
14896 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14897 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14898 or double series encoding as appropriate.
14899 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14900 vec_perm_indices instead of auto_vec_perm_indices.
14901 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14902 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14903 (vect_permute_store_chain): Likewise.
14904 (vect_grouped_load_supported): Likewise.
14905 (vect_permute_load_chain): Likewise.
14906 (vect_shift_permute_load_chain): Likewise.
14907 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14908 (vect_transform_slp_perm_load): Likewise.
14909 (vect_schedule_slp_instance): Likewise.
14910 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14911 (vectorizable_mask_load_store): Likewise.
14912 (vectorizable_bswap): Likewise.
14913 (vectorizable_store): Likewise.
14914 (vectorizable_load): Likewise.
14915 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14916 vec_perm_indices instead of auto_vec_perm_indices. Use
14917 tree_to_vec_perm_builder to read the vector from a tree.
14918 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14919 vec_perm_builder instead of a vec_perm_indices.
14920 (have_whole_vector_shift): Use vec_perm_builder and
14921 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14922 truncation to calc_vec_perm_mask_for_shift.
14923 (vect_create_epilog_for_reduction): Likewise.
14924 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14925 from auto_vec_perm_indices to vec_perm_indices.
14926 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14927 instead of changing individual elements.
14928 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14929 the vector in d.perm.
14930 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14931 from auto_vec_perm_indices to vec_perm_indices.
14932 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14933 instead of changing individual elements.
14934 (arm_vectorize_vec_perm_const): Use new_vector to install
14935 the vector in d.perm.
14936 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14937 Update vec_perm_builder constructor.
14938 (rs6000_expand_interleave): Likewise.
14939 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14940 (rs6000_expand_interleave): Likewise.
14942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14944 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14945 to qimode could truncate the indices.
14946 * optabs.c (expand_vec_perm_var): Likewise.
14948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14950 * Makefile.in (OBJS): Add vec-perm-indices.o.
14951 * vec-perm-indices.h: New file.
14952 * vec-perm-indices.c: Likewise.
14953 * target.h (vec_perm_indices): Replace with a forward class
14955 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14956 * optabs.h: Include vec-perm-indices.h.
14957 (expand_vec_perm): Delete.
14958 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14959 (expand_vec_perm_const): Declare.
14960 * target.def (vec_perm_const_ok): Replace with...
14961 (vec_perm_const): ...this new hook.
14962 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14963 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14964 * doc/tm.texi: Regenerate.
14965 * optabs.def (vec_perm_const): Delete.
14966 * doc/md.texi (vec_perm_const): Likewise.
14967 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14968 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14969 expand_vec_perm for constant permutation vectors. Assert that
14970 the mode of variable permutation vectors is the integer equivalent
14971 of the mode that is being permuted.
14972 * optabs-query.h (selector_fits_mode_p): Declare.
14973 * optabs-query.c: Include vec-perm-indices.h.
14974 (selector_fits_mode_p): New function.
14975 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14976 is defined, instead of checking whether the vec_perm_const_optab
14977 exists. Use targetm.vectorize.vec_perm_const instead of
14978 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14979 fit in the vector mode before using a variable permute.
14980 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14981 vec_perm_indices instead of an rtx.
14982 (expand_vec_perm): Replace with...
14983 (expand_vec_perm_const): ...this new function. Take the selector
14984 as a vec_perm_indices rather than an rtx. Also take the mode of
14985 the selector. Update call to shift_amt_for_vec_perm_mask.
14986 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14987 Use vec_perm_indices::new_expanded_vector to expand the original
14988 selector into bytes. Check whether the indices fit in the vector
14989 mode before using a variable permute.
14990 (expand_vec_perm_var): Make global.
14991 (expand_mult_highpart): Use expand_vec_perm_const.
14992 * fold-const.c: Includes vec-perm-indices.h.
14993 * tree-ssa-forwprop.c: Likewise.
14994 * tree-vect-data-refs.c: Likewise.
14995 * tree-vect-generic.c: Likewise.
14996 * tree-vect-loop.c: Likewise.
14997 * tree-vect-slp.c: Likewise.
14998 * tree-vect-stmts.c: Likewise.
14999 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15001 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15002 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15003 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15004 (aarch64_vectorize_vec_perm_const): ...this new function.
15005 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15006 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15007 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15008 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15009 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15011 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15013 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15014 check for NEON modes.
15015 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15016 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15017 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15018 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15020 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15021 the old VEC_PERM_CONST conditions.
15022 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15023 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15024 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15025 (ia64_vectorize_vec_perm_const_ok): Merge into...
15026 (ia64_vectorize_vec_perm_const): ...this new function.
15027 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15028 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15029 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15030 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15031 * config/mips/mips.c (mips_expand_vec_perm_const)
15032 (mips_vectorize_vec_perm_const_ok): Merge into...
15033 (mips_vectorize_vec_perm_const): ...this new function.
15034 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15035 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15036 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15037 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15038 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15039 (rs6000_expand_vec_perm_const): Delete.
15040 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15042 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15043 (altivec_expand_vec_perm_const_le): Take each operand individually.
15044 Operate on constant selectors rather than rtxes.
15045 (altivec_expand_vec_perm_const): Likewise. Update call to
15046 altivec_expand_vec_perm_const_le.
15047 (rs6000_expand_vec_perm_const): Delete.
15048 (rs6000_vectorize_vec_perm_const_ok): Delete.
15049 (rs6000_vectorize_vec_perm_const): New function.
15050 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15051 an element count and rtx array.
15052 (rs6000_expand_extract_even): Update call accordingly.
15053 (rs6000_expand_interleave): Likewise.
15054 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15055 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15056 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15057 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15058 (rs6000_expand_vec_perm_const): Delete.
15059 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15060 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15061 (altivec_expand_vec_perm_const_le): Take each operand individually.
15062 Operate on constant selectors rather than rtxes.
15063 (altivec_expand_vec_perm_const): Likewise. Update call to
15064 altivec_expand_vec_perm_const_le.
15065 (rs6000_expand_vec_perm_const): Delete.
15066 (rs6000_vectorize_vec_perm_const_ok): Delete.
15067 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15068 reference to the SPE evmerge intructions.
15069 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15070 an element count and rtx array.
15071 (rs6000_expand_extract_even): Update call accordingly.
15072 (rs6000_expand_interleave): Likewise.
15073 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15074 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15076 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15078 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15080 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15081 vector mode and that that mode matches the mode of the data
15083 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15084 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15085 directly using expand_vec_perm_1 when forcing selectors into
15087 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15089 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15091 * optabs-query.h (can_vec_perm_p): Delete.
15092 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15093 * optabs-query.c (can_vec_perm_p): Split into...
15094 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15095 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15096 particular selector is valid.
15097 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15098 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15099 (vect_grouped_load_supported): Likewise.
15100 (vect_shift_permute_load_chain): Likewise.
15101 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15102 (vect_transform_slp_perm_load): Likewise.
15103 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15104 (vectorizable_bswap): Likewise.
15105 (vect_gen_perm_mask_checked): Likewise.
15106 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15107 implementations of variable permutation vectors into account
15108 when deciding which selector to use.
15109 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15110 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15111 with a false third argument.
15112 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15113 to test whether the constant selector is valid and can_vec_perm_var_p
15114 to test whether a variable selector is valid.
15116 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15118 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15119 * optabs-query.c (can_vec_perm_p): Likewise.
15120 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15121 instead of vec_perm_indices.
15122 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15123 (vect_gen_perm_mask_checked): Likewise,
15124 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15125 (vect_gen_perm_mask_checked): Likewise,
15127 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15129 * optabs-query.h (qimode_for_vec_perm): Declare.
15130 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15131 (qimode_for_vec_perm): ...this new function.
15132 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15134 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15136 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15137 does not have a conditional at the top.
15139 2018-01-02 Richard Biener <rguenther@suse.de>
15141 * ipa-inline.c (big_speedup_p): Fix expression.
15143 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15146 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15149 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15153 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15154 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15155 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15156 cond_taken_branch_cost 3->4.
15158 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15160 PR tree-optimization/83581
15161 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15162 TODO_cleanup_cfg if any changes have been made.
15164 PR middle-end/83608
15165 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15166 convert_modes if target mode has the right side, but different mode
15169 PR middle-end/83609
15170 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15171 last argument when extracting from CONCAT. If either from_real or
15172 from_imag is NULL, use expansion through memory. If result is not
15173 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15174 the parts directly to inner mode, if even that fails, use expansion
15177 PR middle-end/83623
15178 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15179 check for bswap in mode rather than HImode and use that in expand_unop
15182 Copyright (C) 2018 Free Software Foundation, Inc.
15184 Copying and distribution of this file, with or without modification,
15185 are permitted in any medium without royalty provided the copyright
15186 notice and this notice are preserved.