1 2018-06-04 Jim Wilson <jimw@sifive.com>
3 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
5 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
6 and maybe_eh_return. Change regno to unsigned int. Use new args to
7 handle EH_RETURN_DATA_REGNO registers properly.
8 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
9 (riscv_expand_epilogue): Update comment. Change argument name and
10 type. Update code to use new name and type. Pass new args to
11 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
13 * config/riscv/riscv.md (NORMAL_RETURN): New.
14 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
15 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
16 (eh_return): Call gen_eh_return_internal and emit barrier.
17 (eh_return_internal): Call riscv_expand_epilogue.
19 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
21 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
22 bit_insertion field and declare can_be_merged_into method.
23 (merged_store_group::can_be_merged_into): New method.
24 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
25 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
26 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
28 2018-06-04 Richard Biener <rguenther@suse.de>
30 PR tree-optimization/85955
31 * builtins.c (fold_builtin_sincos): Convert pointers to
32 destination to appropriate type before dereferencing.
34 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
36 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
38 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
40 * expr.c (expand_expr_real_1): Force the operand into memory if
41 its TYPE_MODE is BLKmode and if there is no integer mode for
42 the number of bits being extracted.
44 2018-06-04 Jakub Jelinek <jakub@redhat.com>
48 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
49 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
51 2018-06-04 Richard Biener <rguenther@suse.de>
53 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
54 (cleanup_tree_cfg_noloop): ... single caller. Do
55 start_recording_case_labels later.
57 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
59 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
60 to _IMMINTRIN_H_INCLUDED.
61 * config/i386/pconfigintrin.h: Ditto.
62 * config/i386/waitpkgintrin.h: Ditto.
63 * config/i386/immintrin.h: Add includes for sgxintrin.h,
64 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
65 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
66 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
67 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
68 waitpkgintrin.h and cldemoteintrin.h.
70 2018-06-04 Richard Biener <rguenther@suse.de>
72 PR tree-optimization/86038
73 * tracer.c (find_best_successor): Check probability for
74 being initialized, bail out if not.
76 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
79 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
80 of bits to ignore when comparing architectures.
82 2018-06-04 Jakub Jelinek <jakub@redhat.com>
84 PR tree-optimization/69615
85 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
86 maximum or minimum of the type, try to merge it also as if
87 range1 is + [-, x - 1] or + [x + 1, -].
90 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
92 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
94 PR tree-optimization/86034
95 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
96 the unsigned bitfield type in a bit insertion sequence if it does not
97 have a larger precision than the bitfield size.
98 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
100 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
102 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
104 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
106 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
107 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
108 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
109 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
111 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
113 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
114 Disable -fdelete-null-pointer-checks for ELF toolchain.
116 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
117 Kito Cheng <kito.cheng@gmail.com>
119 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
120 (nds32le-*-*, nds32be-*-*): Integrate checking process.
121 (nds32*-*-*): Add glibc and uclibc conditions.
122 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
123 (TARGET_EXCEPT_UNWIND_INFO): Define.
124 * config/nds32/elf.h: New file.
125 * config/nds32/linux.h: New file.
126 * config/nds32/nds32-elf.opt: New file.
127 * config/nds32/nds32-linux.opt: New file.
128 * config/nds32/nds32-fp-as-gp.c
129 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
130 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
132 (nds32_asm_file_end): Ditto.
133 (nds32_print_operand): Ditto.
134 (nds32_insert_attributes): Ditto.
135 (nds32_init_libfuncs): New function.
136 (TARGET_HAVE_TLS): Define.
137 (TARGET_INIT_LIBFUNCS): Define.
138 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
140 (TARGET_ELF): Apply different mcmodel setting.
141 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
142 been migrated into elf.h and linux.h files.
143 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
144 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
145 (mcmodel): The content has been migrated into nds32-elf.opt and
146 nds32-linux.opt files.
147 * config/nds32/t-elf: New file.
148 * config/nds32/t-linux: New file.
150 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
151 Shiva Chen <shiva0217@gmail.com>
153 * config/nds32/constants.md (unspec_volatile_element): Add
154 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
155 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
157 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
158 (make_pass_nds32_fp_as_gp): Declare.
159 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
161 (nds32_asm_function_end_prologue): Remove unused asm output.
162 (nds32_asm_function_begin_epilogue): Remove unused asm output.
163 (nds32_asm_file_start): Output necessary fp_as_gp information.
164 (nds32_option_override): Adjust register usage.
165 (nds32_expand_prologue): Consider fp_as_gp situation.
166 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
167 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
170 (simple_return): Ditto.
171 (omit_fp_begin): Output special directive for fp_as_gp.
172 (omit_fp_end): Output special directive for fp_as_gp.
173 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
174 mforbid-fp-as-gp): New options.
176 2018-06-01 Mark Wielaard <mark@klomp.org>
178 * dwarf2out.c (dwarf2out_finish): Remove generation of
181 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
183 * gimple-ssa-store-merging.c: Include gimple-fold.h.
184 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
185 (struct merged_store_group): Add bit_insertion field.
186 (dump_char_array): Use standard hexadecimal format.
187 (merged_store_group::merged_store_group): Set bit_insertion to false.
188 (merged_store_group::apply_stores): Use optimal buffer size. Deal
189 with BIT_INSERT_EXPR stores. Move up code updating the mask and
190 also print the mask in the dump file.
191 (pass_store_merging::gate): Minor tweak.
192 (imm_store_chain_info::coalesce_immediate): Fix wrong association
193 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
194 stores with INTEGER_CST stores.
195 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
196 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
197 and use it throughout. Generate bit insertion sequences if need be.
198 (pass_store_merging::process_store): Remove redundant condition.
199 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
201 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
203 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
204 the 128-bit floating point types. Fix function comment.
206 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
208 * config/aarch64/aarch64-simd.md
209 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
211 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
214 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
216 PR tree-optimization/85989
217 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
219 (backprop::intersect_uses): Check it when deciding whether this
220 is a backedge reference.
221 (backprop::process_block): Add each phi to m_visited_phis
222 after visiting it, then clear it at the end.
224 2018-06-01 Richard Biener <rguenther@suse.de>
226 * tree-vectorizer.h (vect_dr_stmt): New function.
227 (vect_get_load_cost): Adjust.
228 (vect_get_store_cost): Likewise.
229 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
230 Use vect_dr_stmt instead of DR_SMTT.
231 (vect_record_base_alignments): Likewise.
232 (vect_calculate_target_alignment): Likewise.
233 (vect_compute_data_ref_alignment): Likewise and make static.
234 (vect_update_misalignment_for_peel): Likewise.
235 (vect_verify_datarefs_alignment): Likewise.
236 (vector_alignment_reachable_p): Likewise.
237 (vect_get_data_access_cost): Likewise. Pass down
238 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
239 (vect_get_peeling_costs_all_drs): Likewise.
240 (vect_peeling_hash_get_lowest_cost): Likewise.
241 (vect_enhance_data_refs_alignment): Likewise.
242 (vect_find_same_alignment_drs): Likewise.
243 (vect_analyze_data_refs_alignment): Likewise.
244 (vect_analyze_group_access_1): Likewise.
245 (vect_analyze_group_access): Likewise.
246 (vect_analyze_data_ref_access): Likewise.
247 (vect_analyze_data_ref_accesses): Likewise.
248 (vect_vfa_segment_size): Likewise.
249 (vect_small_gap_p): Likewise.
250 (vectorizable_with_step_bound_p): Likewise.
251 (vect_prune_runtime_alias_test_list): Likewise.
252 (vect_analyze_data_refs): Likewise.
253 (vect_supportable_dr_alignment): Likewise.
254 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
255 (vect_gen_prolog_loop_niters): Likewise.
256 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
257 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
259 (vect_recog_mask_conversion_pattern): Likewise.
260 (vect_try_gather_scatter_pattern): Likewise.
261 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
262 to vect_get_store_cost.
263 (vect_get_store_cost): Get stmt_info instead of DR.
264 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
265 (vect_get_load_cost): Get stmt_info instead of DR.
267 2018-06-01 Richard Biener <rguenther@suse.de>
270 * gimple-fold.c (var_decl_component_p): Also allow offsetted
271 vars wrapped in MEM_REFs.
273 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
275 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
276 Fix subreg tests so that we only return a choice between
277 GENERAL_REGS and FP_REGS if the original classes included both.
279 2018-06-01 Richard Biener <rguenther@suse.de>
282 * tree-ssa-structalias.c (get_function_part_constraint):
283 Handle NULL fi->decl.
284 (find_func_aliases_for_call): Properly handle indirect
286 (find_func_clobbers): Likewise.
287 (ipa_pta_execute): Likewise.
288 (create_variable_info_for): For functions that are ifunc_resolver
289 resolve to a varinfo that contains the result of the resolver
291 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
294 2018-05-31 Michael Collison <michael.collison@arm.com>
296 * config/aarch64/aarch64.md:
297 (*fix_to_zero_extenddfdi2): New pattern.
298 * gcc.target/aarch64/fix_extend1.c: New testcase.
300 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
304 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
305 and BUILT_IN_STRNCMP_EQ.
306 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
308 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
309 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
310 (gimple_fold_builtin): Likewise.
311 * tree-ssa-strlen.c (compute_string_length): New function.
312 (determine_min_obsize): New function.
313 (handle_builtin_string_cmp): New function to handle calls to
314 string compare functions.
315 (strlen_optimize_stmt): Add handling to builtin string compare
317 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
318 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
319 * tree.c (build_common_builtin_nodes): Add new defines of
320 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
322 2018-05-31 Jakub Jelinek <jakub@redhat.com>
325 * bb-reorder.c (pass_partition_blocks::gate): Return false for
326 functions with naked attribute.
328 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
330 * config/i386/sse.md (avx_vec_concat<mode>):
331 Substitute concat_tg_mode mode attribute with xtg_mode.
332 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
333 (concat_tg_mode): Remove mode attribute.
335 2018-05-31 Martin Sebor <msebor@redhat.com>
338 * calls.c (alloc_max_size): Correct a logic error/typo.
339 Treat excessive arguments as infinite. Warn for invalid arguments.
340 * doc/invoke.texi (-Walloc-size-larger-than): Update.
342 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
345 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
346 and movx for Haswell.
348 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
349 Cesar Philippidis <cesar@codesourcery.com>
352 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
353 when emitting error on private/firstprivate reductions.
354 * omp-low.c (lower_omp_target): Avoid reference-type processing
355 on pointers for firstprivate clause.
357 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
359 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
362 * config/aarch64/aarch64-simd.md
363 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
364 (aarch64_ld1_x3_<mode>): Likewise
365 (aarch64_st1x2<VALLDIF:mode>): Likewise
366 (aarch64_st1_x2_<mode>): Likewise
367 (aarch64_st1x3<VALLDIF:mode>): Likewise
368 (aarch64_st1_x3_<mode>): Likewise
369 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
370 (vld1_s8_x3): Likewise.
371 (vld1_u16_x3): Likewise.
372 (vld1_s16_x3): Likewise.
373 (vld1_u32_x3): Likewise.
374 (vld1_s32_x3): Likewise.
375 (vld1_u64_x3): Likewise.
376 (vld1_s64_x3): Likewise.
377 (vld1_f16_x3): Likewise.
378 (vld1_f32_x3): Likewise.
379 (vld1_f64_x3): Likewise.
380 (vld1_p8_x3): Likewise.
381 (vld1_p16_x3): Likewise.
382 (vld1_p64_x3): Likewise.
383 (vld1q_u8_x3): Likewise.
384 (vld1q_s8_x3): Likewise.
385 (vld1q_u16_x3): Likewise.
386 (vld1q_s16_x3): Likewise.
387 (vld1q_u32_x3): Likewise.
388 (vld1q_s32_x3): Likewise.
389 (vld1q_u64_x3): Likewise.
390 (vld1q_s64_x3): Likewise.
391 (vld1q_f16_x3): Likewise.
392 (vld1q_f32_x3): Likewise.
393 (vld1q_f64_x3): Likewise.
394 (vld1q_p8_x3): Likewise.
395 (vld1q_p16_x3): Likewise.
396 (vld1q_p64_x3): Likewise.
397 (vst1_s64_x2): Likewise.
398 (vst1_u64_x2): Likewise.
399 (vst1_f64_x2): Likewise.
400 (vst1_s8_x2): Likewise.
401 (vst1_p8_x2): Likewise.
402 (vst1_s16_x2): Likewise.
403 (vst1_p16_x2): Likewise.
404 (vst1_s32_x2): Likewise.
405 (vst1_u8_x2): Likewise.
406 (vst1_u16_x2): Likewise.
407 (vst1_u32_x2): Likewise.
408 (vst1_f16_x2): Likewise.
409 (vst1_f32_x2): Likewise.
410 (vst1_p64_x2): Likewise.
411 (vst1q_s8_x2): Likewise.
412 (vst1q_p8_x2): Likewise.
413 (vst1q_s16_x2): Likewise.
414 (vst1q_p16_x2): Likewise.
415 (vst1q_s32_x2): Likewise.
416 (vst1q_s64_x2): Likewise.
417 (vst1q_u8_x2): Likewise.
418 (vst1q_u16_x2): Likewise.
419 (vst1q_u32_x2): Likewise.
420 (vst1q_u64_x2): Likewise.
421 (vst1q_f16_x2): Likewise.
422 (vst1q_f32_x2): Likewise.
423 (vst1q_f64_x2): Likewise.
424 (vst1q_p64_x2): Likewise.
425 (vst1_s64_x3): Likewise.
426 (vst1_u64_x3): Likewise.
427 (vst1_f64_x3): Likewise.
428 (vst1_s8_x3): Likewise.
429 (vst1_p8_x3): Likewise.
430 (vst1_s16_x3): Likewise.
431 (vst1_p16_x3): Likewise.
432 (vst1_s32_x3): Likewise.
433 (vst1_u8_x3): Likewise.
434 (vst1_u16_x3): Likewise.
435 (vst1_u32_x3): Likewise.
436 (vst1_f16_x3): Likewise.
437 (vst1_f32_x3): Likewise.
438 (vst1_p64_x3): Likewise.
439 (vst1q_s8_x3): Likewise.
440 (vst1q_p8_x3): Likewise.
441 (vst1q_s16_x3): Likewise.
442 (vst1q_p16_x3): Likewise.
443 (vst1q_s32_x3): Likewise.
444 (vst1q_s64_x3): Likewise.
445 (vst1q_u8_x3): Likewise.
446 (vst1q_u16_x3): Likewise.
447 (vst1q_u32_x3): Likewise.
448 (vst1q_u64_x3): Likewise.
449 (vst1q_f16_x3): Likewise.
450 (vst1q_f32_x3): Likewise.
451 (vst1q_f64_x3): Likewise.
452 (vst1q_p64_x3): Likewise.
454 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
456 * config/msp430/msp430.c (msp430_output_labelref): Prepend
457 user_label_prefix to name.
459 * tree-core.h: Update comment about the format of NAME string
460 passed to handler in attribute_spec.
462 * config/msp430/msp430.md: Remove erroneous subreg expression from
463 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
464 zero_extend{q,h}isi2.
466 2018-05-30 Borislav Petkov <bp@suse.de>
468 * doc/extend.texi: Document some architecture specific
469 constraints and sort entries.
471 2018-05-30 Martin Sebor <msebor@redhat.com>
474 * builtins.c (expand_builtin_stpcpy_1): New function.
475 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
476 only if the former succeeds.
478 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
480 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
483 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
485 * doc/invoke.texi (-flinker-output): Document
487 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
489 * passes.c (ipa_write_summaries): Only modify statements if body
491 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
492 incrementally linking.
493 (ipa_passes): Likewise.
494 * lto-cgraph.c (lto_output_node): When incrementally linking do not
495 pass down resolution info.
496 * common.opt (flag_incremental_link): Update info.
497 * gcc.c (plugin specs): Turn flinker-output=* to
498 -plugin-opt=-linker-output-known
499 * toplev.c (compile_file): Also cut compilation when doing incremental
501 * flag-types. (enum lto_partition_model): Add
502 LTO_LINKER_OUTPUT_NOLTOREL.
503 (invoke.texi): Add -flinker-output docs.
504 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
505 link same way as WPA; do not stream in dead initializers.
507 * dwarf2out.c (dwarf2out_die_ref_for_decl,
508 darf2out_register_external_decl): Support incremental link.
510 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
512 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
514 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
516 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
517 it down to simple_object_copy_lto_debug_sections.
518 (run_gcc): Determine incremental LTO link time and configure
519 lto1 into non-wpa mode, disable renaming of debug sections.
521 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
523 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
524 descriptions of various incorrectly documented functions.
526 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
529 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
530 address check not strict.
532 2018-05-30 Richard Biener <rguenther@suse.de>
534 PR tree-optimization/85964
535 * tracer.c (better_p): Drop initialized count check, we only
536 call the function with initialized counts now.
537 (find_best_successor): Do find a best edge if one
538 has uninitialized count.
539 (find_best_predecessor): Likewise. Do BB frequency check only
540 if count is initialized.
542 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
544 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
545 (aarch64_ldrstr_offset_compare): New.
546 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
547 load/store orderings.
548 (aarch64_gen_adjusted_ldpstp): Likewise.
550 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
552 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
553 Check for subset of GENERAL_REGS and FP_REGS.
554 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
557 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
559 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
560 and wi::to_poly_offset. Add the current offset and then check
561 whether the sum fits, rather than using an unchecked addition of
562 a checked term. Check for a shwi rather than a uhwi.
563 * expr.c (get_bit_range): Use tree_to_poly_uint64.
564 (store_constructor): Use poly_int_tree_p.
565 (expand_expr_real_1): Likewise.
566 * function.c (assign_temp): Likewise.
567 * fold-const.c (const_binop): Use poly_int_tree_p and
569 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
571 * ipa-icf-gimple.c (func_checker::compare_operand): Use
572 to_poly_offset for MEM offsets.
573 * ipa-icf.c (sem_variable::equals): Likewise.
574 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
575 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
576 wi::to_poly_offset for BIT_FIELD_REF offsets.
577 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
579 * var-tracking.c (emit_note_insn_var_location): Use
582 2018-05-29 Jim Wilson <jimw@sifive.com>
584 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
586 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
589 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
590 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
592 (sse4_1_round<mode>2): Use nonimmediate_operand
593 for operand 1 predicate.
595 2018-05-29 Martin Sebor <msebor@redhat.com>
596 Richard Biener <rguenther@suse.de>
599 * calls.c (get_size_range): Call determine_value_range instead
601 * tree-vrp.h (determine_value_range): Declared new function.
602 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
604 2018-05-29 Richard Biener <rguenther@suse.de>
606 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
607 sure to use non-pattern stmts for get_earlier_stmt arguments.
608 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
609 called on pattern stmts.
610 (get_later_stmt): Likewise.
612 2018-05-29 Martin Liska <mliska@suse.cz>
614 PR gcov-profile/85759
615 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
618 2018-05-29 Jakub Jelinek <jakub@redhat.com>
620 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
622 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
626 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
627 VEC_PACK_FLOAT_EXPR): New tree codes.
628 * tree-pretty-print.c (op_code_prio): Handle
629 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
630 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
631 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
632 * tree-inline.c (estimate_operator_cost): Likewise.
633 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
634 * fold-const.c (const_binop): Likewise.
635 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
636 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
637 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
638 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
639 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
640 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
641 * expr.c (expand_expr_real_2): Likewise.
642 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
643 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
644 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
646 * optabs.c (expand_widen_pattern_expr): For
647 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
648 sign from result type rather than operand's type.
649 (expand_binop_directly): For vec_packu_float_optab and
650 vec_packs_float_optab allow result type to be different from operand's
652 * optabs-tree.c (optab_for_tree_code): Handle
653 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
654 VEC_PACK_FLOAT_EXPR. Formatting fixes.
655 * tree-vect-generic.c (expand_vector_operations_1): Handle
656 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
658 * tree-vect-stmts.c (supportable_widening_operation): Handle
660 (supportable_narrowing_operation): Handle FLOAT_EXPR.
661 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
662 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
663 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
664 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
666 (vec_pack<floatprefix>_float_<mode>): New expander.
667 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
669 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
670 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
671 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
672 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
673 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
675 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
676 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
677 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
678 VEC_PACK_FLOAT_EXPR): Document.
680 2018-05-29 Richard Biener <rguenther@suse.de>
682 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
684 (stmt_vec_info_vec): Make pointer.
685 (init_stmt_vec_info_vec): Remove.
686 (free_stmt_vec_info_vec): Likewise.
687 (set_stmt_vec_info_vec): New function.
688 (free_stmt_vec_infos): Likewise.
689 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
690 (set_vinfo_for_stmt): Likewise.
691 (get_earlier_stmt): Likewise.
692 (get_later_stmt): Likewise.
693 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
694 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
695 (vec_info::~vec_info): Free stmt_vec_infos.
696 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
697 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
698 (pass_slp_vectorize::execute): Likewise.
699 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
700 (free_stmt_vec_info_vec): Likewise.
701 (set_stmt_vec_info_vec): New function.
702 (free_stmt_vec_infos): Likewise.
703 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
704 the global stmt_vec_info_vec.
705 * tree-parloops.c (gather_scalar_reductions): Use
706 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
709 2018-05-29 Richard Biener <rguenther@suse.de>
711 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
713 2018-05-29 Martin Liska <mliska@suse.cz>
714 David Malcolm <dmalcolm@redhat.com>
716 * vec.c (test_reverse): New.
717 (vec_c_tests): Add new test.
718 * vec.h (vl_ptr>::reverse): New function.
720 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
722 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
724 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
727 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
729 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
731 2018-05-28 Richard Biener <rguenther@suse.de>
733 PR tree-optimization/85933
734 * tree-vect-data-refs.c (vect_record_base_alignments): Only
735 look at stmts marked as vectorizable.
737 2018-05-28 Richard Biener <rguenther@suse.de>
739 PR tree-optimization/85934
740 * tree-vect-generic.c (expand_vector_operations_1): Hoist
741 vector boolean check before scalar optimization.
743 2018-05-28 Jakub Jelinek <jakub@redhat.com>
745 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
748 2018-05-28 Mark Wielaard <mark@klomp.org>
750 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
751 if it is an expression containing a minus sign.
753 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
755 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
757 2018-05-27 Paul Koning <ni1d@arrl.net>
759 * config/pdp11/pdp11.md (truncsihi2): Remove.
761 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
762 Chung-Ju Wu <jasonwucj@gmail.com>
764 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
766 (unaligned_store_dw): Ditto.
767 * config/nds32/nds32-memory-manipulation.c
768 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
769 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
770 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
771 (emit_setmem_word_loop): Rename to ...
772 (emit_setmem_doubleword_loop): ... this.
773 (nds32_gen_dup_4_byte_to_word_value): New function.
774 (nds32_gen_dup_8_byte_to_double_word_value): New function.
775 (nds32_expand_setmem_loop): Refine implementation.
776 (nds32_expand_setmem_loop_v3m): Ditto.
777 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
780 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
782 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
784 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
786 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
787 (nds32_init_machine_status): Initialize machine->attr_naked_p and
788 machine->attr_no_prologue_p.
789 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
790 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
791 (nds32_expand_epilogue): Consider attr_naked_p.
792 (nds32_expand_epilogue_v3pop): Likewise.
793 (nds32_can_use_return_insn): Likewise.
794 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
795 attr_no_prologue_p fields.
796 * config/nds32/nds32.opt (mret-in-naked-func): New option.
798 2018-05-27 Jakub Jelinek <jakub@redhat.com>
801 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
804 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
806 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
808 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
810 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
812 (*<floatsuffix>floatv2div2sf2): Rename to ...
813 (*float<floatunssuffix>v2div2sf2): ... this.
814 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
815 (float<floatunssuffix>v2div2sf2_mask): ... this.
816 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
817 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
818 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
820 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
822 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
824 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
826 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
828 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
830 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
831 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
832 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
833 gen_ufix_truncv8dfv8si2.
834 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
835 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
836 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
837 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
838 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
839 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
840 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
841 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
843 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
847 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
849 2018-05-25 Jim Wilson <jimw@sifive.com>
851 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
852 * config/riscv/riscv.c (struct machine_function): Add
853 interrupt_handler_p and attribute_checked_p fields.
854 (riscv_attribute_table): Add interrupt.
855 (riscv_interrupt_type_p): New.
856 (riscv_save_reg_p): Save extra regs for interrupt handler.
857 (riscv_use_save_libcall): Return false for interrupt handler.
858 (riscv_first_stack_step): Add forward declaration.
859 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
860 for interrupt handler with large frame. Use it for saved reg list.
861 (riscv_expand_prologue): Move flag_stack_usage_info support to
862 eliminate duplication.
863 (riscv_expand_epilogue): Generate mret for interrupt handler.
864 (riscv_epilogue_uses): New.
865 (riscv_can_use_return_insn): Return false for interrupt handler.
866 (riscv_function_ok_for_sibcall): Likewise.
867 (riscv_set_current_function): Add interrupt handler support.
868 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
869 * config/riscv/riscv.md (UNSPECV_MRET): New.
871 (riscv_frflags, riscv_fsflags): Use tab after opcode.
873 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
875 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
877 PR tree-optimization/85712
878 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
879 this candidate has already been replaced in-situ by a copy.
881 2018-05-25 Jason Merrill <jason@redhat.com>
883 PR c++/80485 - inline function non-zero address.
884 * symtab.c (nonzero_address): Check DECL_COMDAT.
886 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
889 * config/alpha/alpha.md (ashlsi3): New insn pattern.
890 (*ashlsi_se): Rename from *ashldi_se. Define as sign
891 extension of SImode operation. Use const123_operand predicate.
893 (*saddl_se_1): Ditto.
895 (*ssubl_se_1): Ditto.
896 * config/alpha/predicates.md (const123_operand): New predicate.
897 * config/alpha/constraints.md (P): Use IN_RANGE.
899 2018-05-25 Richard Biener <rguenther@suse.de>
901 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
903 (ref_maybe_used_by_stmt_p): Likewise.
904 (stmt_may_clobber_ref_p): Likewise.
905 (stmt_may_clobber_ref_p_1): Likewise.
906 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
908 (ref_maybe_used_by_stmt_p): Likewise.
909 (stmt_may_clobber_ref_p): Likewise.
910 (stmt_may_clobber_ref_p_1): Likewise.
911 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
912 the alias oracle to disambiguate DRs with stmts DR analysis
914 (vect_analyze_data_refs): Do not give up on not analyzable
915 DRs for BB vectorization. Remove code truncating the dataref
918 2018-05-25 Jakub Jelinek <jakub@redhat.com>
921 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
922 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
923 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
925 2018-05-25 Richard Biener <rguenther@suse.de>
927 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
928 function, combining stmt data ref gathering and fatal analysis
930 (vect_analyze_data_refs): Remove now redudnant code and simplify.
931 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
932 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
933 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
934 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
936 2018-05-25 Bin Cheng <bin.cheng@arm.com>
938 PR tree-optimization/85720
939 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
940 SCC if all partitions are builtins.
941 (version_loop_by_alias_check): New parameter. Generate cancelable
942 runtime alias check if all partitions are builtins.
943 (distribute_loop): Update call to above function.
945 2018-05-25 Bin Cheng <bin.cheng@arm.com>
947 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
948 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
949 (parm_default_def_partition_arg): Ditto.
950 (set_parm_default_def_partition): Ditto.
951 (get_parm_default_def_partitions): Ditto and make it static.
952 (get_undefined_value_partitions): Ditto and make it static.
953 (remove_ssa_form): Refactor call to init_var_map here.
954 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
955 computation for loop region.
956 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
957 (register_default_def): Delete.
958 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
959 (parm_default_def_partition_arg): Ditto.
960 (set_parm_default_def_partition): Ditto.
961 (get_parm_default_def_partitions): Ditto and make it static.
962 (get_undefined_value_partitions): Ditto and make it static.
963 (coalesce_with_default, coalesce_with_default): Update comment.
964 (create_coalesce_list_for_region): New func factored out from
965 create_outofssa_var_map.
966 (populate_coalesce_list_for_outofssa): New func factored out from
967 create_outofssa_var_map and coalesce_ssa_name.
968 (create_outofssa_var_map): Delete.
969 (coalesce_ssa_name): Refactor to support live range computation.
970 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
971 (get_parm_default_def_partitions): Delete.
972 (get_undefined_value_partitions): Ditto.
973 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
974 computation for loop region.
975 (new_tree_live_info, loe_visit_block): Ditto.
976 (live_worklist, set_var_live_on_entry): Ditto.
977 (calculate_live_on_exit, verify_live_on_entry): Ditto.
978 * tree-ssa-live.h (struct _var_map): New fields.
979 (init_var_map): Change decl.
980 (region_contains_p): New.
982 2018-05-25 Bin Cheng <bin.cheng@arm.com>
984 * tree-ssa-live.h (live_merge_and_clear): Delete.
986 2018-05-25 Richard Biener <rguenther@suse.de>
989 * tree-dump.c (dequeue_and_dump): Remove access to removed
990 operand 2 of a SWITCH_EXPR.
992 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
994 * doc/sourcebuild.texi (vect_double_cond_arith): Include
995 multiplication and division.
996 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
997 (cond_udiv@var{m}, cond_umod@var{m}): Document.
998 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
999 (cond_udiv_optab, cond_umod_optab): New optabs.
1000 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
1001 (IFN_COND_RDIV): New internal functions.
1002 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
1003 TRUNC_MOD_EXPR and RDIV_EXPR.
1004 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
1005 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
1007 (SVE_INT_BINARY): Include mult.
1008 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
1009 (optab, sve_int_op): Handle mult.
1010 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
1012 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
1013 for SVE_INT_BINARY_SD.
1015 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1017 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
1018 (optab, sve_int_op): Handle div and udiv.
1019 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
1020 for SVE_INT_BINARY_SD.
1021 (*<optab><mode>3): New insn for the same.
1023 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1025 * tree-vect-patterns.c: Include predict.h.
1026 (vect_recog_divmod_pattern): Restrict check for division support
1027 to when optimizing for size.
1029 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1031 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
1032 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
1033 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
1034 (gimple_match_op::set_op): Likewise.
1035 (gimple_resimplify4): Declare.
1036 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
1037 (expr::gen_transform): Likewise.
1038 (decision_tree::gen): Generate a simplification routine for 4 operands.
1039 * gimple-match-head.c (gimple_simplify): Add an overload for
1040 4 operands. In the top-level function, handle up to 4 call
1041 arguments and call gimple_resimplify4.
1042 (gimple_resimplify4): New function.
1043 (build_call_internal): Pass a fourth operand.
1044 (maybe_push_to_seq): Likewise.
1045 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
1046 Fold VEC_COND_EXPRs of an operation and a default value into
1047 an IFN_COND_* function if possible.
1048 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1050 (SVE_COND_FP_BINARY): Include them.
1051 (optab, sve_fp_op): Handle them.
1052 (SVE_INT_BINARY_REV): New code iterator.
1053 (SVE_COND_FP_BINARY_REV): New int iterator.
1054 (commutative): New int attribute.
1055 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1057 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1059 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1060 (*cond_<optab><mode>): New patterns for reversed operands.
1062 2018-05-25 Richard Biener <rguenther@suse.de>
1064 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1065 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1066 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1067 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1068 * tree-vect-data-refs.c (everywhere): Adjust users.
1069 * tree-vect-loop.c (everywhere): Likewise.
1070 * tree-vect-slp.c (everywhere): Likewise.
1071 * tree-vect-stmts.c (everywhere): Likewise.
1072 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1074 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1076 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1078 (gcc_cv_as_section_exclude): ... this.
1079 Try Solaris as #exclude syntax.
1080 * configure: Regenerate.
1081 * config.in: Regenerate.
1082 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1084 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1085 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1087 * varasm.c (default_elf_asm_named_section): Don't check if
1088 HAVE_GAS_SECTION_EXCLUDE is defined.
1090 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1092 * doc/md.texi: Update the documentation of the cond_* optabs
1093 to mention the new final operand. Fix GET_MODE_NUNITS call.
1094 Describe the scalar case too.
1095 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1096 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1098 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1099 (get_conditional_internal_fn): Update comment.
1100 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1101 accumulator value as a final argument to conditional functions.
1102 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1103 a define_expand and add an "else" operand. Assert for now that
1104 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1105 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1106 (*cond_<optab><mode>): New patterns.
1107 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1108 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1109 (UNSPEC_COND_EOR): Delete.
1110 (optab): Remove associated mappings.
1111 (SVE_INT_BINARY): New code iterator.
1112 (sve_int_op): Remove int attribute and add "minus" to the code
1114 (SVE_COND_INT_OP): Delete.
1115 (SVE_COND_FP_OP): Rename to...
1116 (SVE_COND_FP_BINARY): ...this.
1118 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1120 * optabs.c (can_reuse_operands_p): New function.
1121 (maybe_legitimize_operands): Try to reuse the results for
1124 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1126 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1127 Add {q} suffix to insn mnemonic.
1129 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1131 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1132 (msp430_warn_func_return): New.
1134 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1136 * fold-const.c (tree_nonzero_bits): New function.
1137 * fold-const.h (tree_nonzero_bits): Likewise.
1138 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1139 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1141 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1145 * varasm.c (assemble_alias): Check ifunc_resolver only on
1148 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1151 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1152 when memory input operand is handled.
1154 2018-05-24 Luis Machado <luis.machado@linaro.org>
1156 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1158 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1160 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1162 * match.pd: Delay FMA folds until after vectorization.
1164 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1167 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1168 address check not strict.
1170 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1172 * gimple-match.h (gimple_match_op): New class.
1173 (mprts_hook): Replace parameters with a gimple_match_op *.
1174 (maybe_build_generic_op): Likewise.
1175 (gimple_simplified_result_is_gimple_val): Replace parameters with
1176 a const gimple_match_op *.
1177 (gimple_simplify): Replace code_helper * and tree * parameters with
1178 a gimple_match_op * parameter.
1179 (gimple_resimplify1): Replace code_helper *, tree and tree *
1180 parameters with a gimple_match_op * parameter.
1181 (gimple_resimplify2): Likewise.
1182 (gimple_resimplify3): Likewise.
1183 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1184 parameters with a gimple_match_op * parameter.
1185 * gimple-match-head.c (gimple_simplify): Change prototypes of
1186 auto-generated functions to take a gimple_match_op * instead of
1187 separate code_helper * and tree * parameters. Make the same
1188 change in the top-level overload and update calls to the
1189 gimple_resimplify routines. Update calls to the auto-generated
1190 functions and to maybe_push_res_to_seq in the publicly-facing
1191 operation-specific gimple_simplify overloads.
1192 (gimple_match_op::MAX_NUM_OPS): Define.
1193 (gimple_resimplify1): Replace rcode and ops with a single res_op
1194 parameter. Update call to gimple_simplify.
1195 (gimple_resimplify2): Likewise.
1196 (gimple_resimplify3): Likewise.
1197 (mprts_hook): Replace parameters with a gimple_match_op *.
1198 (maybe_build_generic_op): Likewise.
1199 (build_call_internal): Replace type, nargs and ops with
1200 a gimple_match_op *.
1201 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1202 with a single gimple_match_op *. Update calls to mprts_hook,
1203 build_call_internal and gimple_simplified_result_is_gimple_val.
1204 Factor out code that is common to the tree_code and combined_fn cases.
1205 * genmatch.c (expr::gen_transform): Replace tem_code and
1206 tem_ops with a gimple_match_op called tem_op. Update calls
1207 to the gimple_resimplify functions and maybe_push_res_to_seq.
1208 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1209 res_ops. Update call to the gimple_resimplify functions.
1210 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1211 (decision_tree::gen): Make the functions take a gimple_match_op *
1212 called res_op instead of separate res_code and res_ops parameters.
1213 Update call accordingly.
1214 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1215 and ops with a single res_op parameter. Update calls to
1216 maybe_build_generic_op and maybe_push_res_to_seq.
1217 (fold_stmt_1): Update calls to gimple_simplify and
1218 replace_stmt_with_simplification.
1219 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1220 and gimple_simplified_result_is_gimple_val.
1221 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1223 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1224 with a gimple_match_op *.
1225 (vn_nary_build_or_lookup): Likewise. Update call to
1226 vn_nary_build_or_lookup_1.
1227 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1228 gimple_match_op *. Update calls to the gimple_resimplify routines
1229 and to gimple_simplified_result_is_gimple_val.
1230 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1231 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1232 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1233 (visit_nary_op): Likewise.
1234 (visit_reference_op_load): Likewise.
1236 2018-05-23 Luis Machado <luis.machado@linaro.org>
1238 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1239 modifier for printing the step amount.
1241 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1244 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1247 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1249 * doc/sourcebuild.texi (Endianness): New subsubsection.
1251 2018-05-23 Luis Machado <luis.machado@linaro.org>
1253 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1254 <prefetch_dynamic_strides>: New const bool field.
1255 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1256 prefetch_dynamic_strides.
1257 (exynosm1_prefetch_tune): Likewise.
1258 (thunderxt88_prefetch_tune): Likewise.
1259 (thunderx_prefetch_tune): Likewise.
1260 (thunderx2t99_prefetch_tune): Likewise.
1261 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1263 (aarch64_override_options_internal): Update to set
1264 PARAM_PREFETCH_DYNAMIC_STRIDES.
1265 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1266 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1267 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1268 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1269 prefetch-dynamic-strides setting.
1271 2018-05-23 Luis Machado <luis.machado@linaro.org>
1273 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1274 <minimum_stride>: New const int field.
1275 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1276 minimum_stride field defaulting to -1.
1277 (exynosm1_prefetch_tune): Likewise.
1278 (thunderxt88_prefetch_tune): Likewise.
1279 (thunderx_prefetch_tune): Likewise.
1280 (thunderx2t99_prefetch_tune): Likewise.
1281 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1282 <default_opt_level>: Set to 3.
1283 (aarch64_override_options_internal): Update to set
1284 PARAM_PREFETCH_MINIMUM_STRIDE.
1285 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1286 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1287 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1288 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1289 stride is constant and is below the minimum stride threshold.
1291 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1293 * config/arm/arm-cpus.in (mode26): Delete.
1294 (armv4): Delete mode26 reference.
1295 * config/arm/arm.c (arm_configure_build_target): Delete use of
1298 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1300 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1302 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1303 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1305 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1306 (floatunsdidf2): Ditto.
1308 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1310 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1311 (fixuns_trunc<mode>si2_avx512f): Ditto.
1312 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1313 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1314 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1316 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1318 PR rtl-optimization/79985
1319 * df-scan.c (df_insn_refs_collect): Remove special case for
1320 global registers and asm statements.
1322 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1324 * extend.texi (Global Register Variables): Rewrite the bullet list.
1325 Note that the register is available for allocation. Note that access
1326 via inline asm must use constraints. Add note about async-signal
1327 handlers. Remove paragraph about automagic register selection.
1329 2018-05-23 Richard Biener <rguenther@suse.de>
1331 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1332 of fixed offset from memset VN.
1334 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1336 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1338 (alloc_cand_and_find_basis): Initialize first_interp field.
1339 (slsr_process_mul): Modify first_interp field.
1340 (slsr_process_add): Likewise.
1341 (slsr_process_cast): Modify first_interp field for each new
1343 (slsr_process_copy): Likewise.
1344 (dump_candidate): Dump first_interp field.
1345 (replace_mult_candidate): Process all interpretations, not just
1347 (replace_rhs_if_not_dup): Likewise.
1348 (replace_one_candidate): Likewise.
1350 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1352 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1354 (aarch64_needs_frame_chain): New function.
1355 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1357 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1360 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1361 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1362 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1363 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1364 as true for strict-align.
1365 (aarch64_can_inline_p): Perform checks even when callee has no
1366 attributes to check for strict alignment.
1367 * doc/extend.texi (AArch64 Function Attributes): Document
1369 * doc/invoke.texi: (AArch64 Options): Likewise.
1371 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1373 PR tree-optimization/85853
1374 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1375 the handling of the root of the node to...
1376 (vect_slp_analyze_node_operations_1): ...this new function,
1377 and run the whole thing with the child nodes' def types
1378 set according to their SLP node's def type.
1380 2018-05-23 Richard Biener <rguenther@suse.de>
1383 * tree-data-ref.c (create_runtime_alias_checks): Defer
1384 and ignore overflow warnings.
1386 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1388 PR tree-optimization/85822
1389 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1392 2018-05-23 Richard Biener <rguenther@suse.de>
1394 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1395 memset constants via native_interpret_expr.
1397 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1400 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1402 (cgraph_node::create_alias): Likewise.
1403 (cgraph_node::get_availability): Check ifunc_resolver instead
1404 of looking up ifunc attribute.
1405 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1406 * varasm.c (do_assemble_alias): Likewise.
1407 (assemble_alias): Likewise.
1408 (default_binds_local_p_3): Likewise.
1409 * cgraph.h (cgraph_node): Add ifunc_resolver.
1410 (cgraph_node::only_called_directly_or_aliased_p): Return false
1412 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1414 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1415 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1416 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1417 instead of looking up ifunc attribute.
1419 2018-05-22 Luis Machado <luis.machado@linaro.org>
1421 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1423 2018-05-22 Martin Sebor <msebor@redhat.com>
1426 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1427 only when expasion succeeds.
1428 (expand_builtin_strcmp): Same.
1429 (expand_builtin_strncmp): Same.
1431 2018-05-22 Martin Sebor <msebor@redhat.com>
1433 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1435 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1436 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1438 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1439 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1441 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1442 Add check for legitimate_address.
1443 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1444 (aarch64_swap_ldrstr_operands): New.
1445 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1448 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1449 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1451 * config/aarch64/aarch64.md: New patterns to generate stp
1453 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1454 single words and double words.
1455 (load_pair_sw, load_pair_dw): Likewise.
1456 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1458 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1460 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1461 for different mode ldpstp and add peephole for merged zero stores.
1463 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1465 (aarch64_gen_store_pair): Rename calls to match new patterns.
1466 (aarch64_gen_load_pair): Rename calls to match new patterns.
1467 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1468 (load_pair<DREG:mode><DREG2:mode>): ... This.
1469 (store_pair<mode>): Rename to...
1470 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1471 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1473 (V_INT_EQUIV): Handle SImode.
1474 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1477 2018-05-22 Martin Sebor <msebor@redhat.com>
1480 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1481 or ajust the presumed bound on an operation to avoid unnecessary
1484 2018-05-22 Martin Sebor <msebor@redhat.com>
1486 PR tree-optimization/85826
1487 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1488 assuming that a DECL necesarily has a constant size.
1490 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1493 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1495 2018-05-22 Richard Biener <rguenther@suse.de>
1497 PR tree-optimization/85834
1498 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1499 non-constant and non-zero memset arguments.
1501 2018-05-22 Martin Liska <mliska@suse.cz>
1504 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1506 2018-05-22 Richard Biener <rguenther@suse.de>
1508 PR tree-optimization/85863
1509 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1510 comparisons when vectype is specified.
1511 (vectorizable_condition): Do not specify vectype for
1512 vect_is_simple_cond when SLP vectorizing.
1514 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1517 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1518 define __ibm128 as long double.
1519 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1520 as a distinct type when IEEE 128-bit support is enabled.
1521 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1522 128-bit types to use the correct functions.
1523 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1524 convert between 128-bit floating point types that have different
1525 modes but the same representation, instead of using gen_lowpart to
1527 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1529 (IFKF_reg): New attributes to give the register constraints for
1531 (extend<mode>tf2_internal): New insns to mark an explicit
1532 conversion between 128-bit floating point types that have a
1533 different mode but share the same representation.
1535 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1537 PR tree-optimization/85814
1538 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1539 a null return from get_strinfo when unsharing the next
1540 strinfo in the chain.
1542 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1545 * varasm.c (weak_finish): Clean up weak_decls.
1547 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1549 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1550 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1551 UNSPEC_UADALP values.
1552 * config/aarch64/iterators.md (ABAL): New int iterator.
1555 (sur): Add mappings for the above.
1556 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1558 (aarch64_<sur>abal<mode>_4): Likewise.
1559 (aarch64_<sur>adalp<mode>_3): Likewise.
1560 (<sur>sadv16qi): New define_expand.
1562 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1564 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1565 (*movdf_internal): Ditto.
1566 (*rcpsf2_sse): Ditto.
1567 (*rsqrtsf2_sse): Ditto.
1568 (*sqrt<mode>2_sse): Ditto.
1570 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1572 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1574 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1575 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1576 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1577 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1579 * config/aarch64/arm_neon.h: Likewise.
1580 * config/aarch64/iterators.md (VQ_I): New.
1582 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1584 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1586 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1589 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1591 * config/nds32/constraints.md (S): New constraint.
1592 * config/nds32/nds32.md (call_internal): Use constraint S.
1593 (call_value_internal): Likewise.
1594 (sibcall_internal): Likewise.
1595 (sibcall_value_internal): Likewise.
1597 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1598 Chung-Ju Wu <jasonwucj@gmail.com>
1600 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1603 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1604 Chung-Ju Wu <jasonwucj@gmail.com>
1606 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1607 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1608 (nds32_rtx_costs_impl): Simplify.
1609 (nds32_address_cost_impl): Simplify.
1610 (nds32_init_rtx_costs): New function.
1611 (nds32_rtx_costs_speed_prefer): Likewise.
1612 (nds32_rtx_costs_size_prefer): Likewise.
1613 (nds32_address_cost_speed_prefer): Likewise.
1614 (nds32_address_cost_speed_fwprop): Likewise.
1615 (nds32_address_cost_size_prefer): Likewise.
1616 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1617 * config/nds32/nds32.c (nds32_option_override): Use
1618 nds32_init_rtx_costs function.
1620 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1622 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1623 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1624 (TARGET_PIPELINE_N8): Likewise.
1625 (TARGET_PIPELINE_N10): Likewise.
1626 (TARGET_PIPELINE_N13): Likewise.
1627 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1629 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1631 * config/nds32/nds32-fpu.md: Update copyright year.
1633 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1635 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1637 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1639 * config/nds32/nds32.c
1640 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1641 * config/nds32/nds32.opt (minline-asm-r15): New option.
1643 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1645 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1647 * config/nds32/nds32.md (abssi2): New pattern.
1649 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1651 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1652 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1653 Merge insn pattern from sse_cvtsi2ss<round_name> and
1654 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1655 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1656 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1657 using SWI48 mode iterator.
1658 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1659 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1660 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1661 pattern from sse_cvttss2si<round_saeonly_name>
1662 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1663 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1664 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1665 using SWI48 mode iterator.
1666 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1667 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1668 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1669 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1670 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1671 using SWI48 mode iterator.
1672 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1673 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1674 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1675 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1676 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1677 SWI48 mode iterator.
1678 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1679 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1680 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1681 pattern from sse_cvttsd2si<round_saeonly_name>
1682 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1684 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1686 * config/nds32/nds32-md-auxiliary.c
1687 (nds32_valid_smw_lwm_base_p): Refine.
1688 (nds32_output_smw_single_word): Refine.
1689 (nds32_output_smw_double_word): New.
1690 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1692 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1694 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1695 (nds32_output_stack_pop): Refine.
1696 (nds32_expand_unaligned_load): Refine.
1697 (nds32_expand_unaligned_store): Refine.
1699 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1700 Chung-Ju Wu <jasonwucj@gmail.com>
1702 * config/nds32/constants.md: Add TP_REGNUM constant.
1703 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1704 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1706 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1707 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1708 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1709 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1710 and PIC code generation.
1711 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1713 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1715 * config/nds32/nds32.md: Support TLS and PIC.
1716 * config/nds32/nds32.c: Support TLS and PIC.
1717 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1718 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1721 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1724 mode with E_ prefix.
1726 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1727 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1730 * config/nds32/nds32-md-auxiliary.c
1731 (symbolic_reference_mentioned_p): New.
1732 (nds32_legitimize_ict_address): New.
1733 (nds32_expand_ict_move): New.
1734 (nds32_indirect_call_referenced_p): New.
1735 (nds32_symbol_binds_local_p): Delete.
1736 (nds32_long_call_p): Modify.
1737 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1738 * config/nds32/nds32-protos.h
1739 (symbolic_reference_mentioned_p): Declare.
1740 (nds32_legitimize_ict_address): Declare.
1741 (nds32_expand_ict_move): Declare.
1742 (nds32_indirect_call_referenced_p): Declare.
1743 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1744 (nds32_relax_group): Use nds32_ict_const_p as condition.
1745 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1746 (nds32_asm_file_start): Output ict_model directive in asm code.
1747 (nds32_legitimate_address_p): Consider indirect call.
1748 (nds32_print_operand): Consider indirect call.
1749 (nds32_print_operand_address): Consider indirect call.
1750 (nds32_insert_attributes): Handle "indirect_call" attribute.
1751 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1752 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1753 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1754 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1755 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1756 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1757 (TARGET_ICT_MODEL_SMALL): Define.
1758 (TARGET_ICT_MODEL_LARGE): Define.
1759 * config/nds32/nds32.md (movsi): Consider ict model.
1760 (call, call_value): Consider ict model.
1761 (sibcall, sibcall_value): Consider ict model.
1762 * config/nds32/nds32.opt (mict-model): New option.
1763 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1766 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1767 Monk Chiang <sh.chiang04@gmail.com>
1768 Jim Wilson <jimw@sifive.com>
1770 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1771 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1772 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1773 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1774 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1775 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1776 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1777 compute save_libcall_adjustment properly.
1778 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1779 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1780 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1781 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1782 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1783 (ABI_SPEC): Handle mabi=ilp32e.
1784 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1785 (RVE): Add RVE mask.
1786 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1787 <-march>: Add rv32e as an example.
1789 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1792 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1793 (intra_create_variable_infos): Handle C++ constructors.
1795 2018-05-18 Martin Liska <mliska@suse.cz>
1797 * passes.def: Remove a redundant pass.
1799 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1802 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1804 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1806 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1808 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1810 (armv2, armv2a, armv3, armv3m): Delete architectures.
1811 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1812 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1813 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1815 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1816 (*mulsidi3adddi): Likewise.
1817 (mulsidi3): Likewise.
1818 (*mulsidi3_nov6): Likewise.
1819 (umulsidi3): Likewise.
1820 (umulsidi3_nov6): Likewise.
1821 (umaddsidi4): Likewise.
1822 (*umulsidi3adddi): Likewise.
1823 (smulsi3_highpart): Likewise.
1824 (*smulsi3_highpart_nov6): Likewise.
1825 (umulsi3_highpart): Likewise.
1826 (*umulsi3_highpart_nov6): Likewise.
1827 * config/arm/arm.h (arm_arch3m): Delete.
1828 * config/arm/arm.c (arm_arch3m): Delete.
1829 (arm_option_override_internal): Update armv3-related comment.
1830 (arm_configure_build_target): Delete use of isa_bit_mode32.
1831 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1832 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1833 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1835 (mulusa3): Likewise.
1836 * config/arm/arm-protos.h (arm_arch3m): Delete.
1837 * config/arm/arm-tables.opt: Regenerate.
1838 * config/arm/arm-tune.md: Likewise.
1839 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1840 deleted architectures.
1842 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1844 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1845 (armv5t, armv5te): New features.
1846 (ARMv5, ARMv5e): Delete fgroups.
1847 (ARMv5t, ARMv5te): Adjust for above changes.
1849 (armv5, armv5e): Delete arches.
1850 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1852 (*call_reg_arm): Likewise.
1853 (*call_value_reg_armv5): Likewise.
1854 (*call_value_reg_arm): Likewise.
1855 (*call_symbol): Likewise.
1856 (*call_value_symbol): Likewise.
1857 (*sibcall_insn): Likewise.
1858 (*sibcall_value_insn): Likewise.
1860 (prefetch): Likewise.
1861 (define_split and define_peephole2 dependent on arm_arch5):
1863 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1865 (TARGET_ARM_QBIT): Likewise.
1866 (TARGET_DSP_MULTIPLY): Likewise.
1867 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1868 (arm_arch5, arm_arch5e): Delete.
1869 (arm_arch5t, arm_arch5te): Declare.
1870 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1871 (arm_arch5t): Declare.
1872 (arm_option_reconfigure_globals): Update for the above.
1873 (arm_options_perform_arch_sanity_checks): Update comment, replace
1874 use of arm_arch5 with arm_arch5t.
1875 (use_return_insn): Likewise.
1876 (arm_emit_call_insn): Likewise.
1877 (output_return_instruction): Likewise.
1878 (arm_final_prescan_insn): Likewise.
1879 (arm_coproc_builtin_available): Likewise.
1880 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1881 arm_arch5e with arm_arch5t and arm_arch5te.
1882 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1883 (arm_arch5t, arm_arch5te): Declare.
1884 * config/arm/arm-tables.opt: Regenerate.
1885 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1886 * config/arm/t-multilib: Likewise.
1887 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1888 instead of arm_arch5.
1889 (*call_reg_thumb1): Likewise.
1890 (*call_value_reg_thumb1_v5): Likewise.
1891 (*call_value_reg_thumb1): Likewise.
1892 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1894 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1896 2018-05-18 Martin Liska <mliska@suse.cz>
1898 PR gcov-profile/84846
1899 * doc/gcov.texi: Document -t option of gcov tool.
1901 2018-05-18 Martin Liska <mliska@suse.cz>
1903 PR gcov-profile/84846
1904 * gcov.c (print_usage): Add new -t option.
1905 (process_args): Handle the option.
1906 (generate_results): Use stdout as output when requested by
1909 2018-05-18 Martin Liska <mliska@suse.cz>
1911 PR gcov-profile/84846
1912 * coverage.c (coverage_init): Write PWD to .gcno file.
1913 * doc/gcov.texi: Document how working directory is printed.
1914 * gcov-dump.c (dump_gcov_file): Print PWD.
1915 * gcov.c (output_intermediate_file): Likewise.
1916 (read_graph_file): Read PWD string.
1917 (output_lines): Print PWD.
1919 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1922 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1923 for retval and return false if all args to phi are zero.
1925 2018-05-18 Richard Biener <rguenther@suse.de>
1927 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1929 (evrp_dom_walker::before_dom_children): Call it.
1931 2018-05-18 Richard Biener <rguenther@suse.de>
1933 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1934 results when processing array refs with variable index.
1936 2018-05-18 Toon Moene <toon@moene.org>
1938 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1939 directly after that of -floop-interchange. Indicate that both
1940 options are enabled by default when specifying -O3.
1942 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1944 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1945 iterator. Delete separate integer-mode vec_set<mode> expander.
1946 (aarch64_simd_vec_setv2di): Delete.
1947 (vec_setv2di): Delete.
1948 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1949 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1950 the "w, r" alternative.
1952 2018-05-18 Martin Liska <mliska@suse.cz>
1954 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1955 * tree-pass.h (make_pass_lower_switch_O0): New function.
1956 * tree-switch-conversion.c (node_has_low_bound): Remove.
1957 (node_has_high_bound): Likewise.
1958 (node_is_bounded): Likewise.
1959 (class pass_lower_switch): Make it a template type and create
1961 (pass_lower_switch::execute): Add template argument.
1962 (make_pass_lower_switch): New function.
1963 (make_pass_lower_switch_O0): New function.
1964 (do_jump_if_equal): Remove.
1965 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1966 all the hard work to tree optimization passes.
1968 2018-05-18 Martin Liska <mliska@suse.cz>
1970 * dbgcnt.c (limit_low): Renamed from limit.
1971 (limit_high): New variable.
1972 (dbg_cnt_is_enabled): Check for upper limit.
1973 (dbg_cnt): Adjust dumping.
1974 (dbg_cnt_set_limit_by_index): Add new argument for high
1976 (dbg_cnt_set_limit_by_name): Likewise.
1977 (dbg_cnt_process_single_pair): Parse new format.
1978 (dbg_cnt_process_opt): Use strtok.
1979 (dbg_cnt_list_all_counters): Remove 'value' and add
1981 * doc/invoke.texi: Document changes.
1983 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1985 * doc/sourcebuild.texi (scalar_all_fma): Document.
1986 * tree.def (FMA_EXPR): Delete.
1987 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1988 * internal-fn.c (ternary_direct): New macro.
1989 (expand_ternary_optab_fn): Likewise.
1990 (direct_ternary_optab_supported_p): Likewise.
1991 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1992 * builtins.c (fold_builtin_fma): Delete.
1993 (fold_builtin_3): Don't call it.
1994 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1995 * expr.c (expand_expr_real_2): Likewise.
1996 * fold-const.c (operand_equal_p): Likewise.
1997 (fold_ternary_loc): Likewise.
1998 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1999 * gimple.c (DEFTREECODE): Likewise.
2000 * gimplify.c (gimplify_expr): Likewise.
2001 * optabs-tree.c (optab_for_tree_code): Likewise.
2002 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
2003 * tree-eh.c (operation_could_trap_p): Likewise.
2004 (stmt_could_throw_1_p): Likewise.
2005 * tree-inline.c (estimate_operator_cost): Likewise.
2006 * tree-pretty-print.c (dump_generic_node): Likewise.
2007 (op_code_prio): Likewise.
2008 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2009 * tree-ssa-operands.c (get_expr_operands): Likewise.
2010 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
2011 * fold-const-call.h (fold_fma): Delete.
2012 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
2013 CFN_FNMA and CFN_FNMS.
2015 * genmatch.c (combined_fn): New enum.
2016 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
2017 (commutative_op): New function.
2018 (commutate): Use it. Handle more than 2 operands.
2019 (dt_operand::gen_gimple_expr): Use commutative_op.
2020 (parser::parse_expr): Allow :c to be used with non-binary
2021 operators if the commutative operand is known.
2022 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
2023 CFN_FMS, CFN_FNMA and CFN_FNMS.
2024 (backprop::process_assign_use): Remove FMA_EXPR handling.
2025 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
2026 (gen_hsa_fma): New function.
2027 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
2028 IFN_FNMA and IFN_FNMS.
2029 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
2030 * gimple-fold.h (follow_all_ssa_edges): Declare.
2031 * gimple-fold.c (follow_all_ssa_edges): New function.
2032 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
2033 gimple_build interface and use follow_all_ssa_edges to fold the result.
2034 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
2035 instead of checking for optabs directly.
2036 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
2037 rather than FMA_EXPRs.
2038 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
2039 call to IFN_FMA instead of an FMA_EXPR.
2041 2018-05-17 Jim Wilson <jimw@sifive.com>
2043 * expr.c (do_tablejump): When converting index to Pmode, if we have a
2044 sign extended promoted subreg, and the range does not have the sign bit
2045 set, then do a sign extend.
2047 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
2048 test, check for sign extended subreg and/or constant operands, and
2049 do a sign extend in that case.
2051 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2053 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2054 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2056 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2057 Change logics_shift_reg to logics_shift_imm.
2058 (thunderx2t99_fp_loadpair_basic): Delete.
2059 (thunderx2t99_fp_storepair_basic): Delete.
2060 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2061 (thunderx2t99_asimd_polynomial): Delete.
2062 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2063 and neon_fp_mul_d_scalar_q.
2064 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2065 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2066 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2067 (thunderx2t99_asimd_lut): Add missing tbl types.
2068 (thunderx2t99_asimd_ext): Delete.
2069 (thunderx2t99_asimd_load1_1_mult): Delete.
2070 (thunderx2t99_asimd_load1_2_mult): Delete.
2071 (thunderx2t99_asimd_load1_ldp): New.
2072 (thunderx2t99_asimd_load1): New.
2073 (thunderx2t99_asimd_load2): Add missing *load2* types.
2074 (thunderx2t99_asimd_load3): New.
2075 (thunderx2t99_asimd_load4): New.
2076 (thunderx2t99_asimd_store1_1_mult): Delete.
2077 (thunderx2t99_asimd_store1_2_mult): Delete.
2078 (thunderx2t99_asimd_store2_mult): Delete.
2079 (thunderx2t99_asimd_store2_onelane): Delete.
2080 (thunderx2t99_asimd_store_stp): New.
2081 (thunderx2t99_asimd_store1): New.
2082 (thunderx2t99_asimd_store2): New.
2083 (thunderx2t99_asimd_store3): New.
2084 (thunderx2t99_asimd_store4): New.
2086 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2088 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2089 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2091 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2092 Segher Boessenkool <segher@kernel.crashing.org>
2095 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2098 2018-05-17 Richard Biener <rguenther@suse.de>
2100 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2101 for pruning loop and prune defs feeding only already visited PHIs.
2103 2018-05-17 Richard Biener <rguenther@suse.de>
2105 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2107 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2108 Richard Biener <rguenther@suse.de>
2110 PR tree-optimization/85793
2111 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2112 for VMAT_ELEMENTWISE.
2114 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2116 * internal-fn.h (lookup_internal_fn): Declare
2117 * internal-fn.c (lookup_internal_fn): New function.
2118 * gimple.c (gimple_build_call_from_tree): Handle calls to
2120 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2121 internal function names.
2122 * tree-pretty-print.c (dump_generic_node): Likewise.
2123 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2125 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2127 * gimple-fold.h (gimple_build): Make the function forms take
2128 combined_fn rather than built_in_function.
2129 (gimple_simplify): Likewise.
2130 * gimple-match-head.c (gimple_simplify): Likewise.
2131 * gimple-fold.c (gimple_build): Likewise.
2132 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2133 rather than gimple_build_call_internal.
2134 (get_initial_defs_for_reduction): Likewise.
2135 (vect_create_epilog_for_reduction): Likewise.
2136 (vectorizable_live_operation): Likewise.
2138 2018-05-17 Martin Liska <mliska@suse.cz>
2140 * gimple-ssa-sprintf.c (format_directive): Do not use
2141 space in between 'G_' and '('.
2143 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2146 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2147 even if the mask is not all ones.
2150 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2152 (ix86_gimple_fold_builtin): Likewise.
2155 * config/i386/i386.c: Include tree-vector-builder.h.
2156 (ix86_vector_shift_count): New function.
2157 (ix86_fold_builtin): Fold shift builtins by scalar count.
2158 (ix86_gimple_fold_builtin): Likewise.
2160 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2161 _mm512_setzero): New intrinsics.
2163 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2164 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2166 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2167 code generation for cases where splatting a value is not useful.
2168 * simplify-rtx.c (simplify_ternary_operation): Simplify
2169 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2170 mode to a vec_concat.
2172 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2174 * config.gcc: Support "goldmont-plus".
2175 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2177 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2178 PROCESSOR_GOLDMONT_PLUS.
2179 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2180 (processor_target_table): Add "goldmont-plus".
2181 (PTA_GOLDMONT_PLUS): Define.
2182 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2183 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2184 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2185 (fold_builtin_cpu): Add "goldmont-plus".
2186 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2187 (ix86_option_override_internal): Add "goldmont-plus".
2188 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2189 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2190 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2191 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2193 2018-05-17 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/85757
2196 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2197 remove defs that only feed that PHI from further processing.
2199 2018-05-16 Jim Wilson <jimw@sifive.com>
2201 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2203 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2205 2018-05-16 Mark Wielaard <mark@klomp.org>
2207 * dwarf2out.c (count_index_strings): New function.
2208 (output_indirect_strings): Call count_index_strings and generate
2209 header for dwarf_version >= 5.
2211 2018-05-16 Mark Wielaard <mark@klomp.org>
2213 * dwarf2out.c (dwarf_FORM): New function.
2214 (set_indirect_string): Use dwarf_FORM.
2215 (reset_indirect_string): Likewise.
2216 (size_of_die): Likewise.
2217 (value_format): Likewise.
2218 (output_die): Likewise.
2219 (add_skeleton_AT_string): Likewise.
2220 (output_macinfo_op): Likewise.
2221 (index_string): Likewise.
2222 (output_index_string_offset): Likewise.
2223 (output_index_string): Likewise.
2224 (count_index_strings): Likewise.
2226 2018-05-16 Carl Love <cel@us.ibm.com>
2228 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2229 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2231 2018-05-16 Martin Jambor <mjambor@suse.cz>
2233 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2234 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2236 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2238 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2239 (fnma<mode>4): Likewise.
2240 (fms<mode>4): Likewise.
2241 (fnms<mode>4): Likewise.
2242 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2243 (aarch64_fnma<mode>4): Likewise.
2244 (aarch64_fms<mode>4): Likewise.
2245 (aarch64_fnms<mode>4): Likewise.
2246 (aarch64_fnmadd<mode>4): Likewise.
2248 2018-05-16 Jason Merrill <jason@redhat.com>
2250 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2252 2018-05-16 Richard Biener <rguenther@suse.de>
2254 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2255 (dump_stmt_cost): Declare.
2256 (add_stmt_cost): Dump cost we add.
2257 (add_stmt_costs): New function.
2258 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2260 (vect_analyze_stmt): Adjust prototype.
2261 (vectorizable_condition): Likewise.
2262 (vectorizable_live_operation): Likewise.
2263 (vectorizable_reduction): Likewise.
2264 (vectorizable_induction): Likewise.
2265 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2266 cost vector to pass to vectorizable_ and record afterwards.
2267 (vect_model_reduction_cost): Take cost vector argument and adjust.
2268 (vect_model_induction_cost): Likewise.
2269 (vectorizable_reduction): Likewise.
2270 (vectorizable_induction): Likewise.
2271 (vectorizable_live_operation): Likewise.
2272 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2273 SLP_TREE_NUMBER_OF_VEC_STMTS.
2274 (vect_analyze_slp_cost_1): Remove.
2275 (vect_analyze_slp_cost): Likewise.
2276 (vect_slp_analyze_node_operations): Take visited args and
2277 a target cost vector. Avoid processing already visited stmt sets.
2278 (vect_slp_analyze_operations): Use a local cost vector to gather
2279 costs and register those of non-discarded instances.
2280 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2281 (vect_schedule_slp_instance): Remove copying of
2282 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2284 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2285 adding cost. Record cost entry location.
2286 (vect_prologue_cost_for_slp_op): Function to compute cost of
2287 a constant or invariant generated for SLP vect in the prologue,
2288 split out from vect_analyze_slp_cost_1.
2289 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2290 (vect_model_promotion_demotion_cost): Likewise.
2291 (vect_model_store_cost): Likewise, make static.
2292 (vect_model_load_cost): Likewise.
2293 (vectorizable_bswap): Add cost vector arg and adjust.
2294 (vectorizable_call): Likewise.
2295 (vectorizable_simd_clone_call): Likewise.
2296 (vectorizable_conversion): Likewise.
2297 (vectorizable_assignment): Likewise.
2298 (vectorizable_shift): Likewise.
2299 (vectorizable_operation): Likewise.
2300 (vectorizable_store): Likewise.
2301 (vectorizable_load): Likewise.
2302 (vectorizable_condition): Likewise.
2303 (vectorizable_comparison): Likewise.
2304 (can_vectorize_live_stmts): Likewise.
2305 (vect_analyze_stmt): Likewise.
2306 (vect_transform_stmt): Adjust calls to vectorizable_*.
2307 * tree-vectorizer.c: Include gimple-pretty-print.h.
2308 (dump_stmt_cost): New function.
2310 2018-05-16 Richard Biener <rguenther@suse.de>
2312 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2313 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2314 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2315 (check_name): New callback.
2316 (dse_classify_store): Track cycles via a visited bitmap of PHI
2317 defs and simplify handling of in-loop and across loop dead stores
2318 and properly fail for loop-variant refs. Handle byte-tracking with
2319 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2322 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2324 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2325 (vect_get_mask_type_for_stmt): Likewise.
2326 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2328 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2329 to determine the statement's vector type and the vector type that
2330 should be used for calculating nunits. Deal with cases in which
2331 the type has to be deferred.
2332 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2333 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2334 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2335 (vect_determine_vf_for_stmt): New functions, split out from...
2336 (vect_determine_vectorization_factor): ...here.
2337 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2338 (vect_get_mask_type_for_stmt): New functions, split out from
2339 vect_determine_vectorization_factor.
2341 2018-05-16 Richard Biener <rguenther@suse.de>
2343 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2344 verify the [VEC_]COND_EXPR embedded comparison.
2346 2018-05-15 Martin Sebor <msebor@redhat.com>
2348 PR tree-optimization/85753
2349 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2350 RECORD_TYPE in addition to ARRAY_TYPE.
2352 2018-05-15 Martin Sebor <msebor@redhat.com>
2355 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2357 2018-05-15 Richard Biener <rguenther@suse.de>
2359 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2360 add by_clobber_p one. Change algorithm to collect all defs
2361 representing uses we need to walk and try reducing them to
2362 a single one before failing.
2363 (dse_dom_walker::dse_optimize_stmt): Adjust.
2365 2018-05-13 Mark Wielaard <mark@klomp.org>
2367 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2368 (size_of_loc_descr): Likewise.
2369 (output_loc_operands): Likewise.
2370 (output_loc_operands_raw): Likewise.
2371 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2372 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2373 (hash_loc_operands): Likewise.
2374 (compare_loc_operands): Likewise.
2376 2018-05-14 Mark Wielaard <mark@klomp.org>
2378 * dwarf2out.c (count_index_addrs): New function.
2379 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2381 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2383 PR tree-optimization/83648
2384 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2385 return value as malloc candidate.
2387 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2390 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2391 as true in call to suggest_attribute.
2393 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2395 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2396 -mreadonly-in-sdata.
2398 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2400 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2402 (aarch64_crypto_aesd_fused): Likewise.
2404 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2406 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2407 (movsi_aarch64): Likewise.
2408 (load_pairsi): Likewise.
2409 (load_pairdi): Likewise.
2410 (store_pairsi): Likewise.
2411 (store_pairdi): Likewise.
2412 (load_pairsf): Likewise.
2413 (load_pairdf): Likewise.
2414 (store_pairsf): Likewise.
2415 (store_pairdf): Likewise.
2416 (zero_extend): Likewise.
2417 (trunc): Swap alternatives.
2418 (fcvt_target): Add '?' to prefer w over r.
2420 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2423 * config/i386/i386.md: Disallow non-commutative arithmetics in
2424 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2425 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2426 in the peephole2 before it.
2428 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2430 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2431 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2432 (ix86_handle_option): Handle -mcldemote.
2433 * config.gcc: New header.
2434 * config/i386/cldemoteintrin.h: New file.
2435 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2436 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2438 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2439 OPTION_MASK_ISA_CLDEMOTE.
2440 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2441 (ix86_valid_target_attribute_inner_p): Ditto.
2442 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2443 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2444 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2445 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2446 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2448 * config/i386/i386.opt: Add -mcldemote.
2449 * config/i386/x86intrin.h: New header.
2450 * doc/invoke.texi: Add -mcldemote.
2452 2018-05-14 Richard Biener <rguenther@suse.de>
2454 * doc/match-and-simplify.texi: Adjust :s documentation.
2456 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2458 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2459 intended memcpy size.
2460 (REORDER_45): Likewise.
2462 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2464 * sort.cc: New file.
2465 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2466 * vec.c (qsort_chk): Use gcc_qsort.
2467 * Makefile.in (OBJS-libcommon): Add sort.o.
2468 (build/sort.o): New target. Use it...
2469 (BUILD_RTL): ... here, and...
2470 (build/gencfn-macros): ... here, and...
2471 (build/genmatch): ... here.
2473 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2474 Chung-Ju Wu <jasonwucj@gmail.com>
2476 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2477 * config/nds32/nds32-graywolf.md: New file.
2478 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2479 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2481 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2482 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2483 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2484 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2485 * config/nds32/pipelines.md: Include n15 settings.
2487 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2488 Chung-Ju Wu <jasonwucj@gmail.com>
2490 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2491 * config/nds32/nds32-n13.md: New file.
2492 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2493 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2495 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2496 * config/nds32/nds32.md (pipeline_model): Add n13.
2497 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2498 * config/nds32/pipelines.md: Include n13 settings.
2500 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2501 Chung-Ju Wu <jasonwucj@gmail.com>
2503 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2504 * config/nds32/nds32-n10.md: New file.
2505 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2506 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2508 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2509 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2510 * config/nds32/nds32.md (pipeline_model): Add n10.
2511 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2512 * config/nds32/pipelines.md: Include n10 settings.
2514 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2515 Kito Cheng <kito.cheng@gmail.com>
2516 Chung-Ju Wu <jasonwucj@gmail.com>
2518 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2519 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2520 Add enum values for DSP extension instructions.
2521 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2523 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2524 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2526 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2527 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2528 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2529 * config/nds32/nds32-intrinsic.md: Likewise.
2530 * config/nds32/nds32_intrinsic.h: Likewise.
2531 * config/nds32/nds32-md-auxiliary.c: Likewise.
2532 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2533 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2534 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2535 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2536 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2537 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2538 TYPE_DMAC in switch statement.
2539 * config/nds32/nds32.c: New checking and implementation for DSP
2540 extension instructions.
2541 * config/nds32/nds32.h: Likewise.
2542 * config/nds32/nds32.md: Likewise.
2543 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2544 * config/nds32/predicates.md: Implement new predicates for DSP
2547 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2549 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2550 Reformat alternatives and attributes so it is easier to identify
2551 which constraints/attributes go with which instruction.
2552 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2553 (mov<mode>_softfloat32, FMOVE64): Likewise.
2554 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2555 (mov<mode>_softfloat64, FMOVE64): Likewise.
2557 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2559 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2561 (Basic PowerPC Built-in Functions): The new name of the
2562 subsection previously known as "PowerPC Built-in Functions".
2563 (Basic PowerPC Built-in Functions Available on all Configurations):
2565 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2566 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2567 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2568 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2570 2018-05-11 Martin Jambor <mjambor@suse.cz>
2573 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2576 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2579 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2581 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2583 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2584 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2585 (ix86_handle_option): Handle -mwaitpkg.
2586 * config.gcc: New header.
2587 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2588 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2589 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2591 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2592 OPTION_MASK_ISA_WAITPKG.
2593 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2594 (ix86_option_override_internal): Add PTA_WAITPKG.
2595 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2596 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2597 IX86_BUILTIN_TPAUSE.
2598 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2599 __builtin_ia32_umwait and __builtin_ia32_tpause.
2600 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2601 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2602 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2603 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2604 UNSPECV_TPAUSE): New.
2605 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2606 * config/i386/i386.opt: Add -mwaitpkg.
2607 * config/i386/waitpkgintrin.h: New file.
2608 * config/i386/x86intrin.h: New header.
2609 * doc/invoke.texi: Add -mwaitpkg.
2611 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2614 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2616 (cortex-m0): Use armv6s-m isa.
2617 (cortex-m0plus): Likewise.
2618 (cortex-m1): Likewise.
2619 (cortex-m0.small-multiply): Likewise.
2620 (cortex-m0plus.small-multiply): Likewise.
2621 (cortex-m1.small-multiply): Likewise.
2623 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2624 Jakub Jelinek <jakub@redhat.com>
2626 PR tree-optimization/85692
2627 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2628 source permute as well.
2630 2018-05-11 Martin Liska <mliska@suse.cz>
2633 * doc/extend.texi: Document LLVM style format for no_sanitize
2636 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2638 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2639 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2640 (mode_supports_vsx_dform_quad): Likewise.
2641 (mode_supports_vmx_dform): Move these functions to be next to the
2642 other mode_supports functions.
2643 (mode_supports_dq_form): Likewise.
2644 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2645 mode_supports_dq_form.
2646 (reg_offset_addressing_ok_p): Likewise.
2647 (offsettable_ok_by_alignment): Likewise.
2648 (rs6000_legitimate_offset_address_p): Likewise.
2649 (legitimate_lo_sum_address_p): Likewise.
2650 (rs6000_legitimize_address): Likewise.
2651 (rs6000_legitimize_reload_address): Likewise.
2652 (rs6000_secondary_reload_inner): Likewise.
2653 (rs6000_preferred_reload_class): Likewise.
2654 (rs6000_output_move_128bit): Likewise.
2656 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2658 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2659 Generate SImode target register for null target.
2660 <case IX86_BUILTIN_XGETBV>: Ditto.
2661 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2662 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2664 2018-05-10 Carl Love <cel@us.ibm.com>
2666 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2667 dcbtt and dcbtstt if operands[2] is 0.
2669 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2672 * config/i386/sse.md (usadv64qi): New expander.
2674 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2676 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2677 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2678 -maltivec=be support.
2679 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2680 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2681 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2682 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2683 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2684 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2685 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2686 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2687 altivec_vsumsws): Adjust.
2688 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2689 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2690 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2692 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2693 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2694 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2695 (altivec_lve<VI_char>x): Delete expand.
2696 (*altivec_lve<VI_char>x_internal): Rename to...
2697 (altivec_lve<VI_char>x): ... this.
2698 (altivec_lvxl_<mode>): Delete expand.
2699 (*altivec_lvxl_<mode>_internal): Rename to ...
2700 (altivec_lvxl_<mode>): ... this.
2701 (altivec_stvxl_<mode>): Delete expand.
2702 (*altivec_stvxl_<mode>_internal): Rename to ...
2703 (altivec_stvxl_<mode>): ... this.
2704 (altivec_stve<VI_char>x): Delete expand.
2705 (*altivec_stve<VI_char>x_internal): Rename to ...
2706 (altivec_stve<VI_char>x): ... this.
2707 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2708 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2709 reduc_plus_scal_<mode>): Adjust.
2710 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2712 (rs6000_cpu_cpp_builtins): Adjust.
2713 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2714 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2715 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2716 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2717 -maltivec=be support.
2718 (rs6000_split_vec_extract_var): Adjust.
2719 (rs6000_split_v4si_init): Adjust.
2720 (swap_selector_for_mode): Delete.
2721 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2722 altivec_expand_stvex_be): Delete.
2723 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2724 -maltivec=be support.
2725 (rs6000_gimple_fold_builtin): Ditto.
2726 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2728 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2729 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2730 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2731 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2732 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2733 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2734 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2735 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2736 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2737 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2738 anonymous split): Adjust.
2739 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2740 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2742 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2744 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2745 when --with-gxx-include-dir is also specified.
2746 * configure: Regenerate.
2748 2018-05-09 Jim Wilson <jimw@sifive.com>
2751 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2752 * config/riscv/t-withmultilib: New.
2753 * config/riscv/withmultilib.h: New.
2754 * doc/install.texi: Document RISC-V --with-multilib-list support.
2756 2018-05-09 Richard Biener <rguenther@suse.de>
2758 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2760 (vect_bb_vectorization_profitable_p): Adjust. Compute
2761 actual scalar cost using the cost vector and the add_stmt_cost
2764 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2766 PR rtl-optimization/85645
2767 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2768 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2770 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2772 PR rtl-optimization/85645
2773 * shrink-wrap.c (spread_components): Return a boolean saying if
2774 anything was changed.
2775 (try_shrink_wrapping_separate): Iterate spread_components until
2776 nothing changes anymore.
2778 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2780 PR rtl-optimization/85645
2781 * regrename.c (build_def_use): Also kill the chains that include the
2782 destination of a REG_CFA_REGISTER note.
2784 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2786 PR rtl-optimization/85645
2787 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2788 insn that has a REG_CFA_REGISTER note.
2790 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2792 * cfgexpand.c (expand_clobber): New function.
2793 (expand_gimple_stmt_1): Use it.
2794 * tree-vect-stmts.c (vect_clobber_variable): New function,
2796 (vectorizable_simd_clone_call): ...here.
2797 (vectorizable_store): Emit a clobber either side of an
2798 IFN_STORE_LANES sequence.
2799 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2801 2018-05-09 Tom de Vries <tom@codesourcery.com>
2804 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2805 (define_insn "trap_if_false"): Add exit after trap.
2807 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2809 PR rtl-optimization/85638
2810 * bb-reorder.c: Include common/common-target.h.
2811 (create_forwarder_block): New function extracted from...
2812 (fix_up_crossing_landing_pad): ...here. Rename into...
2813 (dw2_fix_up_crossing_landing_pad): ...this.
2814 (sjlj_fix_up_crossing_landing_pad): New function.
2815 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2816 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2817 from both partitions and exit the loop after one iteration.
2819 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2822 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2824 (Basic PowerPC Built-in Functions): The new name of the
2825 subsection previously known as "PowerPC Built-in Functions".
2826 (Basic PowerPC Built-in Functions Available on all Configurations):
2828 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2830 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2832 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2834 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2837 2018-05-08 Jim Wilson <jimw@sifive.com>
2839 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2840 (LD_EMUL_SUFFIX): New.
2841 (LINK_SPEC): Use it.
2843 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2845 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2847 (Basic PowerPC Built-in Functions): The new name of the
2848 subsection previously known as "PowerPC Built-in Functions".
2849 (Basic PowerPC Built-in Functions Available on all Configurations):
2851 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2853 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2855 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2857 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2860 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2863 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2864 after cmpelim optimization.
2866 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2868 * config.gcc: Support "goldmont".
2869 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2870 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2872 * config/i386/i386.c (m_GOLDMONT): Define.
2873 (processor_target_table): Add "goldmont".
2874 (PTA_GOLDMONT): Define.
2875 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2876 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2877 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2878 (fold_builtin_cpu): Add "goldmont".
2879 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2880 (ix86_option_override_internal): Add "goldmont".
2881 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2882 (processor_type): Add PROCESSOR_GOLDMONT.
2883 * config/i386/i386.md: Add CPU "glm".
2884 * config/i386/glm.md: New file.
2885 * config/i386/x86-tune.def: Add m_GOLDMONT.
2886 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2888 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2891 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2893 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2894 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2895 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2898 * config/i386/i386.c (ix86_fold_builtin): Handle
2899 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2902 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2903 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2905 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2908 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2909 (check_arch): Likewise.
2910 (check_fpu): Return the result rather than printing it.
2911 (end arch): Fix operator precedence.
2912 (end cpu): Likewise.
2913 (END): Print the result from check_fpu.
2915 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2916 Alan Hayward <alan.hayward@arm.com>
2917 David Sherwood <david.sherwood@arm.com>
2919 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2920 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2921 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2922 (*fcmuo<mode>_and): New patterns.
2924 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2926 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2927 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2928 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2929 (cmp_op, sve_imm_con): New code attributes.
2930 (SVE_COND_INT_CMP, imm_con): Delete.
2931 (cmp_op): Remove above unspecs from int attribute.
2932 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2934 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2935 comparison-specific unspecs.
2936 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2937 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2938 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2939 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2940 (*vec_fcm<cmp_op><mode>): Rename to...
2941 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2942 (*vec_fcmuo<mode>): Rename to...
2943 (*fcmuo<mode>): ...this and adjust likewise.
2944 (*pred_fcm<cmp_op><mode>): New pattern.
2945 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2946 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2948 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2950 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2951 (aarch64_emit_sve_predicated_cond): New function.
2952 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2953 (aarch64_emit_unspec_cond_or): Replace with...
2954 (aarch64_emit_sve_or_conds): ...this new function. Use
2955 aarch64_emit_sve_ptrue_op for the individual comparisons and
2956 aarch64_emit_binop to OR them together.
2957 (aarch64_emit_inverted_unspec_cond): Replace with...
2958 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2959 aarch64_emit_sve_ptrue_op for the comparison and
2960 aarch64_emit_unop to invert the result.
2961 (aarch64_expand_sve_vec_cmp_float): Update after the above
2962 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2964 2018-05-07 Nathan Sidwell <nathan@acm.org>
2966 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2967 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2968 (Backwards Compatibility): Likewise.
2970 2018-05-07 Luis Machado <luis.machado@linaro.org>
2974 2018-05-07 Luis Machado <luis.machado@linaro.org>
2976 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2977 <prefetch_dynamic_strides>: New const bool field.
2978 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2979 prefetch_dynamic_strides.
2980 (exynosm1_prefetch_tune): Likewise.
2981 (thunderxt88_prefetch_tune): Likewise.
2982 (thunderx_prefetch_tune): Likewise.
2983 (thunderx2t99_prefetch_tune): Likewise.
2984 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2985 (aarch64_override_options_internal): Update to set
2986 PARAM_PREFETCH_DYNAMIC_STRIDES.
2987 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2988 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2989 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2990 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2991 prefetch-dynamic-strides setting.
2993 2018-05-07 Luis Machado <luis.machado@linaro.org>
2995 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2996 <minimum_stride>: New const int field.
2997 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2998 minimum_stride field.
2999 (exynosm1_prefetch_tune): Likewise.
3000 (thunderxt88_prefetch_tune): Likewise.
3001 (thunderx_prefetch_tune): Likewise.
3002 (thunderx2t99_prefetch_tune): Likewise.
3003 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3004 (aarch64_override_options_internal): Update to set
3005 PARAM_PREFETCH_MINIMUM_STRIDE.
3006 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3007 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3008 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3009 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3010 stride is constant and is below the minimum stride threshold.
3012 2018-05-07 Luis Machado <luis.machado@linaro.org>
3014 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
3017 2018-05-07 Luis Machado <luis.machado@linaro.org>
3019 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3020 <prefetch_dynamic_strides>: New const bool field.
3021 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3022 prefetch_dynamic_strides.
3023 (exynosm1_prefetch_tune): Likewise.
3024 (thunderxt88_prefetch_tune): Likewise.
3025 (thunderx_prefetch_tune): Likewise.
3026 (thunderx2t99_prefetch_tune): Likewise.
3027 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
3028 (aarch64_override_options_internal): Update to set
3029 PARAM_PREFETCH_DYNAMIC_STRIDES.
3030 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3031 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3032 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3033 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3034 prefetch-dynamic-strides setting.
3036 2018-05-07 Luis Machado <luis.machado@linaro.org>
3038 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3039 <minimum_stride>: New const int field.
3040 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3041 minimum_stride field.
3042 (exynosm1_prefetch_tune): Likewise.
3043 (thunderxt88_prefetch_tune): Likewise.
3044 (thunderx_prefetch_tune): Likewise.
3045 (thunderx2t99_prefetch_tune): Likewise.
3046 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
3047 (aarch64_override_options_internal): Update to set
3048 PARAM_PREFETCH_MINIMUM_STRIDE.
3049 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3050 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3051 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3052 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3053 stride is constant and is below the minimum stride threshold.
3055 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3058 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3059 the type is addressable. Don't force op into register if it has
3062 2018-05-05 Roland McGrath <mcgrathr@google.com>
3065 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3066 any section for which we don't know a specific type it should have,
3067 regardless of name. Previously this was done only for the exact
3068 names ".init_array", ".fini_array", and ".preinit_array".
3069 (default_elf_asm_named_section): Add comment about
3070 relationship with default_section_type_flags and SECTION_NOTYPE.
3071 (get_section): Don't consider it a type conflict if one side has
3072 SECTION_NOTYPE and the other doesn't, as long as neither has the
3073 SECTION_BSS et al used in the default_section_type_flags logic.
3075 2018-05-05 Tom de Vries <tom@codesourcery.com>
3078 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3079 (workaround_barsyncs): New function.
3080 (nvptx_reorg): Use workaround_barsyncs.
3081 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3082 (define_expand "nvptx_membar_cta"): New define_expand.
3083 (define_insn "*nvptx_membar_cta"): New insn.
3085 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3087 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3088 To improve optimization opportunities.
3089 * builtin-types.def: The new needed builtin types for the above.
3091 2018-05-04 Richard Biener <rguenther@suse.de>
3093 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3094 * gimple-ssa-store-merging.c
3095 (imm_store_chain_info::output_merged_store): Remove redundant create,
3096 release split_store vector contents on failure.
3097 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3098 scalar stmt vector on cache hit.
3100 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3102 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3104 * config.gcc (powerpc-xilinx-eabi*): Remove.
3105 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3107 (fusion_addis_mem_combo_load): Ditto.
3108 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3110 (rs6000_cpu_cpp_builtins): Ditto.
3111 * config/rs6000/rs6000-linux.c
3112 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3113 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3114 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3116 (rs6000_setup_reg_addr_masks): Ditto.
3117 (rs6000_init_hard_regno_mode_ok): Ditto.
3118 (rs6000_option_override_internal): Ditto.
3119 (legitimate_lo_sum_address_p): Ditto.
3120 (rs6000_legitimize_address): Ditto.
3121 (rs6000_legitimize_reload_address): Ditto.
3122 (rs6000_legitimate_address_p): Ditto.
3123 (abi_v4_pass_in_fpr): Ditto.
3124 (setup_incoming_varargs): Ditto.
3125 (rs6000_gimplify_va_arg): Ditto.
3126 (rs6000_split_multireg_move): Ditto.
3127 (rs6000_savres_strategy): Ditto.
3128 (rs6000_emit_prologue_components): Ditto.
3129 (rs6000_emit_epilogue_components): Ditto.
3130 (rs6000_emit_prologue): Ditto.
3131 (rs6000_emit_epilogue): Ditto.
3132 (rs6000_elf_file_end): Ditto.
3133 (rs6000_function_value): Ditto.
3134 (rs6000_libcall_value): Ditto.
3135 * config/rs6000/rs6000.h: Ditto.
3136 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3137 (TARGET_MINMAX): ... this. New.
3138 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3139 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3140 (*movsi_internal1_single): Delete.
3141 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3142 mfpu=, mxilinx-fpu): Delete.
3143 * config/rs6000/singlefp.h: Delete.
3144 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3145 * config/rs6000/t-rs6000: Ditto.
3146 * config/rs6000/t-xilinx: Delete.
3147 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3148 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3149 (VStype_simple): Delete.
3150 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3151 * config/rs6000/xfpu.h: Delete.
3152 * config/rs6000/xfpu.md: Delete.
3153 * config/rs6000/xilinx.h: Delete.
3154 * config/rs6000/xilinx.opt: Delete.
3155 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3156 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3158 2018-05-04 Tom de Vries <tom@codesourcery.com>
3161 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3164 2018-05-04 Richard Biener <rguenther@suse.de>
3167 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3168 (expand_complex_div_wide): Likewise.
3169 (expand_complex_operations_1): Likewise.
3170 (expand_complex_libcall): Preserve EH info of the original stmt.
3171 (tree_lower_complex): Handle removed blocks.
3172 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3173 on complex multiplication and division libcall builtins.
3175 2018-05-04 Richard Biener <rguenther@suse.de>
3178 * fold-const.c (negate_expr_p): Restrict negation of operand
3179 zero of a division to when we know that can happen without
3181 (fold_negate_expr_1): Likewise.
3183 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3186 * real.h (real_nextafter): Declare.
3187 * real.c (real_nextafter): New function.
3188 * fold-const-call.c (fold_const_nextafter): New function.
3189 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3190 CASE_CFN_NEXTTOWARD.
3191 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3192 even when arg1_mode is different from arg0_mode.
3194 2018-05-03 Nathan Sidwell <nathan@acm.org>
3196 * doc/extend.texi (Deprecated Features): Remove
3198 (Backwards Compatibility): Likewise.
3199 * doc/invoke.texi (C++ Language Options): Likewise.
3200 (C++ Dialect Options): Likewise.
3202 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3205 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3206 _mm512_mask_mullox_epi64): New intrinsics.
3208 2018-05-03 Tom de Vries <tom@codesourcery.com>
3211 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3212 dump files): Add offload-tree.
3214 2018-05-03 Richard Biener <rguenther@suse.de>
3216 PR tree-optimization/85615
3217 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3218 to loops not nested in BBs loop father to avoid creating multi-entry
3221 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3223 PR tree-optimization/70291
3224 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3225 arguments. Change return type to tree. Emit libcall as a new
3226 statement rather than replacing existing one when inplace_p is true.
3227 (expand_complex_multiplication_components): New function.
3228 (expand_complex_multiplication): Expand floating-point complex
3229 multiplication using the above.
3230 (expand_complex_division): Rename inner_type parameter to type.
3231 Update expand_complex_libcall call-site.
3232 (expand_complex_operations_1): Update expand_complex_multiplication
3233 and expand_complex_division call-sites.
3235 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3238 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3239 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3240 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3241 the highest significant bit of the shift count mask is clear. In
3242 check whether and[sq]i3 is needed verify that all significant bits
3243 of the shift count other than the highest are set.
3245 2018-05-02 Tom de Vries <tom@codesourcery.com>
3248 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3249 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3250 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3251 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3252 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3253 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3254 __builtin_goacc_parlevel_size.
3256 2018-05-02 Richard Biener <rguenther@suse.de>
3258 PR tree-optimization/85597
3259 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3260 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3263 2018-05-02 Tom de Vries <tom@codesourcery.com>
3266 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3267 dump files): Add ltrans-tree.
3269 2018-05-02 Tom de Vries <tom@codesourcery.com>
3272 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3273 dump files): Add wpa-ipa.
3275 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3277 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3278 powerpc*-*-linux*paired* target.
3279 * config/rs6000/750cl.h: Delete.
3280 * config/rs6000/paired.h: Delete.
3281 * config/rs6000/paired.md: Delete.
3282 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3284 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3285 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3286 comment. Remove paired float support.
3287 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3288 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3290 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3291 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3293 * config/rs6000/rs6000.c: Remove paired float support.
3294 (paired_expand_vector_init, paired_expand_vector_move,
3295 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3296 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3297 paired_expand_builtin, paired_expand_predicate_builtin,
3298 paired_init_builtins): Delete.
3299 * config/rs6000/rs6000.h: Remove paired float support.
3300 * config/rs6000/rs6000.md: Remove paired float support.
3301 (move_from_CR_ov_bit): Delete.
3302 * config/rs6000/rs6000.opt (mpaired): Delete.
3303 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3304 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3306 2018-05-02 Richard Biener <rguenther@suse.de>
3309 * gimplify.c (gimplify_save_expr): When in SSA form allow
3310 SAVE_EXPRs to compute to SSA vars.
3312 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3315 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3316 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3317 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3318 clobber operands[2], instead use a new pseudo. Formatting fixes.
3320 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3322 PR tree-optimization/85586
3323 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3324 exit early for statements in the same group if the accesses are
3327 2018-05-02 Tom de Vries <tom@codesourcery.com>
3330 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3333 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3335 PR tree-optimization/85143
3336 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3338 2018-05-01 Tom de Vries <tom@codesourcery.com>
3341 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3342 not found" error message.
3344 2018-05-01 Tom de Vries <tom@codesourcery.com>
3347 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3348 * vec.c (test_ordered_remove_if): New function.
3349 (vec_c_tests): Call test_ordered_remove_if.
3350 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3351 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3352 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3353 VEC_ORDERED_REMOVE_IF.
3355 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3357 PR tree-optimization/82665
3358 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3359 pointer subtraction where arguments come from a memchr call.
3361 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3363 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3364 --push-state --as-needed and --pop-state instead of --as-needed and
3365 --no-as-needed if ld supports it.
3366 * configure: Regenerated.
3369 * doc/install.texi2html: Replace _002d with - and _002a with * in
3370 generated html files using sed.
3372 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3375 * gcc-rich-location.c (blank_line_before_p): New function.
3376 (use_new_line): New function.
3377 (gcc_rich_location::add_fixit_insert_formatted): New function.
3378 * gcc-rich-location.h
3379 (gcc_rich_location::add_fixit_insert_formatted): New function.
3381 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3383 * selftest.c (assert_streq): Rename "expected" and "actual" to
3384 "val1" and "val2". Extend NULL-handling to cover both inputs
3385 symmetrically, while still requiring both to be non-NULL for a pass.
3386 * selftest.h (assert_streq): Rename "expected" and "actual" to
3388 (ASSERT_EQ): Likewise.
3389 (ASSERT_EQ_AT): Likewise.
3390 (ASSERT_KNOWN_EQ): Likewise.
3391 (ASSERT_KNOWN_EQ_AT): Likewise.
3392 (ASSERT_NE): Likewise.
3393 (ASSERT_MAYBE_NE): Likewise.
3394 (ASSERT_MAYBE_NE_AT): Likewise.
3395 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3396 the assertion to pass.
3397 (ASSERT_STREQ_AT): Likewise.
3399 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3401 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3402 interaction with -pie.
3404 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3406 * selftest.h: Fix alphabetization of per-source-file selftest
3409 2018-04-30 Jason Merrill <jason@redhat.com>
3411 PR c++/61982 - dead stores to destroyed objects.
3412 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3415 2018-04-30 Jason Merrill <jason@redhat.com>
3417 * tree.c (build_clobber): New.
3418 * tree.h: Declare it.
3419 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3421 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3423 * diagnostic-show-locus.c (layout::layout): Update for
3424 location_get_source_line returning a char_span.
3425 (struct char_span): Move to input.h.
3426 (struct correction): Update for fields in char_span becoming
3428 (struct source_line): Update for location_get_source_line
3429 returning a char_span.
3430 (layout::print_line): Likewise.
3431 * edit-context.c (edited_file::print_content): Likewise.
3432 (edited_file::print_diff_hunk): Likewise.
3433 (edited_file::print_run_of_changed_lines): Likewise.
3434 (edited_file::get_num_lines): Likewise.
3435 (edited_line::edited_line): Likewise.
3436 * final.c (asm_show_source): Likewise.
3437 * input.c (location_get_source_line): Convert return type
3438 from const char * to char_span, losing the final "line_len"
3440 (dump_location_info): Update for the above.
3441 (get_substring_ranges_for_loc): Likewise. Use a char_span
3442 when handling the literal within the line.
3443 (test_reading_source_line): Update for location_get_source_line
3444 returning a char_span.
3445 * input.h (class char_span): Move here from
3446 diagnostic-show-locus.c, converting from a struct to a class.
3447 Make data members private.
3448 (char_span::operator bool): New.
3449 (char_span::length): New.
3450 (char_span::get_buffer): New.
3451 (char_span::operator[]): New.
3452 (char_span::subspan): Make const.
3453 (char_span::xstrdup): New.
3454 (location_get_source_line): Convert return type from const char *
3455 to char_span, losing the final "line_size" param.
3457 2018-04-30 Jan Hubicka <jh@suse.cz>
3459 * lto-wrapper.c (ltrans_priorities): New static var.
3460 (cmp_priority): New.
3461 (run_gcc): Read priorities and if doing parallel build order
3462 the Makefile by them.
3464 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3466 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3468 2018-04-30 Richard Biener <rguenther@suse.de>
3470 * tree-cfg.c (verify_address): Remove base argument, add
3471 flag whether to check TREE_ADDRESSABLE and do that.
3472 (verify_expr): Remove.
3473 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3474 (verify_gimple_assign_single): Likewise.
3475 (verify_gimple_switch): Likewise.
3476 (verify_expr_location_1): Dereference tp once. Add (disabled)
3477 piece from verify_expr.
3478 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3480 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3482 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3484 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3486 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3487 (small_data_pattern): Likewise.
3488 (arc_rewrite_small_data): Likewise.
3489 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3490 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3491 (get_symbol_alignment): New function.
3492 (legitimate_small_data_address_p): Likewise.
3493 (legitimate_scaled_address): Update, call
3494 legitimate_small_data_address_p.
3495 (output_sdata): New static variable.
3496 (arc_print_operand): Update how we handle small data operands.
3497 (arc_print_operand_address): Likewise.
3498 (arc_legitimate_address_p): Update, use
3499 legitimate_small_data_address_p.
3500 (arc_rewrite_small_data_p): Remove.
3501 (arc_rewrite_small_data_1): Likewise.
3502 (arc_rewrite_small_data): Likewise.
3503 (small_data_pattern): Likewise.
3504 (compact_sda_memory_operand): Update to use
3505 legitimate_small_data_address_p and get_symbol_alignment.
3506 (prepare_move_operands): Don't rewite sdata pattern.
3507 (prepare_extend_operands): Remove.
3508 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3510 (zero_extendqisi2): Likewise.
3511 (zero_extendhisi2): Likewise.
3512 (extendqihi2): Likewise.
3513 (extendqisi2): Likewise.
3514 (extendhisi2): Likewise.
3518 * config/arc/constraints.md (Usd): Change it to memory constraint.
3520 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3522 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3523 as source of std instructions.
3524 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3525 allow 6-bit constants as source for store instructions.
3526 (movdi_insn): Update instruction pattern to allow 6-bit constants
3527 as source for store instructions.
3529 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3531 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3533 2018-04-30 Nathan Sidwell <nathan@acm.org>
3534 Sandra Loosemore <sandra@codesourcery.com>
3536 * dumpfile.c (dump_open): Allow '-' for stdout.
3537 * doc/invoke.texi (Developer Options): Document dump filename
3538 determination early. Document stdin/stdout selection.
3540 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3542 Microblaze Target: PIC data text relative
3544 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3545 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3547 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3548 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3549 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3550 New addressing mode for data-text relative position indepenedent code.
3551 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3552 'ADDRESS_SYMBOLIC_TXT_REL'.
3553 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3554 (microblaze_legitimate_pic_operand): Exclude function calls from
3555 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3556 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3558 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3559 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3560 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3561 for 'address + offset'.
3562 (microblaze_expand_prologue): Add new function prologue call for
3564 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3565 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3566 table in case of TARGET_PIC_DATA_TEXT_REL.
3567 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3568 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3569 Add new macros 'UNSPEC_TEXT',
3570 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3571 + exclude function calls from 'UNSPEC_PLT' in case of data text
3573 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3574 new target hook for generating address diff vector tables in case of
3576 * doc/tm.texi : Regenerate.
3577 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3578 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3579 of addr diff vector generation.
3580 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3581 target hook definition.
3582 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3583 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3584 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3585 option for data text relative.
3587 2018-04-30 Richard Biener <rguenther@suse.de>
3589 * tree-chrec.h (evolution_function_is_constant_p): Remove
3591 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3593 2018-04-30 Richard Biener <rguenther@suse.de>
3596 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3598 2018-04-30 Richard Biener <rguenther@suse.de>
3600 PR tree-optimization/28364
3601 PR tree-optimization/85275
3602 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3603 copying first exit test.
3605 2018-04-28 Mark Wielaard <mark@klomp.org>
3607 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3609 (dwarf_AT): Handle DW_AT_addr_base.
3610 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3612 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3615 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3616 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3617 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3618 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3620 2018-04-28 Richard Biener <rguenther@suse.de>
3622 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3623 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3624 to reflect use. Only add interesting stmts.
3626 2018-04-27 Martin Jambor <mjambor@suse.cz>
3629 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3630 the jump function allows for passing through aggregate values.
3632 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3634 * input.h (in_system_header_at): Convert from macro to inline
3636 (from_macro_expansion_at): Likewise.
3637 (from_macro_definition_at): Likewise.
3639 2018-04-27 Jeff Law <law@redhat.com>
3641 * config.gcc: Mark tile* targets as deprecated/obsolete.
3643 2018-04-27 Richard Biener <rguenther@suse.de>
3645 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3648 2018-04-27 Richard Biener <rguenther@suse.de>
3650 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3652 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3654 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3655 with Yd constraint. Set "preferred_for_speed" attribute from
3656 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3658 (*movdi_internal): Ditto.
3659 (movti_interunit splitters): Remove
3660 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3661 (movdi_interunit splitters): Ditto.
3662 * config/i386/constraints.md (Ye): Remove.
3663 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3665 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3668 * config/aarch64/constraints.md (Usg): Limit to 31.
3671 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3673 PR tree-optimization/85529
3674 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3675 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3676 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3677 zero extension or masking of the MSB bit.
3678 (optimize_range_tests): Add FIRST_BB argument, pass it through
3679 to optimize_range_tests_var_bound.
3680 (maybe_optimize_range_tests, reassociate_bb): Adjust
3681 optimize_range_tests callers.
3683 2018-04-26 Richard Biener <rguenther@suse.de>
3684 Jakub Jelinek <jakub@redhat.com>
3686 * cgraph.h (symbol_table): Just declare debug method here.
3687 * symtab.c (symbol_table::debug): Define.
3689 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3691 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3693 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3695 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3696 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3697 (*movdi_internal): Substitute Yi and Yj constraint with x
3698 and Ym and Yn constraint with y constraint. Update "isa"
3699 attribute and set "preferred_for_speed" attribute from
3700 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3701 (*movsi_internal): Ditto.
3702 (*movdf_internal): Ditto.
3703 (*movsf_internal): Ditto.
3704 (*zero_extendsidi2): Ditto.
3705 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3706 (sse2_loadld): Ditto.
3707 (*vec_extract<ssevecmodelower>_0): Ditto.
3708 (*vec_extractv4si_0_zext_sse4): Ditto.
3709 (vec_concatv2di): Ditto.
3710 (*vec_dup<mode>): Ditto.
3711 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3712 * config/i386/constraints.md (Yi): Remove.
3717 2018-04-26 Nathan Sidwell <nathan@acm.org>
3719 * dumpfile.c (dump_open): New.
3720 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3721 (dump_finish): Detect stdio/stderr by value not name.
3723 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3725 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3727 2018-04-26 Tom de Vries <tom@codesourcery.com>
3730 * config/nvptx/nvptx.c (verify_neutering_jumps)
3731 (verify_neutering_labels): New function
3732 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3734 2018-04-26 Tom de Vries <tom@codesourcery.com>
3737 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3738 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3741 2018-04-26 Richard Biener <rguenther@suse.de>
3742 Tom de Vries <tom@codesourcery.com>
3745 * lto-streamer-out.c (output_function): Fixup loops if required to match
3746 discovery done in the reader.
3748 2018-04-26 Richard Biener <rguenther@suse.de>
3750 PR tree-optimization/85116
3751 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3752 have a loop exit from the single latch predecessor. Remove
3753 case of header with just condition.
3754 (ch_base::copy_headers): Exclude infinite loops from any
3756 (pass_ch::execute): Record exits.
3758 2018-04-26 Richard Biener <rguenther@suse.de>
3760 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3761 prologue cost vector and pass it to vect_get_load_cost.
3762 (vect_get_peeling_costs_all_drs): Likewise.
3763 (vect_peeling_hash_get_lowest_cost): Likewise.
3764 (vect_enhance_data_refs_alignment): Likewise.
3766 2018-04-26 Richard Biener <rguenther@suse.de>
3769 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3770 checking of integer<->pointer conversions.
3771 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3772 sign-/zero-extending pointer types.
3773 (expand_omp_for_static_chunk): Likewise.
3775 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3776 Jean Lee <xiaoyur347@gmail.com>
3778 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3779 (TARGET_ASAN_SHADOW_OFFSET): Define.
3780 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3781 true for -fsanitize=address.
3783 2018-04-25 Mark Wielaard <mark@klomp.org>
3785 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3788 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3790 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3791 than "alu", remove explicit "memory" and "imm_disp" attributes.
3792 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3795 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3796 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3797 gen_lowpart_no_emit.
3799 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3802 * config/i386/i386.c (ix86_expand_builtin): Change memory
3803 operand to XI, extend p0 to Pmode.
3804 * config/i386/i386.md: Change unspec volatile and operand
3805 1 mode to XI, change operand 0 mode to P.
3807 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3809 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3810 GET_MODE_MASK before any checking.
3811 (nds32_can_use_bset_p): Likewise.
3812 (nds32_can_use_btgl_p): Likewise.
3814 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3816 * config/nds32/nds32-doubleword.md: New define_split pattern for
3817 illegal register number.
3819 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3821 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3823 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3825 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3827 2018-04-25 Richard Biener <rguenther@suse.de>
3829 * lto-streamer.h (LTO_major_version): Bump to 8.
3831 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3833 * BASE-VER: Set to 9.0.0.
3835 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3837 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3838 in __abskf2 and __powikf2.
3840 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3843 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3844 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3845 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3846 Use the above on operand 2. Reindent.
3847 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3849 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3852 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3854 * config/i386/i386.opt (mcet): Removed.
3855 * doc/install.texi: Remove -mcet documentation.
3856 * doc/invoke.texi: Likewise.
3858 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3861 * doc/install.texi: Remove -mcet from bootstrap-cet.
3863 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3866 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3867 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3871 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3872 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3873 containing a CONST_VECTOR.
3875 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3877 * doc/install.texi: Update newlib dependency for nvptx.
3879 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3882 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3883 instead of INTVAL when shifting x left.
3885 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3887 PR tree-optimization/85478
3888 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3889 vect_grouped_store_supported for single element vectors.
3891 2018-04-24 Richard Biener <rguenther@suse.de>
3894 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3895 load cost increase to the case of non-constant step.
3897 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3900 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3901 destination if any_malformed_asm.
3903 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3906 * expr.c (store_field): In the bitfield case, if the value comes from
3907 a function call and is returned in registers by means of a PARALLEL,
3908 do not change the mode of the temporary unless BLKmode and VOIDmode.
3910 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3912 PR rtl-optimization/85423
3913 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3914 dependencies to debug insns when the previous insn is non-debug.
3916 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3918 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3919 enums into a single definition.
3920 (fls): Fix predicates and printing.
3923 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3925 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3926 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3927 and short u6 immediate.
3928 (check_if_valid_sleep_operand): Remove.
3929 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3931 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3933 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3934 flag_always_save_lp condition.
3935 * config/nds32/nds32.opt (malways-save-lp): New option.
3937 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3939 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3940 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3941 * config/nds32/nds32.h
3942 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3943 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3945 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3947 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3948 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3950 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3951 Chung-Ju Wu <jasonwucj@gmail.com>
3953 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3955 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3956 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3958 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3960 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3962 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3964 * config/nds32/nds32-protos.h (nds32_data_alignment,
3965 nds32_local_alignment): Declare.
3966 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3967 nds32_local_alignment): New functions.
3968 (TARGET_CONSTANT_ALIGNMENT): Define.
3969 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3971 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3973 * config/nds32/nds32.c
3974 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3975 (TARGET_MODES_TIEABLE_P): Likewise.
3977 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3979 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3982 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3983 Chung-Ju Wu <jasonwucj@gmail.com>
3985 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3986 for unaligned access.
3987 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3989 * config/nds32/nds32-intrinsic.md: Likewise.
3990 * config/nds32/nds32_intrinsic.h: Likewise.
3991 * config/nds32/nds32.h (nds32_builtins): Likewise.
3992 * config/nds32/nds32.opt (munaligned-access): New option.
3993 * config/nds32/nds32.c (nds32_asm_file_start): Display
3994 flag_unaligned_access status.
3996 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3998 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3999 -mno-relax is present.
4000 * config/riscv/linux.h (LINK_SPEC): Ditto.
4002 2018-04-20 Martin Sebor <msebor@redhat.com>
4005 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
4007 (gimple_fold_builtin_stxcpy_chk): Same.
4008 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
4010 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
4013 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
4014 __powikf2 when long double is IEEE 128-bit.
4016 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
4018 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
4019 step to make sure stack always aligned.
4021 2018-04-20 Carl Love <cel@us.ibm.com>
4024 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
4025 size check for arg0.
4027 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
4028 Tom de Vries <tom@codesourcery.com>
4031 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
4032 Emit insns for calls too.
4033 (nvptx_find_par): Always look for worker-level predecessor insn.
4034 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
4036 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
4037 (nvptx_process_pars): Propagate frames for calls.
4039 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
4042 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
4044 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
4045 (ix86_handle_option): Don't handle OPT_mibt.
4046 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
4048 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4050 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4051 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4052 (ix86_target_macros): Define __CET__ with flag_cf_protection
4053 for -fcf-protection.
4054 * config/i386/i386.c (isa2_opts): Remove -mibt.
4055 * config/i386/i386.h (TARGET_IBT): Removed.
4056 (TARGET_IBT_P): Likewise.
4057 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4058 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4059 * config/i386/i386.opt (mcet): Update help message.
4062 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4063 -mcet as an alias for -mshstk.
4065 2018-04-20 Richard Biener <rguenther@suse.de>
4068 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4069 complexity by forcing a single use of the multiply operand.
4071 2018-04-20 Martin Jambor <mjambor@suse.cz>
4074 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4075 recursion dependency to only apply to non-clones.
4077 2018-04-20 Martin Jambor <mjambor@suse.cz>
4080 * ipa-cp.c (create_specialized_node): Check that clones of
4081 self-recursive edges exist during IPA-CP.
4083 2018-04-19 Toon Moene <toon@moene.org>
4085 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4088 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4090 PR tree-optimization/85467
4091 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4092 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4093 VECTOR_CST element to type.
4095 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4098 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4099 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4100 (builtin_longjmp): Likewise.
4101 (save_stack_nonlocal): New pattern.
4102 (restore_stack_nonlocal): Likewise.
4104 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4107 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4108 Replace ASM_OUTPUT_LABEL with fprintf.
4110 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4113 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4114 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4115 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4116 define __IBT__ and __SHSTK__ for -fcf-protection.
4117 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4119 (ix86_trampoline_init): Likewise.
4120 (x86_output_mi_thunk): Likewise.
4121 (ix86_notrack_prefixed_insn_p): Likewise.
4122 (ix86_option_override_internal): Don't disallow -fcf-protection.
4123 * config/i386/i386.md (rdssp<mode>): Also enable for
4125 (incssp<mode>): Likewise.
4126 (nop_endbr): Likewise.
4127 * config/i386/i386.opt (mcet): Change help message to built-in
4131 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4132 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4133 enable CET built-in functions.
4135 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4137 * common/config/i386/i386-common.c
4138 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4139 OPTION_MASK_ISA_MOVDIRI_UNSET,
4140 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4141 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4142 * config.gcc (movdirintrin.h): New header.
4143 * config/i386/cpuid.h (bit_MOVDIRI,
4144 bit_MOVDIR64B): New bits.
4145 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4147 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4148 (VOID, PVOID, PCVOID)): New function types.
4149 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4150 __builtin_ia32_directstoreu_u64,
4151 __builtin_ia32_movdir64b): New builtins.
4152 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4153 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4155 (ix86_valid_target_attribute_inner_p): Ditto.
4156 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4157 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4158 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4159 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4160 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4161 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4162 (movdiri<mode>, movdir64b_<mode>): New.
4163 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4164 * config/i386/immintrin.h: Include movdirintrin.h.
4165 * config/i386/movdirintrin.h: New file.
4166 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4168 2018-04-19 Richard Biener <rguenther@suse.de>
4171 * cfg.c (clear_bb_flags): When loop state says we have
4172 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4174 2018-04-19 Richard Biener <rguenther@suse.de>
4176 PR tree-optimization/84737
4177 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4178 copying restrict info.
4179 (vect_setup_realignment): Use it.
4180 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4181 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4182 the first DR to all generated stores.
4183 (vectorizable_load): Likewise for loads.
4185 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4187 PR tree-optimization/85446
4188 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4189 the integral and pointer types to have the same precision.
4191 * doc/install.texi: Document --disable-cet being the default and
4194 2018-04-18 Martin Liska <mliska@suse.cz>
4196 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4199 2018-04-18 Martin Liska <mliska@suse.cz>
4202 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4205 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4206 arguments if they are comparable.
4208 2018-04-18 Martin Liska <mliska@suse.cz>
4211 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4214 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4217 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4220 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4221 ENDBR after calling __morestack.
4223 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4226 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4227 by using gcc_base_ver to generate a gcc_driver_version, and use
4228 it when generating GCC_DRIVER_NAME.
4229 * configure: Regenerate.
4231 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4234 * config.gcc: Obsolete powerpc*-*-*spe*.
4236 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4239 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4240 (stabstr_D): Change type of unum from unsigned int to
4241 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4244 2018-04-17 Jim Wilson <jimw@sifive.com>
4247 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4248 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4249 Set arg_pointer_offset after using pretend_args_size.
4251 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4253 PR rtl-optimization/85431
4254 * dse.c (record_store): Ignore zero width stores.
4257 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4258 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4259 __builtin_stack_restore rather than after it.
4260 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4261 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4262 argument instead of virtual_dynamic_stack_rtx.
4264 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4266 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4268 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4269 Add note to error message to explain internal mapping of overloaded
4270 built-in function name to non-overloaded built-in function name.
4271 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4274 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4277 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4278 where the inputs overlap with the output.
4280 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4282 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4283 (=v, v) alternative and explicit "memory" attribute.
4284 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4285 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4287 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4288 "sselog1" type instead of "sselog".
4289 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4290 "sselog". Remove explicit "memory" attribute.
4291 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4292 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4294 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4295 "sselog1" type instead of "sselog". Remove explicit "memory"
4297 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4298 use "sselog1" type instead of "sselog". Remove explicit "memory"
4300 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4301 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4303 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4304 "sselog1" type instead of "sselog". Remove explicit "memory"
4306 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4307 use "sselog1" type instead of "sselog". Remove explicit "memory"
4311 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4314 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4317 2018-04-17 Martin Jambor <mjambor@suse.cz>
4320 * ipa-cp.c (create_specialized_node): Call
4321 expand_all_artificial_thunks if necessary.
4323 2018-04-17 Martin Liska <mliska@suse.cz>
4326 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4327 in message, remote space in between '_G' and '('.
4329 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4332 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4333 avx512f_vmcmp<mode>3<round_saeonly_name>,
4334 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4335 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4336 avx512f_rndscale<mode><round_saeonly_name>,
4337 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4338 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4339 Use %<iptr>2 instead of %2 for -masm=intel.
4340 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4341 avx512f_vcvttss2usi<round_saeonly_name>,
4342 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4344 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4345 avx512f_vcvttsd2usi<round_saeonly_name>,
4346 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4347 Use %q1 instead of %1 for -masm=intel.
4348 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4349 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4350 of %3 for -masm=intel.
4351 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4353 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4355 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4357 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4358 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4360 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4362 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4363 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4364 %g1 and one with %0 and %1.
4365 (avx512er_vmrcp28<mode><round_saeonly_name>,
4366 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4368 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4369 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4370 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4371 of %0 and %{%4%} for -masm=intel.
4372 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4373 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4374 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4375 order of %0 and %{%5%}%{z%} for -masm=intel.
4377 2018-04-17 Jan Hubicka <jh@suse.cz>
4380 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4382 2018-04-17 Martin Liska <mliska@suse.cz>
4385 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4386 for target_clone error message. Make default implementation
4387 clone to be a local declaration.
4388 (separate_attrs): Add new argument and check for an empty
4390 (expand_target_clones): Handle it.
4391 (ipa_target_clone): Make redirection just for target_clones
4394 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4395 Tom de Vries <tom@codesourcery.com>
4398 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4399 tiled basic blocks without omp continue statements.
4401 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4404 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4405 vec_extract expression as having side effects to make sure it gets
4408 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4411 * config/i386/i386.c (get_builtin_code_for_version): Check
4414 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4417 * gcc/config.gcc: Support "skylake".
4418 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4420 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4421 (processor_target_table): Add "skylake".
4422 (ix86_option_override_internal): Add "skylake".
4423 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4424 PROCESSOR_CANNONLAKE.
4425 (get_builtin_code_for_version): Fix priority for
4426 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4427 PROCESSOR_SKYLAKE-AVX512.
4428 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4429 (processor_type): Add PROCESSOR_SKYLAKE.
4431 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4432 Jason Merrill <jason@redhat.com>
4435 * convert.c (convert_to_integer_1): Use direct recursion for
4436 enumeral types and types with a precision less than the number
4437 of bits in their mode.
4439 2018-04-16 Julia Koval <julia.koval@intel.com>
4442 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4443 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4445 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4448 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4449 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4450 and -mno-direct-move.
4452 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4455 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4456 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4457 Ensure negative shifts result in {0}.
4459 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4461 PR rtl-optimization/79916
4462 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4463 regs (if any) to define how to gnerate SD moves when LRA is in
4466 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4468 PR rtl-optimization/85393
4469 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4470 * except.c (expand_dw2_landing_pad_for_region): Make static.
4471 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4472 a label and unconditional jump to old_bb, rather than
4473 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4476 PR rtl-optimization/85376
4477 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4478 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4479 instead of a specific value.
4481 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4482 Bin Cheng <bin.cheng@arm.com>
4484 PR tree-optimization/82965
4485 PR tree-optimization/83991
4486 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4487 by_profile_only parameter.
4488 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4489 information if the loop was predicted to iterate too many times.
4490 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4492 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4495 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4498 2018-04-13 Martin Liska <mliska@suse.cz>
4499 Jakub Jelinek <jakub@redhat.com>
4502 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4503 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4504 * builtins.c (expand_builtin_memory_copy_args): Use
4505 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4506 handle dest_addr == pc_rtx.
4508 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4511 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4512 asked to not generate direct moves.
4513 (fix_trunc<mode>si2_stfiwx): Similar.
4514 (fix_trunc<mode>si2_internal): Similar.
4516 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4519 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4520 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4521 lookup if dest in some wider mode is known to be const0_rtx and
4522 if so, record permanent equivalence for it to be ZERO_EXTEND of
4523 the narrower mode destination.
4525 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4527 * lto-streamer-out.c (output_function): Revert 259346.
4528 * omp-expand.c (expand_oacc_for): Likewise.
4530 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4532 PR rtl-optimization/85354
4533 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4534 * sel-sched.c (sel_global_init): ... here.
4536 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4539 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4540 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4541 mode for PE-COFF targets.
4542 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4543 (i386_pe_asm_lto_end): Likewise.
4544 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4545 (TARGET_ASM_LTO_END): Likewise.
4546 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4547 (i386_pe_asm_lto_start): New function.
4548 (i386_pe_asm_lto_end): Likewise.
4550 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4551 Richard Biener <rguenther@suse.de>
4554 * lto-streamer-out.c (output_function): Fix CFG loop state before
4556 * omp-expand.c (expand_oacc_for): Handle calls to internal
4557 functions like regular functions.
4559 2018-04-12 Richard Biener <rguenther@suse.de>
4562 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4563 for the early LTO debug to properly generate references to it
4564 during DIE emission. Do not re-use that for the skeleton for
4566 (dwarf2out_early_finish): Likewise.
4568 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4571 * config/i386/sse.md
4572 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4573 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4574 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4575 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4576 and output is a reg, avoid creating invalid lowpart subreg, but
4577 instead split into a 512-bit move. Don't split if not AVX512VL,
4578 input is xmm16+ reg and output is a mem.
4579 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4580 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4581 xmm16+ reg and output is a mem.
4583 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4585 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4586 also for flag_dwarf2_cfi_asm.
4588 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4590 PR rtl-optimization/85342
4591 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4592 a bool scalar var inside of the loop instead. Don't try to update
4593 recog_data.operand after failed apply_change_group.
4595 2018-04-12 Tom de Vries <tom@codesourcery.com>
4598 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4599 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4600 array with flexible array member as array without given dimension.
4601 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4602 argument for undefined param to true.
4604 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4607 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4608 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4609 from PowerPC section.
4610 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4611 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4612 help text that is too long.
4613 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4614 help text that is too long.
4615 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4616 help text that is too long.
4618 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4620 * config/alpha/alpha.md (stack_probe_internal): Rename
4621 from "probe_stack". Update all callers.
4623 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4625 PR rtl-optimization/84566
4626 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4627 sched_macro_fuse_insns.
4629 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4632 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4633 (compute_block_dependences): ... from here.
4635 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4637 PR tree-optimization/85331
4638 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4639 from int to HOST_WIDE_INT.
4641 2018-04-11 Martin Jambor <mjambor@suse.cz>
4644 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4645 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4646 not the same as the source val.
4647 (cgraph_edge_brings_value_p): New parameter.
4648 (gather_edges_for_value): Pass destination value to
4649 cgraph_edge_brings_value_p.
4650 (perhaps_add_new_callers): Likewise.
4651 (get_info_about_necessary_edges): Likewise and exclude values brought
4652 only by self-recursive edges.
4653 (create_specialized_node): Redirect only clones of self-calling edges.
4654 (+self_recursive_pass_through_p): New function.
4655 (find_more_scalar_values_for_callers_subset): Use it.
4656 (find_aggregate_values_for_callers_subset): Likewise.
4657 (known_aggs_to_agg_replacement_list): Removed.
4658 (decide_whether_version_node): Re-calculate known constants for all
4659 remaining context clones.
4661 2018-04-11 Richard Biener <rguenther@suse.de>
4664 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4665 from early DWARF output.
4666 (dwarf2out_early_finish): Output line info unconditionally into
4667 early DWARF and add reference to it.
4669 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4672 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4673 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4674 other than V2DFmode using iptr mode attribute.
4675 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4677 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4679 PR rtl-optimization/84659
4680 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4682 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4685 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4687 (output_loc_list): Pass address of a dummy size variable even in the
4688 locview handling loop.
4689 (index_location_lists): Add comment on why skip_loc_list_entry can't
4692 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4695 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4698 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4701 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4702 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4703 and -mstring-compare-inline-limit.
4705 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4708 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4709 for stack clash protection in a register whenever we need it to be in
4712 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4714 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4715 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4717 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4720 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4722 (mlong-double-): Ditto.
4723 * config/rs6000/sysv4.opt (msdata=): Ditto.
4724 (mtls-size=): Ditto.
4726 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4728 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4729 erroneous entries for
4730 "vector int vec_ldl (int, long int *)", and
4731 "vector unsigned int vec_ldl (int, unsigned long int *)".
4732 Add comments and entries for
4733 "vector bool char vec_ldl (int, bool char *)",
4734 "vector bool short vec_ldl (int, bool short *)",
4735 "vector bool int vec_ldl (int, bool int *)",
4736 "vector bool long long vec_ldl (int, bool long long *)",
4737 "vector pixel vec_ldl (int, pixel *)",
4738 "vector long long vec_ldl (int, long long *)",
4739 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4740 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4741 type tree bool_long_long_type_node and correct definition of
4742 bool_V2DI_type_node to make reference to this new type tree.
4743 (rs6000_mangle_type): Replace erroneous reference to
4744 bool_long_type_node with bool_long_long_type_node.
4745 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4746 comments to emphasize sign distinctions for char and int types and
4747 replace RS6000_BTI_bool_long constant with
4748 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4749 use of RS6000_BTI_pixel.
4750 (bool_long_type_node): Remove this macro definition.
4751 (bool_long_long_type_node): New macro definition
4753 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4755 PR rtl-optimization/85300
4756 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4757 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4758 simplify_unary_operation fails.
4760 2018-04-10 Martin Liska <mliska@suse.cz>
4762 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4763 cgraph_edge and ipa_ref.
4765 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4769 * config/i386/sse.md
4770 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4771 computation of the VEC_MERGE selector from mask.
4772 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4773 Fix decoding of the VEC_MERGE selector into mask.
4775 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4777 PR tree-optimization/85286
4778 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4780 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4782 * final.c (final_1): Set insn_last_address as well as
4783 insn_current_address.
4785 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4788 * explow.c (emit_stack_probe): Call validize_mem on memory location
4789 before passing it to gen_probe_stack. Create address operand and
4790 legitimize it for the probe_stack_address case.
4792 2018-04-09 Jan Hubicka <jh@suse.cz>
4795 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4796 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4797 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4799 (free_lang_data): Rebuild type inheritance graph.
4801 2018-04-09 Martin Sebor <msebor@redhat.com>
4803 * invoke.texi (-finline-small-functions): Mention other optimization
4805 (-findirect-inlining, -fpartial-inlining): Same.
4806 (-finline-functions-called-once): Same.
4807 (-freorder-blocks-and-partition): Same.
4809 2018-04-09 Jan Hubicka <jh@suse.cz>
4812 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4813 jumps; choose last target that matches the criteria (i.e.
4814 no partition changes for non-crossing jumps).
4815 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4816 support for redirecting crossing jumps to non-crossing.
4818 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4820 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4821 also for naked functions.
4823 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4825 * config/arc/arc.md (add_shift): New pattern.
4826 (add_shift2): Likewise.
4827 (sub_shift): Likewise.
4828 (sub_shift_cmp0_noout): Likewise.
4829 (compare_si_ashiftsi): Likewise.
4830 (xbfu_cmp0_noout): New combine pattern.
4831 (xbfu_cmp0"): Likewise.
4832 (movsi_set_cc_insn): Place the predicable variant first.
4833 (commutative_binary_cmp0_noout): Remove clobber.
4834 (commutative_binary_cmp0): New pattern.
4835 (noncommutative_binary_cmp0): Likewise.
4836 (noncommutative_binary_cmp0_noout): Likewise.
4837 (noncommutative_binary_comparison_result_used): Removed.
4838 (rsub_cmp0): New pattern.
4839 (rsub_cmp0_noout): Likewise.
4840 (extzvsi): Changed, keep only meaningful variants.
4841 (SQH, SEZ): New iterators.
4842 (SQH_postfix): New mode attribute.
4843 (SEZ_prefix): New code attribute.
4844 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4845 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4846 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4848 (noncommutative_operator): Check the availability of barrel
4851 2018-04-09 Richard Biener <rguenther@suse.de>
4853 PR tree-optimization/85284
4854 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4855 Only use the niter constraining form of simple_iv when the exit
4858 2018-04-09 Tom de Vries <tom@codesourcery.com>
4861 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4862 (define_expand "*memory_barrier"): New define_expand.
4863 (define_insn "memory_barrier"): New insn.
4865 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4867 PR rtl-optimization/80463
4868 PR rtl-optimization/83972
4869 PR rtl-optimization/83480
4871 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4872 correct producer for the insn.
4873 (tidy_control_flow): Fixup seqnos in case of debug insns.
4875 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4877 PR rtl-optimization/83913
4879 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4880 different sched-times when merging exprs.
4882 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4884 PR rtl-optimization/83962
4886 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4887 tidy_fallthru_edge and tidy_control_flow.
4889 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4891 PR rtl-optimization/83530
4893 * sel-sched.c (force_next_insn): New global variable.
4894 (remove_insn_for_debug): When force_next_insn is true, also leave only
4895 next insn in the ready list.
4896 (sel_sched_region): When the region wasn't scheduled, make another pass
4897 over it with force_next_insn set to 1.
4899 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4901 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4903 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4904 for interrupt control.
4905 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4906 functions for interrupt control.
4907 * config/nds32/nds32-intrinsic.md: Likewise.
4908 * config/nds32/nds32_intrinsic.h: Likewise.
4909 * config/nds32/nds32.h (nds32_builtins): Likewise.
4911 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4913 * config/nds32/nds32.c (nds32_init_machine_status,
4914 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4915 strict_aligned_p field.
4916 (nds32_expand_to_rtl_hook): New function.
4917 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4918 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4920 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4921 Chung-Ju Wu <jasonwucj@gmail.com>
4923 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4924 * config/nds32/nds32-n7.md: New file.
4925 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4926 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4928 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4929 * config/nds32/nds32.md (pipeline_model): Add n7.
4930 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4931 * config/nds32/pipelines.md: Include n7 settings.
4933 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4934 Chung-Ju Wu <jasonwucj@gmail.com>
4936 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4937 * config/nds32/nds32-e8.md: New file.
4938 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4939 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4941 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4942 * config/nds32/nds32.md (pipeline_model): Add e8.
4943 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4944 * config/nds32/pipelines.md: Include e8 settings.
4946 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4947 Chung-Ju Wu <jasonwucj@gmail.com>
4949 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4950 * config/nds32/nds32-n8.md: New file.
4951 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4952 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4954 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4955 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4956 * config/nds32/nds32.md (pipeline_model): Add n8.
4957 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4958 * config/nds32/pipelines.md: Include n8 settings.
4960 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4961 Chung-Ju Wu <jasonwucj@gmail.com>
4963 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4964 * config/nds32/nds32-n9-2r1w.md: New file.
4965 * config/nds32/nds32-n9-3r2w.md: New file.
4966 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4967 nds32_register_ports): New or modify for cpu n9.
4968 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4970 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4971 * config/nds32/nds32-utils.c: New file.
4972 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4973 TARGET_MUL_SLOW): Define.
4974 * config/nds32/nds32.md (pipeline_model): New attribute.
4975 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4976 New options that support cpu n9.
4977 * config/nds32/pipelines.md: Include n9 settings.
4978 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4980 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4982 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4983 information if necessary.
4984 (output_cond_branch_compare_zero): Likewise.
4985 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4986 (nds32_target_alignment): Refine for alignment.
4987 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4988 (FUNCTION_BOUNDARY): Modify.
4989 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4991 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4993 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4995 * config/nds32/constants.md (unspec_volatile_element): Add values for
4996 TLB operation and data prefetch.
4997 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4998 functions for TLB operation and data prefetch.
4999 * config/nds32/nds32-intrinsic.md: Likewise.
5000 * config/nds32/nds32_intrinsic.h: Likewise.
5001 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
5002 (nds32_print_operand): Likewise.
5003 * config/nds32/nds32.h (nds32_builtins): Likewise.
5005 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
5006 Andrew Pinski <pinsika@gcc.gnu.org>
5009 * match.pd: Use constant_boolean_node of correct type instead of
5010 boolean_true_node or boolean_false_node for simplifying
5011 pointer comparisons to zero.
5013 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5015 PR tree-optimization/80021
5016 * tree.c (verify_type_variant): Make error call in verify_variant_match
5017 translatable and remove final full stop.
5019 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * config/nds32/constants.md (unspec_volatile_element): Add
5022 UNSPEC_VOLATILE_EH_RETURN.
5023 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
5024 nds32_output_stack_pop): Support dwarf exception handling process.
5025 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
5026 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
5027 exception handling process.
5028 (nds32_compute_stack_frame): Likewise.
5029 (nds32_return_addr_rtx): Likewise.
5030 (nds32_initial_elimination_offset): Likewise.
5031 (nds32_expand_prologue): Likewise.
5032 (nds32_expand_epilogue): Likewise.
5033 (nds32_dynamic_chain_address): New function.
5034 * config/nds32/nds32.h (machine_function): Add fields for dwarf
5036 (DYNAMIC_CHAIN_ADDRESS): Define.
5037 (EH_RETURN_DATA_REGNO): Define.
5038 (EH_RETURN_STACKADJ_RTX): Define.
5039 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
5040 patterns for dwarf exception handling.
5042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.h: Clean up obsolete macros.
5046 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5048 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5049 Add enum values for particular instructions.
5050 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5051 particular intrinsic functions.
5052 * config/nds32/nds32-intrinsic.md: Likewise.
5053 * config/nds32/nds32_intrinsic.h: Likewise.
5054 * config/nds32/nds32.h (nds32_builtins): Likewise.
5055 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5056 (btst, ave): New patterns for particular instructions.
5058 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5060 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5061 Add enum values for atomic load/store and memory sync.
5062 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5064 * config/nds32/nds32-intrinsic.md: Likewise.
5065 * config/nds32/nds32_intrinsic.h: Likewise.
5066 * config/nds32/nds32.h (nds32_builtins): Likewise.
5068 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5070 PR tree-optimization/85257
5071 * fold-const.c (native_encode_vector): If not all elts could fit
5072 and off is -1, return 0 rather than offset.
5073 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5074 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5075 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5076 adjust buffer in native_interpret_expr call.
5078 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5080 * config/nds32/constants.md (unspec_volatile_element): Add cache
5081 control enum values.
5082 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5083 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5084 * config/nds32/nds32.c (nds32_cctl_names): New.
5085 (nds32_print_operand): Handle cache control register names.
5086 * config/nds32/nds32.h (nds32_builtins): New enum values.
5087 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5089 * config/nds32/nds32.md (type): Add mmu.
5090 * config/nds32/pipelines.md (simple_insn): Add mmu.
5092 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5094 * config/nds32/nds32.md (type): Remove call.
5095 * config/nds32/pipelines.md (simple_insn): Likewise.
5097 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5099 * config/nds32/constants.md (unspec_volatile_element): Add
5100 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5101 UNSPEC_VOLATILE_FMFCFG.
5102 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5103 description for fmfcfg and fmfcsr.
5104 (bdesc_1arg): Add fmtcsr.
5105 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5106 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5107 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5108 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5109 unspec_fmfcfg): New patterns.
5110 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5111 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5112 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5113 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5114 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5115 __nds32__fmfcfg): Define.
5117 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5119 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5120 intrinsic register names.
5121 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5122 intrinsic register enum values and macros.
5124 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5126 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5127 for load/store addressing form.
5128 (nds32_print_operand_address): Likewise.
5130 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5133 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5134 based on LABEL_REF. Remove useless assertion.
5135 (pic_address_needs_scratch): Fix formatting.
5136 (sparc_legitimize_pic_address): Minor tweaks.
5137 (sparc_delegitimize_address): Adjust assertion accordingly.
5138 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5139 into symbolic_operand.
5140 (movsi_high_pic_label_ref): Likewise.
5141 (movsi_lo_sum_pic_label_ref): Likewise.
5142 (movdi_pic_label_ref): Likewise.
5143 (movdi_high_pic_label_ref): Likewise.
5144 (movdi_lo_sum_pic_label_ref): Likewise.
5146 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5148 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5149 custom LIB_SPEC setup.
5151 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5152 Kito Cheng <kito.cheng@gmail.com>
5154 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5155 * config/riscv/freebsd.h: New.
5157 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5159 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5160 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5163 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5164 Kito Cheng <kito.cheng@gmail.com>
5166 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5167 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5168 * config/nds32/nds32-protos.h (nds32_output_call,
5169 nds32_output_return): Declare.
5170 * config/nds32/nds32.md: Refine all the call and return patterns.
5172 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5175 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5176 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5178 PR rtl-optimization/84872
5179 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5180 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5183 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5185 * expr.c (copy_blkmode_to_reg): Revert 254862.
5186 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5188 2018-04-06 Richard Biener <rguenther@suse.de>
5191 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5192 after seeing a component reference with an adjacent field. Treat
5193 refs to arrays at struct end of external decls similar to
5194 refs to unconstrained commons.
5196 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5199 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5200 look through SAVE_EXPRs with non-side-effects argument. Adjust
5202 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5205 2018-04-06 Richard Biener <rguenther@suse.de>
5208 * alias.c (find_base_term): New wrapper around find_base_term
5209 unwinding CSELIB_VAL_PTR changes.
5210 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5213 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5215 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5217 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5218 constant definitions.
5219 ("nop"): lr 0,0 -> nopr r0
5220 ("nop_lr0", "nop_lr1"): New insn definitions.
5222 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5224 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5225 NDS32_V3PUSH_AVAILABLE_P macro.
5227 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5228 Chung-Ju Wu <jasonwucj@gmail.com>
5230 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5231 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5232 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5233 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5234 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5235 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5236 * config/nds32/constraints.md: New constraints and checking for hard
5237 float configuration.
5238 * config/nds32/iterators.md: New mode iterator and attribute for hard
5239 float configuration.
5240 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5242 * config/nds32/nds32-fpu.md: New file.
5243 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5244 deal with hard float code generation.
5245 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5247 (abi_type, float_reg_number): New enum type.
5248 * config/nds32/nds32-predicates.c: New predicates for hard float.
5249 * config/nds32/nds32-protos.h: Declare functions for hard float.
5250 * config/nds32/nds32.c: Implementation for hard float configuration.
5251 * config/nds32/nds32.h: Definitions for hard float configuration.
5252 * config/nds32/nds32.md: Include hard float machine description and
5253 modify patterns for hard float configuration.
5254 * config/nds32/nds32.opt: New options for hard float configuration.
5255 * config/nds32/predicates.md: New predicates for hard float
5258 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5260 * common/config/nds32/nds32-common.c
5261 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5263 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5266 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5267 CONSTRUCTOR_ELT (ctor, ...)->value.
5269 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5272 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5274 2018-04-05 Tom de Vries <tom@codesourcery.com>
5277 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5280 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5281 Kito Cheng <kito.cheng@gmail.com>
5283 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5284 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5285 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5286 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5288 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5289 Kito Cheng <kito.cheng@gmail.com>
5291 * config/nds32/constraints.md (Ufe): New memory constraint.
5292 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5293 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5294 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5296 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5297 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5299 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5301 * config/nds32/nds32.md: Use optimize_size in the condition for
5302 alu-shift instructions.
5304 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5306 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5308 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5310 * config/nds32/nds32.md (negsi2): Refine pattern.
5312 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5313 Chung-Ju Wu <jasonwucj@gmail.com>
5315 * config/nds32/iterators.md (shift_rotate): New code iterator.
5316 (shift): New code attribute.
5317 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5318 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5319 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5320 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5321 bit-wise operations.
5322 (andsi3, *andsi3): Ditto.
5323 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5324 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5325 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5326 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5327 nds32_ior_operand, nds32_xor_operand): New predicates.
5329 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5331 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5332 (addsi3, subsi3): ... this.
5334 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5336 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5338 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5340 * config/nds32/nds32.md: Adjust indention.
5342 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5344 * config/nds32/nds32.md (feature): New attribute.
5346 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5348 * config/nds32/nds32.md (subtype): New attribute.
5350 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5353 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5354 expansion to perform a bitwise AND of the argument followed by a
5355 boolean negation of the result.
5357 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5359 PR rtl-optimization/84878
5360 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5361 the basic block. Assert the use reference is not artificial and that
5362 it has an associated insn.
5364 2018-04-04 Michael Matz <matz@suse.de>
5366 * builtins.c (compute_objsize): Pass correct operand
5367 to array_at_struct_end_p.
5369 2018-04-04 Richard Biener <rguenther@suse.de>
5372 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5373 from contexts for DINFO_LEVEL_TERSE and below.
5375 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5377 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5378 resiter_operand condition.
5379 * config/nds32/nds32.md (*move<mode>): Ditto.
5381 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5382 Monk Chiang <sh.chiang04@gmail.com>
5384 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5386 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5388 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5390 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5391 Kito Cheng <kito.cheng@gmail.com>
5393 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5394 nds32_cond_code_str, output_cond_branch,
5395 output_cond_branch_compare_zero, nds32_expand_cbranch,
5396 nds32_expand_cstore, nds32_expand_movcc,
5397 nds32_output_cbranchsi4_equality_zero,
5398 nds32_output_cbranchsi4_equality_reg,
5399 nds32_output_cbranchsi4_equality_reg_or_const_int,
5400 nds32_output_cbranchsi4_greater_less_zero: New functions.
5401 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5402 nds32_expand_cstore, nds32_expand_movcc,
5403 nds32_output_cbranchsi4_equality_zero,
5404 nds32_output_cbranchsi4_equality_reg,
5405 nds32_output_cbranchsi4_equality_reg_or_const_int,
5406 nds32_output_cbranchsi4_greater_less_zero): Declare.
5407 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5408 nds32_rimm11s_operand): New predicates.
5409 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5410 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5413 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5415 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5416 * config/nds32/nds32.md: Ditto.
5417 * config/nds32/pipelines.md: Ditto.
5419 2018-04-04 Richard Biener <rguenther@suse.de>
5421 PR tree-optimization/85168
5422 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5423 propagating abnormals.
5425 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5427 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5429 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5430 Kito Cheng <kito.cheng@gmail.com>
5432 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5433 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5434 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5435 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5436 * config/nds32/nds32.md (sibcall_internal): New.
5437 (sibcall_register): Remove.
5438 (sibcall_immediate): Remove.
5439 (sibcall_value_internal): New.
5440 (sibcall_value_register): Remove.
5441 (sibcall_value_immediate): Remove.
5442 * config/nds32/predicates.md (nds32_general_register_operand): New.
5443 (nds32_call_address_operand): New.
5445 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5447 PR rtl-optimization/85167
5448 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5449 bb_defs if *split_p, instead preinitialize it to NULL.
5451 PR tree-optimization/85156
5452 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5453 evaluating the argument multiple times.
5455 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5457 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5459 (_mm_cvtpd_ps): Likewise.
5460 (_mm_cvttpd_epi32): Likewise.
5461 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5462 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5463 vector, pixel, and bool following altivec.h include.
5465 2018-04-03 Martin Sebor <msebor@redhat.com>
5467 * doc/extend.texi (Common Function Attributes): Clarify.
5468 (const attribute): Likewise.
5469 (pure attribute): Likewise.
5471 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5474 * config/i386/i386.c (ix86_expand_vector_set): Use
5475 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5477 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5479 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5480 instructions when changing rounding bits to preserve precision bits
5481 in the x87 control word.
5483 2018-04-03 Martin Liska <mliska@suse.cz>
5485 PR tree-optimization/82491
5486 * rtl.h (strip_offset_and_add): Replace += suboffset with
5487 poly_uint64 () + suboffset.
5489 2018-03-29 Martin Liska <mliska@suse.cz>
5490 Martin Jambor <mjambor@suse.cz>
5493 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5494 param_type is not an integral or pointer type.
5496 2018-04-03 Richard Biener <rguenther@suse.de>
5498 * sese.h (recompute_all_dominators): Remove.
5500 2018-04-02 Martin Sebor <msebor@redhat.com>
5502 * doc/invoke.texi (-Wrestrict): Fix typos.
5504 2018-04-02 Jim Wilson <jimw@sifive.com>
5506 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5507 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5508 (<optab>di3, <optab>si3_extend): Likewise.
5509 (<optab>si3_mask, <optab>si3_mask_1): New.
5510 (<optab>di3_mask, <optab>di3_mask_1): New.
5511 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5512 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5513 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5515 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5517 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5520 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5522 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5523 (nds32_canonicalize_comparison): New function.
5525 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5526 Kito Cheng <kito.cheng@gmail.com>
5527 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5529 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5530 * config/nds32/constants.md (unspec_volatile_element): Add
5531 UNSPEC_VOLATILE_RELAX_GROUP.
5532 * config/nds32/nds32-relax-opt.c: New file.
5533 * config/nds32/nds32-predicates.c
5534 (nds32_symbol_load_store_p): New function.
5535 * config/nds32/nds32-protos.h
5536 (nds32_symbol_load_store_p): Declare function.
5537 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5538 * config/nds32/nds32.c
5539 (nds32_register_pass): New function to register pass.
5540 (nds32_register_passes): New function to register passes.
5541 * config/nds32/nds32.md (relax_group): New pattern.
5542 * config/nds32/nds32.opt (mrelax-hint): New option.
5543 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5545 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5547 * config/nds32/t-nds32: Modify files dependency.
5549 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5551 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5552 (PROFILE_HOOK): Define its implementation.
5554 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5556 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5557 type and 32-bit size.
5559 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5562 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5563 (V_128_256): New mode iterator.
5564 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5565 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5566 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5568 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5569 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5571 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5574 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5575 NaN inputs correctly.
5577 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5580 * config/rs6000/vsx.md (??r): New mode attribute.
5581 (*vsx_mov<mode>_64bit): Use it.
5582 (*vsx_mov<mode>_32bit): Likewise.
5584 2018-03-30 Martin Sebor <msebor@redhat.com>
5586 PR tree-optimization/84818
5587 * builtins.c (check_access): Use warning_n.
5589 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5592 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5594 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5597 2018-03-30 Julia Koval <julia.koval@intel.com>
5600 * x86-tune.def (movx, partial_reg_dependency): Enable for
5603 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5606 * lra-constraints.c (process_alt_operands): Move setting
5607 this_alternative_matches below.
5609 2018-03-29 Martin Liska <mliska@suse.cz>
5612 * doc/invoke.texi: Document how LTO works with debug info.
5613 Describe auto-load support of binutils. Mention 'x86-64'
5614 as valid option value of -march option.
5616 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5618 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5621 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5622 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5623 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5626 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5629 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5630 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5631 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5632 for RS6000_BTM_POWERPC64.
5633 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5634 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5635 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5640 2018-03-28 Carl Love <cel@us.ibm.com>
5643 2017-09-27 Carl Love <cel@us.ibm.com>
5645 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5646 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5647 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5650 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5652 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5653 instead of __vector bool.
5654 (_mm_max_pu8): Likewise.
5655 (_mm_min_pi16): Likewise.
5657 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5660 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5661 (DIVWEUO): Likewise.
5663 (DIVDEUO): Likewise.
5664 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5665 DIVWEUO and DIVDEUO.
5666 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5667 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5668 (div_extend): Likewise.
5669 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5671 (__builtin_divweuo): Likewise.
5672 (__builtin_divdeo): Likewise.
5673 (__builtin_divdeuo): Likewise.
5675 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5678 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5679 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5681 PR tree-optimization/82004
5682 * gimple-match-head.c (optimize_pow_to_exp): New function.
5683 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5684 Don't fold to exp if optimize_pow_to_exp is false.
5686 2018-03-28 Martin Liska <mliska@suse.cz>
5689 * calls.c (initialize_argument_information): Fix trailing space.
5690 * common.opt: Fix typo and provide better explanation for
5691 -fsanitize-coverage option.
5692 * config/i386/i386.opt: Fix typo.
5694 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5695 Martin Liska <mliska@suse.cz>
5698 * gimplify.c (asan_poison_variable): Don't do the check for
5699 gimplify_omp_ctxp here.
5700 (gimplify_decl_expr): Do it here.
5701 (gimplify_target_expr): Likewise.
5703 2018-03-28 Martin Liska <mliska@suse.cz>
5706 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5707 chkp_type_bounds_count if MPX is not enabled.
5709 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5711 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5713 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5716 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5717 function to create the function decl for complex long double
5718 multiply and divide for -mabi=ieeelongdouble.
5719 (init_float128_ieee): Call it.
5721 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5724 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5725 -fcf-protection=branch -mibt.
5726 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5728 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5731 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5733 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5736 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5737 extern array declarations.
5739 2018-03-27 Richard Biener <rguenther@suse.de>
5742 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5743 explicit single_use checks.
5745 2018-03-27 Richard Biener <rguenther@suse.de>
5747 PR tree-optimization/85082
5748 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5751 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5753 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5754 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5755 Turn on fasynchronous-unwind-tables and funwind-tables.
5757 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5760 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5761 (*bmi_blsr_<mode>_ccz): Ditto.
5763 2018-03-26 Tom de Vries <tom@codesourcery.com>
5765 PR tree-optimization/85063
5766 * omp-general.c (offloading_function_p): New function. Factor out
5768 * omp-offload.c (pass_omp_target_link::gate): ... here.
5769 * omp-general.h (offloading_function_p): Declare.
5770 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5771 with attribute omp declare target for offloading functions.
5773 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5775 PR tree-optimization/84005
5776 * tree-data-ref.h (get_base_for_alignment): Declare.
5777 * tree-data-ref.c (get_base_for_alignment_1): New function.
5778 (get_base_for_alignment): Likewise.
5779 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5780 get_base_for_alignment to find a suitable base object, instead
5781 of always using drb->base_address.
5783 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5786 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5787 known size by default.
5789 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5792 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5793 and non BLKmode operands.
5795 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5798 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5799 Clean up attributes.
5801 2018-03-23 Richard Biener <rguenther@suse.de>
5804 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5805 we are going to emit early debug for LTO.
5807 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5810 * function.c (match_asm_constraints_1): Don't optimize if input
5811 doesn't satisfy general_operand predicate for output's mode.
5814 * alias.c (write_dependence_p): Don't require for x_canonicalized
5815 non-VOIDmode if x has VOIDmode.
5818 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5819 just don't try to optimize it rather than assert it never happens.
5821 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5823 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5824 macro expansions for definition of ST_INTERNAL_<mode> and
5825 LD_INTERNAL_<mode> builtins.
5826 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5828 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5830 (altivec_expand_st_builtin): Likewise.
5831 (altivec_expand_builtin): Remove calls to deleted functions.
5832 (rs6000_address_for_altivec): Delete this function.
5833 * config/rs6000/vector.md: Remove expands for
5834 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5836 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5839 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5840 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5841 re-computing once computed.
5842 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5843 (arm_init_machine_status): Initialize
5844 machine->static_chain_stack_bytes.
5846 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5849 * doc/extend.texi: Add four new prototypes for vec_ld.
5850 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5851 definitions for more logical presentation.
5852 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5853 entries for V1TI variants of __builtin_altivec_ld builtin.
5854 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5855 handling of V1TI variant of LVX icode pattern.
5856 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5857 (rs6000_gimple_fold_builtin): Likewise.
5858 (altivec_init_builtins): Add code to define
5859 __builtin_altivec_lvx_v1ti function.
5861 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5864 * function.c (match_asm_constraints_1): Don't do the optimization
5865 if input isn't a REG, SUBREG, MEM or constant.
5867 2018-03-22 Tom de Vries <tom@codesourcery.com>
5869 PR tree-optimization/84956
5870 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5871 bb_has_abnormal_pred.
5873 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5876 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5877 DECL_INITIAL (decl) to decl at the end.
5878 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5881 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5883 * doc/extend.texi (__builtin_tgmath): Document when complex
5884 integer types are treated as _Complex _Float64.
5886 2018-03-21 Tom de Vries <tom@codesourcery.com>
5888 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5890 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5892 PR tree-optimization/84960
5893 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5894 if it is ENTRY block, move them into single succ of ENTRY in that case.
5896 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5898 PR tree-optimization/84811
5899 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5900 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5901 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5902 (known_subrange_p): Update accordingly. Cast each value involved
5903 in the size comparison, rather than casting the result of the
5906 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5908 PR tree-optimization/84982
5909 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5910 by flipping the least significant bit rather than all bits from
5911 bitpos to bitpos + bitsize - 1.
5913 2018-03-21 Nathan Sidwell <nathan@acm.org>
5915 * doc/extend.texi (Deprecated Features): Remove mention of
5916 long-deleted deprecations.
5918 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5921 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5922 * configure: Regenerate.
5924 2018-03-21 Tom de Vries <tom@codesourcery.com>
5926 PR tree-optimization/83126
5927 * tree-parloops.c (num_phis): New function.
5928 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5930 2018-03-21 Nathan Sidwell <nathan@acm.org>
5932 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5933 mention anon-struct/union members and trailing attributes.
5935 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5937 PR tree-optimization/84969
5938 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5939 builtin memset partitions if they set different rhs values.
5941 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5943 PR rtl-optimization/84989
5944 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5945 VEC_DUPLICATE with scalar result mode.
5947 2018-03-21 Martin Liska <mliska@suse.cz>
5950 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5951 not intended return statement.
5953 2018-03-21 Martin Liska <mliska@suse.cz>
5956 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5957 (chkp_find_bound_slots_1): Limit number of iterations.
5959 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5962 * Minor grammar fixes for x86 options.
5964 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5967 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5968 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5970 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5973 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5974 (altivec_lvx_<mode>_1op): Likewise.
5975 (altivec_stvx_<mode>_2op): Likewise.
5976 (altivec_stvx_<mode>_1op): Likewise.
5977 (altivec_lvx_<VM2:mode>): New define_expand.
5978 (altivec_stvx_<VM2:mode>): Likewise.
5979 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5980 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5981 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5982 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5983 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5984 (rs6000_gen_lvx): Likewise.
5985 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5986 (altivec_expand_stv_builtin): Likewise.
5987 (altivec_expand_builtin): Likewise.
5988 * config/rs6000/vector.md: Likewise.
5990 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5993 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5996 2018-03-20 Richard Biener <rguenther@suse.de>
5999 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
6000 sign-conversions as zero, fall back to standard scalar_stmt
6003 2018-03-20 Martin Liska <mliska@suse.cz>
6006 * predict.c (rebuild_frequencies): Handle case when we have
6007 PROFILE_ABSENT, but flag_guess_branch_prob is false.
6009 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6012 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
6013 flag_section_anchors.
6014 * varasm.c (use_blocks_for_decl_p): Remove hack for
6015 dw2_force_const_mem.
6018 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
6020 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
6021 be created, use lowpart_subreg of operands[0] rather than operands[0]
6023 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
6024 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
6025 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
6026 and n constraint instead of aarch64_shift_imm_di and Usd.
6027 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
6028 (*aarch64_<optab>_reg_minus<mode>3): ... this.
6030 2018-03-20 Sudakshina Das <sudi.das@arm.com>
6033 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
6034 to favor GPR over NEON registers.
6035 (<shift>di3_neon): Likewise.
6037 2018-03-20 Tom de Vries <tom@codesourcery.com>
6040 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
6041 (nvptx_process_pars): Emit bar.sync asap and alap.
6043 2018-03-20 Tom de Vries <tom@codesourcery.com>
6046 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
6047 seen_label if seen_label is already set.
6049 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6052 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6053 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6054 Use 1U instead of 1. Formatting fixes.
6057 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6058 instead of TREE_TYPE (s1) for the return value.
6060 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6062 PR tree-optimization/84946
6063 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6064 bitsize + bitsize in poly_uint64 rather than poly_int64.
6067 * dwarf2asm.c: Include fold-const.c.
6068 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6069 of decl rather than decl itself.
6071 PR rtl-optimization/84643
6072 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6074 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6077 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6078 calling assemble_variable.
6080 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6083 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6084 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6086 2018-03-19 Jim Wilson <jimw@sifive.com>
6089 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6090 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6091 (riscv_first_stack_step): Likewise.
6092 (riscv_option_override): Use STACK_BOUNDARY instead of
6094 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6096 (BIGGEST_ALIGNMENT): Set to 128.
6097 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6098 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6101 2018-03-19 Richard Biener <rguenther@suse.de>
6103 PR tree-optimization/84933
6104 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6105 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6107 2018-03-19 Richard Biener <rguenther@suse.de>
6109 PR tree-optimization/84859
6110 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6111 (cond_if_else_store_replacement): Perform sinking operation on
6112 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6113 Generalize what a BB with a single eligible store is.
6115 2018-03-19 Richard Biener <rguenther@suse.de>
6117 PR tree-optimization/84929
6118 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6119 chrec_is_positive against non-chrec arg.
6121 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6124 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6126 2018-03-18 Martin Liska <mliska@suse.cz>
6128 PR rtl-optimization/84635
6129 * regrename.c (build_def_use): Use matches_mode only when
6132 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6134 PR tree-optimization/84913
6135 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6136 vectorize chains of COND_EXPRs.
6138 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6140 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6142 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6144 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6146 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6148 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6150 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6151 Kito Cheng <kito.cheng@gmail.com>
6153 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6154 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6155 (nds32_adjust_reg_alloc_order): New function.
6156 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6158 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6160 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6161 nds32_print_operand, nds32_print_operand_address): Use
6162 HOST_WIDE_INT_PRINT_DEC instead.
6164 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6166 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6168 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6171 * config/i386/i386.c (initial_ix86_tune_features,
6172 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6174 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6175 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6176 rather than 1u << ix86_tune. Formatting fix.
6177 (ix86_option_override_internal): Change ix86_arch_mask from
6178 unsigned int to unsigned HOST_WIDE_INT, initialize to
6179 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6180 (ix86_function_specific_restore): Likewise.
6182 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6185 * postreload.c (reload_combine_recognize_pattern): Perform
6186 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6187 truncate_int_for_mode the result for the destination's mode.
6190 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6191 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6193 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6196 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6199 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6200 regno_allocno_class_array and sorted_pseudos.
6201 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6202 insns where regno is used.
6204 2018-03-16 Martin Liska <mliska@suse.cz>
6207 * multiple_target.c (create_dispatcher_calls): Redirect
6208 reference in the symbol table.
6210 2018-03-16 Martin Liska <mliska@suse.cz>
6213 * multiple_target.c (create_dispatcher_calls): Redirect also
6216 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6220 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6221 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6222 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6224 2018-03-16 Julia Koval <julia.koval@intel.com>
6226 * doc/invoke.texi (Skylake Server): Add CLWB.
6227 Cannonlake): Remove CLWB.
6229 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6231 PR tree-optimization/84841
6232 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6234 (FLOAT_ONE_CONST_TYPE): Define.
6235 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6236 (sort_by_operand_rank): Put entries with higher constant_type last
6237 rather than first to match comments.
6239 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6241 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6244 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6247 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6248 check and use address of target always.
6250 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6253 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6254 (indirect_thunk_bnd_needed): Likewise.
6255 (indirect_thunks_used): Likewise.
6256 (indirect_thunks_bnd_used): Likewise.
6257 (indirect_return_needed): New.
6258 (indirect_return_bnd_needed): Likewise.
6259 (output_indirect_thunk_function): Add a bool argument for
6261 (output_indirect_thunk_function): Don't generate alias for
6262 function return thunk.
6263 (ix86_code_end): Call output_indirect_thunk_function to generate
6264 function return thunks.
6265 (ix86_output_function_return): Set indirect_return_bnd_needed
6266 and indirect_return_needed instead of indirect_thunk_bnd_needed
6267 and indirect_thunk_needed.
6269 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6271 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6272 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6273 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6275 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6276 Paul Hua <paul.hua.gm@gmail.com>
6279 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6281 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6283 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6284 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6287 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6290 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6291 instead of GET_MODE_SIZE when comparing Units.
6293 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6296 * varasm.c (hash_section): Return an unchangeble hash value
6297 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6298 Return !aarch64_can_use_per_function_literal_pools_p ().
6300 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6303 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6304 rather than address of cmode as last argument to prepare_cmp_insn.
6306 2018-03-15 Julia Koval <julia.koval@intel.com>
6308 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6309 F_AVX512VNNI, F_AVX512BITALG): New.
6311 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6314 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6315 insn for floating-point loads and stores.
6317 2018-03-14 Carl Love <cel@us.ibm.com>
6319 * config/rs6000/rs6000-c.c: Add macro definitions for
6320 ALTIVEC_BUILTIN_VEC_PERMXOR.
6321 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6322 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6323 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6324 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6326 * config/doc/extend.texi: Add prototypes for vec_permxor.
6328 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6331 * diagnostic-show-locus.c (class layout_point): Convert m_line
6332 from int to linenum_type.
6333 (line_span::comparator): Use linenum "compare" function when
6334 comparing line numbers.
6335 (test_line_span): New function.
6336 (layout_range::contains_point): Convert param "row" from int to
6338 (layout_range::intersects_line_p): Likewise.
6339 (layout::will_show_line_p): Likewise.
6340 (layout::print_source_line): Likewise.
6341 (layout::should_print_annotation_line_p): Likewise.
6342 (layout::print_annotation_line): Likewise.
6343 (layout::print_leading_fixits): Likewise.
6344 (layout::annotation_line_showed_range_p): Likewise.
6345 (struct line_corrections): Likewise for field m_row.
6346 (line_corrections::line_corrections): Likewise for param "row".
6347 (layout::print_trailing_fixits): Likewise.
6348 (layout::get_state_at_point): Likewise.
6349 (layout::get_x_bound_for_row): Likewise.
6350 (layout::print_line): Likewise.
6351 (diagnostic_show_locus): Likewise for locals "last_line" and
6353 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6354 * input.c (selftest::test_linenum_comparisons): New function.
6355 (selftest::input_c_tests): Call it.
6356 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6357 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6358 * selftest.h (ASSERT_GT): New macro.
6359 (ASSERT_GT_AT): New macro.
6360 (ASSERT_LT): New macro.
6361 (ASSERT_LT_AT): New macro.
6363 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6365 PR rtl-optimization/84780
6366 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6368 2018-03-14 Martin Liska <mliska@suse.cz>
6370 * tree.c (record_node_allocation_statistics): Use
6371 get_stats_node_kind.
6372 (get_stats_node_kind): New function extracted from
6373 record_node_allocation_statistics.
6374 (free_node): Use get_stats_node_kind.
6376 2018-03-14 Richard Biener <rguenther@suse.de>
6378 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6379 that the value-set of ANTIC_IN doesn't grow.
6382 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6384 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6385 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6387 2018-03-14 Julia Koval <julia.koval@intel.com>
6389 * config.gcc (icelake-client, icelake-server): New.
6391 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6392 (initial_ix86_arch_features): Ditto.
6393 (PTA_SKYLAKE): Add SGX.
6394 (PTA_ICELAKE): Remove.
6395 (PTA_ICELAKE_CLIENT): New.
6396 (PTA_ICELAKE_SERVER): New.
6397 (ix86_option_override_internal): Split up icelake on icelake client and
6399 (get_builtin_code_for_version): Ditto.
6400 (fold_builtin_cpu): Ditto.
6401 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6402 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6403 * config/i386/i386.h (processor_type): Ditto.
6404 * doc/invoke.texi: Ditto.
6406 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6409 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6410 INTEGER_CST offset, add it together with bitpos / 8 and
6411 sign extend based on POINTER_SIZE.
6415 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6418 * config/i386/constraints.md (Yc): New register constraint.
6419 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6420 Use Yc constraint for alternative 2 of operand 0. Remove
6421 preferred_for_speed attribute.
6423 2018-03-14 Richard Biener <rguenther@suse.de>
6425 PR tree-optimization/84830
6426 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6427 with the old one to avoid oscillations.
6429 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6432 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6434 (assign_by_spills): Return a flag of reload assignment failure.
6435 Do not process the reload assignment failures. Do not spill other
6436 reload pseudos if they has the same reg class. Update n if
6438 (lra_assign): Add a return arg. Set up from the result of
6439 assign_by_spills call.
6440 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6441 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6442 usage_insns if it is not NULL.
6443 (spill_hard_reg_in_range): New function.
6444 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6445 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6446 function prototypes.
6447 (lra_assign): Change prototype.
6448 * lra.c (lra): Add code to deal with fails by splitting hard reg
6451 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6453 * config/riscv/riscv.opt (mrelax): New option.
6454 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6455 "norelax" when riscv_mrelax is disabled.
6456 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6458 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6461 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6462 reassociation for int modes.
6464 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6466 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6467 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6469 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6470 * config/aarch64/aarch64-sve.md
6471 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6472 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6473 (*extend<mode><Vwide>2): Rename to...
6474 (aarch64_sve_extend<mode><Vwide>2): ...this.
6475 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6476 renaming the old pattern to...
6477 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6479 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6480 define_expand, renaming the old pattern to...
6481 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6482 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6483 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6484 account when deciding which SVE instruction the optab should use.
6485 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6487 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6489 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6490 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6491 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6492 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6493 (tlsdesc_small_<mode>): Turn a define_expand and use
6494 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6495 (tlsdesc_small_advsimd_<mode>): ...this.
6496 (tlsdesc_small_sve_<mode>): New pattern.
6498 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6500 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6501 (UNSPEC_UMUL_HIGHPART): New constants.
6502 (MUL_HIGHPART): New int iteraor.
6503 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6504 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6506 (*<su>mul<mode>3_highpart): New define_insn.
6508 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6511 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6514 2018-03-13 Martin Liska <mliska@suse.cz>
6517 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6519 (sem_item_optimizer::~sem_item_optimizer): Release it.
6520 (sem_item_optimizer::merge_classes): Register variable aliases.
6521 (sem_item_optimizer::fixup_pt_set): New function.
6522 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6523 * ipa-icf.h: Declare new variables and functions.
6525 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6528 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6529 integer_pow2p@2 and test integer_pow2p in condition.
6530 (A < 0 ? C : 0): Similarly for @1.
6533 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6534 characters starting at p contain '\0' character, don't look beyond
6538 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6539 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6542 * reg-stack.c (change_stack): Change update_end var from int to
6543 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6544 also call set_block_for_insn on the newly added insns and rescan.
6547 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6548 on the last operand.
6551 * tree.c (stabilize_reference_1): Return save_expr (e) for
6552 STATEMENT_LIST even if it doesn't have side-effects.
6554 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6556 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6558 2018-03-12 Renlin Li <renlin.li@arm.com>
6560 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6561 aarch64_output_scalar_simd_mov_immediate.
6563 2018-03-12 Martin Sebor <msebor@redhat.com>
6565 PR tree-optimization/83456
6566 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6567 for perfectly overlapping calls to memcpy.
6568 (gimple_fold_builtin_memory_chk): Same.
6569 (gimple_fold_builtin_strcpy): Handle no-warning.
6570 (gimple_fold_builtin_stxcpy_chk): Same.
6571 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6573 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6575 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6576 parameter. Use it for SFmode.
6577 (rs6000_function_arg_advance_1): Adjust.
6578 (rs6000_function_arg): Adjust.
6579 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6581 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6583 PR rtl-optimization/84169
6584 PR rtl-optimization/84780
6585 * combine.c (can_combine_p): Check for a 2-insn combination whether
6586 the destination register is used between the two insns, too.
6588 2018-03-12 Richard Biener <rguenther@suse.de>
6590 PR tree-optimization/84803
6591 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6592 for refs DR analysis didn't process.
6594 2018-03-12 Richard Biener <rguenther@suse.de>
6596 PR tree-optimization/84777
6597 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6598 force-vectorize loops ignore whether we are optimizing for size.
6600 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6602 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6603 (TARGET_MD_ASM_ADJUST): Define.
6605 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6606 Kito Cheng <kito.cheng@gmail.com>
6607 Chung-Ju Wu <jasonwucj@gmail.com>
6609 * config/nds32/nds32.c (nds32_compute_stack_frame,
6610 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6611 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6612 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6613 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6614 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6615 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6616 * config/nds32/nds32.md (prologue, epilogue): Use macro
6617 NDS32_V3PUSH_AVAILABLE_P to do checking.
6619 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6622 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6623 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6624 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6625 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6626 addition of most attributes on !orig_type_die or the attribute not
6627 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6629 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6630 Chung-Ju Wu <jasonwucj@gmail.com>
6632 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6634 * config/nds32/nds32.opt (mvh): New option.
6636 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6637 Chung-Ju Wu <jasonwucj@gmail.com>
6639 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6641 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6642 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6645 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6646 Chung-Ju Wu <jasonwucj@gmail.com>
6648 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6650 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6651 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6653 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6654 Kito Cheng <kito.cheng@gmail.com>
6655 Chung-Ju Wu <jasonwucj@gmail.com>
6657 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6658 UNSPEC_FFMISM and UNSPEC_FLMISM.
6659 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6660 for ffb, ffmism and flmism.
6661 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6662 (unspec_ffmism): Ditto.
6663 (unspec_flmism): Ditto.
6664 (nds32_expand_builtin_impl): Check if string extension is available.
6665 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6666 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6668 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6671 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6674 * lra-assigns.c (assign_by_spills): Return a flag of reload
6675 assignment failure. Do not process the reload assignment
6676 failures. Do not spill other reload pseudos if they has the same
6678 (lra_assign): Add a return arg. Set up from the result of
6679 assign_by_spills call.
6680 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6681 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6682 usage_insns if it is not NULL.
6683 (spill_hard_reg_in_range): New function.
6684 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6685 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6686 function prototypes.
6687 (lra_assign): Change prototype.
6688 * lra.c (lra): Add code to deal with fails by splitting hard reg
6691 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6694 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6696 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6699 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6700 (dw_val_node): Add val_symbolic_view.
6701 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6702 (symview_upper_bound): New.
6703 (new_line_info_table): Initialize symviews_since_reset.
6704 (dwarf2out_source_line): Count symviews_since_reset and set
6705 symview_upper_bound.
6706 (dw_val_equal_p): Handle symview.
6707 (add_AT_symview): New.
6708 (print_dw_val): Handle symview.
6709 (attr_checksum, attr_checksum_ordered): Likewise.
6710 (same_dw_val_p, size_of_die): Likewise.
6711 (value_format, output_die): Likewise.
6712 (add_high_low_attributes): Use add_AT_symview for entry_view.
6713 (dwarf2out_finish): Reset symview_upper_bound, clear
6716 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6719 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6720 Add strict argument and use it.
6721 (rs6000_split_multireg_move): Update for new strict argument.
6722 (mem_operand_gpr): Disallow all non-offsettable addresses.
6723 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6725 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6728 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6729 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6730 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6733 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6734 decl, use remap_type if we want to use the type.
6736 2018-03-09 Martin Sebor <msebor@redhat.com>
6738 PR tree-optimization/84526
6739 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6741 (builtin_access::generic_overlap): Be prepared to handle non-array
6744 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6746 PR rtl-optimization/84682
6747 * lra-constraints.c (process_address_1): Check is_address flag
6748 for address constraints.
6749 (process_alt_operands): Likewise.
6750 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6751 preprocess_constraints.
6752 * recog.h (preprocess_constraints): Add oploc parameter.
6754 * recog.c (preprocess_constraints): Test address_operand for
6755 CT_ADDRESS constraints.
6757 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6760 * lra-assigns.c (assign_by_spills): Return a flag of reload
6761 assignment failure. Do not process the reload assignment
6762 failures. Do not spill other reload pseudos if they has the same
6764 (lra_assign): Add a return arg. Set up from the result of
6765 assign_by_spills call.
6766 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6767 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6768 usage_insns if it is not NULL.
6769 (spill_hard_reg_in_range): New function.
6770 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6771 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6772 function prototypes.
6773 (lra_assign): Change prototype.
6774 * lra.c (lra): Add code to deal with fails by splitting hard reg
6777 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6780 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6781 Accept complain bool parameter. Only emit errors if it is true.
6782 (arm_parse_cpu_option_name): Likewise.
6783 (arm_target_thumb_only): Adjust callers of the above.
6784 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6785 prototype to take a default true bool parameter.
6786 (arm_parse_arch_option_name): Likewise.
6788 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6789 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6793 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6794 * configure: Regenerate.
6795 * configure.ac ("linker --version-script option"): New.
6796 ("linker soname option"): New.
6798 2018-03-09 Richard Biener <rguenther@suse.de>
6800 PR tree-optimization/84775
6801 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6802 immediate uses of predicate stmts and mark them modified.
6805 PR tree-optimization/84178
6806 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6808 (version_loop_for_if_conversion): Delay update_ssa call.
6809 (tree_if_conversion): Delay update_ssa until after predicate
6812 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6815 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6816 when the function accesses prior frames.
6818 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6821 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6822 gen_llsym, otherwise call maybe_gen_llsym.
6825 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6826 has ',' character inside of it.
6828 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6831 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6832 as clobbering CC_REGNUM.
6834 2018-03-08 Richard Biener <rguenther@suse.de>
6837 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6838 (follow_copies_to_constant): Do not follow SSA names registered
6841 2018-03-08 Richard Biener <rguenther@suse.de>
6843 PR tree-optimization/84178
6844 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6846 (version_loop_for_if_conversion): Delay update_ssa call.
6847 (tree_if_conversion): Delay update_ssa until after predicate
6850 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6852 PR tree-optimization/84178
6853 * tree-if-conv.c (release_bb_predicate): Remove the
6854 the assertion that the stmts have NULL use_ops.
6855 Discard the statements, asserting that they haven't
6856 yet been added to a BB.
6858 2018-03-08 Richard Biener <rguenther@suse.de>
6860 PR tree-optimization/84746
6861 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6862 (phi_translate): Pass in destination ANTIC_OUT set.
6863 (phi_translate_1): Likewise. For a simplified result lookup
6864 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6865 (phi_translate_set): Adjust.
6866 (do_pre_regular_insertion): Likewise.
6867 (do_pre_partial_partial_insertion): Likewise.
6869 2018-03-08 Martin Liska <mliska@suse.cz>
6871 PR gcov-profile/84735
6872 * doc/gcov.texi: Document usage of profile files.
6873 * gcov-io.h: Document changes in the format.
6875 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6879 * dwarf2out.c (struct dw_line_info_table): Update comments for
6881 (FORCE_RESET_NEXT_VIEW): New.
6882 (FORCE_RESETTING_VIEW_P): New.
6883 (RESETTING_VIEW_P): Check for -1 too.
6884 (ZERO_VIEW_P): Likewise.
6885 (new_line_info_table): Force-reset next view.
6886 (dwarf2out_begin_function): Likewise.
6887 (dwarf2out_source_line): Simplify zero_view_p initialization.
6888 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6889 view directly. Omit view when omitting .loc at line 0.
6891 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6893 PR tree-optimization/84740
6894 * tree-switch-conversion.c (process_switch): Call build_constructors
6895 only if info.phi_count is non-zero.
6897 PR tree-optimization/84739
6898 * tree-tailcall.c (find_tail_calls): Check call arguments against
6899 DECL_ARGUMENTS (current_function_decl) rather than
6900 DECL_ARGUMENTS (func) when checking for tail recursion.
6902 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6904 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6905 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6906 Volker Reichelt's entry and add entries for people that perform
6907 GCC fuzzy testing and report numerous bugs.
6909 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6912 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6913 readonly data in sdata, if that is disabled.
6914 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6915 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6916 -mreadonly-in-sdata option.
6918 2018-03-07 Martin Sebor <msebor@redhat.com>
6920 PR tree-optimization/84468
6921 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6922 basic block when looking for nul assignment.
6924 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6927 * except.h (output_function_exception_table): Adjust prototype.
6928 * except.c (output_function_exception_table): Remove FNNAME parameter
6929 and add SECTION parameter. Ouput one part of the table at a time.
6930 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6931 the first part of the exception table and emit unwind directives.
6932 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6933 (i386_pe_seh_cold_init): Likewise.
6934 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6935 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6936 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6937 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6938 * config/i386/winnt.c: Include except.h.
6939 (struct seh_frame_state): Add reg_offset, after_prologue and
6940 in_cold_section fields.
6941 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6942 (i386_pe_seh_cold_init): New function.
6943 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6944 to seh->in_cold_section.
6945 (seh_emit_push): Record the offset of the push.
6946 (seh_emit_save): Record the offet of the save.
6947 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6948 Test seh->after_prologue to disregard the epilogue.
6949 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6950 (i386_pe_end_cold_function): New function.
6952 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6955 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6956 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6959 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6961 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6962 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6965 * multiple_target.c: Include tree-inline.h and intl.h.
6966 (expand_target_clones): Diagnose and fail if node->definition and
6967 !tree_versionable_function_p (node->decl).
6969 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6971 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6973 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6974 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6975 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6977 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6980 * combine.c (try_combine): Use reg_or_subregno instead of handling
6981 just paradoxical SUBREGs and REGs.
6983 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6985 * config/arc/arc.c (arc_finalize_pic): Remove function.
6986 (arc_must_save_register): We use single base PIC register, remove
6987 checks to save/restore the PIC register.
6988 (arc_expand_prologue): Likewise.
6989 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6991 (arc_verify_short): Likewise.
6992 (arc_attr_type): Likewise.
6993 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6994 (walk_stores): Likewise.
6995 (arc_address_cost): Make it static.
6996 (arc_verify_short): Likewise.
6997 (branch_dest): Likewise.
6998 (arc_attr_type): Likewise.
6999 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
7000 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
7001 (arc_final_prescan_insn): Remove inserting the nops due to
7002 hardware hazards. It is done in reorg step.
7003 (insn_length_variant_t): Remove.
7004 (insn_length_parameters_t): Likewise.
7005 (arc_insn_length_parameters): Likewise.
7006 (arc_get_insn_variants): Likewise.
7007 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
7009 2018-03-06 Jakub Jelinek <jakub@redhat.com>
7012 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7015 PR tree-optimization/84687
7016 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
7018 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
7020 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7022 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
7023 Rename to ppc_speculation_barrier.
7024 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
7025 __builtin_ppc_speculation_barrier.
7027 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7030 * combine.c (combine_simplify_rtx): Don't try to simplify if
7031 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
7034 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
7036 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
7037 to 32 bytes when compiling for POWER9.
7039 2018-03-05 Jakub Jelinek <jakub@redhat.com>
7042 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
7043 regparm >= 3 with no arg reg available also for calls with
7044 flag_force_indirect_call. Pass decl to ix86_function_regparm.
7047 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
7049 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7051 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7054 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7056 2018-03-05 Richard Biener <rguenther@suse.de>
7058 PR tree-optimization/84486
7059 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7060 When inserting a __builtin_assume_aligned call set the LHS
7061 SSA name alignment info accordingly.
7063 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7065 PR tree-optimization/84114
7066 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7067 Avoid reassociation of FLOAT_MODE addition.
7069 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7071 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7072 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7073 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7074 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7075 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7076 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7077 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7079 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7080 __builtin_ia32_wbinvd): New builtins.
7081 (SPECIAL_ARGS2): New.
7082 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7083 (SPECIAL_ARGS2): New.
7084 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7085 (ix86_valid_target_attribute_inner_p): Ditto.
7086 (ix86_init_mmx_sse_builtins): Add special_args2.
7087 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7088 TARGET_WBNOINVD_P): New.
7089 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7090 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7091 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7092 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7093 * config/i386/pconfigintrin.h: New file.
7094 * config/i386/wbnoinvdintrin.h: Ditto.
7095 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7096 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7098 2018-03-05 Richard Biener <rguenther@suse.de>
7100 PR tree-optimization/84670
7101 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7103 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7104 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7105 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7106 doesn't grow if all successors (recursively) were visited at least
7109 2018-03-05 Richard Biener <rguenther@suse.de>
7111 PR tree-optimization/84650
7112 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7113 if executed in the loop pipeline.
7115 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7117 * doc/configfiles.texi (Configuration Files): Move info about
7118 conditionalizing $target-protos.h to...
7119 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7120 differs from $target-protos.h.
7122 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7123 Chung-Ju Wu <jasonwucj@gmail.com>
7125 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7126 * config/nds32/nds32-multiple.md (setmemsi): Define.
7127 * config/nds32/nds32-memory-manipulation.c
7128 (nds32_gen_dup_4_byte_to_word_value): New.
7129 (emit_setmem_word_loop): New.
7130 (emit_setmem_byte_loop): New.
7131 (nds32_expand_setmem_loop): New.
7132 (nds32_expand_setmem_loop_v3m): New.
7133 (nds32_expand_setmem_unroll): New.
7134 (nds32_expand_setmem): New.
7136 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7137 Chung-Ju Wu <jasonwucj@gmail.com>
7139 * config/nds32/nds32-memory-manipulation.c
7140 (nds32_emit_load_store): New.
7141 (nds32_emit_post_inc_load_store): New.
7142 (nds32_emit_mem_move): New.
7143 (nds32_emit_mem_move_block): New.
7144 (nds32_expand_movmemsi_loop_unknown_size): New.
7145 (nds32_expand_movmemsi_loop_known_size): New.
7146 (nds32_expand_movmemsi_loop): New.
7147 (nds32_expand_movmemsi_unroll): New.
7148 (nds32_expand_movmemqi): Rename ...
7149 (nds32_expand_movmemsi): ... to this.
7150 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7151 (movmemsi): ... to this.
7152 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7153 (nds32_expand_movmemsi): ... to this.
7155 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7156 Monk Chiang <sh.chiang04@gmail.com>
7157 Chung-Ju Wu <jasonwucj@gmail.com>
7159 * config/nds32/nds32-protos.h
7160 (nds32_expand_load_multiple): New arguments.
7161 (nds32_expand_store_multiple): Ditto.
7162 (nds32_valid_multiple_load_store): Rename ...
7163 (nds32_valid_multiple_load_store_p): ... to this.
7164 * config/nds32/nds32-memory-manipulation.c
7165 (nds32_expand_load_multiple): Refine implementation.
7166 (nds32_expand_store_multiple): Ditto.
7167 * config/nds32/nds32-multiple.md
7168 (load_multiple): Update nds32_expand_load_multiple interface.
7169 (store_multiple): Update nds32_expand_store_multiple interface.
7170 * config/nds32/nds32-predicates.c
7171 (nds32_valid_multiple_load_store): Rename ...
7172 (nds32_valid_multiple_load_store_p): ... to this and refine
7174 * config/nds32/predicates.md
7175 (nds32_load_multiple_and_update_address_operation): New predicate.
7176 (nds32_store_multiple_and_update_address_operation): New predicate.
7178 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7179 Chung-Ju Wu <jasonwucj@gmail.com>
7181 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7182 (combo): New attribute.
7183 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7185 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7187 * config/nds32/nds32.opt: Change -mcmodel= default value.
7189 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7190 Monk Chiang <sh.chiang04@gmail.com>
7191 Chung-Ju Wu <jasonwucj@gmail.com>
7193 * config/nds32/constants.md (unspec_element): New enum.
7194 * config/nds32/constraints.md (Umw): New constraint.
7195 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7196 * config/nds32/nds32-intrinsic.md: Likewise.
7197 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7198 (nds32_valid_smw_lwm_base_p): New.
7199 (nds32_output_smw_single_word): New.
7200 (nds32_output_lmw_single_word): New.
7201 (nds32_expand_unaligned_load): New.
7202 (nds32_expand_unaligned_store): New.
7203 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7204 (nds32_output_smw_single_word): Declare.
7205 (nds32_output_lmw_single_word): Declare.
7206 (nds32_expand_unaligned_load): Declare.
7207 (nds32_expand_unaligned_store): Declare.
7208 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7209 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7210 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7211 NDS32_BUILTIN_UASTORE_DW.
7212 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7215 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7216 Kito Cheng <kito.cheng@gmail.com>
7217 Chung-Ju Wu <jasonwucj@gmail.com>
7219 * config/nds32/nds32-intrinsic.c
7220 (nds32_expand_builtin_null_ftype_reg): Delete.
7221 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7222 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7223 (nds32_read_argument): New.
7224 (nds32_legitimize_target): Ditto.
7225 (nds32_legitimize_argument): Ditto.
7226 (nds32_check_constant_argument): Ditto.
7227 (nds32_expand_unop_builtin): Ditto.
7228 (nds32_expand_unopimm_builtin): Ditto.
7229 (nds32_expand_binop_builtin): Ditto.
7230 (nds32_builtin_decl_impl): Ditto.
7231 (builtin_description): Ditto.
7232 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7233 (nds32_init_builtins_impl): Ditto.
7234 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7235 (nds32_builtin_decl): New.
7236 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7237 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7239 2018-03-02 Jeff Law <law@redhat.com>
7241 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7242 (redundant_insn, fill_simple_delay_slots): Likewise.
7243 (fill_slots_from_thread): Likewise.
7244 * resource.c (mark_referenced_resources): Likewise.
7245 (mark_set_resources, find_dead_or_set_registers): Likewise.
7247 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7249 * substring-locations.h (format_warning_va): Formatting fix for
7251 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7253 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7254 * substring-locations.c: Include intl.h.
7255 (format_warning_va): Turned into small wrapper around
7256 format_warning_n_va, renamed to ...
7257 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7258 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7260 (format_warning_at_substring_n): New function.
7261 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7262 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7263 format_warning_at_substring with just a shorter name instead of
7264 const function pointer.
7265 (fmtwarn_n): New function.
7266 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7267 appropriate, get rid of all the fmtstr temporaries, move conditionals
7268 with G_() wrapped string literals directly into fmtwarn arguments,
7269 cast dir.len to (int), formatting fixes.
7271 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7273 * doc/invoke.texi: Remove "Cilk Plus" references.
7275 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7276 Richard Biener <rguenther@suse.de>
7279 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7280 for error or warning attributes if CALL_FROM_THUNK_P is set.
7283 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7286 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7287 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7290 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7291 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7293 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7294 instead of -1U in last predictors element's probability member.
7296 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7299 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7300 arguments if they are comparable.
7302 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7304 PR tree-optimization/84634
7305 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7306 masks and masked_loop_p with a single loop_masks, making sure it's
7307 null for bb vectorization.
7309 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7311 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7312 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7313 loop->force_vectorize to check whether there is no alias.
7315 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7318 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7320 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7322 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7323 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7324 instead of a loop around prev_real_insn.
7325 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7329 * config/i386/i386.c (ix86_print_operand): Use conditional
7330 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7333 2018-03-02 Richard Biener <rguenther@suse.de>
7335 PR tree-optimization/84427
7336 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7337 (bitmap_set_subtract_values): Rewrite to handle multiple
7340 (prune_clobbered_mems): Likewise.
7341 (phi_translate): Take edge instead of pred/phiblock.
7342 (phi_translate_1): Likewise.
7343 (phi_translate_set): Likewise. Insert all translated
7344 exprs for a value into the set, keeping possibly multiple
7345 expressions per value.
7346 (compute_antic_aux): Adjust for phi_translate changes.
7347 When intersecting union the expressions and prune those
7348 not in the final value set, keeping possibly multiple
7349 expressions per value. Do not use value-insertion
7350 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7351 all expressions. Add verification that the value-sets
7352 only shrink during iteration.
7353 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7354 (do_pre_regular_insertion): Likewise.
7355 (do_pre_partial_partial_insertion): Likewise.
7357 2018-03-02 Richard Biener <rguenther@suse.de>
7360 * config/darwin.c (saved_debug_info_level): New static global.
7361 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7362 (darwin_asm_lto_end): Restore debug info generation settings.
7364 2018-03-01 Martin Liska <mliska@suse.cz>
7367 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7370 2018-03-01 Richard Biener <rguenther@suse.de>
7373 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7374 type annotation with LTO when debug was disabled at compile-time.
7376 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7378 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7380 (mips_final_postscan_insn): Likewise.
7382 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7384 PR rtl-optimization/84528
7385 * alias.c (init_alias_target): Add commentary.
7386 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7387 a unique base value if the frame pointer is not eliminated
7388 to the stack pointer.
7390 2018-03-01 Tom de Vries <tom@codesourcery.com>
7392 PR rtl-optimization/83327
7393 * lra-int.h (hard_regs_spilled_into): Declare.
7394 * lra.c (hard_regs_spilled_into): Define.
7395 (init_reg_info): Init hard_regs_spilled_into.
7396 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7397 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7398 (process_bb_lives): Handle hard_regs_spilled_into.
7399 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7400 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7402 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7404 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7405 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7406 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7407 * config/rs6000/aix72.h: New file.
7409 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7411 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7412 instead of warning_at with conditional singular and plural messages
7416 * stor-layout.c (update_alignment_for_field): For
7417 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7418 && !DECL_PACKED (field), do the alignment update, just use
7419 only desired_align instead of MAX (type_align, desired_align)
7421 (place_field): Don't do known_align < desired_align handling
7422 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7423 is non-NULL, instead do it after rli->prev_field handling and
7424 only if not within a bitfield word. For DECL_PACKED (field)
7425 use type_align of BITS_PER_UNIT.
7427 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7429 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7430 superfluous parentheses and trailing spaces.
7432 2018-02-28 Richard Biener <rguenther@suse.de>
7434 PR tree-optimization/84584
7435 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7436 SCoPs with fake exit edge.
7438 2018-02-28 Martin Liska <mliska@suse.cz>
7441 * timevar.c (timer::print): Fix format to properly print 100%
7444 2018-02-28 Richard Biener <rguenther@suse.de>
7447 * genmatch.c (capture_info::walk_match): Do not mark
7448 captured expressions without operands as expr_p given
7449 they act more like predicates and should be subject to
7450 "lost tail" side-effect preserving.
7452 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7454 PR rtl-optimization/81611
7455 * auto-inc-dec.c (attempt_change): Move dead note from
7456 mem_insn if it's the next use of regno
7457 (find_address): Take address use of reg holding
7458 non-incremented value. Add parm to limit search to the named
7460 (merge_in_block): Attempt to use a mem insn that is the next
7461 use of the original regno.
7463 2018-02-27 Martin Sebor <msebor@redhat.com>
7466 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7467 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7469 2018-02-27 Martin Sebor <msebor@redhat.com>
7471 PR translation/84207
7472 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7473 n argument to unsigned HOST_WIDE_INT.
7474 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7475 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7476 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7477 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7479 2018-02-27 Richard Biener <rguenther@suse.de>
7481 PR tree-optimization/84512
7482 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7483 Do not use the estimate returned from record_stmt_cost for
7484 the scalar iteration cost but sum properly using add_stmt_cost.
7486 2018-02-27 Richard Biener <rguenther@suse.de>
7488 PR tree-optimization/84466
7489 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7490 Adjust last change to less strictly validate use operands.
7492 2018-02-27 Martin Liska <mliska@suse.cz>
7494 PR gcov-profile/84548
7495 * gcov.c (process_file): Allow partial overlap and consider it
7496 also as group functions.
7497 (output_lines): Properly calculate range of lines for a group.
7499 2018-02-27 Martin Liska <mliska@suse.cz>
7501 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7502 'ggc' suffixes. Change first column width.
7503 (timer::print): Fix formatting of the column.
7505 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7507 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7508 preserve inline entry blocks for the sake of debug inline
7509 entry point markers alone.
7510 (remove_unused_locals): Suggest in comments a better place to
7511 force the preservation of inline entry blocks that are
7512 otherwise unused, but do not preserve them.
7514 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7516 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7518 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7521 * config/i386/constraints.md (Bs): Replace
7522 ix86_indirect_branch_register with
7523 TARGET_INDIRECT_BRANCH_REGISTER.
7525 * config/i386/i386.md (indirect_jump): Likewise.
7526 (tablejump): Likewise.
7527 (*sibcall_memory): Likewise.
7528 (*sibcall_value_memory): Likewise.
7529 Peepholes of indirect call and jump via memory: Likewise.
7530 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7531 (*sibcall_value_GOT_32): Likewise.
7532 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7533 (GOT_memory_operand): Likewise.
7534 (call_insn_operand): Likewise.
7535 (sibcall_insn_operand): Likewise.
7536 (GOT32_symbol_operand): Likewise.
7537 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7539 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7541 PR rtl-optimization/83496
7542 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7543 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7544 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7545 redundant insn, if any.
7546 (relax_delay_slots): Likewise.
7547 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7549 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7551 PR tree-optimization/83965
7552 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7553 that grouped statements are part of a reduction chain. Return
7554 true if the statement is not marked as a reduction itself but
7556 (vect_recog_dot_prod_pattern): Don't check whether the statement
7557 is part of a group here.
7558 (vect_recog_sad_pattern): Likewise.
7559 (vect_recog_widen_sum_pattern): Likewise.
7561 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7564 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7566 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7569 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7571 (ix86_output_indirect_function_return): New prototype.
7572 (ix86_split_simple_return_pop_internal): Likewise.
7573 * config/i386/i386.c (indirect_return_via_cx): New.
7574 (indirect_return_via_cx_bnd): Likewise.
7575 (indirect_thunk_name): Handle return va CX_REG.
7576 (output_indirect_thunk_function): Create alias for
7577 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7578 (ix86_output_indirect_jmp): Remove the bool argument.
7579 (ix86_output_indirect_function_return): New function.
7580 (ix86_split_simple_return_pop_internal): Likewise.
7581 * config/i386/i386.md (*indirect_jump): Don't pass false
7582 to ix86_output_indirect_jmp.
7583 (*tablejump_1): Likewise.
7584 (simple_return_pop_internal): Change it to define_insn_and_split.
7585 Call ix86_split_simple_return_pop_internal to split it for
7587 (simple_return_indirect_internal): Call
7588 ix86_output_indirect_function_return instead of
7589 ix86_output_indirect_jmp.
7591 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7594 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7595 memset and value initialization afterwards.
7597 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7599 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7601 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7604 * common/config/aarch64/aarch64-common.c
7605 (aarch_option_optimization_table[]): Switch
7606 off fomit-frame-pointer
7608 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7609 Chung-Ju Wu <jasonwucj@gmail.com>
7611 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7613 (store_multiple): Ditto.
7615 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7617 * config.gcc: Add --with-cpu support for nds32 target.
7618 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7619 * config/nds32/nds32.opt: Add -mcpu= option.
7621 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7623 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7624 isel=yes): Warn for these deprecated options.
7626 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7628 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7629 ISA_2_5_MASKS_EMBEDDED.
7631 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7633 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7634 p->max as pointers rather than using iterative_hash_expr.
7636 2018-02-23 Carl Love <cel@us.ibm.com>
7638 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7639 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7641 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7642 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7643 P8V_BUILTIN_VEC_VUNSIGNED2.
7645 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7648 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7649 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7651 * lra-constraints.c (curr_insn_transform): Set up
7652 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7654 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7655 * lra-eliminations.c (spill_pseudos): Ditto.
7656 (process_insn_for_elimination): Ditto.
7657 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7658 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7659 LRA_NON_CLOBBERED_ALT.
7661 2018-02-22 Martin Sebor <msebor@redhat.com>
7663 PR tree-optimization/84480
7664 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7665 to maybe_diag_stxncpy_trunc. Call it.
7666 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7667 from gimple_fold_builtin_strcpy. Print inlining stack.
7668 (handle_builtin_stxncpy): Print inlining stack.
7669 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7671 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7674 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7675 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7676 and -fcheck-pointer-bounds are used together.
7677 (indirect_thunk_prefix): New enum.
7678 (indirect_thunk_need_prefix): New function.
7679 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7680 "_nt" instead of "_bnd" for NOTRACK prefix.
7681 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7682 (output_indirect_thunk_function): Likewise.
7684 (ix86_code_end): Update output_indirect_thunk_function calls.
7685 (ix86_output_indirect_branch_via_reg): Replace
7686 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7687 (ix86_output_indirect_branch_via_push): Likewise.
7688 (ix86_output_function_return): Likewise.
7689 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7690 incompatible with -fcf-protection=branch and
7691 -fcheck-pointer-bounds.
7693 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7696 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7697 Change gcc_assert call to output_operand_lossage.
7699 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7701 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7703 2018-02-22 DJ Delorie <dj@redhat.com>
7704 Sebastian Perta <sebastian.perta@renesas.com>
7705 Oleg Endo <olegendo@gcc.gnu.org>
7707 * config/rx/rx.c (rx_rtx_costs): New function.
7708 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7710 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7712 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7714 2018-02-22 Martin Liska <mliska@suse.cz>
7717 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7718 Add "native" as a possible value.
7720 2018-02-22 Martin Liska <mliska@suse.cz>
7723 * config/i386/i386.c (ix86_option_override_internal):
7724 Add "native" as a possible value for -march and -mtune.
7726 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7729 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7730 to all type variants.
7732 PR tree-optimization/84503
7733 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7734 width as info->bitpos + info->bitsize - start.
7735 (merged_store_group::merge_overlapping): Simplify width computation.
7736 (check_no_overlap): New function.
7737 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7738 start + width and last_order of the group, fail if check_no_overlap
7740 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7741 to group if check_no_overlap fails.
7743 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7745 * config/rs6000/altivec.md: Delete contraint arguments to
7746 define_expand, define_split, and define_peephole2, and in
7747 define_insn_and_split if always unused.
7748 * config/rs6000/darwin.md: Ditto.
7749 * config/rs6000/dfp.md: Ditto.
7750 * config/rs6000/rs6000.md: Ditto.
7751 * config/rs6000/sync.md: Ditto.
7752 * config/rs6000/vector.md: Ditto.
7753 * config/rs6000/vsx.md: Ditto.
7755 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7757 * config/rs6000/altivec.md: Write output control strings as braced
7758 blocks instead of double-quoted strings.
7759 * config/rs6000/darwin.md: Ditto.
7760 * config/rs6000/rs6000.md: Ditto.
7761 * config/rs6000/vector.md: Ditto.
7762 * config/rs6000/vsx.md: Ditto.
7764 2018-02-21 Jason Merrill <jason@redhat.com>
7766 PR c++/84314 - ICE with templates and fastcall attribute.
7767 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7769 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7771 * ipa-cp.c (determine_versionability): Fix comment typos.
7773 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7776 * ipa-cp.c (determine_versionability): Do not version functions caling
7779 2018-02-21 Martin Liska <mliska@suse.cz>
7782 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7783 Add "native" as a possible value.
7784 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7785 the macro when native cpu detection is available.
7787 2018-02-21 Martin Liska <mliska@suse.cz>
7790 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7791 Add "native" as a possible value.
7792 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7793 when native cpu detection is available.
7795 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7796 Martin Sebor <msebor@redhat.com>
7798 PR tree-optimization/84478
7799 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7801 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7802 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7803 support which is conservatively correct, for 2 only stay conservative
7804 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7805 argument to the 2 argument get_range_strlen, adjust 6 arg
7806 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7808 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7809 (gimple_fold_builtin_strlen): Pass true as last argument to
7812 2018-02-20 Martin Sebor <msebor@redhat.com>
7815 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7816 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7817 (builtin_memref::builtin_memref): Factor out parts into
7818 set_base_and_offset and call it.
7820 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7823 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7824 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7825 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7826 search at the associated MODE_INT.
7828 2018-02-20 Jeff Law <law@redhat.com>
7831 PR tree-optimization/81592
7833 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7834 for range data rather than using global data.
7835 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7836 range data rather than using global data.
7837 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7838 pass it to children as needed.
7839 (struct directive::fmtresult): Similarly.
7840 (struct directive::set_width): Similarly.
7841 (struct directive::set_precision): Similarly.
7842 (format_integer, format_directive, parse_directive): Similarly.
7843 (format_none): Accept unnamed vr_values parameter.
7844 (format_percent, format_floating, format_character): Similarly.
7845 (format_string, format_plain): Similarly.
7846 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7847 the EVRP range analyzer for range data rather than using global data.
7848 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7849 gimple-ssa-evrp-analyze.h
7850 (class sprintf_dom_walker): Add after_dom_children member function.
7851 Add evrp_range_analyzer member.
7852 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7853 range analyzer as needed.
7854 (sprintf_dom_walker::after_dom_children): New member function.
7855 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7857 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7858 (evrp_range_analyzer::pop_to_marker): Likewise.
7860 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7862 PR tree-optimization/84419
7863 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7864 with the required type if its current type is compatible but
7867 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7870 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7871 after vectorization.
7873 2018-02-20 Martin Liska <mliska@suse.cz>
7876 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7877 possible values if we don't have a hint.
7879 2018-02-20 Martin Liska <mliska@suse.cz>
7883 * final.c (shorten_branches): Build align_tab array with one
7885 * opts.c (finish_options): Add alignment option limit check.
7886 (MAX_CODE_ALIGN): Likewise.
7887 (MAX_CODE_ALIGN_VALUE): Likewise.
7888 * doc/invoke.texi: Document maximum allowed option value for
7889 all -falign-* options.
7891 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7894 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7895 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7896 * var-tracking.c (emit_note_insn_var_location): Remove all references
7897 to NOTE_INSN_CALL_ARG_LOCATION.
7898 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7899 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7900 Use copy_rtx_if_shared.
7901 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7902 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7903 (dwarf2out_var_location): Remove handling of
7904 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7906 * final.c (final_scan_insn): Remove all references to
7907 NOTE_INSN_CALL_ARG_LOCATION.
7908 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7909 before dumping final insns.
7910 * except.c (emit_note_eh_region_end): Remove all references to
7911 NOTE_INSN_CALL_ARG_LOCATION.
7912 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7913 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7914 * config/arc/arc.c (hwloop_optimize): Likewise.
7915 * config/arm/arm.c (create_fix_barrier): Likewise.
7916 * config/s390/s390.c (s390_chunkify_start): Likewise.
7917 * config/sh/sh.c (find_barrier): Likewise.
7918 * config/i386/i386.c (rest_of_insert_endbranch,
7919 ix86_seh_fixup_eh_fallthru): Likewise.
7920 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7921 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7922 * config/frv/frv.c (frv_function_prologue): Likewise.
7923 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7925 (note_outside_basic_block_p): Remove all references to
7926 NOTE_INSN_CALL_ARG_LOCATION.
7927 * gengtype.c (adjust_field_rtx_def): Likewise.
7928 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7930 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7931 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7934 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7937 PR tree-optimization/84452
7938 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7939 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7942 2018-02-19 Martin Liska <mliska@suse.cz>
7945 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7947 2018-02-19 Martin Liska <mliska@suse.cz>
7948 Richard Sandiford <richard.sandiford@linaro.org>
7950 PR tree-optimization/82491
7951 * gimple-fold.c (get_base_constructor): Make earlier bail out
7954 2018-02-19 Carl Love <cel@us.ibm.com>
7956 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7957 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7959 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7960 P8V_BUILTIN_VEC_NEG.
7962 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7964 * config/rl78/rl78.md (movdf): New define expand.
7966 2018-02-19 Martin Liska <mliska@suse.cz>
7969 * doc/invoke.texi: Fix typo.
7970 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7972 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7974 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7975 handle rs6000_single_float and rs6000_double_float specially for
7978 2018-02-16 Jeff Law <law@redhat.com>
7980 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7983 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7986 * ipa-inline.c (inline_small_functions): Fix a typo.
7988 2018-02-16 Nathan Sidwell <nathan@acm.org>
7990 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7992 2018-02-16 Carl Love <cel@us.ibm.com>
7994 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7995 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7996 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7997 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7998 expansion to P8V_BUILTIN_VEC_FLOAT2.
8000 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
8002 PR rtl-optimization/70023
8003 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
8004 src_regno into account.
8006 2018-02-16 Carl Love <cel@us.ibm.com>
8008 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
8009 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
8010 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
8011 * config/rs6000/rs6000.c: Remove case statements for
8012 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
8013 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
8014 and P9V_BUILTIN_VEC_VINSERT4B.
8015 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
8016 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
8017 * config/rs6000/vsx.md:
8018 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
8021 2018-02-16 Carl Love <cel@us.ibm.com>
8023 * config/rs6000/altivec.h: Add builtin names vec_extract4b
8025 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
8027 * config/rs6000/rs6000-c.c: Add the definitions for
8028 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
8029 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
8030 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
8031 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
8032 definition for insert4b and define insn *insert3b_internal.
8033 * doc/extend.texi: Add documentation for vec_extract4b.
8035 2018-02-16 Nathan Sidwell <nathan@acm.org>
8037 * doc/extend.texi (Backwards Compatibility): Mention friend
8038 injection. Note for-scope is deprecated.
8039 * doc/invoke.texi (-ffriend-injection): Deprecate.
8041 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
8043 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
8044 that moved to I2, also allow destinations that are a paradoxical
8045 subreg (instead of a normal reg).
8047 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8050 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8053 2018-02-16 Richard Biener <rguenther@suse.de>
8055 PR tree-optimization/84037
8056 PR tree-optimization/84016
8058 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8059 Adjust vec_construct for the fact we need additional higher latency
8060 128bit inserts for AVX256 and AVX512 vector builds.
8061 (ix86_add_stmt_cost): Scale vector construction cost for
8064 2018-02-16 Richard Biener <rguenther@suse.de>
8066 PR tree-optimization/84417
8067 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8068 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8069 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8071 2018-02-16 Martin Liska <mliska@suse.cz>
8074 * internal-fn.def (ASAN_CHECK): Set proper flags.
8075 (ASAN_MARK): Likewise.
8077 2018-02-16 Julia Koval <julia.koval@intel.com>
8079 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8080 from PTA_CANNONLAKE.
8082 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8085 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8086 Use ++iter rather than iter++ for std::list iterators.
8087 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8088 defer deleting them until all nodes in the forest are processed. Do
8089 free even leaf nodes. Change to_process into auto_vec.
8092 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8093 * vec.h (vec_default_construct): Use memset instead of placement new
8094 if BROKEN_VALUE_INITIALIZATION is defined.
8095 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8096 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8099 PR rtl-optimization/83723
8100 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8101 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8102 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8104 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8106 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8108 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8110 PR rtl-optimization/81443
8111 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8112 from inner REGs to paradoxical SUBREGs.
8114 2018-02-16 Richard Biener <rguenther@suse.de>
8116 PR tree-optimization/84399
8117 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8118 For operands we can analyze at their definition make sure we can
8119 analyze them at each use as well.
8121 2018-02-16 Richard Biener <rguenther@suse.de>
8123 PR tree-optimization/84190
8124 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8125 volatile accesses if the decl isn't volatile.
8127 2018-02-15 Jason Merrill <jason@redhat.com>
8129 PR c++/84314 - ICE with templates and fastcall attribute.
8130 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8131 TYPE_CANONICAL on an existing type.
8133 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8135 PR tree-optimization/84383
8136 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8137 dstoff nor call operand_equal_p if dstbase is NULL.
8139 PR tree-optimization/84334
8140 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8141 also a CONSTANT_CLASS_P, punt.
8143 2018-02-14 Jim Wilson <jimw@sifive.com>
8145 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8146 first SMALL_OPERAND check. New local min_second_step. Move assert
8147 to where locals are set. Add TARGET_RVC support.
8148 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8150 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8152 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8154 2018-02-14 Martin Sebor <msebor@redhat.com>
8156 PR tree-optimization/83698
8157 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8158 arrays constrain the offset range to their bounds.
8159 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8160 (builtin_access::overlap): Avoid setting the size of overlap if it's
8162 (maybe_diag_overlap): Also consider arrays when deciding what values
8163 of offsets to include in diagnostics.
8165 2018-02-14 Martin Sebor <msebor@redhat.com>
8168 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8169 that correspond to the kind of a declaration.
8171 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8174 * config/pa/pa.md: Load address of PIC label using the linkage table
8175 if the label is nonlocal.
8177 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8179 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8180 warning message if user requests -maltivec=be.
8181 * doc/invoke.texi: Document deprecation of -maltivec=be.
8183 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8186 * config/rs6000/rs6000-c.c: Update definitions for
8187 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8188 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8190 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8193 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8194 add _get_ssp intrinsics. Remove argument from
8195 __builtin_ia32_rdssp[d|q].
8196 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8197 * config/i386/i386-builtin.def: Remove argument from
8198 __builtin_ia32_rdssp[d|q].
8199 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8200 ix86_expand_special_args_builtin for _rdssp[d|q].
8201 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8202 Clear register before usage.
8203 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8204 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8206 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8208 PR tree-optimization/84357
8209 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8210 operand 1 of an ARRAY_REF too.
8212 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8215 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8216 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8218 (set_of_reg): New struct.
8219 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8220 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8221 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8223 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8224 Split into bitclr, bitset, bitinvert patterns if appropriate.
8225 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8226 use rx_fuse_in_memory_bitop.
8227 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8228 to named insn, correct maximum insn length.
8230 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8233 * machmode.def: Define a complex mode for PARTIAL_INT.
8234 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8236 * doc/rtl.texi: Document CSPImode.
8237 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8239 (msp430_hard_regno_nregs_with_padding): Likewise.
8241 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8244 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8246 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8248 PR rtl-optimization/84169
8249 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8250 we generated a parallel as new i3 and we split that to new i2 and i3
8251 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8252 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8253 those to i2, not i1. Partially rewrite this scan code.
8255 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8258 * stor-layout.c (place_field): For variable length fields, adjust
8259 offset_align afterwards not just based on the field's alignment,
8260 but also on the size.
8263 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8264 of exps and logs in the use_exp2 case.
8266 2018-02-13 Jeff Law <law@redhat.com>
8268 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8271 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8274 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8278 * common.opt (gas-loc-support, gas-locview-support): New.
8279 (ginline-points, ginternal-reset-location-views): New.
8280 * doc/invoke.texi: Document them. Use @itemx where intended.
8281 (gvariable-location-views): Adjust.
8282 * target.def (reset_location_view): New.
8283 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8284 (TARGET_RESET_LOCATION_VIEW): New.
8285 * doc/tm.texi: Rebuilt.
8286 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8287 (dwarf2out_default_as_locview_support): New.
8288 (output_asm_line_debug_info): Use option variables.
8289 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8290 (output_loc_list): Likewise.
8291 (add_high_low_attributes): Check option variables.
8292 Don't output entry view attribute in strict mode.
8293 (gen_inlined_subroutine_die): Check option variables.
8294 (dwarf2out_inline_entry): Likewise.
8295 (init_sections_and_labels): Likewise.
8296 (dwarf2out_early_finish): Likewise.
8297 (maybe_reset_location_view): New, from...
8298 (dwarf2out_var_location): ... here. Call it.
8299 * debug.h (dwarf2out_default_as_loc_support): Declare.
8300 (dwarf2out_default_as_locview_support): Declare.
8301 * hooks.c (hook_int_rtx_insn_0): New.
8302 * hooks.h (hook_int_rtx_insn_0): Declare.
8303 * toplev.c (process_options): Take -gas-loc-support and
8304 -gas-locview-support from dwarf2out. Enable
8305 -gvariable-location-views by default only with locview
8306 assembler support. Enable -ginternal-reset-location-views by
8307 default only if the target defines the corresponding hook.
8308 Enable -ginline-points by default if location views are
8309 enabled; force it disabled if statement frontiers are
8311 * tree-inline.c (expand_call_inline): Check option variables.
8312 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8314 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8316 PR tree-optimization/84321
8317 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8318 handling. Also check whether the anti-range contains any values
8319 that satisfy the mask; switch to a VR_RANGE if not.
8321 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8324 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8326 2018-02-13 Martin Jambor <mjambor@suse.cz>
8329 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8330 of call statements, also set location of a load to a temporary.
8332 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8334 * config/rl78/rl78.c (add_vector_labels): New function.
8335 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8336 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8337 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8338 which checks that no arguments are passed.
8339 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8340 * doc/extend.texi: Documentation for the new attribute.
8342 2018-02-13 Andreas Schwab <schwab@suse.de>
8344 * config/riscv/linux.h (CPP_SPEC): Define.
8346 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8349 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8350 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8351 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8352 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8353 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8354 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8355 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8356 temporarily for AES and PCLMUL builtins.
8358 PR tree-optimization/84339
8359 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8360 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8364 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8365 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8366 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8368 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8370 * omp-simd-clone.h: New file.
8371 * omp-simd-clone.c: Include omp-simd-clone.h.
8372 (expand_simd_clones): No longer static.
8373 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8374 cgraph.h and omp-simd-clone.h.
8375 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8376 (vect_recog_widen_shift_pattern): Formatting fix.
8377 (vect_pattern_recog_1): Don't check optab for calls.
8380 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8381 operands[2] into a REG before using gen_lowpart on it.
8383 2018-02-12 Jeff Law <law@redhat.com>
8386 * config/sh/sh.c (find_barrier): Consider a sibling call
8389 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8390 successfully back substituting a reg.
8392 2018-02-12 Richard Biener <rguenther@suse.de>
8394 PR tree-optimization/84037
8395 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8396 parameter, move visited init to caller.
8397 (vect_slp_analyze_operations): Separate cost from validity
8398 check, initialize visited once for all instances.
8399 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8401 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8403 (vect_model_promotion_demotion_cost): Likewise.
8404 (vectorizable_bswap): Guard cost modeling with !slp_node
8405 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8407 (vectorizable_call): Likewise.
8408 (vectorizable_conversion): Likewise.
8409 (vectorizable_assignment): Likewise.
8410 (vectorizable_shift): Likewise.
8411 (vectorizable_operation): Likewise.
8412 (vectorizable_store): Likewise.
8413 (vectorizable_load): Likewise.
8414 (vectorizable_condition): Likewise.
8415 (vectorizable_comparison): Likewise.
8417 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8420 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8421 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8422 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8424 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8427 * params.def (inline-min-speedup): Increase from 8 to 15.
8428 (max-inline-insns-auto): Decrease from 40 to 30.
8429 * ipa-split.c (consider_split): Add some buffer for function to
8430 be considered inlining candidate.
8431 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8434 2018-02-12 Richard Biener <rguenther@suse.de>
8436 PR tree-optimization/84037
8437 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8438 matched stmts if we cannot swap the non-matched ones.
8440 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8442 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8443 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8444 _mm_maskz_scalef_round_ss): New intrinsics.
8445 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8446 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8447 __builtin_ia32_scalefss_round): Remove.
8448 (__builtin_ia32_scalefsd_mask_round,
8449 __builtin_ia32_scalefss_mask_round): New intrinsics.
8450 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8451 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8452 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8453 "<round_constraint>")): Changed to ...
8454 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8455 "<round_scalar_constraint>")): ... this.
8456 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8457 %0, %1, %2<round_op3>}"): Changed to ...
8458 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8459 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8460 %2<round_scalar_mask_op3>}"): ... this.
8461 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8463 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8465 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8466 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8467 (_mm_maskz_sqrt_round_ss): New intrinsics.
8468 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8469 (__builtin_ia32_sqrtsd_mask_round)
8470 (__builtin_ia32_sqrtss_mask_round): New builtins.
8471 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8472 (__builtin_ia32_sqrtss_round): Remove.
8473 (__builtin_ia32_sqrtsd_mask_round)
8474 (__builtin_ia32_sqrtss_mask_round): New builtins.
8475 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8476 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8477 ((match_operand:VF_128 1 "vector_operand"
8478 "xBm,<round_constraint>")): Changed to ...
8479 ((match_operand:VF_128 1 "vector_operand"
8480 "xBm,<round_scalar_constraint>")): ... this.
8481 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8482 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8483 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8484 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8485 %<iptr>1<round_scalar_mask_op3>}): ... this.
8486 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8487 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8489 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8492 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8493 Cast vec_cmpeq result to correct type.
8494 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8495 Cast vec_cmpgt result to correct type.
8497 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8499 * final.c (final_scan_insn_1): Renamed from...
8500 (final_scan_insn): ... this. New wrapper, to recover
8501 seen from the outermost call in recursive ones.
8502 * config/sparc/sparc.c (output_return): Drop seen from call.
8503 (output_sibcall): Likewise.
8504 * config/visium/visium.c (output_branch): Likewise.
8506 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8508 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8511 2018-02-10 Alan Modra <amodra@gmail.com>
8514 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8515 Specify LR as an input.
8517 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8520 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8521 remove_member_access_dummy_vars): New functions.
8522 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8523 lower_omp_1, execute_lower_omp): Use them.
8525 PR rtl-optimization/84308
8526 * shrink-wrap.c (spread_components): Release todo vector.
8528 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8530 PR rtl-optimization/57193
8531 * ira-color.c (struct allocno_color_data): Add member
8532 conflict_allocno_hard_prefs.
8533 (update_conflict_allocno_hard_prefs): New.
8534 (bucket_allocno_compare_func): Add a preference based on
8535 conflict_allocno_hard_prefs.
8536 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8537 (color_allocnos): Remove a dead code. Initiate
8538 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8540 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8543 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8544 constraint from =wa to wa. Avoid a subreg on the output operand,
8545 instead use a pseudo and subreg it in a move.
8546 (p9_xxbrd_<mode>): Changed to ...
8547 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8548 (p9_xxbrd_v2df): New expander.
8549 (p9_xxbrw_<mode>): Changed to ...
8550 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8551 (p9_xxbrw_v4sf): New expander.
8553 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8555 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8557 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8560 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8561 multiply in 32-bit mode.
8562 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8563 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8566 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8568 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8569 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8570 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8571 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8573 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8576 * dwarf2out.c (is_trivial_indirect_ref): New function.
8577 (dwarf2out_late_global_decl): Do not generate a location
8578 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8579 and that are not defined in the current unit.
8581 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8583 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8584 instead of a libcall for UNORDERED.
8586 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8589 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8590 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8592 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8595 * config/s390/s390.c (s390_set_current_function): Invoke
8596 s390_indirect_branch_settings also if fndecl didn't change.
8598 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8600 * config/rs6000/rs6000.md (blockage): Set length to zero.
8602 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8604 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8606 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8609 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8610 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8614 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8615 PARALLEL incoming that failed vt_get_decl_and_offset check.
8618 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8619 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8621 (get_variable_section): For decls in named .bss* sections pass true as
8622 second argument to bss_initializer_p.
8624 2018-02-09 Marek Polacek <polacek@redhat.com>
8625 Jakub Jelinek <jakub@redhat.com>
8628 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8629 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8630 Sync some changes from cxx_fold_indirect_ref.
8632 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8634 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8636 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8637 (BLOCK_INLINE_ENTRY_LABEL): New.
8638 (dwarf2out_var_location): Disregard inline entry markers.
8639 (inline_entry_data): New struct.
8640 (inline_entry_data_hasher): New hashtable type.
8641 (inline_entry_data_hasher::hash): New.
8642 (inline_entry_data_hasher::equal): New.
8643 (inline_entry_data_table): New variable.
8644 (add_high_low_attributes): Add DW_AT_entry_pc and
8645 DW_AT_GNU_entry_view attributes if a pending entry is found
8646 in inline_entry_data_table. Add old entry_pc attribute only
8647 if debug nonbinding markers are disabled.
8648 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8649 markers are enabled.
8650 (block_within_block_p, dwarf2out_inline_entry): New.
8651 (dwarf2out_finish): Check that no entries remained in
8652 inline_entry_data_table.
8653 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8654 (final_scan_insn, notice_source_line): Likewise.
8655 (rest_of_clean_state): Skip inline entry markers.
8656 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8658 * gimple.c (gimple_build_debug_inline_entry): New.
8659 * gimple.h (enum gimple_debug_subcode): Add
8660 GIMPLE_DEBUG_INLINE_ENTRY.
8661 (gimple_build_debug_inline_entry): Declare.
8662 (gimple_debug_inline_entry_p): New.
8663 (gimple_debug_nonbind_marker_p): Adjust.
8664 * insn-notes.def (INLINE_ENTRY): New.
8665 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8666 inline entry marker notes.
8667 (print_insn): Likewise.
8668 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8669 (INSN_DEBUG_MARKER_KIND): Likewise.
8670 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8671 * tree-inline.c (expand_call_inline): Build and insert
8672 debug_inline_entry stmt.
8673 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8674 inline entry blocks early, if nonbind markers are enabled.
8675 (dump_scope_block): Dump fragment info.
8676 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8677 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8678 (gimple_build_debug_inline_entry): New.
8679 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8680 Enable/disable inline entry points too.
8681 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8682 (DEBUG_INSN): Describe inline entry markers.
8684 * common.opt (gvariable-location-views): New.
8685 (gvariable-location-views=incompat5): New.
8686 * config.in: Rebuilt.
8687 * configure: Rebuilt.
8688 * configure.ac: Test assembler for view support.
8689 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8690 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8691 * dwarf2out.c (var_loc_view): New typedef.
8692 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8693 (dwarf2out_locviews_in_attribute): New.
8694 (dwarf2out_locviews_in_loclist): New.
8695 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8696 (enum dw_line_info_opcode): Add LI_adv_address.
8697 (struct dw_line_info_table): Add view.
8698 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8699 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8700 (zero_view_p): New variable.
8701 (ZERO_VIEW_P): New macro.
8702 (output_asm_line_debug_info): New.
8703 (struct var_loc_node): Add view.
8704 (add_AT_view_list, AT_loc_list): New.
8705 (add_var_loc_to_decl): Add view param. Test it against last.
8706 (new_loc_list): Add view params. Record them.
8707 (AT_loc_list_ptr): Handle loc and view lists.
8708 (view_list_to_loc_list_val_node): New.
8709 (print_dw_val): Handle dw_val_class_view_list.
8710 (size_of_die): Likewise.
8711 (value_format): Likewise.
8712 (loc_list_has_views): New.
8713 (gen_llsym): Set vl_symbol too.
8714 (maybe_gen_llsym, skip_loc_list_entry): New.
8715 (dwarf2out_maybe_output_loclist_view_pair): New.
8716 (output_loc_list): Output view list or entries too.
8717 (output_view_list_offset): New.
8718 (output_die): Handle dw_val_class_view_list.
8719 (output_dwarf_version): New.
8720 (output_compilation_unit_header): Use it.
8721 (output_skeleton_debug_sections): Likewise.
8722 (output_rnglists, output_line_info): Likewise.
8723 (output_pubnames, output_aranges): Update version comments.
8724 (output_one_line_info_table): Output view numbers in asm comments.
8725 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8726 Call maybe_gen_llsym.
8727 (loc_list_from_tree_1): Adjust.
8728 (add_AT_location_description): Create view list attribute if
8729 needed, check it's absent otherwise.
8730 (convert_cfa_to_fb_loc_list): Adjust.
8731 (maybe_emit_file): Call output_asm_line_debug_info for test.
8732 (dwarf2out_var_location): Reset views as needed. Precompute
8733 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8734 attribute. Set view.
8735 (new_line_info_table): Reset next view.
8736 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8737 (dwarf2out_source_line): Likewise. Output view resets and labels to
8738 the assembler, or select appropriate line info opcodes.
8739 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8740 (optimize_string_length): Catch it. Adjust.
8741 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8742 dw_val_class_view_list, and remove it if no longer needed.
8743 (hash_loc_list): Hash view numbers.
8744 (loc_list_hasher::equal): Compare them.
8745 (optimize_location_lists): Check whether a view list symbol is
8746 needed, and whether the locview attribute is present, and
8747 whether they match. Remove the locview attribute if no longer
8749 (index_location_lists): Call skip_loc_list_entry for test.
8750 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8751 Use output_dwarf_version.
8752 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8753 (struct dw_val_node): Add val_view_list.
8754 * final.c (SEEN_NEXT_VIEW): New.
8755 (set_next_view_needed): New.
8756 (clear_next_view_needed): New.
8757 (maybe_output_next_view): New.
8758 (final_start_function): Rename to...
8759 (final_start_function_1): ... this. Take pointer to FIRST,
8760 add SEEN parameter. Emit param bindings in the initial view.
8761 (final_start_function): Reintroduce SEEN-less interface.
8762 (final): Rename to...
8763 (final_1): ... this. Take SEEN parameter. Output final pending
8764 next view at the end.
8765 (final): Reintroduce seen-less interface.
8766 (final_scan_insn): Output pending next view before switching
8767 sections or ending a block. Mark the next view as needed when
8768 outputting variable locations. Notify debug backend of section
8769 changes, and of location view changes.
8770 (rest_of_handle_final): Adjust.
8771 * toplev.c (process_options): Autodetect value for debug variable
8772 location views option. Warn on incompat5 without -gdwarf-5.
8773 * doc/invoke.texi (gvariable-location-views): New.
8774 (gvariable-location-views=incompat5): New.
8775 (gno-variable-location-views): New.
8777 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8779 PR tree-optimization/84136
8780 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8781 that the result of find_edge is non-NULL.
8783 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8786 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8787 storing integer register in SImode. Fix cost of 256 and 512
8788 byte aligned SSE register store.
8790 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8792 * config/i386/i386.c (ix86_multiplication_cost): Fix
8793 multiplication cost for TARGET_AVX512DQ.
8795 2018-02-08 Marek Polacek <polacek@redhat.com>
8797 PR tree-optimization/84238
8798 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8801 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8803 PR tree-optimization/84265
8804 * tree-vect-stmts.c (vectorizable_store): Don't treat
8805 VMAT_CONTIGUOUS accesses as grouped.
8806 (vectorizable_load): Likewise.
8808 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8810 PR tree-optimization/81635
8811 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8812 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8813 (test_round_for_mask): New functions.
8814 (wide_int_cc_tests): Call test_round_for_mask.
8815 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8816 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8817 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8818 range returned by get_range_info.
8820 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8823 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8824 * symtab.c: Include builtins.h
8825 (symtab_node::output_to_lto_symbol_table_p): Move here
8826 from lto-streamer-out.c:output_symbol_p.
8827 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8828 (output_symbol_p): Move all logic to symtab.c
8829 (produce_symtab): Update.
8831 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8833 * config/s390/s390-opts.h (enum indirect_branch): Define.
8834 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8835 (s390_indirect_branch_via_thunk)
8836 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8837 (enum s390_indirect_branch_type): Define.
8838 * config/s390/s390.c (struct s390_frame_layout, struct
8839 machine_function): Remove.
8840 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8841 (indirect_branch_table_label_no, indirect_branch_table_name):
8843 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8844 (enum s390_indirect_branch_option): Define.
8845 (s390_return_addr_from_memory): New function.
8846 (s390_handle_string_attribute): New function.
8847 (s390_attribute_table): Add new attribute handler.
8848 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8849 (s390_indirect_branch_via_thunk): New function.
8850 (s390_indirect_branch_via_inline_thunk): New function.
8851 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8852 sibling call optimization for non z10 compiles.
8853 (s390_emit_call): Force indirect branch target to be a single
8854 register. Add r1 clobber for non-z10 compiles.
8855 (s390_emit_epilogue): Emit return jump via return_use expander.
8856 (s390_reorg): Handle JUMP_INSNs as execute targets.
8857 (s390_option_override_internal): Perform validity checks for the
8858 new command line options.
8859 (s390_indirect_branch_attrvalue): New function.
8860 (s390_indirect_branch_settings): New function.
8861 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8862 (s390_output_indirect_thunk_function): New function.
8863 (s390_code_end): Implement target hook.
8864 (s390_case_values_threshold): Implement target hook.
8865 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8867 * config/s390/s390.h (struct s390_frame_layout)
8868 (struct machine_function): Move here from s390.c.
8869 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8870 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8871 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8872 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8873 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8874 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8875 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8876 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8877 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8878 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8879 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8880 (mnemonic attribute): Add values which aren't recognized
8882 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8883 pattern for branch conversion. Fix mnemonic attribute.
8884 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8885 indirect branch via thunk if requested.
8886 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8887 ("*indirect_jump"): Disable for branch conversion using out of
8889 ("indirect_jump_via_thunk<mode>_z10")
8890 ("indirect_jump_via_thunk<mode>")
8891 ("indirect_jump_via_inlinethunk<mode>_z10")
8892 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8893 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8894 ("casesi_jump_via_inlinethunk<mode>_z10")
8895 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8896 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8897 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8898 ("*indirect2_jump"): Disable for branch conversion.
8899 ("casesi_jump"): Turn into expander and expand patterns for branch
8901 ("return_use"): New expander.
8902 ("*return"): Emit return via thunk and rename it to ...
8903 ("*return<mode>"): ... this one.
8904 * config/s390/s390.opt: Add new options and and enum for the
8907 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8909 * lra-constraints.c (match_reload): Unconditionally use
8910 gen_lowpart_SUBREG, rather than selecting between that
8911 and equivalent gen_rtx_SUBREG code.
8913 2018-02-08 Richard Biener <rguenther@suse.de>
8915 PR tree-optimization/84233
8916 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8917 changed flag instead of boguously re-using phi_inserted.
8919 2018-02-08 Martin Jambor <mjambor@suse.cz>
8921 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8922 static local variables.
8924 2018-02-08 Richard Biener <rguenther@suse.de>
8926 PR tree-optimization/84278
8927 * tree-vect-stmts.c (vectorizable_store): When looking for
8928 smaller vector types to perform grouped strided loads/stores
8929 make sure the mode is supported by the target.
8930 (vectorizable_load): Likewise.
8932 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8934 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8935 Increase LDP/STP opportunities by adding adjacent callee-saves.
8937 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8939 PR rtl-optimization/84068
8940 PR rtl-optimization/83459
8941 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8943 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8945 PR tree-optimization/84224
8946 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8947 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8950 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8953 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8954 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8955 count, remove one USE.
8957 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8959 * doc/install.texi (Configuration): Document the
8960 --with-long-double-format={ibm,ieee} PowerPC configuration
8964 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8965 Convert from define_expand to be define_insn_and_split. Rework
8966 float/double/_Float128 conversions to QI/HI/SImode to work with
8967 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8968 conversions to QI/HImode types did a store and then a load to
8969 truncate the value. For conversions to VSX registers, don't split
8970 the insn, instead emit the code directly. Use the code iterator
8971 any_fix to combine signed and unsigned conversions.
8972 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8973 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8974 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8975 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8976 (fix_<mode>di2_hw): Likewise.
8977 (fixuns_<mode>di2_hw): Likewise.
8978 (fix_<mode>si2_hw): Likewise.
8979 (fixuns_<mode>si2_hw): Likewise.
8980 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8981 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8982 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8983 fix<uns>_trunc<SFDF:mode>si2_p8.
8984 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8986 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8987 (fix<uns>_<mode>_mem): Likewise.
8988 (fctiw<u>z_<mode>_mem): Likewise.
8989 (fix<uns>_<mode>_mem): Likewise.
8990 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8991 the register allocator from doing a direct move to the GPRs to do
8992 a store, and instead use the ISA 3.0 store byte/half-word from
8993 vector register instruction. For IEEE 128-bit floating point,
8994 also optimize stores of 32-bit ints.
8995 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8997 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8999 * genextract.c (push_pathstr_operand): New function to support
9001 (walk_rtx): Call push_pathstr_operand.
9002 (print_path): Support [a-zA-Z].
9004 2018-02-07 Richard Biener <rguenther@suse.de>
9006 PR tree-optimization/84037
9007 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
9008 (cse_and_gimplify_to_preheader): Declare.
9009 (vect_get_place_in_interleaving_chain): Likewise.
9010 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
9012 (_loop_vec_info::~_loop_vec_info): Delete it.
9013 (cse_and_gimplify_to_preheader): New function.
9014 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
9015 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
9016 (vectorizable_load): Likewise. For grouped stores always base
9017 the IV on the first element.
9018 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
9019 condition before gimplifying.
9021 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9023 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
9024 *DIV_EXPR and *MOD_EXPR.
9026 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
9029 * config/i386/i386.c (ix86_option_override_internal): Mask out
9030 the CF_SET bit when checking -fcf-protection.
9032 2018-02-07 Tom de Vries <tom@codesourcery.com>
9035 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
9038 2018-02-07 Richard Biener <rguenther@suse.de>
9040 PR tree-optimization/84204
9041 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
9044 PR tree-optimization/84205
9045 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
9046 special-case isl_ast_op_zdiv_r.
9048 PR tree-optimization/84223
9049 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9050 Only add conditions from within the region.
9051 (gather_bbs::after_dom_children): Adjust.
9053 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9056 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9057 * config/avr/avr.md: Only post-reload split REG-REG moves if
9058 either register is GENERAL_REG_P.
9060 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9062 PR tree-optimization/84235
9063 * tree-ssa-scopedtables.c
9064 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9065 if the subtraction is performed in floating point type where NaNs are
9066 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9067 build 1. Formatting fix.
9069 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9072 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9073 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9074 and skip it regardless of bb boundaries. Use CALL_P macro,
9075 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9076 unnecessarily, formatting fix.
9078 2018-02-06 Michael Collison <michael.collison@arm.com>
9080 * config/arm/thumb2.md:
9081 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9082 (*thumb_mov_notscc): Ditto.
9084 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9087 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9088 unsigned_fix, not "s".
9090 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9092 * configure.ac (gcc_fn_eh_frame_ro): New function.
9093 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9094 correct .eh_frame permissions.
9095 * configure: Regenerate.
9097 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9099 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9102 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9104 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9105 Display warning message for -mno-speculate-indirect-jumps.
9107 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9109 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9111 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9113 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9115 PR tree-optimization/84225
9116 * tree-eh.c (find_trapping_overflow): Only call
9117 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9119 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9122 * config/i386/i386.c: Reimplement the check of possible options
9123 -mibt/-mshstk conbination. Change error messages.
9124 * doc/invoke.texi: Fix a typo: remove extra '='.
9126 2018-02-06 Marek Polacek <polacek@redhat.com>
9128 PR tree-optimization/84228
9129 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9131 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9134 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9135 emitted arch directives.
9136 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9137 __ARM_FEATURE_COPROC before changing architectures.
9139 2018-02-06 Richard Biener <rguenther@suse.de>
9141 * config/i386/i386.c (print_reg): Fix typo.
9142 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9144 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9146 * configure: Regenerate.
9148 2018-02-05 Martin Sebor <msebor@redhat.com>
9150 PR tree-optimization/83369
9151 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9154 2018-02-05 Martin Liska <mliska@suse.cz>
9156 * doc/invoke.texi: Cherry-pick upstream r323995.
9158 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9160 * ira.c (ira_init_register_move_cost): Adjust comment.
9162 2018-02-05 Martin Liska <mliska@suse.cz>
9164 PR gcov-profile/84137
9165 * doc/gcov.texi: Fix typo in documentation.
9167 2018-02-05 Martin Liska <mliska@suse.cz>
9169 PR gcov-profile/83879
9170 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9171 using dlopen functionality.
9173 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9175 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9176 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9177 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9178 _mm_maskz_range_round_ss): New intrinsics.
9179 (__builtin_ia32_rangesd128_round)
9180 (__builtin_ia32_rangess128_round): Remove.
9181 (__builtin_ia32_rangesd128_mask_round,
9182 __builtin_ia32_rangess128_mask_round): New builtins.
9183 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9184 __builtin_ia32_rangess128_round): Remove.
9185 (__builtin_ia32_rangesd128_mask_round,
9186 __builtin_ia32_rangess128_mask_round): New builtins.
9187 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9188 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9189 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9190 "<round_saeonly_constraint>")): Changed to ...
9191 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9192 "<round_saeonly_scalar_constraint>")): ... this.
9193 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9194 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9195 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9196 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9197 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9199 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9201 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9203 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9204 Remove all values except native, 8540 and 8548.
9206 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9208 * config/i386/i386.c (ix86_output_function_return): Pass
9209 INVALID_REGNUM, instead of -1, as invalid register number to
9210 indirect_thunk_name and output_indirect_thunk.
9212 2018-02-02 Julia Koval <julia.koval@intel.com>
9214 * config.gcc: Add -march=icelake.
9215 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9216 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9217 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9218 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9219 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9220 (processor_target_table): Add icelake.
9221 (ix86_option_override_internal): Handle new PTAs.
9222 (get_builtin_code_for_version): Handle icelake.
9223 (M_INTEL_COREI7_ICELAKE): New.
9224 (fold_builtin_cpu): Handle icelake.
9225 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9226 * doc/invoke.texi: Add -march=icelake.
9228 2018-02-02 Julia Koval <julia.koval@intel.com>
9230 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9231 to wide_int_bitmask.
9232 * wide-int-bitmask.h: New.
9234 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9237 * config/i386/i386.md: Replace Pmode with word_mode in
9238 builtin_setjmp_setup and builtin_longjmp to support x32.
9240 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9244 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9246 (rs6000_supported_cpu_names): New static variable.
9247 (linux_cpu_translation_table): Likewise.
9248 (elf_platform) <cpu>: Define new static variable and use it.
9249 Translate kernel AT_PLATFORM name to canonical name if needed.
9250 Error if platform name is unknown.
9252 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9255 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9257 2018-02-01 Jeff Law <law@redhat.com>
9260 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9261 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9262 the scratch if RELEASE_VIA_POP is false.
9263 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9264 If we have to save a temporary register, decrement SIZE appropriately.
9265 Pass new arguments to release_scratch_register_on_entry.
9266 (ix86_adjust_stack_and_probe): Likewise.
9267 (ix86_emit_probe_stack_range): Pass new arguments to
9268 release_scratch_register_on_entry.
9270 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9272 PR rtl-optimization/84157
9273 * combine.c (change_zero_ext): Use REG_P predicate in
9274 front of HARD_REGISTER_P predicate.
9276 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9278 * config/avr/avr.c (avr_option_override): Move disabling of
9279 -fdelete-null-pointer-checks to...
9280 * common/config/avr/avr-common.c (avr_option_optimization_table):
9283 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9285 PR tree-optimization/81635
9286 * tree-data-ref.c (split_constant_offset_1): For types that
9287 wrap on overflow, try to use range info to prove that wrapping
9290 2018-02-01 Renlin Li <renlin.li@arm.com>
9293 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9295 (aarch64_register_move_cost): Likewise.
9296 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9298 (REG_CLASS_NAMES): Likewise.
9299 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9300 TAILCALL_ADDR_REGS. Remove IP registers.
9301 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9303 2018-02-01 Richard Biener <rguenther@suse.de>
9305 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9306 for specifying RPO order and allow NULL for that.
9307 * domwalk.c (dom_walker::dom_walker): Likewise.
9308 (dom_walker::walk): Handle NULL RPO order.
9309 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9311 (rewrite_update_dom_walker): Likewise.
9312 (mark_def_dom_walker): Likewise.
9314 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9316 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9317 (aarch64_maybe_expand_sve_subreg_move): Declare.
9318 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9319 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9321 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9322 that are semantically a reverse operation.
9323 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9324 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9325 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9327 (aarch64_can_change_mode_class): For big-endian, forbid changes
9328 between two SVE modes if they have different element sizes.
9330 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9332 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9333 the TImode handling for big-endian targets.
9335 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9337 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9338 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9340 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9341 Remove BSWAP handing for big-endian targets and use the form of
9342 LD1RQ appropariate for the mode.
9344 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9346 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9347 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9350 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9353 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9354 check for operands that need to go through aarch64_sve_reload_be.
9356 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9358 PR tree-optimization/81661
9359 PR tree-optimization/84117
9360 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9361 * tree-eh.c: Include gimplify.h.
9362 (find_trapping_overflow, replace_trapping_overflow,
9363 rewrite_to_non_trapping_overflow): New functions.
9364 * tree-vect-loop.c: Include tree-eh.h.
9365 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9366 * tree-data-ref.c: Include tree-eh.h.
9367 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9369 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9371 PR rtl-optimization/84123
9372 * combine.c (change_zero_ext): Check if hard register satisfies
9373 can_change_dest_mode before calling gen_lowpart_SUBREG.
9375 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9378 * ira.c (ira_init_register_move_cost): Remove assert.
9380 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9382 PR rtl-optimization/84071
9383 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9384 * doc/tm.texi: Regenerate.
9386 2018-01-31 Richard Biener <rguenther@suse.de>
9388 PR tree-optimization/84132
9389 * tree-data-ref.c (analyze_miv_subscript): Properly
9390 check whether evolution_function_is_affine_multivariate_p
9391 before calling gcd_of_steps_may_divide_p.
9393 2018-01-31 Julia Koval <julia.koval@intel.com>
9396 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9397 * config/i386/i386.md (rdpid_rex64) New.
9398 (rdpid): Make 32bit only.
9400 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9403 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9404 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9406 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9409 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9411 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9413 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9415 PR rtl-optimization/84071
9416 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9417 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9419 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9421 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9422 (arc_attribute_table): Add 'aux' attribute.
9423 (arc_in_small_data_p): Consider aux like variables.
9424 (arc_is_aux_reg_p): New function.
9425 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9426 (arc_get_aux_arg): New function.
9427 (prepare_move_operands): Handle aux-register access.
9428 (arc_handle_aux_attribute): New function.
9429 * doc/extend.texi (ARC Variable attributes): Add subsection.
9431 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9433 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9434 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9435 (arc_attribute_table): Add 'uncached' attribute.
9436 (arc_print_operand): Print '.di' flag for uncached memory
9438 (arc_in_small_data_p): Do not consider for small data the uncached
9440 (arc_is_uncached_mem_p): New function.
9441 * config/arc/predicates.md (compact_store_memory_operand): Check
9442 for uncached memory accesses.
9443 (nonvol_nonimm_operand): Likewise.
9444 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9446 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9449 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9450 falign-loops=): Add Optimization flag.
9452 2018-01-30 Jeff Law <law@redhat.com>
9455 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9456 INT_REGISTERS_SAVED. Check it prior to calling
9457 get_scratch_register_on_entry.
9458 (ix86_adjust_stack_and_probe): Similarly.
9459 (ix86_emit_probe_stack_range): Similarly.
9460 (ix86_expand_prologue): Corresponding changes.
9462 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9465 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9466 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9468 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9471 * lra-constraints.c (curr_insn_transform): Process AND in the
9474 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9476 PR rtl-optimization/83986
9477 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9478 dependence against last_pending_memory_flush in addition to
9481 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9483 PR tree-optimization/81611
9484 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9487 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9490 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9493 2018-01-30 Richard Biener <rguenther@suse.de>
9494 Jakub Jelinek <jakub@redhat.com>
9496 PR tree-optimization/84111
9497 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9498 inner loops added during recursion, as they don't have up-to-date
9501 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9504 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9505 (can_inline_edge_by_limits_p): ... here.
9506 (can_early_inline_edge_p, check_callers,
9507 update_caller_keys, update_callee_keys, recursive_inlining,
9508 add_new_edges_to_heap, speculation_useful_p,
9509 inline_small_functions,
9510 inline_small_functions, flatten_function,
9511 inline_to_all_callers_1): Update.
9513 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9515 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9518 2018-01-30 Richard Biener <rguenther@suse.de>
9520 PR tree-optimization/83008
9521 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9522 invariant and constant vector uses in stmts when they need
9525 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9528 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9529 * configure: Regenerate.
9531 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9533 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9535 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9536 Use gen_rtx_REG rather than gen_lowpart.
9538 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9540 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9541 rather than 0 when creating partial subregs.
9543 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9545 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9548 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9551 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9552 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9553 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9554 flags. This restores the settings used before the 2017-07-24.
9555 Turning off pre increment/decrement/modify allows IVOPTS to
9556 optimize DF/SF loops where the index is an int.
9558 2018-01-29 Richard Biener <rguenther@suse.de>
9559 Kelvin Nilsen <kelvin@gcc.gnu.org>
9562 * tree-vect-stmts.c (vectorizable_call): Don't call
9563 targetm.vectorize_builtin_md_vectorized_function if callee is
9566 2018-01-22 Carl Love <cel@us.ibm.com>
9568 * doc/extend.tex: Fix typo in second arg in
9569 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9571 2018-01-29 Richard Biener <rguenther@suse.de>
9573 PR tree-optimization/84086
9574 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9575 (flush_ssaname_freelist): When SSA names were released reset
9576 the SCEV hash table.
9578 2018-01-29 Richard Biener <rguenther@suse.de>
9580 PR tree-optimization/84057
9581 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9582 removed paths when removing edges.
9584 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9586 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9587 -mfunction-return=@var{choice}.
9589 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9592 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9594 (layout::print_source_line): Likewise.
9595 (test_get_line_width_without_trailing_whitespace): Add test cases.
9597 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9600 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9603 2018-01-26 Jim Wilson <jimw@sifive.com>
9605 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9607 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9610 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9612 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9613 and CMP + SUB-immediate -> SUBS.
9615 2018-01-26 Martin Sebor <msebor@redhat.com>
9617 PR tree-optimization/83896
9618 * tree-ssa-strlen.c (get_string_len): Rename...
9619 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9620 Avoid assuming length is constant.
9621 (handle_char_store): Use HOST_WIDE_INT for string length.
9623 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9626 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9627 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9629 2018-01-26 Richard Biener <rguenther@suse.de>
9631 PR rtl-optimization/84003
9632 * dse.c (record_store): Only record redundant stores when
9633 the earlier store aliases at least all accesses the later one does.
9635 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9637 PR rtl-optimization/83985
9638 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9639 REG_CFA_RESTORE insns.
9640 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9641 REG_CFA_RESTORE insns here.
9644 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9645 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9647 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9649 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9650 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9651 (arc_init): Likewise.
9652 (arc_override_options): Likewise.
9653 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9655 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9657 * config/arc/arc.h (TARGET_DBNZ): Define.
9658 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9659 properly set the tune attribute.
9660 (dbnz): Use TARGET_DBNZ guard.
9661 * config/arc/arc.opt (mtune): Add core3 option.
9663 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9665 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9666 recognize new pic like addresses.
9667 (arc_delegitimize_address): Clean up.
9669 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9671 * config/arc/arc-arches.def: Option mrf16 valid for all
9673 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9674 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9675 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9676 * config/arc/arc-tables.opt: Regenerate.
9677 * config/arc/arc.c (arc_conditional_register_usage): Handle
9678 reduced register file case.
9679 (arc_file_start): Set must have build attributes.
9680 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9682 * config/arc/arc.opt (mrf16): Add new option.
9683 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9684 * config/arc/genmultilib.awk: Handle new mrf16 option.
9685 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9686 * config/arc/t-multilib: Regenerate.
9687 * doc/invoke.texi (ARC Options): Document mrf16 option.
9689 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9691 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9692 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9693 (arc_attribute_table): Add 'secure_call' attribute.
9694 (arc_print_operand): Print secure call operand.
9695 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9697 (arc_is_secure_call_p): New function. * config/arc/arc.md
9698 (call_i): Add support for sjli instruction.
9699 (call_value_i): Likewise.
9700 * config/arc/constraints.md (Csc): New constraint.
9702 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9703 John Eric Martin <John.Martin@emmicro-us.com>
9705 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9706 * config/arc/arc.c (_arc_jli_section): New struct.
9707 (arc_jli_section): New type.
9708 (rc_jli_sections): New static variable.
9709 (arc_handle_jli_attribute): New function.
9710 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9711 (arc_file_end): New function.
9712 (TARGET_ASM_FILE_END): Define.
9713 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9714 (arc_add_jli_section): New function.
9715 (jli_call_scan): Likewise.
9716 (arc_reorg): Call jli_call_scan.
9717 (arc_output_addsi): Remove 'S' from printing asm operand.
9718 (arc_is_jli_call_p): New function.
9719 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9721 (movhi_insn): Likewise.
9722 (movsi_insn): Likewise.
9723 (movsi_set_cc_insn): Likewise.
9724 (loadqi_update): Likewise.
9725 (load_zeroextendqisi_update): Likewise.
9726 (load_signextendqisi_update): Likewise.
9727 (loadhi_update): Likewise.
9728 (load_zeroextendhisi_update): Likewise.
9729 (load_signextendhisi_update): Likewise.
9730 (loadsi_update): Likewise.
9731 (loadsf_update): Likewise.
9732 (movsicc_insn): Likewise.
9733 (bset_insn): Likewise.
9734 (bxor_insn): Likewise.
9735 (bclr_insn): Likewise.
9736 (bmsk_insn): Likewise.
9737 (bicsi3_insn): Likewise.
9738 (cmpsi_cc_c_insn): Likewise.
9739 (movsi_ne): Likewise.
9740 (movsi_cond_exec): Likewise.
9741 (clrsbsi2): Likewise.
9752 (call_i): Remove 'S' asm letter, add jli instruction.
9753 (call_value_i): Likewise.
9754 * config/arc/arc.op (mjli-always): New option.
9755 * config/arc/constraints.md (Cji): New constraint.
9756 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9758 (subsf3_fpx): Likewise.
9759 (mulsf3_fpx): Likewise.
9760 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9762 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9764 * doc/invoke.texi (ARC): Document mjli-always option.
9766 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9768 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9769 avoid addition with 0 and use incw and decw where possible.
9771 2018-01-26 Richard Biener <rguenther@suse.de>
9773 PR tree-optimization/81082
9774 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9775 association if it requires casting to unsigned.
9776 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9777 from fold_plusminus_mult_expr to catch important cases late when
9778 range info is available.
9780 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9782 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9783 * configure.ac (hidden_linkonce): New test.
9784 * configure: Regenerate.
9785 * config.in: Regenerate.
9787 2018-01-26 Julia Koval <julia.koval@intel.com>
9789 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9790 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9791 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9792 _mm_mask_bitshuffle_epi64_mask): Fix type.
9793 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9794 USI_FTYPE_V4DI_V4DI_USI): Remove.
9795 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9796 __builtin_ia32_vpshufbitqmb256_mask,
9797 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9798 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9799 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9801 2018-01-26 Alan Modra <amodra@gmail.com>
9804 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9805 UNSPEC_VBPERMQ. Sort other unspecs.
9807 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9809 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9811 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9814 * predict.c (drop_profile): Do not push/pop cfun; update also
9816 (handle_missing_profiles): Fix logic looking for zero profiles.
9818 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9821 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9822 on functions with #pragma omp declare simd or functions with simd
9824 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9825 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9826 Remove trailing \n from warning_at calls.
9828 2018-01-25 Tom de Vries <tom@codesourcery.com>
9831 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9832 for neutered workers.
9834 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9837 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9838 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9840 2018-01-24 Jeff Law <law@redhat.com>
9843 * i386.c (get_probe_interval): Move to earlier point.
9844 (ix86_compute_frame_layout): If -fstack-clash-protection and
9845 the frame is larger than the probe interval, then use pushes
9846 to save registers rather than reg->mem moves.
9847 (ix86_expand_prologue): Remove conditional for int_registers_saved
9850 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9853 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9854 min/max for never referenced object.
9856 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9859 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9861 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9862 attributes from DECL_ATTRIBUTES (decl) without affecting
9863 DECL_ATTRIBUTES (current_function_decl).
9864 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9865 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9867 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9869 PR tree-optimization/83979
9870 * fold-const.c (fold_comparison): Use constant_boolean_node
9871 instead of boolean_{true,false}_node.
9873 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9875 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9878 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9880 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9881 Simplify the clause that sets the length attribute.
9882 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9883 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9884 clause that sets the length attribute.
9885 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9887 2018-01-24 Tom de Vries <tom@codesourcery.com>
9890 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9891 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9892 Add strict parameter.
9893 (prevent_branch_around_nothing): Insert dummy insn between branch to
9894 label and label with no ptx insn inbetween.
9895 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9897 2018-01-24 Tom de Vries <tom@codesourcery.com>
9900 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9901 for neutered threads in warp.
9902 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9904 2018-01-24 Richard Biener <rguenther@suse.de>
9906 PR tree-optimization/83176
9907 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9910 2018-01-24 Richard Biener <rguenther@suse.de>
9912 PR tree-optimization/82819
9913 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9914 code generating pluses that are no-ops in the target precision.
9916 2018-01-24 Richard Biener <rguenther@suse.de>
9919 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9921 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9923 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9924 to merge probabilities.
9925 * predict.c (probably_never_executed): Also mark as cold functions
9926 with global 0 profile and guessed local profile.
9927 * profile-count.c (profile_probability::combine_with_count): New
9929 * profile-count.h (profile_probability::operator*,
9930 profile_probability::operator*=, profile_probability::operator/,
9931 profile_probability::operator/=): Reduce precision to adjusted
9932 and set value to guessed on contradictory divisions.
9933 (profile_probability::combine_with_freq): Remove.
9934 (profile_probability::combine_wiht_count): Declare.
9935 (profile_count::force_nonzero):: Set to adjusted.
9936 (profile_count::probability_in):: Set quality to adjusted.
9937 * tree-ssa-tail-merge.c (replace_block_by): Use
9940 2018-01-23 Andrew Waterman <andrew@sifive.com>
9941 Jim Wilson <jimw@sifive.com>
9943 * config/riscv/riscv.c (riscv_stack_boundary): New.
9944 (riscv_option_override): Set riscv_stack_boundary. Handle
9945 riscv_preferred_stack_boundary_arg.
9946 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9947 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9948 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9949 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9950 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9951 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9953 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9956 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9957 of struct ix86_frame.
9958 (ix86_expand_epilogue): Likewise. Add a local variable for
9959 the reg_save_offset field in struct ix86_frame.
9961 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9963 PR tree-optimization/82604
9964 * tree-loop-distribution.c (enum partition_kind): New enum item
9965 PKIND_PARTIAL_MEMSET.
9966 (partition_builtin_p): Support above new enum item.
9967 (generate_code_for_partition): Ditto.
9968 (compute_access_range): Differentiate cases that equality can be
9969 proven at all loops, the innermost loops or no loops.
9970 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9971 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9972 (finalize_partitions, distribute_loop): Don't fuse partition of
9973 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9974 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9977 2018-01-23 Martin Liska <mliska@suse.cz>
9979 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9980 order to ignore the predictor.
9981 (PRED_POLYMORPHIC_CALL): Likewise.
9982 (PRED_RECURSIVE_CALL): Likewise.
9984 2018-01-23 Martin Liska <mliska@suse.cz>
9986 * tree-profile.c (tree_profiling): Print function header to
9987 aware reader which function we are working on.
9988 * value-prof.c (gimple_find_values_to_profile): Do not print
9989 not interesting value histograms.
9991 2018-01-23 Martin Liska <mliska@suse.cz>
9993 * profile-count.h (enum profile_quality): Add
9994 profile_uninitialized as the first value. Do not number values
9995 as they are zero based.
9996 (profile_count::verify): Update sanity check.
9997 (profile_probability::verify): Likewise.
9999 2018-01-23 Nathan Sidwell <nathan@acm.org>
10001 * doc/invoke.texi (ffor-scope): Deprecate.
10003 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10005 PR tree-optimization/83510
10006 * domwalk.c (set_all_edges_as_executable): New function.
10007 (dom_walker::dom_walker): Convert bool param
10008 "skip_unreachable_blocks" to enum reachability. Move setup of
10009 edge flags to set_all_edges_as_executable and only do it when
10010 reachability is REACHABLE_BLOCKS.
10011 * domwalk.h (enum dom_walker::reachability): New enum.
10012 (dom_walker::dom_walker): Convert bool param
10013 "skip_unreachable_blocks" to enum reachability.
10014 (set_all_edges_as_executable): New decl.
10015 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
10016 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
10018 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
10019 but converting true to REACHABLE_BLOCKS.
10020 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
10022 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
10023 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
10024 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
10026 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
10027 if check_all_array_refs will be called.
10029 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10031 * tree.c (selftest::test_location_wrappers): Add more test
10034 2018-01-23 David Malcolm <dmalcolm@redhat.com>
10036 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
10037 (selftest::test_bit_in_range): Likewise.
10039 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10042 * doc/sourcebuild.texi (vect_float): Say that the selector
10043 only describes the situation when -funsafe-math-optimizations is on.
10044 (vect_float_strict): Document.
10046 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
10048 PR tree-optimization/83965
10049 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10050 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10051 instead of checking only for a reduction.
10052 (vect_recog_widen_sum_pattern): Likewise.
10054 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10056 * predict.c (probably_never_executed): Only use precise profile info.
10057 (compute_function_frequency): Skip after inlining hack since we now
10058 have quality checking.
10060 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10062 * profile-count.h (profile_probability::very_unlikely,
10063 profile_probability::unlikely, profile_probability::even): Set
10064 precision to guessed.
10066 2018-01-23 Richard Biener <rguenther@suse.de>
10068 PR tree-optimization/83963
10069 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10070 Properly terminate dominator walk when crossing the exit edge not
10071 when visiting its source block.
10073 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10076 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10077 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10079 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10081 PR tree-optimization/83957
10082 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10083 semicolon after for body surrounded by braces.
10085 PR tree-optimization/83081
10086 * profile-count.h (profile_probability::split): New method.
10087 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10088 Use profile_probability::split.
10089 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10090 when splitting a single conditional jump into 2.
10092 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10094 PR tree-optimization/69452
10095 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10098 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10100 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10101 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10102 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10104 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10106 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10107 * config/rl78/rl78.md (movdi): New define_expand.
10108 * config/rl78/rl78.c (rl78_split_movdi): New function.
10110 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10113 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10115 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10116 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10117 128-bit to produce an UNSPEC move to get the double word with the
10118 signbit and then a shift directly to do signbit.
10119 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10120 implementation with a new version that just does either a direct
10121 move or a regular move. Move memory interface to separate insns.
10122 Move insns so they are next to the expander.
10123 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10124 with signbit move. Split big and little endian case.
10125 (signbit<mode>2_dm_mem_le): Likewise.
10126 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10127 (signbit<mode>2_dm2): Likewise.
10129 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10131 * config/rl78/rl78.md (anddi3): New define_expand.
10133 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10135 * config/rl78/rl78.md (umindi3): New define_expand.
10137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10139 * config/rl78/rl78.md (smindi3): New define_expand.
10141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10143 * config/rl78/rl78.md (smaxdi3): New define_expand.
10145 2018-01-22 Carl Love <cel@us.ibm.com>
10147 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10148 LVX_V1TI): Add macro expansion.
10149 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10150 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10151 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10152 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10153 Change check to determine if the instruction is a byte reversing
10154 entry. Fix typo in comment.
10155 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10156 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10157 Add def_builtin calls for new builtins.
10158 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10159 Add define_insn expansion.
10161 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10163 * config/rl78/rl78.md (umaxdi3): New define_expand.
10165 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10167 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10168 for non-QImode registers.
10170 2018-01-22 Richard Biener <rguenther@suse.de>
10172 PR tree-optimization/83963
10173 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10174 including the loop exit block.
10175 (scop_detection::merge_sese): Likewise.
10176 (scop_detection::add_scop): Do it here instead.
10178 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10180 * doc/sourcebuild.texi (arm_softfloat): Document.
10182 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10185 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10186 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10187 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10189 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10190 David Edelsohn <dje.gcc@gmail.com>
10193 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10194 Change "crset eq" to "crset 2".
10195 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10196 (*call_indirect_aix<mode>_nospec): Likewise.
10197 (*call_value_indirect_aix<mode>_nospec): Likewise.
10198 (*call_indirect_elfv2<mode>_nospec): Likewise.
10199 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10200 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10201 change assembly output from . to $.
10202 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10203 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10204 (*tablejump<mode>_internal1_nospec): Likewise.
10206 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10209 * config/sh/sh_optimize_sett_clrt.cc:
10210 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10212 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10214 PR tree-optimization/83940
10215 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10216 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10217 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10218 use it to pass back the definition type.
10219 (vect_check_store_rhs): Likewise rhs_dt_out.
10220 (vect_build_gather_load_calls): Add a mask_dt argument and use
10221 it instead of a call to vect_is_simple_use.
10222 (vectorizable_store): Update calls to vect_check_load_store_mask
10223 and vect_check_store_rhs. Use the dt returned by the latter instead
10224 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10225 instead of calls to vect_is_simple_use. Pass the scalar rather
10226 than the vector operand to vect_is_simple_use when handling
10227 second and subsequent copies of an rhs value.
10228 (vectorizable_load): Update calls to vect_check_load_store_mask
10229 and vect_build_gather_load_calls. Use the cached mask_dt and
10230 gs_info.offset_dt instead of calls to vect_is_simple_use.
10232 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10234 PR middle-end/83945
10235 * tree-emutls.c: Include gimplify.h.
10236 (lower_emutls_2): New function.
10237 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10238 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10239 it before further processing.
10242 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10243 UINTVAL (trueop1) instead of INTVAL (op1).
10245 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10249 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10250 INCOMING_FRAME_SP_OFFSET if not defined.
10251 (scan_trace): Add ENTRY argument. If true and
10252 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10253 emit a note to adjust the CFA offset.
10254 (create_cfi_notes): Adjust scan_trace callers.
10255 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10256 INCOMING_FRAME_SP_OFFSET in the CIE.
10257 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10258 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10260 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10261 * doc/tm.texi: Regenerated.
10263 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10265 PR rtl-optimization/83147
10266 * lra-constraints.c (remove_inheritance_pseudos): Use
10267 lra_substitute_pseudo_within_insn.
10269 2018-01-19 Tom de Vries <tom@codesourcery.com>
10270 Cesar Philippidis <cesar@codesourcery.com>
10273 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10275 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10278 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10279 spaces for function labels.
10281 2018-01-19 Martin Liska <mliska@suse.cz>
10283 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10284 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10285 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10286 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10287 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10288 (PRED_CONST_RETURN): Change from 69 to 65.
10289 (PRED_NULL_RETURN): Change from 91 to 71.
10290 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10291 (PRED_LOOP_GUARD): Change from 66 to 73.
10293 2018-01-19 Martin Liska <mliska@suse.cz>
10295 * predict.c (predict_insn_def): Add new assert.
10296 (struct branch_predictor): Change type to signed integer.
10297 (test_prediction_value_range): Amend test to cover
10298 PROB_UNINITIALIZED.
10299 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10300 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10301 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10302 (PRED_LOOP_IV_COMPARE): Likewise.
10303 * predict.h (PROB_UNINITIALIZED): Define new constant.
10305 2018-01-19 Martin Liska <mliska@suse.cz>
10307 * predict.c (dump_prediction): Add new format for
10308 analyze_brprob.py script which is enabled with -details
10310 * profile-count.h (precise_p): New function.
10312 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10314 PR tree-optimization/83922
10315 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10316 there are no statements that need masking.
10317 (vect_active_double_reduction_p): New function.
10318 (vect_analyze_loop_operations): Use it when handling phis that
10319 are not in the loop header.
10321 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10323 PR tree-optimization/83914
10324 * tree-vect-loop.c (vectorizable_induction): Don't convert
10325 init_expr or apply the peeling adjustment for inductions
10326 that are nested within the vectorized loop.
10328 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10330 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10333 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10337 * function.h (gimplify_parameters): Add gimple_seq * argument.
10338 * function.c: Include gimple.h and options.h.
10339 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10340 for the added local temporaries if needed.
10341 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10342 if there are any parameter cleanups, wrap whole body into a
10343 try/finally with the cleanups.
10345 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10348 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10349 Use GET_MODE_CLASS for scalar floating point.
10351 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10355 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10356 Fix call of call_cgraph_insertion_hooks.
10358 2018-01-18 Martin Sebor <msebor@redhat.com>
10360 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10362 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10365 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10368 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10371 * common.opt: (-ffile-prefix-map): New option.
10372 * opts.c (common_handle_option): Defer it.
10373 * opts-global.c (handle_common_deferred_options): Handle it.
10374 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10375 * file-prefix-map.h: New file.
10376 (remap_debug_filename, add_debug_prefix_map): ...here.
10377 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10378 * final.c (debug_prefix_map, add_debug_prefix_map
10379 remap_debug_filename): Move to...
10380 * file-prefix-map.c: New file.
10381 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10382 generalize, get rid of alloca(), use strrchr() instead of strchr().
10383 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10384 Implement in terms of add_prefix_map().
10385 (remap_macro_filename, remap_debug_filename): Implement in term of
10387 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10388 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10389 * dbxout.c: Include file-prefix-map.h.
10390 * varasm.c: Likewise.
10391 * vmsdbgout.c: Likewise.
10392 * xcoffout.c: Likewise.
10393 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10394 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10395 * doc/invoke.texi (-ffile-prefix-map): Document.
10396 (-fdebug-prefix-map): Update description.
10398 2018-01-18 Martin Liska <mliska@suse.cz>
10400 * config/i386/i386.c (indirect_thunk_name): Document that also
10402 (output_indirect_thunk): Document why both instructions
10403 (pause and lfence) are generated.
10405 2018-01-18 Richard Biener <rguenther@suse.de>
10407 PR tree-optimization/83887
10408 * graphite-scop-detection.c
10409 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10410 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10411 (scop_detection::merge_sese): Re-implement with a flood-fill
10412 algorithm that properly finds a SESE region if it exists.
10414 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10417 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10418 pointer_diff optimizations use view_convert instead of convert.
10420 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10422 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10423 Generate different code for -mno-speculate-indirect-jumps.
10424 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10425 (*call_indirect_aix<mode>): Disable for
10426 -mno-speculate-indirect-jumps.
10427 (*call_indirect_aix<mode>_nospec): New define_insn.
10428 (*call_value_indirect_aix<mode>): Disable for
10429 -mno-speculate-indirect-jumps.
10430 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10431 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10432 -mno-speculate-indirect-jumps.
10433 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10435 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10437 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10438 long double type, set the flags for noting the default long double
10439 type, even if we don't pass or return a long double type.
10441 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10444 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10447 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10449 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10450 support for merge[hl].
10451 (fold_mergehl_helper): New helper function.
10452 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10453 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10454 (altivec_vmrglw_direct): Add xxmrglw insn.
10456 2018-01-17 Andrew Waterman <andrew@sifive.com>
10458 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10459 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10461 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10464 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10465 call the lto_location_cache before reading the
10466 DECL_SOURCE_LOCATION of the types.
10468 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10469 Richard Sandiford <richard.sandiford@linaro.org>
10471 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10472 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10473 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10474 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10475 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10477 * config/aarch64/constraints.md (aarch64_movti_operand):
10479 * config/aarch64/predicates.md (Uti): Add new constraint.
10481 2018-01-17 Carl Love <cel@us.ibm.com>
10483 * config/rs6000/vsx.md (define_expand xl_len_r,
10484 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10485 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10487 (define_expand, define_insn): Move the shift left from the
10488 define_insn to the define_expand for lxvl and stxvl instructions.
10489 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10490 and XL_LEN_R definitions to PURE.
10492 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10494 * config/i386/i386.c (indirect_thunk_name): Declare regno
10495 as unsigned int. Compare regno with INVALID_REGNUM.
10496 (output_indirect_thunk): Ditto.
10497 (output_indirect_thunk_function): Ditto.
10498 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10499 in the call to output_indirect_thunk_function.
10501 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10503 PR middle-end/83884
10504 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10505 rather than the size of inner_type to determine the stack slot size
10506 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10508 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10511 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10514 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10516 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10517 endian Linux systems to optionally enable multilibs for selecting
10518 the long double type if the user configured an explicit type.
10519 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10520 have no long double multilibs if not defined.
10521 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10522 warn if the user used -mabi={ieee,ibm}longdouble and we built
10523 multilibs for long double.
10524 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10525 appropriate multilib option.
10526 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10528 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10529 for building long double multilibs.
10530 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10532 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10534 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10537 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10539 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10542 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10545 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10548 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10550 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10551 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10553 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10555 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10556 different rtl trees depending on TARGET_64BIT.
10557 (rs6000_gen_lvx): Likewise.
10559 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10561 * config/visium/visium.md (nop): Tweak comment.
10562 (hazard_nop): Likewise.
10564 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10566 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10567 -mspeculate-indirect-jumps.
10568 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10569 for -mno-speculate-indirect-jumps.
10570 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10571 (*call_value_indirect_elfv2<mode>): Disable for
10572 -mno-speculate-indirect-jumps.
10573 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10574 (indirect_jump): Emit different RTL for
10575 -mno-speculate-indirect-jumps.
10576 (*indirect_jump<mode>): Disable for
10577 -mno-speculate-indirect-jumps.
10578 (*indirect_jump<mode>_nospec): New define_insn.
10579 (tablejump): Emit different RTL for
10580 -mno-speculate-indirect-jumps.
10581 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10582 (tablejumpsi_nospec): New define_expand.
10583 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10584 (tablejumpdi_nospec): New define_expand.
10585 (*tablejump<mode>_internal1): Disable for
10586 -mno-speculate-indirect-jumps.
10587 (*tablejump<mode>_internal1_nospec): New define_insn.
10588 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10591 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10593 * caller-save.c (insert_save): Drop unnecessary parameter. All
10596 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10597 Richard Biener <rguenth@suse.de>
10600 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10601 return early, inline manually is_gimple_sizepos. Make sure if we
10602 call gimplify_expr we don't end up with a gimple constant.
10603 * tree.c (variably_modified_type_p): Don't return true for
10604 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10605 * gimplify.h (is_gimple_sizepos): Remove.
10607 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10609 PR tree-optimization/83857
10610 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10611 vectorizable_live_operation for pure SLP statements.
10612 (vectorizable_live_operation): Handle PHIs.
10614 2018-01-16 Richard Biener <rguenther@suse.de>
10616 PR tree-optimization/83867
10617 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10618 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10620 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10623 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10624 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10625 If off is not INTEGER_CST, issue a may not be aligned warning
10626 rather than isn't aligned. Use isn%'t rather than isn't.
10627 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10629 <case MULT_EXPR>: Improve the case when bottom and one of the
10630 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10631 operand, in that case check if the other operand is multiple of
10632 bottom divided by the INTEGER_CST operand.
10634 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10637 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10638 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10639 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10640 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10641 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10642 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10643 (pa_function_arg_size): New function.
10645 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10647 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10648 in a separate statement.
10650 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10652 PR tree-optimization/83847
10653 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10654 group gathers and scatters.
10656 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10658 PR rtl-optimization/86620
10659 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10661 PR rtl-optimization/83213
10662 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10663 to last if both are JUMP_INSNs.
10665 PR tree-optimization/83843
10666 * gimple-ssa-store-merging.c
10667 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10668 store_immediate_info for bswap/nop orig_stores.
10670 2018-01-15 Andrew Waterman <andrew@sifive.com>
10672 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10674 <UDIV>: Increase cost if !TARGET_DIV.
10676 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10678 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10679 (define_attr "cr_logical_3op"): New.
10680 (cceq_ior_compare): Adjust.
10681 (cceq_ior_compare_complement): Adjust.
10682 (*cceq_rev_compare): Adjust.
10683 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10684 (is_cracked_insn): Adjust.
10685 (insn_must_be_first_in_group): Adjust.
10686 * config/rs6000/40x.md: Adjust.
10687 * config/rs6000/440.md: Adjust.
10688 * config/rs6000/476.md: Adjust.
10689 * config/rs6000/601.md: Adjust.
10690 * config/rs6000/603.md: Adjust.
10691 * config/rs6000/6xx.md: Adjust.
10692 * config/rs6000/7450.md: Adjust.
10693 * config/rs6000/7xx.md: Adjust.
10694 * config/rs6000/8540.md: Adjust.
10695 * config/rs6000/cell.md: Adjust.
10696 * config/rs6000/e300c2c3.md: Adjust.
10697 * config/rs6000/e500mc.md: Adjust.
10698 * config/rs6000/e500mc64.md: Adjust.
10699 * config/rs6000/e5500.md: Adjust.
10700 * config/rs6000/e6500.md: Adjust.
10701 * config/rs6000/mpc.md: Adjust.
10702 * config/rs6000/power4.md: Adjust.
10703 * config/rs6000/power5.md: Adjust.
10704 * config/rs6000/power6.md: Adjust.
10705 * config/rs6000/power7.md: Adjust.
10706 * config/rs6000/power8.md: Adjust.
10707 * config/rs6000/power9.md: Adjust.
10708 * config/rs6000/rs64.md: Adjust.
10709 * config/rs6000/titan.md: Adjust.
10711 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10713 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10714 ix86_indirect_branch_register logic.
10716 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10718 * config/i386/constraints.md (Bs): Update
10719 ix86_indirect_branch_register check. Don't check
10720 ix86_indirect_branch_register with GOT_memory_operand.
10722 * config/i386/predicates.md (GOT_memory_operand): Don't check
10723 ix86_indirect_branch_register here.
10724 (GOT32_symbol_operand): Likewise.
10726 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10728 * config/i386/predicates.md (constant_call_address_operand):
10729 Rewrite ix86_indirect_branch_register logic.
10730 (sibcall_insn_operand): Likewise.
10732 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10734 * config/i386/constraints.md (Bs): Replace
10735 ix86_indirect_branch_thunk_register with
10736 ix86_indirect_branch_register.
10738 * config/i386/i386.md (indirect_jump): Likewise.
10739 (tablejump): Likewise.
10740 (*sibcall_memory): Likewise.
10741 (*sibcall_value_memory): Likewise.
10742 Peepholes of indirect call and jump via memory: Likewise.
10743 * config/i386/i386.opt: Likewise.
10744 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10745 (GOT_memory_operand): Likewise.
10746 (call_insn_operand): Likewise.
10747 (sibcall_insn_operand): Likewise.
10748 (GOT32_symbol_operand): Likewise.
10750 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10752 PR middle-end/83837
10753 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10754 type rather than type addr's type points to.
10755 (expand_omp_atomic_mutex): Likewise.
10756 (expand_omp_atomic): Likewise.
10758 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10761 * config/i386/i386.c (output_indirect_thunk_function): Use
10762 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10763 for __x86_return_thunk.
10765 2018-01-15 Richard Biener <rguenther@suse.de>
10767 PR middle-end/83850
10768 * expmed.c (extract_bit_field_1): Fix typo.
10770 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10773 * config/arm/iterators.md (VF): New mode iterator.
10774 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10775 Remove integer-related logic from pattern.
10776 (neon_vabd<mode>_3): Likewise.
10778 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10780 PR middle-end/82694
10781 * common.opt (fstrict-overflow): No longer an alias.
10782 (fwrapv-pointer): New option.
10783 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10784 also for pointer types based on flag_wrapv_pointer.
10785 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10786 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10787 opts->x_flag_wrapv got set.
10788 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10789 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10790 POINTER_TYPE_OVERFLOW_UNDEFINED.
10791 * match.pd: Likewise in address comparison pattern.
10792 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10794 2018-01-15 Richard Biener <rguenther@suse.de>
10797 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10798 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10799 Reset type names to their identifier if their TYPE_DECL doesn't
10800 have linkage (and thus is used for ODR and devirt).
10801 (save_debug_info_for_decl): Remove.
10802 (save_debug_info_for_type): Likewise.
10803 (add_tree_to_fld_list): Adjust.
10804 * tree-pretty-print.c (dump_generic_node): Make dumping of
10805 type names more robust.
10807 2018-01-15 Richard Biener <rguenther@suse.de>
10809 * BASE-VER: Bump to 8.0.1.
10811 2018-01-14 Martin Sebor <msebor@redhat.com>
10814 * builtins.c (check_access): Avoid warning when the no-warning bit
10817 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10819 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10820 * ira-color (allocno_hard_regs_compare): Likewise.
10822 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10825 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10826 Use .pushsection/.popsection.
10828 2018-01-14 Martin Sebor <msebor@redhat.com>
10831 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10833 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10835 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10836 entry from extra_headers.
10837 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10838 extra_headers, make the list bitwise identical to the i?86-*-* one.
10840 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10842 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10843 -mcmodel=large with -mindirect-branch=thunk,
10844 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10845 -mfunction-return=thunk-extern.
10846 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10847 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10848 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10850 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10852 * config/i386/i386.c (print_reg): Print the name of the full
10853 integer register without '%'.
10854 (ix86_print_operand): Handle 'V'.
10855 * doc/extend.texi: Document 'V' modifier.
10857 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10859 * config/i386/constraints.md (Bs): Disallow memory operand for
10860 -mindirect-branch-register.
10862 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10863 (GOT_memory_operand): Likewise.
10864 (call_insn_operand): Likewise.
10865 (sibcall_insn_operand): Likewise.
10866 (GOT32_symbol_operand): Likewise.
10867 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10868 for -mindirect-branch-register.
10869 (tablejump): Likewise.
10870 (*sibcall_memory): Likewise.
10871 (*sibcall_value_memory): Likewise.
10872 Disallow peepholes of indirect call and jump via memory for
10873 -mindirect-branch-register.
10874 (*call_pop): Replace m with Bw.
10875 (*call_value_pop): Likewise.
10876 (*sibcall_pop_memory): Replace m with Bs.
10877 * config/i386/i386.opt (mindirect-branch-register): New option.
10878 * doc/invoke.texi: Document -mindirect-branch-register option.
10880 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10882 * config/i386/i386-protos.h (ix86_output_function_return): New.
10883 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10884 set function_return_type.
10885 (indirect_thunk_name): Add ret_p to indicate thunk for function
10887 (output_indirect_thunk_function): Pass false to
10888 indirect_thunk_name.
10889 (ix86_output_indirect_branch_via_reg): Likewise.
10890 (ix86_output_indirect_branch_via_push): Likewise.
10891 (output_indirect_thunk_function): Create alias for function
10892 return thunk if regno < 0.
10893 (ix86_output_function_return): New function.
10894 (ix86_handle_fndecl_attribute): Handle function_return.
10895 (ix86_attribute_table): Add function_return.
10896 * config/i386/i386.h (machine_function): Add
10897 function_return_type.
10898 * config/i386/i386.md (simple_return_internal): Use
10899 ix86_output_function_return.
10900 (simple_return_internal_long): Likewise.
10901 * config/i386/i386.opt (mfunction-return=): New option.
10902 (indirect_branch): Mention -mfunction-return=.
10903 * doc/extend.texi: Document function_return function attribute.
10904 * doc/invoke.texi: Document -mfunction-return= option.
10906 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10908 * config/i386/i386-opts.h (indirect_branch): New.
10909 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10910 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10911 with local indirect jump when converting indirect call and jump.
10912 (ix86_set_indirect_branch_type): New.
10913 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10914 (indirectlabelno): New.
10915 (indirect_thunk_needed): Likewise.
10916 (indirect_thunk_bnd_needed): Likewise.
10917 (indirect_thunks_used): Likewise.
10918 (indirect_thunks_bnd_used): Likewise.
10919 (INDIRECT_LABEL): Likewise.
10920 (indirect_thunk_name): Likewise.
10921 (output_indirect_thunk): Likewise.
10922 (output_indirect_thunk_function): Likewise.
10923 (ix86_output_indirect_branch_via_reg): Likewise.
10924 (ix86_output_indirect_branch_via_push): Likewise.
10925 (ix86_output_indirect_branch): Likewise.
10926 (ix86_output_indirect_jmp): Likewise.
10927 (ix86_code_end): Call output_indirect_thunk_function if needed.
10928 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10930 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10931 (ix86_attribute_table): Add indirect_branch.
10932 * config/i386/i386.h (machine_function): Add indirect_branch_type
10933 and has_local_indirect_jump.
10934 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10936 (tablejump): Likewise.
10937 (*indirect_jump): Use ix86_output_indirect_jmp.
10938 (*tablejump_1): Likewise.
10939 (simple_return_indirect_internal): Likewise.
10940 * config/i386/i386.opt (mindirect-branch=): New option.
10941 (indirect_branch): New.
10944 (thunk-inline): Likewise.
10945 (thunk-extern): Likewise.
10946 * doc/extend.texi: Document indirect_branch function attribute.
10947 * doc/invoke.texi: Document -mindirect-branch= option.
10949 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10952 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10954 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10956 * ipa-inline.c (want_inline_small_function_p): Return false if
10957 inlining has already failed with CIF_FINAL_ERROR.
10958 (update_caller_keys): Call want_inline_small_function_p before
10960 (update_callee_keys): Likewise.
10962 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10964 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10966 (rs6000_quadword_masked_address_p): Likewise.
10967 (quad_aligned_load_p): Likewise.
10968 (quad_aligned_store_p): Likewise.
10969 (const_load_sequence_p): Add comment to describe the outer-most loop.
10970 (mimic_memory_attributes_and_flags): New function.
10971 (rs6000_gen_stvx): Likewise.
10972 (replace_swapped_aligned_store): Likewise.
10973 (rs6000_gen_lvx): Likewise.
10974 (replace_swapped_aligned_load): Likewise.
10975 (replace_swapped_load_constant): Capitalize argument name in
10976 comment describing this function.
10977 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10978 and stores that access quad-word aligned addresses and replace
10979 with stvx or lvx instructions when appropriate.
10980 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10981 New function prototype.
10982 (rs6000_quadword_masked_address_p): Likewise.
10983 (rs6000_gen_lvx): Likewise.
10984 (rs6000_gen_stvx): Likewise.
10985 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10986 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10987 when memory address is aligned.
10988 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10989 this split to select lvx instruction when memory address is aligned.
10990 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10991 instruction when memory address is aligned.
10992 (*vsx_le_perm_load_v16qi): Likewise.
10993 (four unnamed splitters): Modify to select the stvx instruction
10994 when memory is aligned.
10996 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10998 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10999 which appears in the queue multiple times.
11001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11002 Alan Hayward <alan.hayward@arm.com>
11003 David Sherwood <david.sherwood@arm.com>
11005 * tree-vectorizer.h (vec_lower_bound): New structure.
11006 (_loop_vec_info): Add check_nonzero and lower_bounds.
11007 (LOOP_VINFO_CHECK_NONZERO): New macro.
11008 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
11009 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
11010 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
11011 fields. Make seg_len the distance travelled, not including the
11013 (dr_direction_indicator): Declare.
11014 (dr_zero_step_indicator): Likewise.
11015 (dr_known_forward_stride_p): Likewise.
11016 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
11018 (runtime_alias_check_p): Allow runtime alias checks with
11020 (operator ==): Compare access_size and align.
11021 (prune_runtime_alias_test_list): Rework for new distinction between
11022 the access_size and seg_len.
11023 (create_intersect_range_checks_index): Likewise. Cope with polynomial
11025 (get_segment_min_max): New function.
11026 (create_intersect_range_checks): Use it.
11027 (dr_step_indicator): New function.
11028 (dr_direction_indicator): Likewise.
11029 (dr_zero_step_indicator): Likewise.
11030 (dr_known_forward_stride_p): Likewise.
11031 * tree-loop-distribution.c (data_ref_segment_size): Return
11032 DR_STEP * (niters - 1).
11033 (compute_alias_check_pairs): Update call to the dr_with_seg_len
11035 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
11036 (vect_preserves_scalar_order_p): New function, split out from...
11037 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
11038 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
11039 (vect_vfa_access_size): New function.
11040 (vect_vfa_align): Likewise.
11041 (vect_compile_time_alias): Take access_size_a and access_b arguments.
11042 (dump_lower_bound): New function.
11043 (vect_check_lower_bound): Likewise.
11044 (vect_small_gap_p): Likewise.
11045 (vectorizable_with_step_bound_p): Likewise.
11046 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
11047 depencies if the vectorization factor is 1. Convert the checks
11048 for nonzero steps into checks on the bounds of DR_STEP. Try using
11049 a bunds check for variable steps if the minimum required step is
11050 relatively small. Update calls to the dr_with_seg_len
11051 constructor and to vect_compile_time_alias.
11052 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11054 (vect_loop_versioning): Call it.
11055 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11057 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11060 Alan Hayward <alan.hayward@arm.com>
11061 David Sherwood <david.sherwood@arm.com>
11063 * doc/sourcebuild.texi (vect_scatter_store): Document.
11064 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11066 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11068 * genopinit.c (main): Add supports_vec_scatter_store and
11069 supports_vec_scatter_store_cached to target_optabs.
11070 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11071 IFN_MASK_SCATTER_STORE.
11072 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11074 * internal-fn.h (internal_store_fn_p): Declare.
11075 (internal_fn_stored_value_index): Likewise.
11076 * internal-fn.c (scatter_store_direct): New macro.
11077 (expand_scatter_store_optab_fn): New function.
11078 (direct_scatter_store_optab_supported_p): New macro.
11079 (internal_store_fn_p): New function.
11080 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11081 IFN_MASK_SCATTER_STORE.
11082 (internal_fn_mask_index): Likewise.
11083 (internal_fn_stored_value_index): New function.
11084 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11085 for scatter stores.
11086 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11087 * optabs-query.c (supports_vec_scatter_store_p): New function.
11088 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11089 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11090 true for scatter stores.
11091 (vect_gather_scatter_fn_p): Handle scatter stores too.
11092 (vect_check_gather_scatter): Consider using scatter stores if
11093 supports_vec_scatter_store_p.
11094 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11095 scatter stores too.
11096 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11097 internal_fn_stored_value_index.
11098 (check_load_store_masking): Handle scatter stores too.
11099 (vect_get_store_rhs): Make public.
11100 (vectorizable_call): Use internal_store_fn_p.
11101 (vectorizable_store): Handle scatter store internal functions.
11102 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11103 when deciding whether the end of the group has been reached.
11104 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11105 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11106 (mask_scatter_store<mode>): New insns.
11108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11109 Alan Hayward <alan.hayward@arm.com>
11110 David Sherwood <david.sherwood@arm.com>
11112 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11113 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11114 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11116 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11117 Use vect_truncate_gather_scatter_offset if we can't treat the
11118 operation as a normal gather load or scatter store.
11119 (get_group_load_store_type): Take the gather_scatter_info
11120 as argument. Try using a gather load or scatter store for
11121 single-element groups.
11122 (get_load_store_type): Update calls to get_group_load_store_type
11123 and vect_use_strided_gather_scatters_p.
11125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11126 Alan Hayward <alan.hayward@arm.com>
11127 David Sherwood <david.sherwood@arm.com>
11129 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11130 optional tree argument.
11131 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11133 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11134 but continue to use the current value as a fallback.
11135 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11136 to compare the updates.
11137 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11138 (get_load_store_type): Use it when handling a strided access.
11139 (vect_get_strided_load_store_ops): New function.
11140 (vect_get_data_ptr_increment): Likewise.
11141 (vectorizable_load): Handle strided gather loads. Always pass
11142 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * doc/md.texi (gather_load@var{m}): Document.
11149 (mask_gather_load@var{m}): Likewise.
11150 * genopinit.c (main): Add supports_vec_gather_load and
11151 supports_vec_gather_load_cached to target_optabs.
11152 * optabs-tree.c (init_tree_optimization_optabs): Use
11153 ggc_cleared_alloc to allocate target_optabs.
11154 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11155 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11157 * internal-fn.h (internal_load_fn_p): Declare.
11158 (internal_gather_scatter_fn_p): Likewise.
11159 (internal_fn_mask_index): Likewise.
11160 (internal_gather_scatter_fn_supported_p): Likewise.
11161 * internal-fn.c (gather_load_direct): New macro.
11162 (expand_gather_load_optab_fn): New function.
11163 (direct_gather_load_optab_supported_p): New macro.
11164 (direct_internal_fn_optab): New function.
11165 (internal_load_fn_p): Likewise.
11166 (internal_gather_scatter_fn_p): Likewise.
11167 (internal_fn_mask_index): Likewise.
11168 (internal_gather_scatter_fn_supported_p): Likewise.
11169 * optabs-query.c (supports_at_least_one_mode_p): New function.
11170 (supports_vec_gather_load_p): Likewise.
11171 * optabs-query.h (supports_vec_gather_load_p): Declare.
11172 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11173 and memory_type field.
11174 (NUM_PATTERNS): Bump to 15.
11175 * tree-vect-data-refs.c: Include internal-fn.h.
11176 (vect_gather_scatter_fn_p): New function.
11177 (vect_describe_gather_scatter_call): Likewise.
11178 (vect_check_gather_scatter): Try using internal functions for
11179 gather loads. Recognize existing calls to a gather load function.
11180 (vect_analyze_data_refs): Consider using gather loads if
11181 supports_vec_gather_load_p.
11182 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11183 (vect_get_gather_scatter_offset_type): Likewise.
11184 (vect_convert_mask_for_vectype): Likewise.
11185 (vect_add_conversion_to_patterm): Likewise.
11186 (vect_try_gather_scatter_pattern): Likewise.
11187 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11188 (vect_vect_recog_func_ptrs): Add it.
11189 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11190 internal_fn_mask_index and internal_gather_scatter_fn_p.
11191 (check_load_store_masking): Take the gather_scatter_info as an
11192 argument and handle gather loads.
11193 (vect_get_gather_scatter_ops): New function.
11194 (vectorizable_call): Check internal_load_fn_p.
11195 (vectorizable_load): Likewise. Handle gather load internal
11197 (vectorizable_store): Update call to check_load_store_masking.
11198 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11199 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11200 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11201 (aarch64_gather_scale_operand_d): New predicates.
11202 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11203 (mask_gather_load<mode>): New insns.
11205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11206 Alan Hayward <alan.hayward@arm.com>
11207 David Sherwood <david.sherwood@arm.com>
11209 * optabs.def (fold_left_plus_optab): New optab.
11210 * doc/md.texi (fold_left_plus_@var{m}): Document.
11211 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11212 * internal-fn.c (fold_left_direct): Define.
11213 (expand_fold_left_optab_fn): Likewise.
11214 (direct_fold_left_optab_supported_p): Likewise.
11215 * fold-const-call.c (fold_const_fold_left): New function.
11216 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11217 * tree-parloops.c (valid_reduction_p): New function.
11218 (gather_scalar_reductions): Use it.
11219 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11220 (vect_finish_replace_stmt): Declare.
11221 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11222 (needs_fold_left_reduction_p): New function, split out from...
11223 (vect_is_simple_reduction): ...here. Accept reductions that
11224 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11225 (vect_force_simple_reduction): Also store the reduction type in
11226 the assignment's STMT_VINFO_REDUC_TYPE.
11227 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11228 (merge_with_identity): New function.
11229 (vect_expand_fold_left): Likewise.
11230 (vectorize_fold_left_reduction): Likewise.
11231 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11232 scalar phi in place for it. Check for target support and reject
11233 cases that would reassociate the operation. Defer the transform
11234 phase to vectorize_fold_left_reduction.
11235 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11236 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11237 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11239 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11241 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11242 call to ifc_temp_var.
11244 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11245 Alan Hayward <alan.hayward@arm.com>
11246 David Sherwood <david.sherwood@arm.com>
11248 * target.def (legitimize_address_displacement): Take the original
11249 offset as a poly_int.
11250 * targhooks.h (default_legitimize_address_displacement): Update
11252 * targhooks.c (default_legitimize_address_displacement): Likewise.
11253 * doc/tm.texi: Regenerate.
11254 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11255 as an argument, moving assert of ad->disp == ad->disp_term to...
11256 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11257 Try calling targetm.legitimize_address_displacement before expanding
11258 the address rather than afterwards, and adjust for the new interface.
11259 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11260 Match the new hook interface. Handle SVE addresses.
11261 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11262 new hook interface.
11264 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11266 * Makefile.in (OBJS): Add early-remat.o.
11267 * target.def (select_early_remat_modes): New hook.
11268 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11269 * doc/tm.texi: Regenerate.
11270 * targhooks.h (default_select_early_remat_modes): Declare.
11271 * targhooks.c (default_select_early_remat_modes): New function.
11272 * timevar.def (TV_EARLY_REMAT): New timevar.
11273 * passes.def (pass_early_remat): New pass.
11274 * tree-pass.h (make_pass_early_remat): Declare.
11275 * early-remat.c: New file.
11276 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11278 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11281 Alan Hayward <alan.hayward@arm.com>
11282 David Sherwood <david.sherwood@arm.com>
11284 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11285 vfm1 with a bound_epilog parameter.
11286 (vect_do_peeling): Update calls accordingly, and move the prologue
11287 call earlier in the function. Treat the base bound_epilog as 0 for
11288 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11289 this base when peeling for gaps.
11290 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11291 with fully-masked loops.
11292 (vect_estimate_min_profitable_iters): Handle the single peeled
11293 iteration in that case.
11295 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11296 Alan Hayward <alan.hayward@arm.com>
11297 David Sherwood <david.sherwood@arm.com>
11299 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11300 single-element interleaving even if the size is not a power of 2.
11301 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11302 accesses for single-element interleaving if the group size is
11305 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11306 Alan Hayward <alan.hayward@arm.com>
11307 David Sherwood <david.sherwood@arm.com>
11309 * doc/md.texi (fold_extract_last_@var{m}): Document.
11310 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11311 * optabs.def (fold_extract_last_optab): New optab.
11312 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11313 * internal-fn.c (fold_extract_direct): New macro.
11314 (expand_fold_extract_optab_fn): Likewise.
11315 (direct_fold_extract_optab_supported_p): Likewise.
11316 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11317 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11318 EXTRACT_LAST_REDUCTION.
11319 (get_initial_def_for_reduction): Do not create an initial vector
11320 for EXTRACT_LAST_REDUCTION reductions.
11321 (vectorizable_reduction): Leave the scalar phi in place for
11322 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11323 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11324 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11325 transform phase to vectorizable_condition.
11326 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11328 (vect_finish_stmt_generation): ...here.
11329 (vect_finish_replace_stmt): New function.
11330 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11331 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11333 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11335 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11336 Alan Hayward <alan.hayward@arm.com>
11337 David Sherwood <david.sherwood@arm.com>
11339 * doc/md.texi (extract_last_@var{m}): Document.
11340 * optabs.def (extract_last_optab): New optab.
11341 * internal-fn.def (EXTRACT_LAST): New internal function.
11342 * internal-fn.c (cond_unary_direct): New macro.
11343 (expand_cond_unary_optab_fn): Likewise.
11344 (direct_cond_unary_optab_supported_p): Likewise.
11345 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11346 loops using EXTRACT_LAST.
11347 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11348 (extract_last_<mode>): ...this optab.
11349 (vec_extract<mode><Vel>): Update accordingly.
11351 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11352 Alan Hayward <alan.hayward@arm.com>
11353 David Sherwood <david.sherwood@arm.com>
11355 * target.def (empty_mask_is_expensive): New hook.
11356 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11357 * doc/tm.texi: Regenerate.
11358 * targhooks.h (default_empty_mask_is_expensive): Declare.
11359 * targhooks.c (default_empty_mask_is_expensive): New function.
11360 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11361 if the target says that empty masks are expensive.
11362 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11364 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11366 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11367 Alan Hayward <alan.hayward@arm.com>
11368 David Sherwood <david.sherwood@arm.com>
11370 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11371 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11372 (vect_use_loop_mask_for_alignment_p): New function.
11373 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11374 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11375 niters_skip argument. Make sure that the first niters_skip elements
11376 of the first iteration are inactive.
11377 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11378 Update call to vect_set_loop_masks_directly.
11379 (get_misalign_in_elems): New function, split out from...
11380 (vect_gen_prolog_loop_niters): ...here.
11381 (vect_update_init_of_dr): Take a code argument that specifies whether
11382 the adjustment should be added or subtracted.
11383 (vect_update_init_of_drs): Likewise.
11384 (vect_prepare_for_masked_peels): New function.
11385 (vect_do_peeling): Skip prologue peeling if we're using a mask
11386 instead. Update call to vect_update_inits_of_drs.
11387 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11389 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11390 alignment. Do not include the number of peeled iterations in
11391 the minimum threshold in that case.
11392 (vectorizable_induction): Adjust the start value down by
11393 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11394 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11395 Take the number of skipped iterations into account when calculating
11397 * tree-vect-stmts.c (vect_gen_while_not): New function.
11399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11400 Alan Hayward <alan.hayward@arm.com>
11401 David Sherwood <david.sherwood@arm.com>
11403 * doc/sourcebuild.texi (vect_fully_masked): Document.
11404 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11405 default value to 0.
11406 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11408 (vect_analyze_loop_2): ...here. Don't check the vectorization
11409 factor against the number of loop iterations if the loop is
11412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11413 Alan Hayward <alan.hayward@arm.com>
11414 David Sherwood <david.sherwood@arm.com>
11416 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11417 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11418 (dump_groups): Update accordingly.
11419 (iv_use::mem_type): New member variable.
11420 (address_p): New function.
11421 (record_use): Add a mem_type argument and initialize the new
11423 (record_group_use): Add a mem_type argument. Use address_p.
11424 Remove obsolete null checks of base_object. Update call to record_use.
11425 (find_interesting_uses_op): Update call to record_group_use.
11426 (find_interesting_uses_cond): Likewise.
11427 (find_interesting_uses_address): Likewise.
11428 (get_mem_type_for_internal_fn): New function.
11429 (find_address_like_use): Likewise.
11430 (find_interesting_uses_stmt): Try find_address_like_use before
11431 calling find_interesting_uses_op.
11432 (addr_offset_valid_p): Use the iv mem_type field as the type
11433 of the addressed memory.
11434 (add_autoinc_candidates): Likewise.
11435 (get_address_cost): Likewise.
11436 (split_small_address_groups_p): Use address_p.
11437 (split_address_groups): Likewise.
11438 (add_iv_candidate_for_use): Likewise.
11439 (autoinc_possible_for_pair): Likewise.
11440 (rewrite_groups): Likewise.
11441 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11442 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11443 (get_alias_ptr_type_for_ptr_address): New function.
11444 (rewrite_use_address): Rewrite address uses in calls that were
11445 identified by find_address_like_use.
11447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11448 Alan Hayward <alan.hayward@arm.com>
11449 David Sherwood <david.sherwood@arm.com>
11451 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11453 * gimple-expr.h (is_gimple_addressable: Likewise.
11454 * gimple-expr.c (is_gimple_address): Likewise.
11455 * internal-fn.c (expand_call_mem_ref): New function.
11456 (expand_mask_load_optab_fn): Use it.
11457 (expand_mask_store_optab_fn): Likewise.
11459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11460 Alan Hayward <alan.hayward@arm.com>
11461 David Sherwood <david.sherwood@arm.com>
11463 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11464 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11465 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11466 (cond_umax@var{mode}): Document.
11467 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11468 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11469 (cond_umin_optab, cond_umax_optab): New optabs.
11470 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11471 (COND_IOR, COND_XOR): New internal functions.
11472 * internal-fn.h (get_conditional_internal_fn): Declare.
11473 * internal-fn.c (cond_binary_direct): New macro.
11474 (expand_cond_binary_optab_fn): Likewise.
11475 (direct_cond_binary_optab_supported_p): Likewise.
11476 (get_conditional_internal_fn): New function.
11477 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11478 Cope with reduction statements that are vectorized as calls rather
11480 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11481 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11482 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11483 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11484 (UNSPEC_COND_EOR): New unspecs.
11485 (optab): Add mappings for them.
11486 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11487 (sve_int_op, sve_fp_op): New int attributes.
11489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11490 Alan Hayward <alan.hayward@arm.com>
11491 David Sherwood <david.sherwood@arm.com>
11493 * optabs.def (while_ult_optab): New optab.
11494 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11495 * internal-fn.def (WHILE_ULT): New internal function.
11496 * internal-fn.h (direct_internal_fn_supported_p): New override
11497 that takes two types as argument.
11498 * internal-fn.c (while_direct): New macro.
11499 (expand_while_optab_fn): New function.
11500 (convert_optab_supported_p): Likewise.
11501 (direct_while_optab_supported_p): New macro.
11502 * wide-int.h (wi::udiv_ceil): New function.
11503 * tree-vectorizer.h (rgroup_masks): New structure.
11504 (vec_loop_masks): New typedef.
11505 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11506 and fully_masked_p.
11507 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11508 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11509 (vect_max_vf): New function.
11510 (slpeel_make_loop_iterate_ntimes): Delete.
11511 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11512 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11513 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11514 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11515 internal-fn.h, stor-layout.h and optabs-query.h.
11516 (vect_set_loop_mask): New function.
11517 (add_preheader_seq): Likewise.
11518 (add_header_seq): Likewise.
11519 (interleave_supported_p): Likewise.
11520 (vect_maybe_permute_loop_masks): Likewise.
11521 (vect_set_loop_masks_directly): Likewise.
11522 (vect_set_loop_condition_masked): Likewise.
11523 (vect_set_loop_condition_unmasked): New function, split out from
11524 slpeel_make_loop_iterate_ntimes.
11525 (slpeel_make_loop_iterate_ntimes): Rename to..
11526 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11527 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11528 (vect_do_peeling): Update call accordingly.
11529 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11531 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11532 mask_compare_type, can_fully_mask_p and fully_masked_p.
11533 (release_vec_loop_masks): New function.
11534 (_loop_vec_info): Use it to free the loop masks.
11535 (can_produce_all_loop_masks_p): New function.
11536 (vect_get_max_nscalars_per_iter): Likewise.
11537 (vect_verify_full_masking): Likewise.
11538 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11539 retries, and free the mask rgroups before retrying. Check loop-wide
11540 reasons for disallowing fully-masked loops. Make the final decision
11541 about whether use a fully-masked loop or not.
11542 (vect_estimate_min_profitable_iters): Do not assume that peeling
11543 for the number of iterations will be needed for fully-masked loops.
11544 (vectorizable_reduction): Disable fully-masked loops.
11545 (vectorizable_live_operation): Likewise.
11546 (vect_halve_mask_nunits): New function.
11547 (vect_double_mask_nunits): Likewise.
11548 (vect_record_loop_mask): Likewise.
11549 (vect_get_loop_mask): Likewise.
11550 (vect_transform_loop): Handle the case in which the final loop
11551 iteration might handle a partial vector. Call vect_set_loop_condition
11552 instead of slpeel_make_loop_iterate_ntimes.
11553 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11554 (check_load_store_masking): New function.
11555 (prepare_load_store_mask): Likewise.
11556 (vectorizable_store): Handle fully-masked loops.
11557 (vectorizable_load): Likewise.
11558 (supportable_widening_operation): Use vect_halve_mask_nunits for
11560 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11561 (vect_gen_while): New function.
11562 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11563 (aarch64_uqdec<mode>): New insn.
11565 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11566 Alan Hayward <alan.hayward@arm.com>
11567 David Sherwood <david.sherwood@arm.com>
11569 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11570 (reduc_xor_scal_optab): New optabs.
11571 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11572 (reduc_xor_scal_@var{m}): Document.
11573 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11574 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11575 internal functions.
11576 * fold-const-call.c (fold_const_call): Handle them.
11577 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11578 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11579 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11580 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11581 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11582 (UNSPEC_XORV): New unspecs.
11583 (optab): Add entries for them.
11584 (BITWISEV): New int iterator.
11585 (bit_reduc_op): New int attributes.
11587 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11588 Alan Hayward <alan.hayward@arm.com>
11589 David Sherwood <david.sherwood@arm.com>
11591 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11592 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11593 * optabs.def (vec_shl_insert_optab): New optab.
11594 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11595 (duplicate_and_interleave): Likewise.
11596 * tree-vect-loop.c: Include internal-fn.h.
11597 (neutral_op_for_slp_reduction): New function, split out from
11598 get_initial_defs_for_reduction.
11599 (get_initial_def_for_reduction): Handle option 2 for variable-length
11600 vectors by loading the neutral value into a vector and then shifting
11601 the initial value into element 0.
11602 (get_initial_defs_for_reduction): Replace the code argument with
11603 the neutral value calculated by neutral_op_for_slp_reduction.
11604 Use gimple_build_vector for constant-length vectors.
11605 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11606 but the first group_size elements have a neutral value.
11607 Use duplicate_and_interleave otherwise.
11608 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11609 Update call to get_initial_defs_for_reduction. Handle SLP
11610 reductions for variable-length vectors by creating one vector
11611 result for each scalar result, with the elements associated
11612 with other scalar results stubbed out with the neutral value.
11613 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11614 Require IFN_VEC_SHL_INSERT for double reductions on
11615 variable-length vectors, or SLP reductions that have
11616 a neutral value. Require can_duplicate_and_interleave_p
11617 support for variable-length unchained SLP reductions if there
11618 is no neutral value, such as for MIN/MAX reductions. Also require
11619 the number of vector elements to be a multiple of the number of
11620 SLP statements when doing variable-length unchained SLP reductions.
11621 Update call to vect_create_epilog_for_reduction.
11622 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11623 and remove initial values.
11624 (duplicate_and_interleave): Make public.
11625 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11626 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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-slp.c: Include gimple-fold.h and internal-fn.h
11633 (can_duplicate_and_interleave_p): New function.
11634 (vect_get_and_check_slp_defs): Take the vector of statements
11635 rather than just the current one. Remove excess parentheses.
11636 Restriction rejectinon of vect_constant_def and vect_external_def
11637 for variable-length vectors to boolean types, or types for which
11638 can_duplicate_and_interleave_p is false.
11639 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11640 (duplicate_and_interleave): New function.
11641 (vect_get_constant_vectors): Use gimple_build_vector for
11642 constant-length vectors and suitable variable-length constant
11643 vectors. Use duplicate_and_interleave for other variable-length
11644 vectors. Don't defer the update when inserting new statements.
11646 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11647 Alan Hayward <alan.hayward@arm.com>
11648 David Sherwood <david.sherwood@arm.com>
11650 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11651 min_profitable_iters doesn't go negative.
11653 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11654 Alan Hayward <alan.hayward@arm.com>
11655 David Sherwood <david.sherwood@arm.com>
11657 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11658 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11659 * optabs.def (vec_mask_load_lanes_optab): New optab.
11660 (vec_mask_store_lanes_optab): Likewise.
11661 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11662 (MASK_STORE_LANES): Likewise.
11663 * internal-fn.c (mask_load_lanes_direct): New macro.
11664 (mask_store_lanes_direct): Likewise.
11665 (expand_mask_load_optab_fn): Handle masked operations.
11666 (expand_mask_load_lanes_optab_fn): New macro.
11667 (expand_mask_store_optab_fn): Handle masked operations.
11668 (expand_mask_store_lanes_optab_fn): New macro.
11669 (direct_mask_load_lanes_optab_supported_p): Likewise.
11670 (direct_mask_store_lanes_optab_supported_p): Likewise.
11671 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11673 (vect_load_lanes_supported): Likewise.
11674 * tree-vect-data-refs.c (strip_conversion): New function.
11675 (can_group_stmts_p): Likewise.
11676 (vect_analyze_data_ref_accesses): Use it instead of checking
11677 for a pair of assignments.
11678 (vect_store_lanes_supported): Take a masked_p parameter.
11679 (vect_load_lanes_supported): Likewise.
11680 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11681 vect_store_lanes_supported and vect_load_lanes_supported.
11682 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11683 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11684 parameter. Don't allow gaps for masked accesses.
11685 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11686 and vect_load_lanes_supported.
11687 (get_load_store_type): Take a masked_p parameter and update
11688 call to get_group_load_store_type.
11689 (vectorizable_store): Update call to get_load_store_type.
11690 Handle IFN_MASK_STORE_LANES.
11691 (vectorizable_load): Update call to get_load_store_type.
11692 Handle IFN_MASK_LOAD_LANES.
11694 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11695 Alan Hayward <alan.hayward@arm.com>
11696 David Sherwood <david.sherwood@arm.com>
11698 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11700 * config/aarch64/aarch64-protos.h
11701 (aarch64_sve_struct_memory_operand_p): Declare.
11702 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11703 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11704 (VPRED, vpred): Handle SVE structure modes.
11705 * config/aarch64/constraints.md (Utx): New constraint.
11706 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11707 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11708 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11709 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11710 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11711 structure modes. Split into pieces after RA.
11712 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11713 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11715 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11716 SVE structure modes.
11717 (aarch64_classify_address): Likewise.
11718 (sizetochar): Move earlier in file.
11719 (aarch64_print_operand): Handle SVE register lists.
11720 (aarch64_array_mode): New function.
11721 (aarch64_sve_struct_memory_operand_p): Likewise.
11722 (TARGET_ARRAY_MODE): Redefine.
11724 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11725 Alan Hayward <alan.hayward@arm.com>
11726 David Sherwood <david.sherwood@arm.com>
11728 * target.def (array_mode): New target hook.
11729 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11730 * doc/tm.texi: Regenerate.
11731 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11732 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11733 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11734 targetm.array_mode.
11735 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11739 Alan Hayward <alan.hayward@arm.com>
11740 David Sherwood <david.sherwood@arm.com>
11742 * fold-const.c (fold_binary_loc): Check the argument types
11743 rather than the result type when testing for a vector operation.
11745 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11747 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11748 * doc/tm.texi: Regenerate.
11750 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11751 Alan Hayward <alan.hayward@arm.com>
11752 David Sherwood <david.sherwood@arm.com>
11754 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11755 (sve): Document new AArch64 extension.
11756 * doc/md.texi (w): Extend the description of the AArch64
11757 constraint to include SVE vectors.
11758 (Upl, Upa): Document new AArch64 predicate constraints.
11759 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11761 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11762 (msve-vector-bits=): New option.
11763 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11764 SVE when these are disabled.
11765 (sve): New extension.
11766 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11767 modes. Adjust their number of units based on aarch64_sve_vg.
11768 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11769 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11770 aarch64_addr_query_type.
11771 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11772 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11773 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11774 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11775 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11776 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11777 (aarch64_simd_imm_zero_p): Delete.
11778 (aarch64_check_zero_based_sve_index_immediate): Declare.
11779 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11780 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11781 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11782 (aarch64_sve_float_mul_immediate_p): Likewise.
11783 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11784 rather than an rtx.
11785 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11786 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11787 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11788 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11789 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11790 (aarch64_regmode_natural_size): Likewise.
11791 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11792 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11794 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11795 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11796 for VG and the SVE predicate registers.
11797 (V_ALIASES): Add a "z"-prefixed alias.
11798 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11799 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11800 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11801 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11802 (REG_CLASS_NAMES): Add entries for them.
11803 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11804 and the predicate registers.
11805 (aarch64_sve_vg): Declare.
11806 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11807 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11808 (REGMODE_NATURAL_SIZE): Define.
11809 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11811 * config/aarch64/aarch64.c: Include cfgrtl.h.
11812 (simd_immediate_info): Add a constructor for series vectors,
11813 and an associated step field.
11814 (aarch64_sve_vg): New variable.
11815 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11816 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11817 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11818 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11819 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11820 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11821 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11822 (aarch64_get_mask_mode): New functions.
11823 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11824 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11825 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11826 predicate modes and predicate registers. Explicitly restrict
11827 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11828 to store a vector mode if it is recognized by
11829 aarch64_classify_vector_mode.
11830 (aarch64_regmode_natural_size): New function.
11831 (aarch64_hard_regno_caller_save_mode): Return the original mode
11833 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11834 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11835 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11836 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11838 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11839 does not overlap dest if the function is frame-related. Handle
11841 (aarch64_split_add_offset): New function.
11842 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11843 them aarch64_add_offset.
11844 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11845 and update call to aarch64_sub_sp.
11846 (aarch64_add_cfa_expression): New function.
11847 (aarch64_expand_prologue): Pass extra temporary registers to the
11848 functions above. Handle the case in which we need to emit new
11849 DW_CFA_expressions for registers that were originally saved
11850 relative to the stack pointer, but now have to be expressed
11851 relative to the frame pointer.
11852 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11854 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11855 IP0 and IP1 values for SVE frames.
11856 (aarch64_expand_vec_series): New function.
11857 (aarch64_expand_sve_widened_duplicate): Likewise.
11858 (aarch64_expand_sve_const_vector): Likewise.
11859 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11860 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11861 into the register, rather than emitting a SET directly.
11862 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11863 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11864 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11865 (offset_9bit_signed_scaled_p): New functions.
11866 (aarch64_replicate_bitmask_imm): New function.
11867 (aarch64_bitmask_imm): Use it.
11868 (aarch64_cannot_force_const_mem): Reject expressions involving
11869 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11870 (aarch64_classify_index): Handle SVE indices, by requiring
11871 a plain register index with a scale that matches the element size.
11872 (aarch64_classify_address): Handle SVE addresses. Assert that
11873 the mode of the address is VOIDmode or an integer mode.
11874 Update call to aarch64_classify_symbol.
11875 (aarch64_classify_symbolic_expression): Update call to
11876 aarch64_classify_symbol.
11877 (aarch64_const_vec_all_in_range_p): New function.
11878 (aarch64_print_vector_float_operand): Likewise.
11879 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11880 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11881 and the FP immediates 1.0 and 0.5.
11882 (aarch64_print_address_internal): Handle SVE addresses.
11883 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11884 (aarch64_regno_regclass): Handle predicate registers.
11885 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11887 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11888 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11889 (aarch64_convert_sve_vector_bits): New function.
11890 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11891 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11892 rather than an rtx.
11893 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11894 Handle SVE vector and predicate modes. Accept VL-based constants
11895 that need only one temporary register, and VL offsets that require
11896 no temporary registers.
11897 (aarch64_conditional_register_usage): Mark the predicate registers
11898 as fixed if SVE isn't available.
11899 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11900 Return true for SVE vector and predicate modes.
11901 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11902 rather than an unsigned int. Handle SVE modes.
11903 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11905 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11907 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11908 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11909 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11910 (aarch64_sve_float_mul_immediate_p): New functions.
11911 (aarch64_sve_valid_immediate): New function.
11912 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11913 Explicitly reject structure modes. Check for INDEX constants.
11914 Handle PTRUE and PFALSE constants.
11915 (aarch64_check_zero_based_sve_index_immediate): New function.
11916 (aarch64_simd_imm_zero_p): Delete.
11917 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11918 vector modes. Accept constants in the range of CNT[BHWD].
11919 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11920 ask for an Advanced SIMD mode.
11921 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11922 (aarch64_simd_vector_alignment): Handle SVE predicates.
11923 (aarch64_vectorize_preferred_vector_alignment): New function.
11924 (aarch64_simd_vector_alignment_reachable): Use it instead of
11926 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11927 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11929 (MAX_VECT_LEN): Delete.
11930 (expand_vec_perm_d): Add a vec_flags field.
11931 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11932 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11933 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11935 (aarch64_evpc_rev): Rename to...
11936 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11937 (aarch64_evpc_rev_global): New function.
11938 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11939 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11941 (aarch64_evpc_sve_tbl): New function.
11942 (aarch64_expand_vec_perm_const_1): Update after rename of
11943 aarch64_evpc_rev. Handle SVE permutes too, trying
11944 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11945 than aarch64_evpc_tbl.
11946 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11947 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11948 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11949 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11950 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11951 (aarch64_expand_sve_vcond): New functions.
11952 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11953 of aarch64_vector_mode_p.
11954 (aarch64_dwarf_poly_indeterminate_value): New function.
11955 (aarch64_compute_pressure_classes): Likewise.
11956 (aarch64_can_change_mode_class): Likewise.
11957 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11958 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11959 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11960 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11961 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11962 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11963 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11964 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11966 (Dn, Dl, Dr): Accept const as well as const_vector.
11967 (Dz): Likewise. Compare against CONST0_RTX.
11968 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11969 of "vector" where appropriate.
11970 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11971 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11972 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11973 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11974 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11975 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11976 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11977 (v_int_equiv): Extend to SVE modes.
11978 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11980 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11981 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11982 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11983 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11984 (SVE_COND_FP_CMP): New int iterators.
11985 (perm_hilo): Handle the new unpack unspecs.
11986 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11988 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11989 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11990 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11991 (aarch64_equality_operator, aarch64_constant_vector_operand)
11992 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11993 (aarch64_sve_nonimmediate_operand): Likewise.
11994 (aarch64_sve_general_operand): Likewise.
11995 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11996 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11997 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11998 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11999 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
12000 (aarch64_sve_float_arith_immediate): Likewise.
12001 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
12002 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
12003 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
12004 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
12005 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
12006 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
12007 (aarch64_sve_float_arith_operand): Likewise.
12008 (aarch64_sve_float_arith_with_sub_operand): Likewise.
12009 (aarch64_sve_float_mul_operand): Likewise.
12010 (aarch64_sve_vec_perm_operand): Likewise.
12011 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
12012 (aarch64_mov_operand): Accept const_poly_int and const_vector.
12013 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
12014 as well as const_vector.
12015 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
12016 in file. Use CONST0_RTX and CONSTM1_RTX.
12017 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
12018 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
12019 Use aarch64_simd_imm_zero.
12020 * config/aarch64/aarch64-sve.md: New file.
12021 * config/aarch64/aarch64.md: Include it.
12022 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
12023 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
12024 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
12025 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
12026 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
12027 (sve): New attribute.
12028 (enabled): Disable instructions with the sve attribute unless
12030 (movqi, movhi): Pass CONST_POLY_INT operaneds through
12031 aarch64_expand_mov_immediate.
12032 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
12033 CNT[BHSD] immediates.
12034 (movti): Split CONST_POLY_INT moves into two halves.
12035 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
12036 Split additions that need a temporary here if the destination
12037 is the stack pointer.
12038 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
12039 (*add<mode>3_poly_1): New instruction.
12040 (set_clobber_cc): New expander.
12042 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12044 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
12045 parameter and use it instead of GET_MODE_SIZE (innermode). Use
12046 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
12047 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
12048 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
12049 Change innermode from fixed_mode_size to machine_mode.
12050 (simplify_subreg): Update call accordingly. Handle a constant-sized
12051 subreg of a variable-length CONST_VECTOR.
12053 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12054 Alan Hayward <alan.hayward@arm.com>
12055 David Sherwood <david.sherwood@arm.com>
12057 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12058 (add_offset_to_base): New function, split out from...
12059 (create_mem_ref): ...here. When handling a scale other than 1,
12060 check first whether the address is valid without the offset.
12061 Add it into the base if so, leaving the index and scale as-is.
12063 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12066 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12067 fold_for_warn before checking if arg2 is INTEGER_CST.
12069 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12071 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12072 (store_multiple_operation): Delete.
12073 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12074 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12075 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12076 guarded by TARGET_STRING.
12077 (rs6000_output_load_multiple): Delete.
12078 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12079 OPTION_MASK_STRING / TARGET_STRING handling.
12080 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12081 (const rs6000_opt_masks) <"string">: Change mask to 0.
12082 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12083 (MASK_STRING): Delete.
12084 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12086 (load_multiple): Delete.
12093 (store_multiple): Delete.
12100 (movmemsi_8reg): Delete.
12101 (corresponding unnamed define_insn): Delete.
12102 (movmemsi_6reg): Delete.
12103 (corresponding unnamed define_insn): Delete.
12104 (movmemsi_4reg): Delete.
12105 (corresponding unnamed define_insn): Delete.
12106 (movmemsi_2reg): Delete.
12107 (corresponding unnamed define_insn): Delete.
12108 (movmemsi_1reg): Delete.
12109 (corresponding unnamed define_insn): Delete.
12110 * config/rs6000/rs6000.opt (mno-string): New.
12111 (mstring): Replace by deprecation warning stub.
12112 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12114 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12116 * regrename.c (regrename_do_replace): If replacing the same
12117 reg multiple times, try to reuse last created gen_raw_REG.
12120 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12121 main to workaround a bug in GDB.
12123 2018-01-12 Tom de Vries <tom@codesourcery.com>
12126 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12128 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12130 PR rtl-optimization/80481
12131 * ira-color.c (get_cap_member): New function.
12132 (allocnos_conflict_by_live_ranges_p): Use it.
12133 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12134 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12136 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12139 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12140 (*saddl_se_1): Ditto.
12141 (*ssubsi_1): Ditto.
12142 (*ssubl_se_1): Ditto.
12144 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12146 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12147 rather than wi::to_widest for DR_INITs.
12148 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12149 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12150 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12152 (vect_analyze_group_access_1): Note that here.
12154 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12156 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12157 polynomial type sizes.
12159 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12161 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12162 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12163 (gimple_add_tmp_var): Likewise.
12165 2018-01-12 Martin Liska <mliska@suse.cz>
12167 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12168 (gimple_alloc_sizes): Likewise.
12169 (dump_gimple_statistics): Use PRIu64 in printf format.
12170 * gimple.h: Change uint64_t to int.
12172 2018-01-12 Martin Liska <mliska@suse.cz>
12174 * tree-core.h: Use uint64_t instead of int.
12175 * tree.c (tree_node_counts): Likewise.
12176 (tree_node_sizes): Likewise.
12177 (dump_tree_statistics): Use PRIu64 in printf format.
12179 2018-01-12 Martin Liska <mliska@suse.cz>
12181 * Makefile.in: As qsort_chk is implemented in vec.c, add
12182 vec.o to linkage of gencfn-macros.
12183 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12184 passing the info to record_node_allocation_statistics.
12185 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12187 * ggc-common.c (struct ggc_usage): Add operator== and use
12188 it in operator< and compare function.
12189 * mem-stats.h (struct mem_usage): Likewise.
12190 * vec.c (struct vec_usage): Remove operator< and compare
12191 function. Can be simply inherited.
12193 2018-01-12 Martin Jambor <mjambor@suse.cz>
12196 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12197 * tree-ssa-math-opts.c: Include domwalk.h.
12198 (convert_mult_to_fma_1): New function.
12199 (fma_transformation_info): New type.
12200 (fma_deferring_state): Likewise.
12201 (cancel_fma_deferring): New function.
12202 (result_of_phi): Likewise.
12203 (last_fma_candidate_feeds_initial_phi): Likewise.
12204 (convert_mult_to_fma): Added deferring logic, split actual
12205 transformation to convert_mult_to_fma_1.
12206 (math_opts_dom_walker): New type.
12207 (math_opts_dom_walker::after_dom_children): New method, body moved
12208 here from pass_optimize_widening_mul::execute, added deferring logic
12210 (pass_optimize_widening_mul::execute): Moved most of code to
12211 math_opts_dom_walker::after_dom_children.
12212 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12213 * config/i386/i386.c (ix86_option_override_internal): Added
12214 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12216 2018-01-12 Richard Biener <rguenther@suse.de>
12219 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12220 inline instance vars.
12222 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12225 * config/rx/rx.c (rx_is_restricted_memory_address):
12226 Handle SUBREG case.
12228 2018-01-12 Richard Biener <rguenther@suse.de>
12230 PR tree-optimization/80846
12231 * target.def (split_reduction): New target hook.
12232 * targhooks.c (default_split_reduction): New function.
12233 * targhooks.h (default_split_reduction): Declare.
12234 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12235 target requests first reduce vectors by combining low and high
12237 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12238 (get_vectype_for_scalar_type_and_size): Export.
12239 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12240 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12241 * doc/tm.texi: Regenerate.
12242 * config/i386/i386.c (ix86_split_reduction): Implement
12243 TARGET_VECTORIZE_SPLIT_REDUCTION.
12245 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12248 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12249 in PIC mode except for TARGET_VXWORKS_RTP.
12250 * config/sparc/sparc.c: Include cfgrtl.h.
12251 (TARGET_INIT_PIC_REG): Define.
12252 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12253 (sparc_pic_register_p): New predicate.
12254 (sparc_legitimate_address_p): Use it.
12255 (sparc_legitimize_pic_address): Likewise.
12256 (sparc_delegitimize_address): Likewise.
12257 (sparc_mode_dependent_address_p): Likewise.
12258 (gen_load_pcrel_sym): Remove 4th parameter.
12259 (load_got_register): Adjust call to above. Remove obsolete stuff.
12260 (sparc_expand_prologue): Do not call load_got_register here.
12261 (sparc_flat_expand_prologue): Likewise.
12262 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12263 (sparc_use_pseudo_pic_reg): New function.
12264 (sparc_init_pic_reg): Likewise.
12265 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12266 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12268 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12270 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12271 Add item for branch_cost.
12273 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12275 PR rtl-optimization/83565
12276 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12277 not extend the result to a larger mode for rotate operations.
12278 (num_sign_bit_copies1): Likewise.
12280 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12283 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12285 Use values-Xc.o for -pedantic.
12286 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12288 2018-01-12 Martin Liska <mliska@suse.cz>
12291 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12293 (possible_polymorphic_call_targets): Use it.
12294 (ipa_devirt): Likewise.
12296 2018-01-12 Martin Liska <mliska@suse.cz>
12298 * profile-count.h (enum profile_quality): Use 0 as invalid
12299 enum value of profile_quality.
12301 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12303 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12304 -mext-string options.
12306 2018-01-12 Richard Biener <rguenther@suse.de>
12308 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12309 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12310 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12312 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12314 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12316 * configure.ac (--with-long-double-format): Add support for the
12317 configuration option to change the default long double format on
12319 * config.gcc (powerpc*-linux*-*): Likewise.
12320 * configure: Regenerate.
12321 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12322 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12323 used without modification.
12325 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12327 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12328 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12329 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12330 MISC_BUILTIN_SPEC_BARRIER.
12331 (rs6000_init_builtins): Likewise.
12332 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12334 (speculation_barrier): New define_insn.
12335 * doc/extend.texi: Document __builtin_speculation_barrier.
12337 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12340 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12341 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12342 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12344 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12345 integral modes instead of "ss" and "sd".
12346 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12347 vectors with 32-bit and 64-bit elements.
12348 (vecdupssescalarmodesuffix): New mode attribute.
12349 (vec_dup<mode>): Use it.
12351 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12354 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12355 frame if argument is passed on stack.
12357 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12360 * ree.c (combine_reaching_defs): Optimize also
12361 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12362 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12364 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12366 PR middle-end/83189
12367 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12369 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12371 PR middle-end/83718
12372 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12373 after they are computed.
12375 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12377 PR tree-optimization/83695
12378 * gimple-loop-linterchange.cc
12379 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12380 reset cached scev information after interchange.
12381 (pass_linterchange::execute): Remove call to scev_reset_htab.
12383 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12385 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12386 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12387 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12388 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12389 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12390 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12391 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12392 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12393 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12394 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12395 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12396 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12397 (V_lane_reg): Likewise.
12398 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12400 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12401 (vfmal_lane_low<mode>_intrinsic,
12402 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12403 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12404 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12405 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12406 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12407 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12409 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12411 * config/arm/arm-cpus.in (fp16fml): New feature.
12412 (ALL_SIMD): Add fp16fml.
12413 (armv8.2-a): Add fp16fml as an option.
12414 (armv8.3-a): Likewise.
12415 (armv8.4-a): Add fp16fml as part of fp16.
12416 * config/arm/arm.h (TARGET_FP16FML): Define.
12417 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12419 * config/arm/arm-modes.def (V2HF): Define.
12420 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12421 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12422 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12423 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12424 vfmsl_low, vfmsl_high): New set of builtins.
12425 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12426 (vfml_op): New code attribute.
12427 (VFMLHALVES): New int iterator.
12428 (VFML, VFMLSEL): New mode attributes.
12429 (V_reg): Define mapping for V2HF.
12430 (V_hi, V_lo): New mode attributes.
12431 (VF_constraint): Likewise.
12432 (vfml_half, vfml_half_selector): New int attributes.
12433 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12435 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12436 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12438 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12439 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12440 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12441 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12443 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12444 Document new effective target and option set.
12446 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12448 * config/arm/arm-cpus.in (armv8_4): New feature.
12449 (ARMv8_4a): New fgroup.
12450 (armv8.4-a): New arch.
12451 * config/arm/arm-tables.opt: Regenerate.
12452 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12453 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12454 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12455 Add matching rules for -march=armv8.4-a and extensions.
12456 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12458 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12461 * config/rx/rx.md (BW): New mode attribute.
12462 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12464 2018-01-11 Richard Biener <rguenther@suse.de>
12466 PR tree-optimization/83435
12467 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12468 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12469 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12471 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12472 Alan Hayward <alan.hayward@arm.com>
12473 David Sherwood <david.sherwood@arm.com>
12475 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12477 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12478 (aarch64_print_address_internal): Use it to check for a zero offset.
12480 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12481 Alan Hayward <alan.hayward@arm.com>
12482 David Sherwood <david.sherwood@arm.com>
12484 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12485 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12486 Return a poly_int64 rather than a HOST_WIDE_INT.
12487 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12488 rather than a HOST_WIDE_INT.
12489 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12490 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12491 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12492 final_offset from HOST_WIDE_INT to poly_int64.
12493 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12494 to_constant when getting the number of units in an Advanced SIMD
12496 (aarch64_builtin_vectorized_function): Check for a constant number
12498 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12500 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12501 attribute instead of GET_MODE_NUNITS.
12502 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12503 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12504 GET_MODE_SIZE for fixed-size registers.
12505 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12506 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12507 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12508 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12509 (aarch64_print_operand, aarch64_print_address_internal)
12510 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12511 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12512 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12513 Handle polynomial GET_MODE_SIZE.
12514 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12515 wider than SImode without modification.
12516 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12517 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12518 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12519 passing and returning SVE modes.
12520 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12521 rather than GEN_INT.
12522 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12523 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12524 (aarch64_allocate_and_probe_stack_space): Likewise.
12525 (aarch64_layout_frame): Cope with polynomial offsets.
12526 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12527 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12528 polynomial offsets.
12529 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12530 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12531 poly_int64 rather than a HOST_WIDE_INT.
12532 (aarch64_get_separate_components, aarch64_process_components)
12533 (aarch64_expand_prologue, aarch64_expand_epilogue)
12534 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12535 (aarch64_anchor_offset): New function, split out from...
12536 (aarch64_legitimize_address): ...here.
12537 (aarch64_builtin_vectorization_cost): Handle polynomial
12538 TYPE_VECTOR_SUBPARTS.
12539 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12541 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12542 number of elements from the PARALLEL rather than the mode.
12543 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12544 rather than GET_MODE_BITSIZE.
12545 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12546 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12547 (aarch64_expand_vec_perm_const_1): Handle polynomial
12548 d->perm.length () and d->perm elements.
12549 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12550 Apply to_constant to d->perm elements.
12551 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12552 polynomial CONST_VECTOR_NUNITS.
12553 (aarch64_move_pointer): Take amount as a poly_int64 rather
12555 (aarch64_progress_pointer): Avoid temporary variable.
12556 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12557 the mode attribute instead of GET_MODE.
12559 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12560 Alan Hayward <alan.hayward@arm.com>
12561 David Sherwood <david.sherwood@arm.com>
12563 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12564 x exists before using it.
12565 (aarch64_add_constant_internal): Rename to...
12566 (aarch64_add_offset_1): ...this. Replace regnum with separate
12567 src and dest rtxes. Handle the case in which they're different,
12568 including when the offset is zero. Replace scratchreg with an rtx.
12569 Use 2 additions if there is no spare register into which we can
12570 move a 16-bit constant.
12571 (aarch64_add_constant): Delete.
12572 (aarch64_add_offset): Replace reg with separate src and dest
12573 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12574 Use aarch64_add_offset_1.
12575 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12576 an rtx rather than an int. Take the delta as a poly_int64
12577 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12578 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12579 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12580 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12581 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12582 and aarch64_add_sp.
12583 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12584 aarch64_add_constant.
12586 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12588 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12589 Use scalar_float_mode.
12591 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12593 * config/aarch64/aarch64-simd.md
12594 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12595 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12596 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12597 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12598 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12599 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12600 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12601 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12602 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12603 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12605 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12608 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12609 targ_options->x_arm_arch_string is non NULL.
12611 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12613 * config/aarch64/aarch64.h
12614 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12616 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12619 * expmed.c (emit_store_flag_force): Swap if const op0
12620 and change VOIDmode to mode of op0.
12622 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12624 PR rtl-optimization/83761
12625 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12626 than bytes to mode_for_size.
12628 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12630 PR middle-end/83189
12631 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12632 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12635 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12637 PR middle-end/83575
12638 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12639 when in layout mode.
12640 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12641 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12644 2018-01-10 Michael Collison <michael.collison@arm.com>
12646 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12647 * config/aarch64/aarch64-option-extension.def: Add
12648 AARCH64_OPT_EXTENSION of 'fp16fml'.
12649 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12650 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12651 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12652 * config/aarch64/constraints.md (Ui7): New constraint.
12653 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12654 (VFMLA_SEL_W): Ditto.
12657 (VFMLA16_LOW): New int iterator.
12658 (VFMLA16_HIGH): Ditto.
12659 (UNSPEC_FMLAL): New unspec.
12660 (UNSPEC_FMLSL): Ditto.
12661 (UNSPEC_FMLAL2): Ditto.
12662 (UNSPEC_FMLSL2): Ditto.
12663 (f16mac): New code attribute.
12664 * config/aarch64/aarch64-simd-builtins.def
12665 (aarch64_fmlal_lowv2sf): Ditto.
12666 (aarch64_fmlsl_lowv2sf): Ditto.
12667 (aarch64_fmlalq_lowv4sf): Ditto.
12668 (aarch64_fmlslq_lowv4sf): Ditto.
12669 (aarch64_fmlal_highv2sf): Ditto.
12670 (aarch64_fmlsl_highv2sf): Ditto.
12671 (aarch64_fmlalq_highv4sf): Ditto.
12672 (aarch64_fmlslq_highv4sf): Ditto.
12673 (aarch64_fmlal_lane_lowv2sf): Ditto.
12674 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12675 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12676 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12677 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12678 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12679 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12680 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12681 (aarch64_fmlal_lane_highv2sf): Ditto.
12682 (aarch64_fmlsl_lane_highv2sf): Ditto.
12683 (aarch64_fmlal_laneq_highv2sf): Ditto.
12684 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12685 (aarch64_fmlalq_lane_highv4sf): Ditto.
12686 (aarch64_fmlsl_lane_highv4sf): Ditto.
12687 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12688 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12689 * config/aarch64/aarch64-simd.md:
12690 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12691 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12692 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12693 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12694 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12695 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12696 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12697 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12698 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12699 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12700 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12701 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12702 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12703 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12704 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12705 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12706 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12707 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12708 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12709 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12710 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12711 (vfmlsl_low_u32): Ditto.
12712 (vfmlalq_low_u32): Ditto.
12713 (vfmlslq_low_u32): Ditto.
12714 (vfmlal_high_u32): Ditto.
12715 (vfmlsl_high_u32): Ditto.
12716 (vfmlalq_high_u32): Ditto.
12717 (vfmlslq_high_u32): Ditto.
12718 (vfmlal_lane_low_u32): Ditto.
12719 (vfmlsl_lane_low_u32): Ditto.
12720 (vfmlal_laneq_low_u32): Ditto.
12721 (vfmlsl_laneq_low_u32): Ditto.
12722 (vfmlalq_lane_low_u32): Ditto.
12723 (vfmlslq_lane_low_u32): Ditto.
12724 (vfmlalq_laneq_low_u32): Ditto.
12725 (vfmlslq_laneq_low_u32): Ditto.
12726 (vfmlal_lane_high_u32): Ditto.
12727 (vfmlsl_lane_high_u32): Ditto.
12728 (vfmlal_laneq_high_u32): Ditto.
12729 (vfmlsl_laneq_high_u32): Ditto.
12730 (vfmlalq_lane_high_u32): Ditto.
12731 (vfmlslq_lane_high_u32): Ditto.
12732 (vfmlalq_laneq_high_u32): Ditto.
12733 (vfmlslq_laneq_high_u32): Ditto.
12734 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12735 (AARCH64_FL_FOR_ARCH8_4): New.
12736 (AARCH64_ISA_F16FML): New ISA flag.
12737 (TARGET_F16FML): New feature flag for fp16fml.
12738 (doc/invoke.texi): Document new fp16fml option.
12740 2018-01-10 Michael Collison <michael.collison@arm.com>
12742 * config/aarch64/aarch64-builtins.c:
12743 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12744 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12745 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12746 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12747 (AARCH64_ISA_SHA3): New ISA flag.
12748 (TARGET_SHA3): New feature flag for sha3.
12749 * config/aarch64/iterators.md (sha512_op): New int attribute.
12750 (CRYPTO_SHA512): New int iterator.
12751 (UNSPEC_SHA512H): New unspec.
12752 (UNSPEC_SHA512H2): Ditto.
12753 (UNSPEC_SHA512SU0): Ditto.
12754 (UNSPEC_SHA512SU1): Ditto.
12755 * config/aarch64/aarch64-simd-builtins.def
12756 (aarch64_crypto_sha512hqv2di): New builtin.
12757 (aarch64_crypto_sha512h2qv2di): Ditto.
12758 (aarch64_crypto_sha512su0qv2di): Ditto.
12759 (aarch64_crypto_sha512su1qv2di): Ditto.
12760 (aarch64_eor3qv8hi): Ditto.
12761 (aarch64_rax1qv2di): Ditto.
12762 (aarch64_xarqv2di): Ditto.
12763 (aarch64_bcaxqv8hi): Ditto.
12764 * config/aarch64/aarch64-simd.md:
12765 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12766 (aarch64_crypto_sha512su0qv2di): Ditto.
12767 (aarch64_crypto_sha512su1qv2di): Ditto.
12768 (aarch64_eor3qv8hi): Ditto.
12769 (aarch64_rax1qv2di): Ditto.
12770 (aarch64_xarqv2di): Ditto.
12771 (aarch64_bcaxqv8hi): Ditto.
12772 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12773 (vsha512h2q_u64): Ditto.
12774 (vsha512su0q_u64): Ditto.
12775 (vsha512su1q_u64): Ditto.
12776 (veor3q_u16): Ditto.
12777 (vrax1q_u64): Ditto.
12778 (vxarq_u64): Ditto.
12779 (vbcaxq_u16): Ditto.
12780 * config/arm/types.md (crypto_sha512): New type attribute.
12781 (crypto_sha3): Ditto.
12782 (doc/invoke.texi): Document new sha3 option.
12784 2018-01-10 Michael Collison <michael.collison@arm.com>
12786 * config/aarch64/aarch64-builtins.c:
12787 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12788 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12789 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12790 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12791 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12792 (AARCH64_ISA_SM4): New ISA flag.
12793 (TARGET_SM4): New feature flag for sm4.
12794 * config/aarch64/aarch64-simd-builtins.def
12795 (aarch64_sm3ss1qv4si): Ditto.
12796 (aarch64_sm3tt1aq4si): Ditto.
12797 (aarch64_sm3tt1bq4si): Ditto.
12798 (aarch64_sm3tt2aq4si): Ditto.
12799 (aarch64_sm3tt2bq4si): Ditto.
12800 (aarch64_sm3partw1qv4si): Ditto.
12801 (aarch64_sm3partw2qv4si): Ditto.
12802 (aarch64_sm4eqv4si): Ditto.
12803 (aarch64_sm4ekeyqv4si): Ditto.
12804 * config/aarch64/aarch64-simd.md:
12805 (aarch64_sm3ss1qv4si): Ditto.
12806 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12807 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12808 (aarch64_sm4eqv4si): Ditto.
12809 (aarch64_sm4ekeyqv4si): Ditto.
12810 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12811 (sm3part_op): Ditto.
12812 (CRYPTO_SM3TT): Ditto.
12813 (CRYPTO_SM3PART): Ditto.
12814 (UNSPEC_SM3SS1): New unspec.
12815 (UNSPEC_SM3TT1A): Ditto.
12816 (UNSPEC_SM3TT1B): Ditto.
12817 (UNSPEC_SM3TT2A): Ditto.
12818 (UNSPEC_SM3TT2B): Ditto.
12819 (UNSPEC_SM3PARTW1): Ditto.
12820 (UNSPEC_SM3PARTW2): Ditto.
12821 (UNSPEC_SM4E): Ditto.
12822 (UNSPEC_SM4EKEY): Ditto.
12823 * config/aarch64/constraints.md (Ui2): New constraint.
12824 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12825 * config/arm/types.md (crypto_sm3): New type attribute.
12826 (crypto_sm4): Ditto.
12827 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12828 (vsm3tt1aq_u32): Ditto.
12829 (vsm3tt1bq_u32): Ditto.
12830 (vsm3tt2aq_u32): Ditto.
12831 (vsm3tt2bq_u32): Ditto.
12832 (vsm3partw1q_u32): Ditto.
12833 (vsm3partw2q_u32): Ditto.
12834 (vsm4eq_u32): Ditto.
12835 (vsm4ekeyq_u32): Ditto.
12836 (doc/invoke.texi): Document new sm4 option.
12838 2018-01-10 Michael Collison <michael.collison@arm.com>
12840 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12841 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12842 (AARCH64_FL_FOR_ARCH8_4): New.
12843 (AARCH64_FL_V8_4): New flag.
12844 (doc/invoke.texi): Document new armv8.4-a option.
12846 2018-01-10 Michael Collison <michael.collison@arm.com>
12848 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12849 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12850 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12851 * config/aarch64/aarch64-option-extension.def: Add
12852 AARCH64_OPT_EXTENSION of 'sha2'.
12853 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12854 (crypto): Disable sha2 and aes if crypto disabled.
12855 (crypto): Enable aes and sha2 if enabled.
12856 (simd): Disable sha2 and aes if simd disabled.
12857 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12859 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12860 (TARGET_SHA2): New feature flag for sha2.
12861 (TARGET_AES): New feature flag for aes.
12862 * config/aarch64/aarch64-simd.md:
12863 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12864 conditional on TARGET_AES.
12865 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12866 (aarch64_crypto_sha1hsi): Make pattern conditional
12868 (aarch64_crypto_sha1hv4si): Ditto.
12869 (aarch64_be_crypto_sha1hv4si): Ditto.
12870 (aarch64_crypto_sha1su1v4si): Ditto.
12871 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12872 (aarch64_crypto_sha1su0v4si): Ditto.
12873 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12874 (aarch64_crypto_sha256su0v4si): Ditto.
12875 (aarch64_crypto_sha256su1v4si): Ditto.
12876 (doc/invoke.texi): Document new aes and sha2 options.
12878 2018-01-10 Martin Sebor <msebor@redhat.com>
12880 PR tree-optimization/83781
12881 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12884 2018-01-11 Martin Sebor <msebor@gmail.com>
12885 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12887 PR tree-optimization/83501
12888 PR tree-optimization/81703
12890 * tree-ssa-strlen.c (get_string_cst): Rename...
12891 (get_string_len): ...to this. Handle global constants.
12892 (handle_char_store): Adjust.
12894 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12895 Jim Wilson <jimw@sifive.com>
12897 * config/riscv/riscv-protos.h (riscv_output_return): New.
12898 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12899 (riscv_attribute_table, riscv_output_return),
12900 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12901 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12902 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12904 (riscv_expand_prologue): Add early return for naked function.
12905 (riscv_expand_epilogue): Likewise.
12906 (riscv_function_ok_for_sibcall): Return false for naked function.
12907 (riscv_set_current_function): New.
12908 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12909 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12910 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12911 * doc/extend.texi (RISC-V Function Attributes): New.
12913 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12915 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12916 check for 128-bit long double before checking TCmode.
12917 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12918 128-bit long doubles before checking TFmode or TCmode.
12919 (FLOAT128_IBM_P): Likewise.
12921 2018-01-10 Martin Sebor <msebor@redhat.com>
12923 PR tree-optimization/83671
12924 * builtins.c (c_strlen): Unconditionally return zero for the empty
12926 Use -Warray-bounds for warnings.
12927 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12928 for non-constant array indices with COMPONENT_REF, arrays of
12929 arrays, and pointers to arrays.
12930 (gimple_fold_builtin_strlen): Determine and set length range for
12931 non-constant character arrays.
12933 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12935 PR middle-end/81897
12936 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12939 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12941 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12943 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12946 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12947 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12948 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12949 indexed_or_indirect_operand predicate.
12950 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12951 (*vsx_le_perm_load_v8hi): Likewise.
12952 (*vsx_le_perm_load_v16qi): Likewise.
12953 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12954 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12955 (*vsx_le_perm_store_v8hi): Likewise.
12956 (*vsx_le_perm_store_v16qi): Likewise.
12957 (eight unnamed splitters): Likewise.
12959 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12961 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12962 * config/rs6000/emmintrin.h: Likewise.
12963 * config/rs6000/mmintrin.h: Likewise.
12964 * config/rs6000/xmmintrin.h: Likewise.
12966 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12969 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12971 * tree.c (tree_nop_conversion): Return true for location wrapper
12973 (maybe_wrap_with_location): New function.
12974 (selftest::check_strip_nops): New function.
12975 (selftest::test_location_wrappers): New function.
12976 (selftest::tree_c_tests): Call it.
12977 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12978 (maybe_wrap_with_location): New decl.
12979 (EXPR_LOCATION_WRAPPER_P): New macro.
12980 (location_wrapper_p): New inline function.
12981 (tree_strip_any_location_wrapper): New inline function.
12983 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12986 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12987 stack_realign_offset for the largest alignment of stack slot
12989 (ix86_find_max_used_stack_alignment): New function.
12990 (ix86_finalize_stack_frame_flags): Use it. Set
12991 max_used_stack_alignment if we don't realign stack.
12992 * config/i386/i386.h (machine_function): Add
12993 max_used_stack_alignment.
12995 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12997 * config/arm/arm.opt (-mbranch-cost): New option.
12998 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
13001 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
13004 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
13005 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
13007 2018-01-10 Richard Biener <rguenther@suse.de>
13010 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
13011 early out so it also covers the case where we have a non-NULL
13014 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13016 PR tree-optimization/83753
13017 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
13018 for non-strided grouped accesses if the number of elements is 1.
13020 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
13023 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
13024 * i386.h (TARGET_USE_GATHER): Define.
13025 * x86-tune.def (X86_TUNE_USE_GATHER): New.
13027 2018-01-10 Martin Liska <mliska@suse.cz>
13030 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
13031 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
13033 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
13034 CLEANUP_NO_PARTITIONING is not set.
13036 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
13038 * doc/rtl.texi: Remove documentation of (const ...) wrappers
13039 for vectors, as a partial revert of r254296.
13040 * rtl.h (const_vec_p): Delete.
13041 (const_vec_duplicate_p): Don't test for vector CONSTs.
13042 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
13043 * expmed.c (make_tree): Likewise.
13046 * common.md (E, F): Use CONSTANT_P instead of checking for
13048 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
13049 checking for CONST_VECTOR.
13051 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13053 PR middle-end/83575
13054 * predict.c (force_edge_cold): Handle in more sane way edges
13055 with no prediction.
13057 2018-01-09 Carl Love <cel@us.ibm.com>
13059 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13061 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13062 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13063 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13064 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13065 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13066 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13067 * config/rs6000/rs6000-protos.h: Add extern defition for
13068 rs6000_generate_float2_double_code.
13069 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13071 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13072 (float2_v2df): Add define_expand.
13074 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13077 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13078 op_mode in the force_to_mode call.
13080 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13082 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13083 instead of checking each element individually.
13084 (aarch64_evpc_uzp): Likewise.
13085 (aarch64_evpc_zip): Likewise.
13086 (aarch64_evpc_ext): Likewise.
13087 (aarch64_evpc_rev): Likewise.
13088 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13089 instead of checking each element individually. Return true without
13091 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13092 whether all selected elements come from the same input, instead of
13093 checking each element individually. Remove calls to gen_rtx_REG,
13094 start_sequence and end_sequence and instead assert that no rtl is
13097 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13099 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13100 order of HIGH and CONST checks.
13102 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13104 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13105 if the destination isn't an SSA_NAME.
13107 2018-01-09 Richard Biener <rguenther@suse.de>
13109 PR tree-optimization/83668
13110 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13112 (canonicalize_loop_form): ... here, renamed from ...
13113 (canonicalize_loop_closed_ssa_form): ... this and amended to
13114 swap successor edges for loop exit blocks to make us use
13115 the RPO order we need for initial schedule generation.
13117 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13119 PR tree-optimization/64811
13120 * match.pd: When optimizing comparisons with Inf, avoid
13121 introducing or losing exceptions from comparisons with NaN.
13123 2018-01-09 Martin Liska <mliska@suse.cz>
13126 * asan.c (shadow_mem_size): Add gcc_assert.
13128 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13130 Don't save registers in main().
13133 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13134 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13135 * config/avr/avr.c (avr_set_current_function): Don't error if
13136 naked, OS_task or OS_main are specified at the same time.
13137 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13139 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13141 * common/config/avr/avr-common.c (avr_option_optimization_table):
13142 Switch on -mmain-is-OS_task for optimizing compilations.
13144 2018-01-09 Richard Biener <rguenther@suse.de>
13146 PR tree-optimization/83572
13147 * graphite.c: Include cfganal.h.
13148 (graphite_transform_loops): Connect infinite loops to exit
13149 and remove fake edges at the end.
13151 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13153 * ipa-inline.c (edge_badness): Revert accidental checkin.
13155 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13158 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13159 symbols; not inline clones.
13161 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13164 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13165 hard registers. Formatting fixes.
13167 PR preprocessor/83722
13168 * gcc.c (try_generate_repro): Pass
13169 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13170 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13173 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13174 Kito Cheng <kito.cheng@gmail.com>
13176 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13177 (riscv_leaf_function_p): Delete.
13178 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13180 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13182 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13184 (do_ifelse): New function.
13185 (do_isel): New function.
13186 (do_sub3): New function.
13187 (do_add3): New function.
13188 (do_load_mask_compare): New function.
13189 (do_overlap_load_compare): New function.
13190 (expand_compare_loop): New function.
13191 (expand_block_compare): Call expand_compare_loop() when appropriate.
13192 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13193 option description.
13194 (-mblock-compare-inline-loop-limit): New option.
13196 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13199 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13200 Reverse order of second and third operands in first alternative.
13201 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13202 of first and second elements in UNSPEC_VPERMR vector.
13203 (altivec_expand_vec_perm_le): Likewise.
13205 2018-01-08 Jeff Law <law@redhat.com>
13207 PR rtl-optimizatin/81308
13208 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13209 (process_switch): If group_case_labels makes a change, then set
13211 (pass_convert_switch::execute): If a switch is converted, then
13212 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13214 PR rtl-optimization/81308
13215 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13218 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13220 PR target/83663 - Revert r255946
13221 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13222 generation for cases where splatting a value is not useful.
13223 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13224 across a vec_duplicate and a paradoxical subreg forming a vector
13225 mode to a vec_concat.
13227 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13229 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13230 -march=armv8.3-a variants.
13231 * config/arm/t-multilib: Likewise.
13232 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13234 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13236 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13238 (cceq_ior_compare_complement): Give it a name so I can use it, and
13239 change boolean_or_operator predicate to boolean_operator so it can
13240 be used to generate a crand.
13241 (eqne): New code iterator.
13242 (bd/bd_neg): New code_attrs.
13243 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13244 a single define_insn.
13245 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13246 decrement (bdnzt/bdnzf/bdzt/bdzf).
13247 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13248 with the new names of the branch decrement patterns, and added the
13249 names of the branch decrement conditional patterns.
13251 2018-01-08 Richard Biener <rguenther@suse.de>
13253 PR tree-optimization/83563
13254 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13257 2018-01-08 Richard Biener <rguenther@suse.de>
13259 PR middle-end/83713
13260 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13262 2018-01-08 Richard Biener <rguenther@suse.de>
13264 PR tree-optimization/83685
13265 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13266 references to abnormals.
13268 2018-01-08 Richard Biener <rguenther@suse.de>
13271 * dwarf2out.c (output_indirect_strings): Handle empty
13272 skeleton_debug_str_hash.
13273 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13275 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13277 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13278 (emit_store_direct): Likewise.
13279 (arc_trampoline_adjust_address): Likewise.
13280 (arc_asm_trampoline_template): New function.
13281 (arc_initialize_trampoline): Use asm_trampoline_template.
13282 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13283 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13284 * config/arc/arc.md (flush_icache): Delete pattern.
13286 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13288 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13289 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13292 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13295 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13296 by not USED_FOR_TARGET.
13297 (make_pass_resolve_sw_modes): Likewise.
13299 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13301 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13304 2018-01-08 Richard Biener <rguenther@suse.de>
13306 PR middle-end/83580
13307 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13309 2018-01-08 Richard Biener <rguenther@suse.de>
13311 PR middle-end/83517
13312 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13314 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13316 PR middle-end/81897
13317 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13318 basic blocks with a small number of successors.
13319 (convert_control_dep_chain_into_preds): Improve handling of
13321 (dump_predicates): Split apart into...
13322 (dump_pred_chain): ...here...
13323 (dump_pred_info): ...and here.
13324 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13325 (can_chain_union_be_invalidated_p): Improve check for invalidation
13327 (uninit_uses_cannot_happen): Avoid unnecessary if
13328 convert_control_dep_chain_into_preds yielded nothing.
13330 2018-01-06 Martin Sebor <msebor@redhat.com>
13332 PR tree-optimization/83640
13333 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13334 subtracting negative offset from size.
13335 (builtin_access::overlap): Adjust offset bounds of the access to fall
13336 within the size of the object if possible.
13338 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13340 PR rtl-optimization/83699
13341 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13342 extract_bit_field_as_subreg to cases in which the extracted
13343 value is also a vector.
13345 * lra-constraints.c (process_alt_operands): Test for the equivalence
13346 substitutions when detecting a possible reload cycle.
13348 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13351 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13352 by default if flag_selective_schedling{,2}. Formatting fixes.
13354 PR rtl-optimization/83682
13355 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13356 if it has non-VECTOR_MODE element mode.
13357 (vec_duplicate_p): Likewise.
13359 PR middle-end/83694
13360 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13361 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13363 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13366 * config/i386/i386-builtin.def
13367 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13368 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13369 Require also OPTION_MASK_ISA_AVX512F in addition to
13370 OPTION_MASK_ISA_GFNI.
13371 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13372 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13373 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13374 to OPTION_MASK_ISA_GFNI.
13375 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13376 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13377 OPTION_MASK_ISA_AVX512BW.
13378 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13379 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13380 addition to OPTION_MASK_ISA_GFNI.
13381 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13382 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13383 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13384 to OPTION_MASK_ISA_GFNI.
13385 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13386 a requirement for all ISAs rather than any of them with a few
13388 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13390 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13391 bitmasks to be enabled with 3 exceptions, instead of requiring any
13392 enabled ISA with lots of exceptions.
13393 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13394 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13395 Change avx512bw in isa attribute to avx512f.
13396 * config/i386/sgxintrin.h: Add license boilerplate.
13397 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13398 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13399 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13400 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13402 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13403 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13404 temporarily sse2 rather than sse if not enabled already.
13407 * config/i386/sse.md (VI248_VLBW): Rename to ...
13408 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13409 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13410 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13411 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13412 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13413 mode iterator instead of VI248_VLBW.
13415 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13417 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13418 (record_modified): Skip clobbers; add debug output.
13419 (param_change_prob): Use sreal frequencies.
13421 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13423 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13424 punt for user-aligned variables.
13426 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13428 * tree-chrec.c (chrec_contains_symbols): Return true for
13431 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13434 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13435 of (x|y) == x for BICS pattern.
13437 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13439 PR tree-optimization/83605
13440 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13441 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13444 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13446 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13447 * config/epiphany/rtems.h: New file.
13449 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13450 Uros Bizjak <ubizjak@gmail.com>
13453 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13454 QIreg_operand instead of register_operand predicate.
13455 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13456 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13457 comments instead of -fmitigate[-_]rop.
13459 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13462 * cgraphunit.c (symbol_table::compile): Switch to text_section
13463 before calling assembly_start debug hook.
13464 * run-rtl-passes.c (run_rtl_passes): Likewise.
13467 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13469 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13470 range_int_cst_p rather than !symbolic_range_p before calling
13471 extract_range_from_multiplicative_op_1.
13473 2018-01-04 Jeff Law <law@redhat.com>
13475 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13476 redundant test in assertion.
13478 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13480 * doc/rtl.texi: Document machine_mode wrapper classes.
13482 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13484 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13485 using tree_to_uhwi.
13487 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13489 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13490 the VEC_PERM_EXPR fold to fail.
13492 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13495 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13496 to switched_sections.
13498 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13501 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13502 test for d.testing.
13504 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13507 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13508 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13510 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13513 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13514 is BLKmode and bitpos not zero or mode change is needed.
13516 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13519 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13522 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13525 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13526 instead of MULT rtx. Update all corresponding splitters.
13527 (*saddl_se): Ditto.
13528 (*ssub<modesuffix>): Ditto.
13529 (*ssubl_se): Ditto.
13530 (*cmp_sadd_di): Update split patterns.
13531 (*cmp_sadd_si): Ditto.
13532 (*cmp_sadd_sidi): Ditto.
13533 (*cmp_ssub_di): Ditto.
13534 (*cmp_ssub_si): Ditto.
13535 (*cmp_ssub_sidi): Ditto.
13536 * config/alpha/predicates.md (const23_operand): New predicate.
13537 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13538 Look for ASHIFT, not MULT inner operand.
13539 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13541 2018-01-04 Martin Liska <mliska@suse.cz>
13543 PR gcov-profile/83669
13544 * gcov.c (output_intermediate_file): Add version to intermediate
13546 * doc/gcov.texi: Document new field 'version' in intermediate
13547 file format. Fix location of '-k' option of gcov command.
13549 2018-01-04 Martin Liska <mliska@suse.cz>
13552 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13554 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13556 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13558 2018-01-03 Martin Sebor <msebor@redhat.com>
13560 PR tree-optimization/83655
13561 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13562 checking calls with invalid arguments.
13564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13566 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13567 (vectorizable_mask_load_store): Delete.
13568 (vectorizable_call): Return false for masked loads and stores.
13569 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13570 instead of gimple_assign_rhs1.
13571 (vectorizable_load): Handle IFN_MASK_LOAD.
13572 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13576 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13578 (vectorizable_mask_load_store): ...here.
13579 (vectorizable_load): ...and here.
13581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13583 * tree-vect-stmts.c (vect_build_all_ones_mask)
13584 (vect_build_zero_merge_argument): New functions, split out from...
13585 (vectorizable_load): ...here.
13587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13589 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13591 (vectorizable_mask_load_store): ...here.
13592 (vectorizable_store): ...and here.
13594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13596 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13598 (vectorizable_mask_load_store): ...here.
13600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13602 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13603 (vect_model_store_cost): Take a vec_load_store_type instead of a
13605 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13606 (vect_model_store_cost): Take a vec_load_store_type instead of a
13608 (vectorizable_mask_load_store): Update accordingly.
13609 (vectorizable_store): Likewise.
13610 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13614 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13615 IFN_MASK_LOAD calls here rather than...
13616 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13619 Alan Hayward <alan.hayward@arm.com>
13620 David Sherwood <david.sherwood@arm.com>
13622 * expmed.c (extract_bit_field_1): For vector extracts,
13623 fall back to extract_bit_field_as_subreg if vec_extract
13626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13627 Alan Hayward <alan.hayward@arm.com>
13628 David Sherwood <david.sherwood@arm.com>
13630 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13631 they are variable or constant sized.
13632 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13633 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13640 handling COND_EXPRs with boolean comparisons, try to find a better
13641 basis for the mask type than the boolean itself.
13643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13645 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13646 is calculated and how it can be overridden.
13647 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13648 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13650 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13654 Alan Hayward <alan.hayward@arm.com>
13655 David Sherwood <david.sherwood@arm.com>
13657 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13658 Remove the mode argument.
13659 (aarch64_simd_valid_immediate): Remove the mode and inverse
13661 * config/aarch64/iterators.md (bitsize): New iterator.
13662 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13663 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13664 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13665 aarch64_simd_valid_immediate.
13666 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13667 (aarch64_reg_or_bic_imm): Likewise.
13668 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13669 with an insn_type enum and msl with a modifier_type enum.
13670 Replace element_width with a scalar_mode. Change the shift
13671 to unsigned int. Add constructors for scalar_float_mode and
13672 scalar_int_mode elements.
13673 (aarch64_vect_float_const_representable_p): Delete.
13674 (aarch64_can_const_movi_rtx_p)
13675 (aarch64_simd_scalar_immediate_valid_for_move)
13676 (aarch64_simd_make_constant): Update call to
13677 aarch64_simd_valid_immediate.
13678 (aarch64_advsimd_valid_immediate_hs): New function.
13679 (aarch64_advsimd_valid_immediate): Likewise.
13680 (aarch64_simd_valid_immediate): Remove mode and inverse
13681 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13682 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13683 and aarch64_float_const_representable_p on the result.
13684 (aarch64_output_simd_mov_immediate): Remove mode argument.
13685 Update call to aarch64_simd_valid_immediate and use of
13686 simd_immediate_info.
13687 (aarch64_output_scalar_simd_mov_immediate): Update call
13690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13691 Alan Hayward <alan.hayward@arm.com>
13692 David Sherwood <david.sherwood@arm.com>
13694 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13695 (mode_nunits): Likewise CONST_MODE_NUNITS.
13696 * machmode.def (ADJUST_NUNITS): Document.
13697 * genmodes.c (mode_data::need_nunits_adj): New field.
13698 (blank_mode): Update accordingly.
13699 (adj_nunits): New variable.
13700 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13702 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13703 listed in adj_nunits.
13704 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13705 listed in adj_nunits. Don't emit case statements for such modes.
13706 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13707 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13708 nothing if adj_nunits is nonnull.
13709 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13710 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13711 (emit_mode_fbit): Update use of print_maybe_const_decl.
13712 (emit_move_size): Likewise. Treat the array as non-const
13714 (emit_mode_adjustments): Handle adj_nunits.
13716 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13718 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13719 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13720 (VECTOR_MODES): Use it.
13721 (make_vector_modes): Take the prefix as an argument.
13723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13724 Alan Hayward <alan.hayward@arm.com>
13725 David Sherwood <david.sherwood@arm.com>
13727 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13728 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13729 for MODE_VECTOR_BOOL.
13730 * machmode.def (VECTOR_BOOL_MODE): Document.
13731 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13732 (make_vector_bool_mode): New function.
13733 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13735 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13736 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13738 * stor-layout.c (int_mode_for_mode): Likewise.
13739 * tree.c (build_vector_type_for_mode): Likewise.
13740 * varasm.c (output_constant_pool_2): Likewise.
13741 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13742 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13743 for MODE_VECTOR_BOOL.
13744 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13745 of mode class checks.
13746 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13747 instead of a list of mode class checks.
13748 (expand_vector_scalar_condition): Likewise.
13749 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13751 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13752 Alan Hayward <alan.hayward@arm.com>
13753 David Sherwood <david.sherwood@arm.com>
13755 * machmode.h (mode_size): Change from unsigned short to
13757 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13758 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13759 or if measurement_type is not polynomial.
13760 (fixed_size_mode::includes_p): Check for constant-sized modes.
13761 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13762 return a poly_uint16 rather than an unsigned short.
13763 (emit_mode_size): Change the type of mode_size from unsigned short
13764 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13765 (emit_mode_adjustments): Cope with polynomial vector sizes.
13766 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13768 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13770 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13771 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13772 * caller-save.c (setup_save_areas): Likewise.
13773 (replace_reg_with_saved_mem): Likewise.
13774 * calls.c (emit_library_call_value_1): Likewise.
13775 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13776 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13777 (gen_lowpart_for_combine): Likewise.
13778 * convert.c (convert_to_integer_1): Likewise.
13779 * cse.c (equiv_constant, cse_insn): Likewise.
13780 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13781 (cselib_subst_to_values): Likewise.
13782 * dce.c (word_dce_process_block): Likewise.
13783 * df-problems.c (df_word_lr_mark_ref): Likewise.
13784 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13785 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13786 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13787 (rtl_for_decl_location): Likewise.
13788 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13789 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13790 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13791 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13792 (expand_expr_real_1): Likewise.
13793 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13794 (pad_below): Likewise.
13795 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13796 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13797 * ira.c (get_subreg_tracking_sizes): Likewise.
13798 * ira-build.c (ira_create_allocno_objects): Likewise.
13799 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13800 (ira_sort_regnos_for_alter_reg): Likewise.
13801 * ira-costs.c (record_operand_costs): Likewise.
13802 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13803 (resolve_simple_move): Likewise.
13804 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13805 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13806 (lra_constraints): Likewise.
13807 (CONST_POOL_OK_P): Reject variable-sized modes.
13808 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13809 (add_pseudo_to_slot, lra_spill): Likewise.
13810 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13811 * optabs-query.c (get_best_extraction_insn): Likewise.
13812 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13813 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13814 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13815 * recog.c (offsettable_address_addr_space_p): Likewise.
13816 * regcprop.c (maybe_mode_change): Likewise.
13817 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13818 * regrename.c (build_def_use): Likewise.
13819 * regstat.c (dump_reg_info): Likewise.
13820 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13821 (find_reloads, find_reloads_subreg_address): Likewise.
13822 * reload1.c (eliminate_regs_1): Likewise.
13823 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13824 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13825 (simplify_binary_operation_1, simplify_subreg): Likewise.
13826 * targhooks.c (default_function_arg_padding): Likewise.
13827 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13828 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13829 (verify_gimple_assign_ternary): Likewise.
13830 * tree-inline.c (estimate_move_cost): Likewise.
13831 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13832 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13833 (get_address_cost_ainc): Likewise.
13834 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13835 (vect_supportable_dr_alignment): Likewise.
13836 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13837 (vectorizable_reduction): Likewise.
13838 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13839 (vectorizable_operation, vectorizable_load): Likewise.
13840 * tree.c (build_same_sized_truth_vector_type): Likewise.
13841 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13842 * var-tracking.c (emit_note_insn_var_location): Likewise.
13843 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13844 (ADDR_VEC_ALIGN): Likewise.
13846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13847 Alan Hayward <alan.hayward@arm.com>
13848 David Sherwood <david.sherwood@arm.com>
13850 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13852 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13853 or if measurement_type is polynomial.
13854 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13855 * combine.c (make_extraction): Likewise.
13856 * dse.c (find_shift_sequence): Likewise.
13857 * dwarf2out.c (mem_loc_descriptor): Likewise.
13858 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13859 (extract_bit_field, extract_low_bits): Likewise.
13860 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13861 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13862 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13863 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13864 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13865 * reload.c (find_reloads): Likewise.
13866 * reload1.c (alter_reg): Likewise.
13867 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13868 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13869 * tree-if-conv.c (predicate_mem_writes): Likewise.
13870 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13871 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13872 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13873 * valtrack.c (dead_debug_insert_temp): Likewise.
13874 * varasm.c (mergeable_constant_section): Likewise.
13875 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13878 Alan Hayward <alan.hayward@arm.com>
13879 David Sherwood <david.sherwood@arm.com>
13881 * expr.c (expand_assignment): Cope with polynomial mode sizes
13882 when assigning to a CONCAT.
13884 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13885 Alan Hayward <alan.hayward@arm.com>
13886 David Sherwood <david.sherwood@arm.com>
13888 * machmode.h (mode_precision): Change from unsigned short to
13890 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13892 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13893 or if measurement_type is not polynomial.
13894 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13895 in which the mode is already known to be a scalar_int_mode.
13896 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13897 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13899 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13900 for GET_MODE_PRECISION.
13901 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13902 for GET_MODE_PRECISION.
13903 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13905 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13906 (expand_field_assignment, make_extraction): Likewise.
13907 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13908 (get_last_value): Likewise.
13909 * convert.c (convert_to_integer_1): Likewise.
13910 * cse.c (cse_insn): Likewise.
13911 * expr.c (expand_expr_real_1): Likewise.
13912 * lra-constraints.c (simplify_operand_subreg): Likewise.
13913 * optabs-query.c (can_atomic_load_p): Likewise.
13914 * optabs.c (expand_atomic_load): Likewise.
13915 (expand_atomic_store): Likewise.
13916 * ree.c (combine_reaching_defs): Likewise.
13917 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13918 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13919 * tree.h (type_has_mode_precision_p): Likewise.
13920 * ubsan.c (instrument_si_overflow): Likewise.
13922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13923 Alan Hayward <alan.hayward@arm.com>
13924 David Sherwood <david.sherwood@arm.com>
13926 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13927 polynomial numbers of units.
13928 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13929 (valid_vector_subparts_p): New function.
13930 (build_vector_type): Remove temporary shim and take the number
13931 of units as a poly_uint64 rather than an int.
13932 (build_opaque_vector_type): Take the number of units as a
13933 poly_uint64 rather than an int.
13934 * tree.c (build_vector_from_ctor): Handle polynomial
13935 TYPE_VECTOR_SUBPARTS.
13936 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13937 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13938 (build_vector_from_val): If the number of units is variable,
13939 use build_vec_duplicate_cst for constant operands and
13940 VEC_DUPLICATE_EXPR otherwise.
13941 (make_vector_type): Remove temporary is_constant ().
13942 (build_vector_type, build_opaque_vector_type): Take the number of
13943 units as a poly_uint64 rather than an int.
13944 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13946 * cfgexpand.c (expand_debug_expr): Likewise.
13947 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13948 (store_constructor, expand_expr_real_1): Likewise.
13949 (const_scalar_mask_from_tree): Likewise.
13950 * fold-const-call.c (fold_const_reduction): Likewise.
13951 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13952 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13953 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13954 (fold_relational_const): Likewise.
13955 (native_interpret_vector): Likewise. Change the size from an
13956 int to an unsigned int.
13957 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13958 TYPE_VECTOR_SUBPARTS.
13959 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13960 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13961 duplicating a non-constant operand into a variable-length vector.
13962 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13963 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13964 * ipa-icf.c (sem_variable::equals): Likewise.
13965 * match.pd: Likewise.
13966 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13967 * print-tree.c (print_node): Likewise.
13968 * stor-layout.c (layout_type): Likewise.
13969 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13970 * tree-cfg.c (verify_gimple_comparison): Likewise.
13971 (verify_gimple_assign_binary): Likewise.
13972 (verify_gimple_assign_ternary): Likewise.
13973 (verify_gimple_assign_single): Likewise.
13974 * tree-pretty-print.c (dump_generic_node): Likewise.
13975 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13976 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13977 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13978 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13979 (vect_shift_permute_load_chain): Likewise.
13980 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13981 (expand_vector_condition, optimize_vector_constructor): Likewise.
13982 (lower_vec_perm, get_compute_type): Likewise.
13983 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13984 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13985 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13986 (vect_recog_mask_conversion_pattern): Likewise.
13987 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13988 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13989 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13990 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13991 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13992 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13993 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13994 (supportable_widening_operation): Likewise.
13995 (supportable_narrowing_operation): Likewise.
13996 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13998 * varasm.c (output_constant): Likewise.
14000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14001 Alan Hayward <alan.hayward@arm.com>
14002 David Sherwood <david.sherwood@arm.com>
14004 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
14005 so that both the length == 3 and length != 3 cases set up their
14006 own permute vectors. Add comments explaining why we know the
14007 number of elements is constant.
14008 (vect_permute_load_chain): Likewise.
14010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14011 Alan Hayward <alan.hayward@arm.com>
14012 David Sherwood <david.sherwood@arm.com>
14014 * machmode.h (mode_nunits): Change from unsigned char to
14016 (ONLY_FIXED_SIZE_MODES): New macro.
14017 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
14018 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
14019 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
14021 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
14022 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
14023 or if measurement_type is not polynomial.
14024 * genmodes.c (ZERO_COEFFS): New macro.
14025 (emit_mode_nunits_inline): Make mode_nunits_inline return a
14027 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
14028 Use ZERO_COEFFS when emitting initializers.
14029 * data-streamer.h (bp_pack_poly_value): New function.
14030 (bp_unpack_poly_value): Likewise.
14031 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
14032 for GET_MODE_NUNITS.
14033 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
14034 for GET_MODE_NUNITS.
14035 * tree.c (make_vector_type): Remove temporary shim and make
14036 the real function take the number of units as a poly_uint64
14037 rather than an int.
14038 (build_vector_type_for_mode): Handle polynomial nunits.
14039 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
14040 * emit-rtl.c (const_vec_series_p_1): Likewise.
14041 (gen_rtx_CONST_VECTOR): Likewise.
14042 * fold-const.c (test_vec_duplicate_folding): Likewise.
14043 * genrecog.c (validate_pattern): Likewise.
14044 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
14045 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
14046 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
14047 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
14048 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
14049 * rtlanal.c (subreg_get_info): Likewise.
14050 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14051 (vect_grouped_load_supported): Likewise.
14052 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14053 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14054 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14055 (simplify_const_unary_operation, simplify_binary_operation_1)
14056 (simplify_const_binary_operation, simplify_ternary_operation)
14057 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14058 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14059 instead of CONST_VECTOR_NUNITS.
14060 * varasm.c (output_constant_pool_2): Likewise.
14061 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14062 explicit-encoded elements in the XVEC for variable-length vectors.
14064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14066 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14069 Alan Hayward <alan.hayward@arm.com>
14070 David Sherwood <david.sherwood@arm.com>
14072 * coretypes.h (fixed_size_mode): Declare.
14073 (fixed_size_mode_pod): New typedef.
14074 * builtins.h (target_builtins::x_apply_args_mode)
14075 (target_builtins::x_apply_result_mode): Change type to
14076 fixed_size_mode_pod.
14077 * builtins.c (apply_args_size, apply_result_size, result_vector)
14078 (expand_builtin_apply_args_1, expand_builtin_apply)
14079 (expand_builtin_return): Update accordingly.
14081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14083 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14084 * cselib.c (cselib_hash_rtx): Likewise.
14085 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14086 CONST_VECTOR encoding.
14088 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14089 Jeff Law <law@redhat.com>
14092 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14093 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14094 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14095 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14098 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14099 explicitly probe *sp in a noreturn function if there were any callee
14100 register saves or frame pointer is needed.
14102 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14105 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14106 BLKmode for ternary, binary or unary expressions.
14109 * var-tracking.c (delete_vta_debug_insn): New inline function.
14110 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14111 insns from get_insns () to NULL instead of each bb separately.
14112 Use delete_vta_debug_insn. No longer static.
14113 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14114 delete_vta_debug_insns callers.
14115 * rtl.h (delete_vta_debug_insns): Declare.
14116 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14117 instead of variable_tracking_main.
14119 2018-01-03 Martin Sebor <msebor@redhat.com>
14121 PR tree-optimization/83603
14122 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14123 arguments past the endof the argument list in functions declared
14124 without a prototype.
14125 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14126 Avoid checking when arguments are null.
14128 2018-01-03 Martin Sebor <msebor@redhat.com>
14131 * doc/extend.texi (attribute const): Fix a typo.
14132 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14133 issuing -Wsuggest-attribute for void functions.
14135 2018-01-03 Martin Sebor <msebor@redhat.com>
14137 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14138 offset_int::from instead of wide_int::to_shwi.
14139 (maybe_diag_overlap): Remove assertion.
14140 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14141 * gimple-ssa-sprintf.c (format_directive): Same.
14142 (parse_directive): Same.
14143 (sprintf_dom_walker::compute_format_length): Same.
14144 (try_substitute_return_value): Same.
14146 2018-01-03 Jeff Law <law@redhat.com>
14148 PR middle-end/83654
14149 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14150 non-constant residual for zero at runtime and avoid probing in
14151 that case. Reorganize code for trailing problem to mirror handling
14154 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14156 PR tree-optimization/83501
14157 * tree-ssa-strlen.c (get_string_cst): New.
14158 (handle_char_store): Call get_string_cst.
14160 2018-01-03 Martin Liska <mliska@suse.cz>
14162 PR tree-optimization/83593
14163 * tree-ssa-strlen.c: Include tree-cfg.h.
14164 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14165 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14166 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14168 (strlen_dom_walker::before_dom_children): Call
14169 gimple_purge_dead_eh_edges. Dump tranformation with details
14171 (strlen_dom_walker::before_dom_children): Update call by adding
14172 new argument cleanup_eh.
14173 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14175 2018-01-03 Martin Liska <mliska@suse.cz>
14178 * cif-code.def (VARIADIC_THUNK): New enum value.
14179 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14182 2018-01-03 Jan Beulich <jbeulich@suse.com>
14184 * sse.md (mov<mode>_internal): Tighten condition for when to use
14185 vmovdqu<ssescalarsize> for TI and OI modes.
14187 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14189 Update copyright years.
14191 2018-01-03 Martin Liska <mliska@suse.cz>
14194 * ipa-visibility.c (function_and_variable_visibility): Skip
14195 functions with noipa attribure.
14197 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14199 * gcc.c (process_command): Update copyright notice dates.
14200 * gcov-dump.c (print_version): Ditto.
14201 * gcov.c (print_version): Ditto.
14202 * gcov-tool.c (print_version): Ditto.
14203 * gengtype.c (create_file): Ditto.
14204 * doc/cpp.texi: Bump @copying's copyright year.
14205 * doc/cppinternals.texi: Ditto.
14206 * doc/gcc.texi: Ditto.
14207 * doc/gccint.texi: Ditto.
14208 * doc/gcov.texi: Ditto.
14209 * doc/install.texi: Ditto.
14210 * doc/invoke.texi: Ditto.
14212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14214 * vector-builder.h (vector_builder::m_full_nelts): Change from
14215 unsigned int to poly_uint64.
14216 (vector_builder::full_nelts): Update prototype accordingly.
14217 (vector_builder::new_vector): Likewise.
14218 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14219 (vector_builder::operator ==): Likewise.
14220 (vector_builder::finalize): Likewise.
14221 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14222 Take the number of elements as a poly_uint64 rather than an
14224 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14225 from unsigned int to poly_uint64.
14226 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14227 (vec_perm_indices::new_vector): Likewise.
14228 (vec_perm_indices::length): Likewise.
14229 (vec_perm_indices::nelts_per_input): Likewise.
14230 (vec_perm_indices::input_nelts): Likewise.
14231 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14232 number of elements per input as a poly_uint64 rather than an
14233 unsigned int. Use the original encoding for variable-length
14234 vectors, rather than clamping each individual element.
14235 For the second and subsequent elements in each pattern,
14236 clamp the step and base before clamping their sum.
14237 (vec_perm_indices::series_p): Handle polynomial element counts.
14238 (vec_perm_indices::all_in_range_p): Likewise.
14239 (vec_perm_indices_to_tree): Likewise.
14240 (vec_perm_indices_to_rtx): Likewise.
14241 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14242 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14243 (tree_vector_builder::new_binary_operation): Handle polynomial
14244 element counts. Return false if we need to know the number
14245 of elements at compile time.
14246 * fold-const.c (fold_vec_perm): Punt if the number of elements
14247 isn't known at compile time.
14249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14251 * vec-perm-indices.h (vec_perm_builder): Change element type
14252 from HOST_WIDE_INT to poly_int64.
14253 (vec_perm_indices::element_type): Update accordingly.
14254 (vec_perm_indices::clamp): Handle polynomial element_types.
14255 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14256 (vec_perm_indices::all_in_range_p): Likewise.
14257 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14259 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14260 polynomial vec_perm_indices element types.
14261 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14262 * fold-const.c (fold_vec_perm): Likewise.
14263 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14264 * tree-vect-generic.c (lower_vec_perm): Likewise.
14265 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14266 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14267 element type to HOST_WIDE_INT.
14269 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14270 Alan Hayward <alan.hayward@arm.com>
14271 David Sherwood <david.sherwood@arm.com>
14273 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14274 rather than an int. Use plus_constant.
14275 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14276 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
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 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14283 a HOST_WIDE_INT to a poly_int64.
14285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14286 Alan Hayward <alan.hayward@arm.com>
14287 David Sherwood <david.sherwood@arm.com>
14289 * calls.c (load_register_parameters): Cope with polynomial
14290 mode sizes. Require a constant size for BLKmode parameters
14291 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14292 forces a parameter to be padded at the lsb end in order to
14293 fill a complete number of words, require the parameter size
14294 to be ordered wrt UNITS_PER_WORD.
14296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14297 Alan Hayward <alan.hayward@arm.com>
14298 David Sherwood <david.sherwood@arm.com>
14300 * reload1.c (spill_stack_slot_width): Change element type
14301 from unsigned int to poly_uint64_pod.
14302 (alter_reg): Treat mode sizes as polynomial.
14304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14305 Alan Hayward <alan.hayward@arm.com>
14306 David Sherwood <david.sherwood@arm.com>
14308 * reload.c (complex_word_subreg_p): New function.
14309 (reload_inner_reg_of_subreg, push_reload): Use it.
14311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14312 Alan Hayward <alan.hayward@arm.com>
14313 David Sherwood <david.sherwood@arm.com>
14315 * lra-constraints.c (process_alt_operands): Reject matched
14316 operands whose sizes aren't ordered.
14317 (match_reload): Refer to this check here.
14319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14320 Alan Hayward <alan.hayward@arm.com>
14321 David Sherwood <david.sherwood@arm.com>
14323 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14324 that the mode size is in the set {1, 2, 4, 8, 16}.
14326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14327 Alan Hayward <alan.hayward@arm.com>
14328 David Sherwood <david.sherwood@arm.com>
14330 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14331 Use plus_constant instead of gen_rtx_PLUS.
14333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14334 Alan Hayward <alan.hayward@arm.com>
14335 David Sherwood <david.sherwood@arm.com>
14337 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14338 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14339 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14340 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14341 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14342 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14343 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14344 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14345 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14346 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14348 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14349 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14350 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14351 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14352 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14353 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14354 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14355 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14356 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14357 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14359 * expr.c (emit_move_resolve_push): Treat the input and result
14360 of PUSH_ROUNDING as a poly_int64.
14361 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14362 (emit_push_insn): Likewise.
14363 * lra-eliminations.c (mark_not_eliminable): Likewise.
14364 * recog.c (push_operand): Likewise.
14365 * reload1.c (elimination_effects): Likewise.
14366 * rtlanal.c (nonzero_bits1): Likewise.
14367 * calls.c (store_one_arg): Likewise. Require the padding to be
14368 known at compile time.
14370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14371 Alan Hayward <alan.hayward@arm.com>
14372 David Sherwood <david.sherwood@arm.com>
14374 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14375 Use plus_constant instead of gen_rtx_PLUS.
14377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14378 Alan Hayward <alan.hayward@arm.com>
14379 David Sherwood <david.sherwood@arm.com>
14381 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14382 rather than an int.
14384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14385 Alan Hayward <alan.hayward@arm.com>
14386 David Sherwood <david.sherwood@arm.com>
14388 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14389 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14390 via stack temporaries. Treat the mode size as polynomial too.
14392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14393 Alan Hayward <alan.hayward@arm.com>
14394 David Sherwood <david.sherwood@arm.com>
14396 * expr.c (expand_expr_real_2): When handling conversions involving
14397 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14398 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14399 as a poly_uint64 too.
14401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14402 Alan Hayward <alan.hayward@arm.com>
14403 David Sherwood <david.sherwood@arm.com>
14405 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14408 Alan Hayward <alan.hayward@arm.com>
14409 David Sherwood <david.sherwood@arm.com>
14411 * combine.c (can_change_dest_mode): Handle polynomial
14412 REGMODE_NATURAL_SIZE.
14413 * expmed.c (store_bit_field_1): Likewise.
14414 * expr.c (store_constructor): Likewise.
14415 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14416 and polynomial REGMODE_NATURAL_SIZE.
14417 (gen_lowpart_common): Likewise.
14418 * reginfo.c (record_subregs_of_mode): Likewise.
14419 * rtlanal.c (read_modify_subreg_p): Likewise.
14421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14422 Alan Hayward <alan.hayward@arm.com>
14423 David Sherwood <david.sherwood@arm.com>
14425 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14426 numbers of elements.
14428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14429 Alan Hayward <alan.hayward@arm.com>
14430 David Sherwood <david.sherwood@arm.com>
14432 * match.pd: Cope with polynomial numbers of vector elements.
14434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14435 Alan Hayward <alan.hayward@arm.com>
14436 David Sherwood <david.sherwood@arm.com>
14438 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14439 in a POINTER_PLUS_EXPR.
14441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14442 Alan Hayward <alan.hayward@arm.com>
14443 David Sherwood <david.sherwood@arm.com>
14445 * omp-simd-clone.c (simd_clone_subparts): New function.
14446 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14447 (ipa_simd_modify_function_body): Likewise.
14449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14450 Alan Hayward <alan.hayward@arm.com>
14451 David Sherwood <david.sherwood@arm.com>
14453 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14454 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14455 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14456 (expand_vector_condition, vector_element): Likewise.
14457 (subparts_gt): New function.
14458 (get_compute_type): Use subparts_gt.
14459 (count_type_subparts): Delete.
14460 (expand_vector_operations_1): Use subparts_gt instead of
14461 count_type_subparts.
14463 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14464 Alan Hayward <alan.hayward@arm.com>
14465 David Sherwood <david.sherwood@arm.com>
14467 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14468 (vect_compile_time_alias): ...this new function. Do the calculation
14469 on poly_ints rather than trees.
14470 (vect_prune_runtime_alias_test_list): Update call accordingly.
14472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14473 Alan Hayward <alan.hayward@arm.com>
14474 David Sherwood <david.sherwood@arm.com>
14476 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14478 (vect_schedule_slp_instance): Likewise.
14480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14481 Alan Hayward <alan.hayward@arm.com>
14482 David Sherwood <david.sherwood@arm.com>
14484 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14485 constant and extern definitions for variable-length vectors.
14486 (vect_get_constant_vectors): Note that the number of units
14487 is known to be constant.
14489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14490 Alan Hayward <alan.hayward@arm.com>
14491 David Sherwood <david.sherwood@arm.com>
14493 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14494 of units as polynomial. Choose between WIDE and NARROW based
14497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14498 Alan Hayward <alan.hayward@arm.com>
14499 David Sherwood <david.sherwood@arm.com>
14501 * tree-vect-stmts.c (simd_clone_subparts): New function.
14502 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14505 Alan Hayward <alan.hayward@arm.com>
14506 David Sherwood <david.sherwood@arm.com>
14508 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14509 vectors as polynomial. Use build_index_vector for
14510 IFN_GOMP_SIMD_LANE.
14512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14513 Alan Hayward <alan.hayward@arm.com>
14514 David Sherwood <david.sherwood@arm.com>
14516 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14517 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14518 for variable-length vectors.
14519 (vectorizable_mask_load_store): Treat the number of units as
14520 polynomial, asserting that it is constant if the condition has
14521 already been enforced.
14522 (vectorizable_store, vectorizable_load): Likewise.
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-vect-loop.c (vectorizable_live_operation): Treat the number
14529 of units as polynomial. Punt if we can't tell at compile time
14530 which vector contains the final result.
14532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14533 Alan Hayward <alan.hayward@arm.com>
14534 David Sherwood <david.sherwood@arm.com>
14536 * tree-vect-loop.c (vectorizable_induction): Treat the number
14537 of units as polynomial. Punt on SLP inductions. Use an integer
14538 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14539 cast of such a series for variable-length floating-point
14542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14543 Alan Hayward <alan.hayward@arm.com>
14544 David Sherwood <david.sherwood@arm.com>
14546 * tree.h (build_index_vector): Declare.
14547 * tree.c (build_index_vector): New function.
14548 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14549 of units as polynomial, forcibly converting it to a constant if
14550 vectorizable_reduction has already enforced the condition.
14551 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14552 to create a {1,2,3,...} vector.
14553 (vectorizable_reduction): Treat the number of units as polynomial.
14554 Choose vectype_in based on the largest scalar element size rather
14555 than the smallest number of units. Enforce the restrictions
14558 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14559 Alan Hayward <alan.hayward@arm.com>
14560 David Sherwood <david.sherwood@arm.com>
14562 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14563 number of units as polynomial.
14565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14566 Alan Hayward <alan.hayward@arm.com>
14567 David Sherwood <david.sherwood@arm.com>
14569 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14570 * target.def (autovectorize_vector_sizes): Return the vector sizes
14571 by pointer, using vector_sizes rather than a bitmask.
14572 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14573 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14574 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14576 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14577 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14578 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14579 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14580 * omp-general.c (omp_max_vf): Likewise.
14581 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14582 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14583 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14584 * tree-vect-slp.c (vect_slp_bb): Likewise.
14585 * doc/tm.texi: Regenerate.
14586 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14588 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14589 the vector size as a poly_uint64 rather than an unsigned int.
14590 (current_vector_size): Change from an unsigned int to a poly_uint64.
14591 (get_vectype_for_scalar_type): Update accordingly.
14592 * tree.h (build_truth_vector_type): Take the size and number of
14593 units as a poly_uint64 rather than an unsigned int.
14594 (build_vector_type): Add a temporary overload that takes
14595 the number of units as a poly_uint64 rather than an unsigned int.
14596 * tree.c (make_vector_type): Likewise.
14597 (build_truth_vector_type): Take the number of units as a poly_uint64
14598 rather than an unsigned int.
14600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14601 Alan Hayward <alan.hayward@arm.com>
14602 David Sherwood <david.sherwood@arm.com>
14604 * target.def (get_mask_mode): Take the number of units and length
14605 as poly_uint64s rather than unsigned ints.
14606 * targhooks.h (default_get_mask_mode): Update accordingly.
14607 * targhooks.c (default_get_mask_mode): Likewise.
14608 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14609 * doc/tm.texi: Regenerate.
14611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14612 Alan Hayward <alan.hayward@arm.com>
14613 David Sherwood <david.sherwood@arm.com>
14615 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14616 * omp-general.c (omp_max_vf): Likewise.
14617 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14618 (expand_omp_simd): Handle polynomial safelen.
14619 * omp-low.c (omplow_simd_context): Add a default constructor.
14620 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14621 (lower_rec_simd_input_clauses): Update accordingly.
14622 (lower_rec_input_clauses): Likewise.
14624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14625 Alan Hayward <alan.hayward@arm.com>
14626 David Sherwood <david.sherwood@arm.com>
14628 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14629 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14630 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14631 (vect_analyze_slp_cost): Likewise.
14632 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14633 (vect_model_load_cost): Likewise.
14635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14636 Alan Hayward <alan.hayward@arm.com>
14637 David Sherwood <david.sherwood@arm.com>
14639 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14640 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14641 from an unsigned int * to a poly_uint64_pod *.
14642 (calculate_unrolling_factor): New function.
14643 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14646 Alan Hayward <alan.hayward@arm.com>
14647 David Sherwood <david.sherwood@arm.com>
14649 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14650 from an unsigned int to a poly_uint64.
14651 (_loop_vec_info::slp_unrolling_factor): Likewise.
14652 (_loop_vec_info::vectorization_factor): Change from an int
14654 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14655 (vect_get_num_vectors): New function.
14656 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14657 (vect_get_num_copies): Use vect_get_num_vectors.
14658 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14659 to an unsigned int *.
14660 (vect_analyze_data_refs): Change min_vf from an int * to a
14662 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14663 than an unsigned HOST_WIDE_INT.
14664 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14665 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14666 to an unsigned int *.
14667 (vect_analyze_data_ref_dependences): Likewise.
14668 (vect_compute_data_ref_alignment): Handle polynomial vf.
14669 (vect_enhance_data_refs_alignment): Likewise.
14670 (vect_prune_runtime_alias_test_list): Likewise.
14671 (vect_shift_permute_load_chain): Likewise.
14672 (vect_supportable_dr_alignment): Likewise.
14673 (dependence_distance_ge_vf): Take the vectorization factor as a
14674 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14675 (vect_analyze_data_refs): Change min_vf from an int * to a
14677 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14678 vfm1 as a poly_uint64 rather than an int. Make the same change
14679 for the returned bound_scalar.
14680 (vect_gen_vector_loop_niters): Handle polynomial vf.
14681 (vect_do_peeling): Likewise. Update call to
14682 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14683 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14685 * tree-vect-loop.c (vect_determine_vectorization_factor)
14686 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14687 (vect_get_known_peeling_cost): Likewise.
14688 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14689 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14690 (vect_transform_loop): Likewise. Use the lowest possible VF when
14691 updating the upper bounds of the loop.
14692 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14693 rather than an int.
14694 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14695 polynomial unroll factors.
14696 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14697 (vect_make_slp_decision): Likewise.
14698 (vect_supported_load_permutation_p): Likewise, and polynomial
14700 (vect_analyze_slp_cost): Handle polynomial vf.
14701 (vect_slp_analyze_node_operations): Likewise.
14702 (vect_slp_analyze_bb_1): Likewise.
14703 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14704 than an unsigned HOST_WIDE_INT.
14705 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14706 (vectorizable_load): Handle polynomial vf.
14707 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14709 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14712 Alan Hayward <alan.hayward@arm.com>
14713 David Sherwood <david.sherwood@arm.com>
14715 * match.pd: Handle bit operations involving three constants
14716 and try to fold one pair.
14718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14720 * tree-vect-loop-manip.c: Include gimple-fold.h.
14721 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14722 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14723 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14724 Add a path that uses a step of VF instead of 1, but disable it
14726 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14727 and niters_no_overflow parameters. Update calls to
14728 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14729 Create a new SSA name if the latter choses to use a ste other
14730 than zero, and return it via niters_vector_mult_vf_var.
14731 * tree-vect-loop.c (vect_transform_loop): Update calls to
14732 vect_do_peeling, vect_gen_vector_loop_niters and
14733 slpeel_make_loop_iterate_ntimes.
14734 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14735 (vect_gen_vector_loop_niters): Update declarations after above changes.
14737 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14739 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14740 128-bit round to integer instructions.
14741 (ceil<mode>2): Likewise.
14742 (btrunc<mode>2): Likewise.
14743 (round<mode>2): Likewise.
14745 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14747 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14748 unaligned VSX load/store on P8/P9.
14749 (expand_block_clear): Allow the use of unaligned VSX
14750 load/store on P8/P9.
14752 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14754 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14756 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14757 swap associated with both a load and a store.
14759 2018-01-02 Andrew Waterman <andrew@sifive.com>
14761 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14762 * config/riscv/riscv.md (clear_cache): Use it.
14764 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14766 * web.c: Remove out-of-date comment.
14768 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14770 * expr.c (fixup_args_size_notes): Check that any existing
14771 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14772 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14773 (emit_single_push_insn): ...here.
14775 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14777 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14778 (const_vector_encoded_nelts): New function.
14779 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14780 (const_vector_int_elt, const_vector_elt): Declare.
14781 * emit-rtl.c (const_vector_int_elt_1): New function.
14782 (const_vector_elt): Likewise.
14783 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14784 of CONST_VECTOR_ELT.
14786 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14788 * expr.c: Include rtx-vector-builder.h.
14789 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14790 directly on the tree encoding.
14791 (const_vector_from_tree): Likewise.
14792 * optabs.c: Include rtx-vector-builder.h.
14793 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14794 sequence of "u" values.
14795 * vec-perm-indices.c: Include rtx-vector-builder.h.
14796 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14797 directly on the vec_perm_indices encoding.
14799 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14801 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14802 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14803 * rtx-vector-builder.h: New file.
14804 * rtx-vector-builder.c: Likewise.
14805 * rtl.h (rtx_def::u2): Add a const_vector field.
14806 (CONST_VECTOR_NPATTERNS): New macro.
14807 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14808 (CONST_VECTOR_DUPLICATE_P): Likewise.
14809 (CONST_VECTOR_STEPPED_P): Likewise.
14810 (CONST_VECTOR_ENCODED_ELT): Likewise.
14811 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14812 (unwrap_const_vec_duplicate): Likewise.
14813 (const_vec_series_p): Check for a non-duplicated vector encoding.
14814 Say that the function only returns true for integer vectors.
14815 * emit-rtl.c: Include rtx-vector-builder.h.
14816 (gen_const_vec_duplicate_1): Delete.
14817 (gen_const_vector): Call gen_const_vec_duplicate instead of
14818 gen_const_vec_duplicate_1.
14819 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14820 (gen_const_vec_duplicate): Use rtx_vector_builder.
14821 (gen_const_vec_series): Likewise.
14822 (gen_rtx_CONST_VECTOR): Likewise.
14823 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14824 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14825 Build a new vector rather than modifying a CONST_VECTOR in-place.
14826 (handle_special_swappables): Update call accordingly.
14827 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14828 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14829 Build a new vector rather than modifying a CONST_VECTOR in-place.
14830 (handle_special_swappables): Update call accordingly.
14832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14834 * simplify-rtx.c (simplify_const_binary_operation): Use
14835 CONST_VECTOR_ELT instead of XVECEXP.
14837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14839 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14840 the selector elements to be different from the data elements
14841 if the selector is a VECTOR_CST.
14842 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14843 ssizetype for the selector.
14845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14847 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14848 before testing each element individually.
14849 * tree-vect-generic.c (lower_vec_perm): Likewise.
14851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14853 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14854 * selftest-run-tests.c (selftest::run_tests): Call it.
14855 * vector-builder.h (vector_builder::operator ==): New function.
14856 (vector_builder::operator !=): Likewise.
14857 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14858 (vec_perm_indices::all_from_input_p): New function.
14859 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14860 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14861 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14862 instead of reading the VECTOR_CST directly. Detect whether both
14863 vector inputs are the same before constructing the vec_perm_indices,
14864 and update the number of inputs argument accordingly. Use the
14865 utility functions added above. Only construct sel2 if we need to.
14867 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14869 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14870 the broadcast of the low byte.
14871 (expand_mult_highpart): Use an explicit encoding for the permutes.
14872 * optabs-query.c (can_mult_highpart_p): Likewise.
14873 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14874 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14875 (vectorizable_bswap): Likewise.
14876 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14877 explicit encoding for the power-of-2 permutes.
14878 (vect_permute_store_chain): Likewise.
14879 (vect_grouped_load_supported): Likewise.
14880 (vect_permute_load_chain): Likewise.
14882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14884 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14885 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14886 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14887 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14888 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14889 (vect_gen_perm_mask_any): Likewise.
14891 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14893 * int-vector-builder.h: New file.
14894 * vec-perm-indices.h: Include int-vector-builder.h.
14895 (vec_perm_indices): Redefine as an int_vector_builder.
14896 (auto_vec_perm_indices): Delete.
14897 (vec_perm_builder): Redefine as a stand-alone class.
14898 (vec_perm_indices::vec_perm_indices): New function.
14899 (vec_perm_indices::clamp): Likewise.
14900 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14901 (vec_perm_indices::new_vector): New function.
14902 (vec_perm_indices::new_expanded_vector): Update for new
14903 vec_perm_indices class.
14904 (vec_perm_indices::rotate_inputs): New function.
14905 (vec_perm_indices::all_in_range_p): Operate directly on the
14906 encoded form, without computing elided elements.
14907 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14908 encoding. Update for new vec_perm_indices class.
14909 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14910 the given vec_perm_builder.
14911 (expand_vec_perm_var): Update vec_perm_builder constructor.
14912 (expand_mult_highpart): Use vec_perm_builder instead of
14913 auto_vec_perm_indices.
14914 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14915 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14916 or double series encoding as appropriate.
14917 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14918 vec_perm_indices instead of auto_vec_perm_indices.
14919 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14920 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14921 (vect_permute_store_chain): Likewise.
14922 (vect_grouped_load_supported): Likewise.
14923 (vect_permute_load_chain): Likewise.
14924 (vect_shift_permute_load_chain): Likewise.
14925 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14926 (vect_transform_slp_perm_load): Likewise.
14927 (vect_schedule_slp_instance): Likewise.
14928 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14929 (vectorizable_mask_load_store): Likewise.
14930 (vectorizable_bswap): Likewise.
14931 (vectorizable_store): Likewise.
14932 (vectorizable_load): Likewise.
14933 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14934 vec_perm_indices instead of auto_vec_perm_indices. Use
14935 tree_to_vec_perm_builder to read the vector from a tree.
14936 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14937 vec_perm_builder instead of a vec_perm_indices.
14938 (have_whole_vector_shift): Use vec_perm_builder and
14939 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14940 truncation to calc_vec_perm_mask_for_shift.
14941 (vect_create_epilog_for_reduction): Likewise.
14942 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14943 from auto_vec_perm_indices to vec_perm_indices.
14944 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14945 instead of changing individual elements.
14946 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14947 the vector in d.perm.
14948 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14949 from auto_vec_perm_indices to vec_perm_indices.
14950 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14951 instead of changing individual elements.
14952 (arm_vectorize_vec_perm_const): Use new_vector to install
14953 the vector in d.perm.
14954 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14955 Update vec_perm_builder constructor.
14956 (rs6000_expand_interleave): Likewise.
14957 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14958 (rs6000_expand_interleave): Likewise.
14960 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14962 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14963 to qimode could truncate the indices.
14964 * optabs.c (expand_vec_perm_var): Likewise.
14966 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14968 * Makefile.in (OBJS): Add vec-perm-indices.o.
14969 * vec-perm-indices.h: New file.
14970 * vec-perm-indices.c: Likewise.
14971 * target.h (vec_perm_indices): Replace with a forward class
14973 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14974 * optabs.h: Include vec-perm-indices.h.
14975 (expand_vec_perm): Delete.
14976 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14977 (expand_vec_perm_const): Declare.
14978 * target.def (vec_perm_const_ok): Replace with...
14979 (vec_perm_const): ...this new hook.
14980 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14981 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14982 * doc/tm.texi: Regenerate.
14983 * optabs.def (vec_perm_const): Delete.
14984 * doc/md.texi (vec_perm_const): Likewise.
14985 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14986 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14987 expand_vec_perm for constant permutation vectors. Assert that
14988 the mode of variable permutation vectors is the integer equivalent
14989 of the mode that is being permuted.
14990 * optabs-query.h (selector_fits_mode_p): Declare.
14991 * optabs-query.c: Include vec-perm-indices.h.
14992 (selector_fits_mode_p): New function.
14993 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14994 is defined, instead of checking whether the vec_perm_const_optab
14995 exists. Use targetm.vectorize.vec_perm_const instead of
14996 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14997 fit in the vector mode before using a variable permute.
14998 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14999 vec_perm_indices instead of an rtx.
15000 (expand_vec_perm): Replace with...
15001 (expand_vec_perm_const): ...this new function. Take the selector
15002 as a vec_perm_indices rather than an rtx. Also take the mode of
15003 the selector. Update call to shift_amt_for_vec_perm_mask.
15004 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
15005 Use vec_perm_indices::new_expanded_vector to expand the original
15006 selector into bytes. Check whether the indices fit in the vector
15007 mode before using a variable permute.
15008 (expand_vec_perm_var): Make global.
15009 (expand_mult_highpart): Use expand_vec_perm_const.
15010 * fold-const.c: Includes vec-perm-indices.h.
15011 * tree-ssa-forwprop.c: Likewise.
15012 * tree-vect-data-refs.c: Likewise.
15013 * tree-vect-generic.c: Likewise.
15014 * tree-vect-loop.c: Likewise.
15015 * tree-vect-slp.c: Likewise.
15016 * tree-vect-stmts.c: Likewise.
15017 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
15019 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
15020 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
15021 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
15022 (aarch64_vectorize_vec_perm_const): ...this new function.
15023 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15024 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15025 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
15026 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
15027 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15028 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15029 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
15031 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
15032 check for NEON modes.
15033 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
15034 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
15035 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
15036 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
15038 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
15039 the old VEC_PERM_CONST conditions.
15040 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
15041 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
15042 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
15043 (ia64_vectorize_vec_perm_const_ok): Merge into...
15044 (ia64_vectorize_vec_perm_const): ...this new function.
15045 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
15046 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
15047 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
15048 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
15049 * config/mips/mips.c (mips_expand_vec_perm_const)
15050 (mips_vectorize_vec_perm_const_ok): Merge into...
15051 (mips_vectorize_vec_perm_const): ...this new function.
15052 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15053 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15054 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15055 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15056 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15057 (rs6000_expand_vec_perm_const): Delete.
15058 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
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.
15068 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15069 an element count and rtx array.
15070 (rs6000_expand_extract_even): Update call accordingly.
15071 (rs6000_expand_interleave): Likewise.
15072 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15073 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15074 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15075 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15076 (rs6000_expand_vec_perm_const): Delete.
15077 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15078 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15079 (altivec_expand_vec_perm_const_le): Take each operand individually.
15080 Operate on constant selectors rather than rtxes.
15081 (altivec_expand_vec_perm_const): Likewise. Update call to
15082 altivec_expand_vec_perm_const_le.
15083 (rs6000_expand_vec_perm_const): Delete.
15084 (rs6000_vectorize_vec_perm_const_ok): Delete.
15085 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15086 reference to the SPE evmerge intructions.
15087 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15088 an element count and rtx array.
15089 (rs6000_expand_extract_even): Update call accordingly.
15090 (rs6000_expand_interleave): Likewise.
15091 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15092 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15094 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15096 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15098 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15099 vector mode and that that mode matches the mode of the data
15101 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15102 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15103 directly using expand_vec_perm_1 when forcing selectors into
15105 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15107 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15109 * optabs-query.h (can_vec_perm_p): Delete.
15110 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15111 * optabs-query.c (can_vec_perm_p): Split into...
15112 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15113 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15114 particular selector is valid.
15115 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15116 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15117 (vect_grouped_load_supported): Likewise.
15118 (vect_shift_permute_load_chain): Likewise.
15119 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15120 (vect_transform_slp_perm_load): Likewise.
15121 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15122 (vectorizable_bswap): Likewise.
15123 (vect_gen_perm_mask_checked): Likewise.
15124 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15125 implementations of variable permutation vectors into account
15126 when deciding which selector to use.
15127 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15128 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15129 with a false third argument.
15130 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15131 to test whether the constant selector is valid and can_vec_perm_var_p
15132 to test whether a variable selector is valid.
15134 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15136 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15137 * optabs-query.c (can_vec_perm_p): Likewise.
15138 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15139 instead of vec_perm_indices.
15140 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15141 (vect_gen_perm_mask_checked): Likewise,
15142 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15143 (vect_gen_perm_mask_checked): Likewise,
15145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15147 * optabs-query.h (qimode_for_vec_perm): Declare.
15148 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15149 (qimode_for_vec_perm): ...this new function.
15150 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15152 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15154 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15155 does not have a conditional at the top.
15157 2018-01-02 Richard Biener <rguenther@suse.de>
15159 * ipa-inline.c (big_speedup_p): Fix expression.
15161 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15164 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15167 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15171 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15172 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15173 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15174 cond_taken_branch_cost 3->4.
15176 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15178 PR tree-optimization/83581
15179 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15180 TODO_cleanup_cfg if any changes have been made.
15182 PR middle-end/83608
15183 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15184 convert_modes if target mode has the right side, but different mode
15187 PR middle-end/83609
15188 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15189 last argument when extracting from CONCAT. If either from_real or
15190 from_imag is NULL, use expansion through memory. If result is not
15191 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15192 the parts directly to inner mode, if even that fails, use expansion
15195 PR middle-end/83623
15196 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15197 check for bswap in mode rather than HImode and use that in expand_unop
15200 Copyright (C) 2018 Free Software Foundation, Inc.
15202 Copying and distribution of this file, with or without modification,
15203 are permitted in any medium without royalty provided the copyright
15204 notice and this notice are preserved.