1 2018-06-04 Richard Biener <rguenther@suse.de>
3 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4 (cleanup_tree_cfg_noloop): ... single caller. Do
5 start_recording_case_labels later.
7 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
9 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10 to _IMMINTRIN_H_INCLUDED.
11 * config/i386/pconfigintrin.h: Ditto.
12 * config/i386/waitpkgintrin.h: Ditto.
13 * config/i386/immintrin.h: Add includes for sgxintrin.h,
14 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
15 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
16 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
18 waitpkgintrin.h and cldemoteintrin.h.
20 2018-06-04 Richard Biener <rguenther@suse.de>
22 PR tree-optimization/86038
23 * tracer.c (find_best_successor): Check probability for
24 being initialized, bail out if not.
26 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
29 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
30 of bits to ignore when comparing architectures.
32 2018-06-04 Jakub Jelinek <jakub@redhat.com>
34 PR tree-optimization/69615
35 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
36 maximum or minimum of the type, try to merge it also as if
37 range1 is + [-, x - 1] or + [x + 1, -].
40 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
42 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
44 PR tree-optimization/86034
45 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
46 the unsigned bitfield type in a bit insertion sequence if it does not
47 have a larger precision than the bitfield size.
48 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
50 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
52 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
54 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
56 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
57 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
58 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
59 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
61 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
63 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
64 Disable -fdelete-null-pointer-checks for ELF toolchain.
66 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
67 Kito Cheng <kito.cheng@gmail.com>
69 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
70 (nds32le-*-*, nds32be-*-*): Integrate checking process.
71 (nds32*-*-*): Add glibc and uclibc conditions.
72 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
73 (TARGET_EXCEPT_UNWIND_INFO): Define.
74 * config/nds32/elf.h: New file.
75 * config/nds32/linux.h: New file.
76 * config/nds32/nds32-elf.opt: New file.
77 * config/nds32/nds32-linux.opt: New file.
78 * config/nds32/nds32-fp-as-gp.c
79 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
80 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
82 (nds32_asm_file_end): Ditto.
83 (nds32_print_operand): Ditto.
84 (nds32_insert_attributes): Ditto.
85 (nds32_init_libfuncs): New function.
86 (TARGET_HAVE_TLS): Define.
87 (TARGET_INIT_LIBFUNCS): Define.
88 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
90 (TARGET_ELF): Apply different mcmodel setting.
91 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
92 been migrated into elf.h and linux.h files.
93 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
94 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
95 (mcmodel): The content has been migrated into nds32-elf.opt and
96 nds32-linux.opt files.
97 * config/nds32/t-elf: New file.
98 * config/nds32/t-linux: New file.
100 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
101 Shiva Chen <shiva0217@gmail.com>
103 * config/nds32/constants.md (unspec_volatile_element): Add
104 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
105 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
107 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
108 (make_pass_nds32_fp_as_gp): Declare.
109 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
111 (nds32_asm_function_end_prologue): Remove unused asm output.
112 (nds32_asm_function_begin_epilogue): Remove unused asm output.
113 (nds32_asm_file_start): Output necessary fp_as_gp information.
114 (nds32_option_override): Adjust register usage.
115 (nds32_expand_prologue): Consider fp_as_gp situation.
116 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
117 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
120 (simple_return): Ditto.
121 (omit_fp_begin): Output special directive for fp_as_gp.
122 (omit_fp_end): Output special directive for fp_as_gp.
123 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
124 mforbid-fp-as-gp): New options.
126 2018-06-01 Mark Wielaard <mark@klomp.org>
128 * dwarf2out.c (dwarf2out_finish): Remove generation of
131 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
133 * gimple-ssa-store-merging.c: Include gimple-fold.h.
134 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
135 (struct merged_store_group): Add bit_insertion field.
136 (dump_char_array): Use standard hexadecimal format.
137 (merged_store_group::merged_store_group): Set bit_insertion to false.
138 (merged_store_group::apply_stores): Use optimal buffer size. Deal
139 with BIT_INSERT_EXPR stores. Move up code updating the mask and
140 also print the mask in the dump file.
141 (pass_store_merging::gate): Minor tweak.
142 (imm_store_chain_info::coalesce_immediate): Fix wrong association
143 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
144 stores with INTEGER_CST stores.
145 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
146 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
147 and use it throughout. Generate bit insertion sequences if need be.
148 (pass_store_merging::process_store): Remove redundant condition.
149 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
151 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
153 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
154 the 128-bit floating point types. Fix function comment.
156 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
158 * config/aarch64/aarch64-simd.md
159 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
161 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
164 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
166 PR tree-optimization/85989
167 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
169 (backprop::intersect_uses): Check it when deciding whether this
170 is a backedge reference.
171 (backprop::process_block): Add each phi to m_visited_phis
172 after visiting it, then clear it at the end.
174 2018-06-01 Richard Biener <rguenther@suse.de>
176 * tree-vectorizer.h (vect_dr_stmt): New function.
177 (vect_get_load_cost): Adjust.
178 (vect_get_store_cost): Likewise.
179 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
180 Use vect_dr_stmt instead of DR_SMTT.
181 (vect_record_base_alignments): Likewise.
182 (vect_calculate_target_alignment): Likewise.
183 (vect_compute_data_ref_alignment): Likewise and make static.
184 (vect_update_misalignment_for_peel): Likewise.
185 (vect_verify_datarefs_alignment): Likewise.
186 (vector_alignment_reachable_p): Likewise.
187 (vect_get_data_access_cost): Likewise. Pass down
188 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
189 (vect_get_peeling_costs_all_drs): Likewise.
190 (vect_peeling_hash_get_lowest_cost): Likewise.
191 (vect_enhance_data_refs_alignment): Likewise.
192 (vect_find_same_alignment_drs): Likewise.
193 (vect_analyze_data_refs_alignment): Likewise.
194 (vect_analyze_group_access_1): Likewise.
195 (vect_analyze_group_access): Likewise.
196 (vect_analyze_data_ref_access): Likewise.
197 (vect_analyze_data_ref_accesses): Likewise.
198 (vect_vfa_segment_size): Likewise.
199 (vect_small_gap_p): Likewise.
200 (vectorizable_with_step_bound_p): Likewise.
201 (vect_prune_runtime_alias_test_list): Likewise.
202 (vect_analyze_data_refs): Likewise.
203 (vect_supportable_dr_alignment): Likewise.
204 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
205 (vect_gen_prolog_loop_niters): Likewise.
206 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
207 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
209 (vect_recog_mask_conversion_pattern): Likewise.
210 (vect_try_gather_scatter_pattern): Likewise.
211 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
212 to vect_get_store_cost.
213 (vect_get_store_cost): Get stmt_info instead of DR.
214 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
215 (vect_get_load_cost): Get stmt_info instead of DR.
217 2018-06-01 Richard Biener <rguenther@suse.de>
220 * gimple-fold.c (var_decl_component_p): Also allow offsetted
221 vars wrapped in MEM_REFs.
223 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
225 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
226 Fix subreg tests so that we only return a choice between
227 GENERAL_REGS and FP_REGS if the original classes included both.
229 2018-06-01 Richard Biener <rguenther@suse.de>
232 * tree-ssa-structalias.c (get_function_part_constraint):
233 Handle NULL fi->decl.
234 (find_func_aliases_for_call): Properly handle indirect
236 (find_func_clobbers): Likewise.
237 (ipa_pta_execute): Likewise.
238 (create_variable_info_for): For functions that are ifunc_resolver
239 resolve to a varinfo that contains the result of the resolver
241 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
244 2018-05-31 Michael Collison <michael.collison@arm.com>
246 * config/aarch64/aarch64.md:
247 (*fix_to_zero_extenddfdi2): New pattern.
248 * gcc.target/aarch64/fix_extend1.c: New testcase.
250 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
254 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
255 and BUILT_IN_STRNCMP_EQ.
256 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
258 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
259 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
260 (gimple_fold_builtin): Likewise.
261 * tree-ssa-strlen.c (compute_string_length): New function.
262 (determine_min_obsize): New function.
263 (handle_builtin_string_cmp): New function to handle calls to
264 string compare functions.
265 (strlen_optimize_stmt): Add handling to builtin string compare
267 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
268 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
269 * tree.c (build_common_builtin_nodes): Add new defines of
270 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
272 2018-05-31 Jakub Jelinek <jakub@redhat.com>
275 * bb-reorder.c (pass_partition_blocks::gate): Return false for
276 functions with naked attribute.
278 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
280 * config/i386/sse.md (avx_vec_concat<mode>):
281 Substitute concat_tg_mode mode attribute with xtg_mode.
282 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
283 (concat_tg_mode): Remove mode attribute.
285 2018-05-31 Martin Sebor <msebor@redhat.com>
288 * calls.c (alloc_max_size): Correct a logic error/typo.
289 Treat excessive arguments as infinite. Warn for invalid arguments.
290 * doc/invoke.texi (-Walloc-size-larger-than): Update.
292 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
295 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
296 and movx for Haswell.
298 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
299 Cesar Philippidis <cesar@codesourcery.com>
302 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
303 when emitting error on private/firstprivate reductions.
304 * omp-low.c (lower_omp_target): Avoid reference-type processing
305 on pointers for firstprivate clause.
307 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
309 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
312 * config/aarch64/aarch64-simd.md
313 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
314 (aarch64_ld1_x3_<mode>): Likewise
315 (aarch64_st1x2<VALLDIF:mode>): Likewise
316 (aarch64_st1_x2_<mode>): Likewise
317 (aarch64_st1x3<VALLDIF:mode>): Likewise
318 (aarch64_st1_x3_<mode>): Likewise
319 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
320 (vld1_s8_x3): Likewise.
321 (vld1_u16_x3): Likewise.
322 (vld1_s16_x3): Likewise.
323 (vld1_u32_x3): Likewise.
324 (vld1_s32_x3): Likewise.
325 (vld1_u64_x3): Likewise.
326 (vld1_s64_x3): Likewise.
327 (vld1_f16_x3): Likewise.
328 (vld1_f32_x3): Likewise.
329 (vld1_f64_x3): Likewise.
330 (vld1_p8_x3): Likewise.
331 (vld1_p16_x3): Likewise.
332 (vld1_p64_x3): Likewise.
333 (vld1q_u8_x3): Likewise.
334 (vld1q_s8_x3): Likewise.
335 (vld1q_u16_x3): Likewise.
336 (vld1q_s16_x3): Likewise.
337 (vld1q_u32_x3): Likewise.
338 (vld1q_s32_x3): Likewise.
339 (vld1q_u64_x3): Likewise.
340 (vld1q_s64_x3): Likewise.
341 (vld1q_f16_x3): Likewise.
342 (vld1q_f32_x3): Likewise.
343 (vld1q_f64_x3): Likewise.
344 (vld1q_p8_x3): Likewise.
345 (vld1q_p16_x3): Likewise.
346 (vld1q_p64_x3): Likewise.
347 (vst1_s64_x2): Likewise.
348 (vst1_u64_x2): Likewise.
349 (vst1_f64_x2): Likewise.
350 (vst1_s8_x2): Likewise.
351 (vst1_p8_x2): Likewise.
352 (vst1_s16_x2): Likewise.
353 (vst1_p16_x2): Likewise.
354 (vst1_s32_x2): Likewise.
355 (vst1_u8_x2): Likewise.
356 (vst1_u16_x2): Likewise.
357 (vst1_u32_x2): Likewise.
358 (vst1_f16_x2): Likewise.
359 (vst1_f32_x2): Likewise.
360 (vst1_p64_x2): Likewise.
361 (vst1q_s8_x2): Likewise.
362 (vst1q_p8_x2): Likewise.
363 (vst1q_s16_x2): Likewise.
364 (vst1q_p16_x2): Likewise.
365 (vst1q_s32_x2): Likewise.
366 (vst1q_s64_x2): Likewise.
367 (vst1q_u8_x2): Likewise.
368 (vst1q_u16_x2): Likewise.
369 (vst1q_u32_x2): Likewise.
370 (vst1q_u64_x2): Likewise.
371 (vst1q_f16_x2): Likewise.
372 (vst1q_f32_x2): Likewise.
373 (vst1q_f64_x2): Likewise.
374 (vst1q_p64_x2): Likewise.
375 (vst1_s64_x3): Likewise.
376 (vst1_u64_x3): Likewise.
377 (vst1_f64_x3): Likewise.
378 (vst1_s8_x3): Likewise.
379 (vst1_p8_x3): Likewise.
380 (vst1_s16_x3): Likewise.
381 (vst1_p16_x3): Likewise.
382 (vst1_s32_x3): Likewise.
383 (vst1_u8_x3): Likewise.
384 (vst1_u16_x3): Likewise.
385 (vst1_u32_x3): Likewise.
386 (vst1_f16_x3): Likewise.
387 (vst1_f32_x3): Likewise.
388 (vst1_p64_x3): Likewise.
389 (vst1q_s8_x3): Likewise.
390 (vst1q_p8_x3): Likewise.
391 (vst1q_s16_x3): Likewise.
392 (vst1q_p16_x3): Likewise.
393 (vst1q_s32_x3): Likewise.
394 (vst1q_s64_x3): Likewise.
395 (vst1q_u8_x3): Likewise.
396 (vst1q_u16_x3): Likewise.
397 (vst1q_u32_x3): Likewise.
398 (vst1q_u64_x3): Likewise.
399 (vst1q_f16_x3): Likewise.
400 (vst1q_f32_x3): Likewise.
401 (vst1q_f64_x3): Likewise.
402 (vst1q_p64_x3): Likewise.
404 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
406 * config/msp430/msp430.c (msp430_output_labelref): Prepend
407 user_label_prefix to name.
409 * tree-core.h: Update comment about the format of NAME string
410 passed to handler in attribute_spec.
412 * config/msp430/msp430.md: Remove erroneous subreg expression from
413 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
414 zero_extend{q,h}isi2.
416 2018-05-30 Borislav Petkov <bp@suse.de>
418 * doc/extend.texi: Document some architecture specific
419 constraints and sort entries.
421 2018-05-30 Martin Sebor <msebor@redhat.com>
424 * builtins.c (expand_builtin_stpcpy_1): New function.
425 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
426 only if the former succeeds.
428 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
430 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
433 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
435 * doc/invoke.texi (-flinker-output): Document
437 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
439 * passes.c (ipa_write_summaries): Only modify statements if body
441 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
442 incrementally linking.
443 (ipa_passes): Likewise.
444 * lto-cgraph.c (lto_output_node): When incrementally linking do not
445 pass down resolution info.
446 * common.opt (flag_incremental_link): Update info.
447 * gcc.c (plugin specs): Turn flinker-output=* to
448 -plugin-opt=-linker-output-known
449 * toplev.c (compile_file): Also cut compilation when doing incremental
451 * flag-types. (enum lto_partition_model): Add
452 LTO_LINKER_OUTPUT_NOLTOREL.
453 (invoke.texi): Add -flinker-output docs.
454 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
455 link same way as WPA; do not stream in dead initializers.
457 * dwarf2out.c (dwarf2out_die_ref_for_decl,
458 darf2out_register_external_decl): Support incremental link.
460 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
462 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
464 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
466 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
467 it down to simple_object_copy_lto_debug_sections.
468 (run_gcc): Determine incremental LTO link time and configure
469 lto1 into non-wpa mode, disable renaming of debug sections.
471 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
473 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
474 descriptions of various incorrectly documented functions.
476 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
479 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
480 address check not strict.
482 2018-05-30 Richard Biener <rguenther@suse.de>
484 PR tree-optimization/85964
485 * tracer.c (better_p): Drop initialized count check, we only
486 call the function with initialized counts now.
487 (find_best_successor): Do find a best edge if one
488 has uninitialized count.
489 (find_best_predecessor): Likewise. Do BB frequency check only
490 if count is initialized.
492 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
494 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
495 (aarch64_ldrstr_offset_compare): New.
496 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
497 load/store orderings.
498 (aarch64_gen_adjusted_ldpstp): Likewise.
500 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
502 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
503 Check for subset of GENERAL_REGS and FP_REGS.
504 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
507 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
509 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
510 and wi::to_poly_offset. Add the current offset and then check
511 whether the sum fits, rather than using an unchecked addition of
512 a checked term. Check for a shwi rather than a uhwi.
513 * expr.c (get_bit_range): Use tree_to_poly_uint64.
514 (store_constructor): Use poly_int_tree_p.
515 (expand_expr_real_1): Likewise.
516 * function.c (assign_temp): Likewise.
517 * fold-const.c (const_binop): Use poly_int_tree_p and
519 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
521 * ipa-icf-gimple.c (func_checker::compare_operand): Use
522 to_poly_offset for MEM offsets.
523 * ipa-icf.c (sem_variable::equals): Likewise.
524 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
525 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
526 wi::to_poly_offset for BIT_FIELD_REF offsets.
527 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
529 * var-tracking.c (emit_note_insn_var_location): Use
532 2018-05-29 Jim Wilson <jimw@sifive.com>
534 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
536 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
539 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
540 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
542 (sse4_1_round<mode>2): Use nonimmediate_operand
543 for operand 1 predicate.
545 2018-05-29 Martin Sebor <msebor@redhat.com>
546 Richard Biener <rguenther@suse.de>
549 * calls.c (get_size_range): Call determine_value_range instead
551 * tree-vrp.h (determine_value_range): Declared new function.
552 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
554 2018-05-29 Richard Biener <rguenther@suse.de>
556 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
557 sure to use non-pattern stmts for get_earlier_stmt arguments.
558 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
559 called on pattern stmts.
560 (get_later_stmt): Likewise.
562 2018-05-29 Martin Liska <mliska@suse.cz>
564 PR gcov-profile/85759
565 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
568 2018-05-29 Jakub Jelinek <jakub@redhat.com>
570 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
572 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
576 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
577 VEC_PACK_FLOAT_EXPR): New tree codes.
578 * tree-pretty-print.c (op_code_prio): Handle
579 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
580 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
581 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
582 * tree-inline.c (estimate_operator_cost): Likewise.
583 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
584 * fold-const.c (const_binop): Likewise.
585 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
586 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
587 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
588 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
589 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
590 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
591 * expr.c (expand_expr_real_2): Likewise.
592 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
593 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
594 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
596 * optabs.c (expand_widen_pattern_expr): For
597 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
598 sign from result type rather than operand's type.
599 (expand_binop_directly): For vec_packu_float_optab and
600 vec_packs_float_optab allow result type to be different from operand's
602 * optabs-tree.c (optab_for_tree_code): Handle
603 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
604 VEC_PACK_FLOAT_EXPR. Formatting fixes.
605 * tree-vect-generic.c (expand_vector_operations_1): Handle
606 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
608 * tree-vect-stmts.c (supportable_widening_operation): Handle
610 (supportable_narrowing_operation): Handle FLOAT_EXPR.
611 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
612 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
613 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
614 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
616 (vec_pack<floatprefix>_float_<mode>): New expander.
617 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
619 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
620 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
621 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
622 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
623 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
625 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
626 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
627 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
628 VEC_PACK_FLOAT_EXPR): Document.
630 2018-05-29 Richard Biener <rguenther@suse.de>
632 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
634 (stmt_vec_info_vec): Make pointer.
635 (init_stmt_vec_info_vec): Remove.
636 (free_stmt_vec_info_vec): Likewise.
637 (set_stmt_vec_info_vec): New function.
638 (free_stmt_vec_infos): Likewise.
639 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
640 (set_vinfo_for_stmt): Likewise.
641 (get_earlier_stmt): Likewise.
642 (get_later_stmt): Likewise.
643 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
644 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
645 (vec_info::~vec_info): Free stmt_vec_infos.
646 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
647 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
648 (pass_slp_vectorize::execute): Likewise.
649 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
650 (free_stmt_vec_info_vec): Likewise.
651 (set_stmt_vec_info_vec): New function.
652 (free_stmt_vec_infos): Likewise.
653 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
654 the global stmt_vec_info_vec.
655 * tree-parloops.c (gather_scalar_reductions): Use
656 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
659 2018-05-29 Richard Biener <rguenther@suse.de>
661 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
663 2018-05-29 Martin Liska <mliska@suse.cz>
664 David Malcolm <dmalcolm@redhat.com>
666 * vec.c (test_reverse): New.
667 (vec_c_tests): Add new test.
668 * vec.h (vl_ptr>::reverse): New function.
670 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
672 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
674 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
677 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
679 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
681 2018-05-28 Richard Biener <rguenther@suse.de>
683 PR tree-optimization/85933
684 * tree-vect-data-refs.c (vect_record_base_alignments): Only
685 look at stmts marked as vectorizable.
687 2018-05-28 Richard Biener <rguenther@suse.de>
689 PR tree-optimization/85934
690 * tree-vect-generic.c (expand_vector_operations_1): Hoist
691 vector boolean check before scalar optimization.
693 2018-05-28 Jakub Jelinek <jakub@redhat.com>
695 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
698 2018-05-28 Mark Wielaard <mark@klomp.org>
700 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
701 if it is an expression containing a minus sign.
703 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
705 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
707 2018-05-27 Paul Koning <ni1d@arrl.net>
709 * config/pdp11/pdp11.md (truncsihi2): Remove.
711 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
712 Chung-Ju Wu <jasonwucj@gmail.com>
714 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
716 (unaligned_store_dw): Ditto.
717 * config/nds32/nds32-memory-manipulation.c
718 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
719 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
720 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
721 (emit_setmem_word_loop): Rename to ...
722 (emit_setmem_doubleword_loop): ... this.
723 (nds32_gen_dup_4_byte_to_word_value): New function.
724 (nds32_gen_dup_8_byte_to_double_word_value): New function.
725 (nds32_expand_setmem_loop): Refine implementation.
726 (nds32_expand_setmem_loop_v3m): Ditto.
727 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
730 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
732 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
734 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
736 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
737 (nds32_init_machine_status): Initialize machine->attr_naked_p and
738 machine->attr_no_prologue_p.
739 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
740 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
741 (nds32_expand_epilogue): Consider attr_naked_p.
742 (nds32_expand_epilogue_v3pop): Likewise.
743 (nds32_can_use_return_insn): Likewise.
744 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
745 attr_no_prologue_p fields.
746 * config/nds32/nds32.opt (mret-in-naked-func): New option.
748 2018-05-27 Jakub Jelinek <jakub@redhat.com>
751 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
754 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
756 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
758 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
760 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
762 (*<floatsuffix>floatv2div2sf2): Rename to ...
763 (*float<floatunssuffix>v2div2sf2): ... this.
764 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
765 (float<floatunssuffix>v2div2sf2_mask): ... this.
766 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
767 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
768 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
770 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
772 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
774 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
776 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
778 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
780 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
781 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
782 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
783 gen_ufix_truncv8dfv8si2.
784 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
785 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
786 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
787 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
788 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
789 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
790 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
791 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
793 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
797 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
799 2018-05-25 Jim Wilson <jimw@sifive.com>
801 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
802 * config/riscv/riscv.c (struct machine_function): Add
803 interrupt_handler_p and attribute_checked_p fields.
804 (riscv_attribute_table): Add interrupt.
805 (riscv_interrupt_type_p): New.
806 (riscv_save_reg_p): Save extra regs for interrupt handler.
807 (riscv_use_save_libcall): Return false for interrupt handler.
808 (riscv_first_stack_step): Add forward declaration.
809 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
810 for interrupt handler with large frame. Use it for saved reg list.
811 (riscv_expand_prologue): Move flag_stack_usage_info support to
812 eliminate duplication.
813 (riscv_expand_epilogue): Generate mret for interrupt handler.
814 (riscv_epilogue_uses): New.
815 (riscv_can_use_return_insn): Return false for interrupt handler.
816 (riscv_function_ok_for_sibcall): Likewise.
817 (riscv_set_current_function): Add interrupt handler support.
818 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
819 * config/riscv/riscv.md (UNSPECV_MRET): New.
821 (riscv_frflags, riscv_fsflags): Use tab after opcode.
823 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
825 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
827 PR tree-optimization/85712
828 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
829 this candidate has already been replaced in-situ by a copy.
831 2018-05-25 Jason Merrill <jason@redhat.com>
833 PR c++/80485 - inline function non-zero address.
834 * symtab.c (nonzero_address): Check DECL_COMDAT.
836 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
839 * config/alpha/alpha.md (ashlsi3): New insn pattern.
840 (*ashlsi_se): Rename from *ashldi_se. Define as sign
841 extension of SImode operation. Use const123_operand predicate.
843 (*saddl_se_1): Ditto.
845 (*ssubl_se_1): Ditto.
846 * config/alpha/predicates.md (const123_operand): New predicate.
847 * config/alpha/constraints.md (P): Use IN_RANGE.
849 2018-05-25 Richard Biener <rguenther@suse.de>
851 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
853 (ref_maybe_used_by_stmt_p): Likewise.
854 (stmt_may_clobber_ref_p): Likewise.
855 (stmt_may_clobber_ref_p_1): Likewise.
856 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
858 (ref_maybe_used_by_stmt_p): Likewise.
859 (stmt_may_clobber_ref_p): Likewise.
860 (stmt_may_clobber_ref_p_1): Likewise.
861 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
862 the alias oracle to disambiguate DRs with stmts DR analysis
864 (vect_analyze_data_refs): Do not give up on not analyzable
865 DRs for BB vectorization. Remove code truncating the dataref
868 2018-05-25 Jakub Jelinek <jakub@redhat.com>
871 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
872 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
873 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
875 2018-05-25 Richard Biener <rguenther@suse.de>
877 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
878 function, combining stmt data ref gathering and fatal analysis
880 (vect_analyze_data_refs): Remove now redudnant code and simplify.
881 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
882 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
883 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
884 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
886 2018-05-25 Bin Cheng <bin.cheng@arm.com>
888 PR tree-optimization/85720
889 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
890 SCC if all partitions are builtins.
891 (version_loop_by_alias_check): New parameter. Generate cancelable
892 runtime alias check if all partitions are builtins.
893 (distribute_loop): Update call to above function.
895 2018-05-25 Bin Cheng <bin.cheng@arm.com>
897 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
898 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
899 (parm_default_def_partition_arg): Ditto.
900 (set_parm_default_def_partition): Ditto.
901 (get_parm_default_def_partitions): Ditto and make it static.
902 (get_undefined_value_partitions): Ditto and make it static.
903 (remove_ssa_form): Refactor call to init_var_map here.
904 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
905 computation for loop region.
906 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
907 (register_default_def): Delete.
908 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
909 (parm_default_def_partition_arg): Ditto.
910 (set_parm_default_def_partition): Ditto.
911 (get_parm_default_def_partitions): Ditto and make it static.
912 (get_undefined_value_partitions): Ditto and make it static.
913 (coalesce_with_default, coalesce_with_default): Update comment.
914 (create_coalesce_list_for_region): New func factored out from
915 create_outofssa_var_map.
916 (populate_coalesce_list_for_outofssa): New func factored out from
917 create_outofssa_var_map and coalesce_ssa_name.
918 (create_outofssa_var_map): Delete.
919 (coalesce_ssa_name): Refactor to support live range computation.
920 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
921 (get_parm_default_def_partitions): Delete.
922 (get_undefined_value_partitions): Ditto.
923 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
924 computation for loop region.
925 (new_tree_live_info, loe_visit_block): Ditto.
926 (live_worklist, set_var_live_on_entry): Ditto.
927 (calculate_live_on_exit, verify_live_on_entry): Ditto.
928 * tree-ssa-live.h (struct _var_map): New fields.
929 (init_var_map): Change decl.
930 (region_contains_p): New.
932 2018-05-25 Bin Cheng <bin.cheng@arm.com>
934 * tree-ssa-live.h (live_merge_and_clear): Delete.
936 2018-05-25 Richard Biener <rguenther@suse.de>
939 * tree-dump.c (dequeue_and_dump): Remove access to removed
940 operand 2 of a SWITCH_EXPR.
942 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
944 * doc/sourcebuild.texi (vect_double_cond_arith): Include
945 multiplication and division.
946 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
947 (cond_udiv@var{m}, cond_umod@var{m}): Document.
948 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
949 (cond_udiv_optab, cond_umod_optab): New optabs.
950 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
951 (IFN_COND_RDIV): New internal functions.
952 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
953 TRUNC_MOD_EXPR and RDIV_EXPR.
954 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
955 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
957 (SVE_INT_BINARY): Include mult.
958 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
959 (optab, sve_int_op): Handle mult.
960 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
962 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
963 for SVE_INT_BINARY_SD.
965 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
967 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
968 (optab, sve_int_op): Handle div and udiv.
969 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
970 for SVE_INT_BINARY_SD.
971 (*<optab><mode>3): New insn for the same.
973 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
975 * tree-vect-patterns.c: Include predict.h.
976 (vect_recog_divmod_pattern): Restrict check for division support
977 to when optimizing for size.
979 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
981 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
982 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
983 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
984 (gimple_match_op::set_op): Likewise.
985 (gimple_resimplify4): Declare.
986 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
987 (expr::gen_transform): Likewise.
988 (decision_tree::gen): Generate a simplification routine for 4 operands.
989 * gimple-match-head.c (gimple_simplify): Add an overload for
990 4 operands. In the top-level function, handle up to 4 call
991 arguments and call gimple_resimplify4.
992 (gimple_resimplify4): New function.
993 (build_call_internal): Pass a fourth operand.
994 (maybe_push_to_seq): Likewise.
995 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
996 Fold VEC_COND_EXPRs of an operation and a default value into
997 an IFN_COND_* function if possible.
998 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
1000 (SVE_COND_FP_BINARY): Include them.
1001 (optab, sve_fp_op): Handle them.
1002 (SVE_INT_BINARY_REV): New code iterator.
1003 (SVE_COND_FP_BINARY_REV): New int iterator.
1004 (commutative): New int attribute.
1005 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
1007 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
1009 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
1010 (*cond_<optab><mode>): New patterns for reversed operands.
1012 2018-05-25 Richard Biener <rguenther@suse.de>
1014 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
1015 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
1016 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
1017 (STMT_VINFO_GROUPED_ACCESS): Adjust.
1018 * tree-vect-data-refs.c (everywhere): Adjust users.
1019 * tree-vect-loop.c (everywhere): Likewise.
1020 * tree-vect-slp.c (everywhere): Likewise.
1021 * tree-vect-stmts.c (everywhere): Likewise.
1022 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
1024 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1026 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
1028 (gcc_cv_as_section_exclude): ... this.
1029 Try Solaris as #exclude syntax.
1030 * configure: Regenerate.
1031 * config.in: Regenerate.
1032 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
1034 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
1035 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1037 * varasm.c (default_elf_asm_named_section): Don't check if
1038 HAVE_GAS_SECTION_EXCLUDE is defined.
1040 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1042 * doc/md.texi: Update the documentation of the cond_* optabs
1043 to mention the new final operand. Fix GET_MODE_NUNITS call.
1044 Describe the scalar case too.
1045 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1046 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1048 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1049 (get_conditional_internal_fn): Update comment.
1050 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1051 accumulator value as a final argument to conditional functions.
1052 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1053 a define_expand and add an "else" operand. Assert for now that
1054 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1055 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1056 (*cond_<optab><mode>): New patterns.
1057 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1058 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1059 (UNSPEC_COND_EOR): Delete.
1060 (optab): Remove associated mappings.
1061 (SVE_INT_BINARY): New code iterator.
1062 (sve_int_op): Remove int attribute and add "minus" to the code
1064 (SVE_COND_INT_OP): Delete.
1065 (SVE_COND_FP_OP): Rename to...
1066 (SVE_COND_FP_BINARY): ...this.
1068 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1070 * optabs.c (can_reuse_operands_p): New function.
1071 (maybe_legitimize_operands): Try to reuse the results for
1074 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1076 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1077 Add {q} suffix to insn mnemonic.
1079 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1081 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1082 (msp430_warn_func_return): New.
1084 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1086 * fold-const.c (tree_nonzero_bits): New function.
1087 * fold-const.h (tree_nonzero_bits): Likewise.
1088 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1089 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1091 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1095 * varasm.c (assemble_alias): Check ifunc_resolver only on
1098 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1101 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1102 when memory input operand is handled.
1104 2018-05-24 Luis Machado <luis.machado@linaro.org>
1106 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1108 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1110 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1112 * match.pd: Delay FMA folds until after vectorization.
1114 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1117 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1118 address check not strict.
1120 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1122 * gimple-match.h (gimple_match_op): New class.
1123 (mprts_hook): Replace parameters with a gimple_match_op *.
1124 (maybe_build_generic_op): Likewise.
1125 (gimple_simplified_result_is_gimple_val): Replace parameters with
1126 a const gimple_match_op *.
1127 (gimple_simplify): Replace code_helper * and tree * parameters with
1128 a gimple_match_op * parameter.
1129 (gimple_resimplify1): Replace code_helper *, tree and tree *
1130 parameters with a gimple_match_op * parameter.
1131 (gimple_resimplify2): Likewise.
1132 (gimple_resimplify3): Likewise.
1133 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1134 parameters with a gimple_match_op * parameter.
1135 * gimple-match-head.c (gimple_simplify): Change prototypes of
1136 auto-generated functions to take a gimple_match_op * instead of
1137 separate code_helper * and tree * parameters. Make the same
1138 change in the top-level overload and update calls to the
1139 gimple_resimplify routines. Update calls to the auto-generated
1140 functions and to maybe_push_res_to_seq in the publicly-facing
1141 operation-specific gimple_simplify overloads.
1142 (gimple_match_op::MAX_NUM_OPS): Define.
1143 (gimple_resimplify1): Replace rcode and ops with a single res_op
1144 parameter. Update call to gimple_simplify.
1145 (gimple_resimplify2): Likewise.
1146 (gimple_resimplify3): Likewise.
1147 (mprts_hook): Replace parameters with a gimple_match_op *.
1148 (maybe_build_generic_op): Likewise.
1149 (build_call_internal): Replace type, nargs and ops with
1150 a gimple_match_op *.
1151 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1152 with a single gimple_match_op *. Update calls to mprts_hook,
1153 build_call_internal and gimple_simplified_result_is_gimple_val.
1154 Factor out code that is common to the tree_code and combined_fn cases.
1155 * genmatch.c (expr::gen_transform): Replace tem_code and
1156 tem_ops with a gimple_match_op called tem_op. Update calls
1157 to the gimple_resimplify functions and maybe_push_res_to_seq.
1158 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1159 res_ops. Update call to the gimple_resimplify functions.
1160 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1161 (decision_tree::gen): Make the functions take a gimple_match_op *
1162 called res_op instead of separate res_code and res_ops parameters.
1163 Update call accordingly.
1164 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1165 and ops with a single res_op parameter. Update calls to
1166 maybe_build_generic_op and maybe_push_res_to_seq.
1167 (fold_stmt_1): Update calls to gimple_simplify and
1168 replace_stmt_with_simplification.
1169 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1170 and gimple_simplified_result_is_gimple_val.
1171 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1173 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1174 with a gimple_match_op *.
1175 (vn_nary_build_or_lookup): Likewise. Update call to
1176 vn_nary_build_or_lookup_1.
1177 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1178 gimple_match_op *. Update calls to the gimple_resimplify routines
1179 and to gimple_simplified_result_is_gimple_val.
1180 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1181 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1182 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1183 (visit_nary_op): Likewise.
1184 (visit_reference_op_load): Likewise.
1186 2018-05-23 Luis Machado <luis.machado@linaro.org>
1188 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1189 modifier for printing the step amount.
1191 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1194 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1197 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1199 * doc/sourcebuild.texi (Endianness): New subsubsection.
1201 2018-05-23 Luis Machado <luis.machado@linaro.org>
1203 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1204 <prefetch_dynamic_strides>: New const bool field.
1205 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1206 prefetch_dynamic_strides.
1207 (exynosm1_prefetch_tune): Likewise.
1208 (thunderxt88_prefetch_tune): Likewise.
1209 (thunderx_prefetch_tune): Likewise.
1210 (thunderx2t99_prefetch_tune): Likewise.
1211 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1213 (aarch64_override_options_internal): Update to set
1214 PARAM_PREFETCH_DYNAMIC_STRIDES.
1215 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1216 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1217 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1218 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1219 prefetch-dynamic-strides setting.
1221 2018-05-23 Luis Machado <luis.machado@linaro.org>
1223 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1224 <minimum_stride>: New const int field.
1225 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1226 minimum_stride field defaulting to -1.
1227 (exynosm1_prefetch_tune): Likewise.
1228 (thunderxt88_prefetch_tune): Likewise.
1229 (thunderx_prefetch_tune): Likewise.
1230 (thunderx2t99_prefetch_tune): Likewise.
1231 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1232 <default_opt_level>: Set to 3.
1233 (aarch64_override_options_internal): Update to set
1234 PARAM_PREFETCH_MINIMUM_STRIDE.
1235 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1236 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1237 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1238 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1239 stride is constant and is below the minimum stride threshold.
1241 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1243 * config/arm/arm-cpus.in (mode26): Delete.
1244 (armv4): Delete mode26 reference.
1245 * config/arm/arm.c (arm_configure_build_target): Delete use of
1248 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1250 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1252 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1253 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1255 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1256 (floatunsdidf2): Ditto.
1258 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1260 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1261 (fixuns_trunc<mode>si2_avx512f): Ditto.
1262 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1263 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1264 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1266 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1268 PR rtl-optimization/79985
1269 * df-scan.c (df_insn_refs_collect): Remove special case for
1270 global registers and asm statements.
1272 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1274 * extend.texi (Global Register Variables): Rewrite the bullet list.
1275 Note that the register is available for allocation. Note that access
1276 via inline asm must use constraints. Add note about async-signal
1277 handlers. Remove paragraph about automagic register selection.
1279 2018-05-23 Richard Biener <rguenther@suse.de>
1281 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1282 of fixed offset from memset VN.
1284 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1286 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1288 (alloc_cand_and_find_basis): Initialize first_interp field.
1289 (slsr_process_mul): Modify first_interp field.
1290 (slsr_process_add): Likewise.
1291 (slsr_process_cast): Modify first_interp field for each new
1293 (slsr_process_copy): Likewise.
1294 (dump_candidate): Dump first_interp field.
1295 (replace_mult_candidate): Process all interpretations, not just
1297 (replace_rhs_if_not_dup): Likewise.
1298 (replace_one_candidate): Likewise.
1300 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1302 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1304 (aarch64_needs_frame_chain): New function.
1305 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1307 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1310 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1311 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1312 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1313 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1314 as true for strict-align.
1315 (aarch64_can_inline_p): Perform checks even when callee has no
1316 attributes to check for strict alignment.
1317 * doc/extend.texi (AArch64 Function Attributes): Document
1319 * doc/invoke.texi: (AArch64 Options): Likewise.
1321 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1323 PR tree-optimization/85853
1324 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1325 the handling of the root of the node to...
1326 (vect_slp_analyze_node_operations_1): ...this new function,
1327 and run the whole thing with the child nodes' def types
1328 set according to their SLP node's def type.
1330 2018-05-23 Richard Biener <rguenther@suse.de>
1333 * tree-data-ref.c (create_runtime_alias_checks): Defer
1334 and ignore overflow warnings.
1336 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1338 PR tree-optimization/85822
1339 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1342 2018-05-23 Richard Biener <rguenther@suse.de>
1344 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1345 memset constants via native_interpret_expr.
1347 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1350 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1352 (cgraph_node::create_alias): Likewise.
1353 (cgraph_node::get_availability): Check ifunc_resolver instead
1354 of looking up ifunc attribute.
1355 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1356 * varasm.c (do_assemble_alias): Likewise.
1357 (assemble_alias): Likewise.
1358 (default_binds_local_p_3): Likewise.
1359 * cgraph.h (cgraph_node): Add ifunc_resolver.
1360 (cgraph_node::only_called_directly_or_aliased_p): Return false
1362 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1364 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1365 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1366 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1367 instead of looking up ifunc attribute.
1369 2018-05-22 Luis Machado <luis.machado@linaro.org>
1371 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1373 2018-05-22 Martin Sebor <msebor@redhat.com>
1376 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1377 only when expasion succeeds.
1378 (expand_builtin_strcmp): Same.
1379 (expand_builtin_strncmp): Same.
1381 2018-05-22 Martin Sebor <msebor@redhat.com>
1383 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1385 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1386 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1388 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1389 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1391 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1392 Add check for legitimate_address.
1393 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1394 (aarch64_swap_ldrstr_operands): New.
1395 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1398 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1399 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1401 * config/aarch64/aarch64.md: New patterns to generate stp
1403 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1404 single words and double words.
1405 (load_pair_sw, load_pair_dw): Likewise.
1406 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1408 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1410 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1411 for different mode ldpstp and add peephole for merged zero stores.
1413 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1415 (aarch64_gen_store_pair): Rename calls to match new patterns.
1416 (aarch64_gen_load_pair): Rename calls to match new patterns.
1417 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1418 (load_pair<DREG:mode><DREG2:mode>): ... This.
1419 (store_pair<mode>): Rename to...
1420 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1421 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1423 (V_INT_EQUIV): Handle SImode.
1424 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1427 2018-05-22 Martin Sebor <msebor@redhat.com>
1430 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1431 or ajust the presumed bound on an operation to avoid unnecessary
1434 2018-05-22 Martin Sebor <msebor@redhat.com>
1436 PR tree-optimization/85826
1437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1438 assuming that a DECL necesarily has a constant size.
1440 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1443 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1445 2018-05-22 Richard Biener <rguenther@suse.de>
1447 PR tree-optimization/85834
1448 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1449 non-constant and non-zero memset arguments.
1451 2018-05-22 Martin Liska <mliska@suse.cz>
1454 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1456 2018-05-22 Richard Biener <rguenther@suse.de>
1458 PR tree-optimization/85863
1459 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1460 comparisons when vectype is specified.
1461 (vectorizable_condition): Do not specify vectype for
1462 vect_is_simple_cond when SLP vectorizing.
1464 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1467 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1468 define __ibm128 as long double.
1469 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1470 as a distinct type when IEEE 128-bit support is enabled.
1471 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1472 128-bit types to use the correct functions.
1473 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1474 convert between 128-bit floating point types that have different
1475 modes but the same representation, instead of using gen_lowpart to
1477 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1479 (IFKF_reg): New attributes to give the register constraints for
1481 (extend<mode>tf2_internal): New insns to mark an explicit
1482 conversion between 128-bit floating point types that have a
1483 different mode but share the same representation.
1485 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1487 PR tree-optimization/85814
1488 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1489 a null return from get_strinfo when unsharing the next
1490 strinfo in the chain.
1492 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1495 * varasm.c (weak_finish): Clean up weak_decls.
1497 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1499 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1500 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1501 UNSPEC_UADALP values.
1502 * config/aarch64/iterators.md (ABAL): New int iterator.
1505 (sur): Add mappings for the above.
1506 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1508 (aarch64_<sur>abal<mode>_4): Likewise.
1509 (aarch64_<sur>adalp<mode>_3): Likewise.
1510 (<sur>sadv16qi): New define_expand.
1512 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1514 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1515 (*movdf_internal): Ditto.
1516 (*rcpsf2_sse): Ditto.
1517 (*rsqrtsf2_sse): Ditto.
1518 (*sqrt<mode>2_sse): Ditto.
1520 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1522 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1524 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1525 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1526 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1527 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1529 * config/aarch64/arm_neon.h: Likewise.
1530 * config/aarch64/iterators.md (VQ_I): New.
1532 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1534 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1536 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1539 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1541 * config/nds32/constraints.md (S): New constraint.
1542 * config/nds32/nds32.md (call_internal): Use constraint S.
1543 (call_value_internal): Likewise.
1544 (sibcall_internal): Likewise.
1545 (sibcall_value_internal): Likewise.
1547 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1548 Chung-Ju Wu <jasonwucj@gmail.com>
1550 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1553 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1554 Chung-Ju Wu <jasonwucj@gmail.com>
1556 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1557 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1558 (nds32_rtx_costs_impl): Simplify.
1559 (nds32_address_cost_impl): Simplify.
1560 (nds32_init_rtx_costs): New function.
1561 (nds32_rtx_costs_speed_prefer): Likewise.
1562 (nds32_rtx_costs_size_prefer): Likewise.
1563 (nds32_address_cost_speed_prefer): Likewise.
1564 (nds32_address_cost_speed_fwprop): Likewise.
1565 (nds32_address_cost_size_prefer): Likewise.
1566 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1567 * config/nds32/nds32.c (nds32_option_override): Use
1568 nds32_init_rtx_costs function.
1570 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1572 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1573 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1574 (TARGET_PIPELINE_N8): Likewise.
1575 (TARGET_PIPELINE_N10): Likewise.
1576 (TARGET_PIPELINE_N13): Likewise.
1577 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1579 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1581 * config/nds32/nds32-fpu.md: Update copyright year.
1583 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1585 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1587 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1589 * config/nds32/nds32.c
1590 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1591 * config/nds32/nds32.opt (minline-asm-r15): New option.
1593 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1595 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1597 * config/nds32/nds32.md (abssi2): New pattern.
1599 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1601 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1602 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1603 Merge insn pattern from sse_cvtsi2ss<round_name> and
1604 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1605 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1606 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1607 using SWI48 mode iterator.
1608 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1609 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1610 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1611 pattern from sse_cvttss2si<round_saeonly_name>
1612 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1613 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1614 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1615 using SWI48 mode iterator.
1616 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1617 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1618 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1619 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1620 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1621 using SWI48 mode iterator.
1622 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1623 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1624 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1625 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1626 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1627 SWI48 mode iterator.
1628 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1629 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1630 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1631 pattern from sse_cvttsd2si<round_saeonly_name>
1632 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1634 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1636 * config/nds32/nds32-md-auxiliary.c
1637 (nds32_valid_smw_lwm_base_p): Refine.
1638 (nds32_output_smw_single_word): Refine.
1639 (nds32_output_smw_double_word): New.
1640 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1642 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1644 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1645 (nds32_output_stack_pop): Refine.
1646 (nds32_expand_unaligned_load): Refine.
1647 (nds32_expand_unaligned_store): Refine.
1649 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1650 Chung-Ju Wu <jasonwucj@gmail.com>
1652 * config/nds32/constants.md: Add TP_REGNUM constant.
1653 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1654 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1656 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1657 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1658 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1659 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1660 and PIC code generation.
1661 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1663 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1665 * config/nds32/nds32.md: Support TLS and PIC.
1666 * config/nds32/nds32.c: Support TLS and PIC.
1667 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1668 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1671 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1673 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1674 mode with E_ prefix.
1676 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1677 Chung-Ju Wu <jasonwucj@gmail.com>
1679 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1680 * config/nds32/nds32-md-auxiliary.c
1681 (symbolic_reference_mentioned_p): New.
1682 (nds32_legitimize_ict_address): New.
1683 (nds32_expand_ict_move): New.
1684 (nds32_indirect_call_referenced_p): New.
1685 (nds32_symbol_binds_local_p): Delete.
1686 (nds32_long_call_p): Modify.
1687 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1688 * config/nds32/nds32-protos.h
1689 (symbolic_reference_mentioned_p): Declare.
1690 (nds32_legitimize_ict_address): Declare.
1691 (nds32_expand_ict_move): Declare.
1692 (nds32_indirect_call_referenced_p): Declare.
1693 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1694 (nds32_relax_group): Use nds32_ict_const_p as condition.
1695 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1696 (nds32_asm_file_start): Output ict_model directive in asm code.
1697 (nds32_legitimate_address_p): Consider indirect call.
1698 (nds32_print_operand): Consider indirect call.
1699 (nds32_print_operand_address): Consider indirect call.
1700 (nds32_insert_attributes): Handle "indirect_call" attribute.
1701 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1702 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1703 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1704 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1705 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1706 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1707 (TARGET_ICT_MODEL_SMALL): Define.
1708 (TARGET_ICT_MODEL_LARGE): Define.
1709 * config/nds32/nds32.md (movsi): Consider ict model.
1710 (call, call_value): Consider ict model.
1711 (sibcall, sibcall_value): Consider ict model.
1712 * config/nds32/nds32.opt (mict-model): New option.
1713 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1716 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1717 Monk Chiang <sh.chiang04@gmail.com>
1718 Jim Wilson <jimw@sifive.com>
1720 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1721 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1722 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1723 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1724 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1725 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1726 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1727 compute save_libcall_adjustment properly.
1728 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1729 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1730 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1731 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1732 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1733 (ABI_SPEC): Handle mabi=ilp32e.
1734 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1735 (RVE): Add RVE mask.
1736 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1737 <-march>: Add rv32e as an example.
1739 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1742 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1743 (intra_create_variable_infos): Handle C++ constructors.
1745 2018-05-18 Martin Liska <mliska@suse.cz>
1747 * passes.def: Remove a redundant pass.
1749 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1752 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1754 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1756 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1758 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1760 (armv2, armv2a, armv3, armv3m): Delete architectures.
1761 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1762 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1763 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1765 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1766 (*mulsidi3adddi): Likewise.
1767 (mulsidi3): Likewise.
1768 (*mulsidi3_nov6): Likewise.
1769 (umulsidi3): Likewise.
1770 (umulsidi3_nov6): Likewise.
1771 (umaddsidi4): Likewise.
1772 (*umulsidi3adddi): Likewise.
1773 (smulsi3_highpart): Likewise.
1774 (*smulsi3_highpart_nov6): Likewise.
1775 (umulsi3_highpart): Likewise.
1776 (*umulsi3_highpart_nov6): Likewise.
1777 * config/arm/arm.h (arm_arch3m): Delete.
1778 * config/arm/arm.c (arm_arch3m): Delete.
1779 (arm_option_override_internal): Update armv3-related comment.
1780 (arm_configure_build_target): Delete use of isa_bit_mode32.
1781 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1782 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1783 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1785 (mulusa3): Likewise.
1786 * config/arm/arm-protos.h (arm_arch3m): Delete.
1787 * config/arm/arm-tables.opt: Regenerate.
1788 * config/arm/arm-tune.md: Likewise.
1789 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1790 deleted architectures.
1792 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1794 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1795 (armv5t, armv5te): New features.
1796 (ARMv5, ARMv5e): Delete fgroups.
1797 (ARMv5t, ARMv5te): Adjust for above changes.
1799 (armv5, armv5e): Delete arches.
1800 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1802 (*call_reg_arm): Likewise.
1803 (*call_value_reg_armv5): Likewise.
1804 (*call_value_reg_arm): Likewise.
1805 (*call_symbol): Likewise.
1806 (*call_value_symbol): Likewise.
1807 (*sibcall_insn): Likewise.
1808 (*sibcall_value_insn): Likewise.
1810 (prefetch): Likewise.
1811 (define_split and define_peephole2 dependent on arm_arch5):
1813 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1815 (TARGET_ARM_QBIT): Likewise.
1816 (TARGET_DSP_MULTIPLY): Likewise.
1817 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1818 (arm_arch5, arm_arch5e): Delete.
1819 (arm_arch5t, arm_arch5te): Declare.
1820 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1821 (arm_arch5t): Declare.
1822 (arm_option_reconfigure_globals): Update for the above.
1823 (arm_options_perform_arch_sanity_checks): Update comment, replace
1824 use of arm_arch5 with arm_arch5t.
1825 (use_return_insn): Likewise.
1826 (arm_emit_call_insn): Likewise.
1827 (output_return_instruction): Likewise.
1828 (arm_final_prescan_insn): Likewise.
1829 (arm_coproc_builtin_available): Likewise.
1830 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1831 arm_arch5e with arm_arch5t and arm_arch5te.
1832 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1833 (arm_arch5t, arm_arch5te): Declare.
1834 * config/arm/arm-tables.opt: Regenerate.
1835 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1836 * config/arm/t-multilib: Likewise.
1837 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1838 instead of arm_arch5.
1839 (*call_reg_thumb1): Likewise.
1840 (*call_value_reg_thumb1_v5): Likewise.
1841 (*call_value_reg_thumb1): Likewise.
1842 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1844 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1846 2018-05-18 Martin Liska <mliska@suse.cz>
1848 PR gcov-profile/84846
1849 * doc/gcov.texi: Document -t option of gcov tool.
1851 2018-05-18 Martin Liska <mliska@suse.cz>
1853 PR gcov-profile/84846
1854 * gcov.c (print_usage): Add new -t option.
1855 (process_args): Handle the option.
1856 (generate_results): Use stdout as output when requested by
1859 2018-05-18 Martin Liska <mliska@suse.cz>
1861 PR gcov-profile/84846
1862 * coverage.c (coverage_init): Write PWD to .gcno file.
1863 * doc/gcov.texi: Document how working directory is printed.
1864 * gcov-dump.c (dump_gcov_file): Print PWD.
1865 * gcov.c (output_intermediate_file): Likewise.
1866 (read_graph_file): Read PWD string.
1867 (output_lines): Print PWD.
1869 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1872 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1873 for retval and return false if all args to phi are zero.
1875 2018-05-18 Richard Biener <rguenther@suse.de>
1877 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1879 (evrp_dom_walker::before_dom_children): Call it.
1881 2018-05-18 Richard Biener <rguenther@suse.de>
1883 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1884 results when processing array refs with variable index.
1886 2018-05-18 Toon Moene <toon@moene.org>
1888 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1889 directly after that of -floop-interchange. Indicate that both
1890 options are enabled by default when specifying -O3.
1892 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1894 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1895 iterator. Delete separate integer-mode vec_set<mode> expander.
1896 (aarch64_simd_vec_setv2di): Delete.
1897 (vec_setv2di): Delete.
1898 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1899 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1900 the "w, r" alternative.
1902 2018-05-18 Martin Liska <mliska@suse.cz>
1904 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1905 * tree-pass.h (make_pass_lower_switch_O0): New function.
1906 * tree-switch-conversion.c (node_has_low_bound): Remove.
1907 (node_has_high_bound): Likewise.
1908 (node_is_bounded): Likewise.
1909 (class pass_lower_switch): Make it a template type and create
1911 (pass_lower_switch::execute): Add template argument.
1912 (make_pass_lower_switch): New function.
1913 (make_pass_lower_switch_O0): New function.
1914 (do_jump_if_equal): Remove.
1915 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1916 all the hard work to tree optimization passes.
1918 2018-05-18 Martin Liska <mliska@suse.cz>
1920 * dbgcnt.c (limit_low): Renamed from limit.
1921 (limit_high): New variable.
1922 (dbg_cnt_is_enabled): Check for upper limit.
1923 (dbg_cnt): Adjust dumping.
1924 (dbg_cnt_set_limit_by_index): Add new argument for high
1926 (dbg_cnt_set_limit_by_name): Likewise.
1927 (dbg_cnt_process_single_pair): Parse new format.
1928 (dbg_cnt_process_opt): Use strtok.
1929 (dbg_cnt_list_all_counters): Remove 'value' and add
1931 * doc/invoke.texi: Document changes.
1933 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1935 * doc/sourcebuild.texi (scalar_all_fma): Document.
1936 * tree.def (FMA_EXPR): Delete.
1937 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1938 * internal-fn.c (ternary_direct): New macro.
1939 (expand_ternary_optab_fn): Likewise.
1940 (direct_ternary_optab_supported_p): Likewise.
1941 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1942 * builtins.c (fold_builtin_fma): Delete.
1943 (fold_builtin_3): Don't call it.
1944 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1945 * expr.c (expand_expr_real_2): Likewise.
1946 * fold-const.c (operand_equal_p): Likewise.
1947 (fold_ternary_loc): Likewise.
1948 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1949 * gimple.c (DEFTREECODE): Likewise.
1950 * gimplify.c (gimplify_expr): Likewise.
1951 * optabs-tree.c (optab_for_tree_code): Likewise.
1952 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1953 * tree-eh.c (operation_could_trap_p): Likewise.
1954 (stmt_could_throw_1_p): Likewise.
1955 * tree-inline.c (estimate_operator_cost): Likewise.
1956 * tree-pretty-print.c (dump_generic_node): Likewise.
1957 (op_code_prio): Likewise.
1958 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1959 * tree-ssa-operands.c (get_expr_operands): Likewise.
1960 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1961 * fold-const-call.h (fold_fma): Delete.
1962 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1963 CFN_FNMA and CFN_FNMS.
1965 * genmatch.c (combined_fn): New enum.
1966 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1967 (commutative_op): New function.
1968 (commutate): Use it. Handle more than 2 operands.
1969 (dt_operand::gen_gimple_expr): Use commutative_op.
1970 (parser::parse_expr): Allow :c to be used with non-binary
1971 operators if the commutative operand is known.
1972 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1973 CFN_FMS, CFN_FNMA and CFN_FNMS.
1974 (backprop::process_assign_use): Remove FMA_EXPR handling.
1975 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1976 (gen_hsa_fma): New function.
1977 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1978 IFN_FNMA and IFN_FNMS.
1979 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1980 * gimple-fold.h (follow_all_ssa_edges): Declare.
1981 * gimple-fold.c (follow_all_ssa_edges): New function.
1982 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1983 gimple_build interface and use follow_all_ssa_edges to fold the result.
1984 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1985 instead of checking for optabs directly.
1986 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1987 rather than FMA_EXPRs.
1988 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1989 call to IFN_FMA instead of an FMA_EXPR.
1991 2018-05-17 Jim Wilson <jimw@sifive.com>
1993 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1994 sign extended promoted subreg, and the range does not have the sign bit
1995 set, then do a sign extend.
1997 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1998 test, check for sign extended subreg and/or constant operands, and
1999 do a sign extend in that case.
2001 2018-05-17 Steve Ellcey <sellcey@cavium.com>
2003 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
2004 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
2006 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
2007 Change logics_shift_reg to logics_shift_imm.
2008 (thunderx2t99_fp_loadpair_basic): Delete.
2009 (thunderx2t99_fp_storepair_basic): Delete.
2010 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
2011 (thunderx2t99_asimd_polynomial): Delete.
2012 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
2013 and neon_fp_mul_d_scalar_q.
2014 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
2015 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
2016 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
2017 (thunderx2t99_asimd_lut): Add missing tbl types.
2018 (thunderx2t99_asimd_ext): Delete.
2019 (thunderx2t99_asimd_load1_1_mult): Delete.
2020 (thunderx2t99_asimd_load1_2_mult): Delete.
2021 (thunderx2t99_asimd_load1_ldp): New.
2022 (thunderx2t99_asimd_load1): New.
2023 (thunderx2t99_asimd_load2): Add missing *load2* types.
2024 (thunderx2t99_asimd_load3): New.
2025 (thunderx2t99_asimd_load4): New.
2026 (thunderx2t99_asimd_store1_1_mult): Delete.
2027 (thunderx2t99_asimd_store1_2_mult): Delete.
2028 (thunderx2t99_asimd_store2_mult): Delete.
2029 (thunderx2t99_asimd_store2_onelane): Delete.
2030 (thunderx2t99_asimd_store_stp): New.
2031 (thunderx2t99_asimd_store1): New.
2032 (thunderx2t99_asimd_store2): New.
2033 (thunderx2t99_asimd_store3): New.
2034 (thunderx2t99_asimd_store4): New.
2036 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2038 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2039 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2041 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2042 Segher Boessenkool <segher@kernel.crashing.org>
2045 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2048 2018-05-17 Richard Biener <rguenther@suse.de>
2050 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2051 for pruning loop and prune defs feeding only already visited PHIs.
2053 2018-05-17 Richard Biener <rguenther@suse.de>
2055 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2057 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2058 Richard Biener <rguenther@suse.de>
2060 PR tree-optimization/85793
2061 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2062 for VMAT_ELEMENTWISE.
2064 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2066 * internal-fn.h (lookup_internal_fn): Declare
2067 * internal-fn.c (lookup_internal_fn): New function.
2068 * gimple.c (gimple_build_call_from_tree): Handle calls to
2070 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2071 internal function names.
2072 * tree-pretty-print.c (dump_generic_node): Likewise.
2073 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2075 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2077 * gimple-fold.h (gimple_build): Make the function forms take
2078 combined_fn rather than built_in_function.
2079 (gimple_simplify): Likewise.
2080 * gimple-match-head.c (gimple_simplify): Likewise.
2081 * gimple-fold.c (gimple_build): Likewise.
2082 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2083 rather than gimple_build_call_internal.
2084 (get_initial_defs_for_reduction): Likewise.
2085 (vect_create_epilog_for_reduction): Likewise.
2086 (vectorizable_live_operation): Likewise.
2088 2018-05-17 Martin Liska <mliska@suse.cz>
2090 * gimple-ssa-sprintf.c (format_directive): Do not use
2091 space in between 'G_' and '('.
2093 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2096 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2097 even if the mask is not all ones.
2100 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2102 (ix86_gimple_fold_builtin): Likewise.
2105 * config/i386/i386.c: Include tree-vector-builder.h.
2106 (ix86_vector_shift_count): New function.
2107 (ix86_fold_builtin): Fold shift builtins by scalar count.
2108 (ix86_gimple_fold_builtin): Likewise.
2110 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2111 _mm512_setzero): New intrinsics.
2113 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2114 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2116 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2117 code generation for cases where splatting a value is not useful.
2118 * simplify-rtx.c (simplify_ternary_operation): Simplify
2119 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2120 mode to a vec_concat.
2122 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2124 * config.gcc: Support "goldmont-plus".
2125 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2127 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2128 PROCESSOR_GOLDMONT_PLUS.
2129 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2130 (processor_target_table): Add "goldmont-plus".
2131 (PTA_GOLDMONT_PLUS): Define.
2132 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2133 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2134 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2135 (fold_builtin_cpu): Add "goldmont-plus".
2136 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2137 (ix86_option_override_internal): Add "goldmont-plus".
2138 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2139 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2140 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2141 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2143 2018-05-17 Richard Biener <rguenther@suse.de>
2145 PR tree-optimization/85757
2146 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2147 remove defs that only feed that PHI from further processing.
2149 2018-05-16 Jim Wilson <jimw@sifive.com>
2151 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2153 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2155 2018-05-16 Mark Wielaard <mark@klomp.org>
2157 * dwarf2out.c (count_index_strings): New function.
2158 (output_indirect_strings): Call count_index_strings and generate
2159 header for dwarf_version >= 5.
2161 2018-05-16 Mark Wielaard <mark@klomp.org>
2163 * dwarf2out.c (dwarf_FORM): New function.
2164 (set_indirect_string): Use dwarf_FORM.
2165 (reset_indirect_string): Likewise.
2166 (size_of_die): Likewise.
2167 (value_format): Likewise.
2168 (output_die): Likewise.
2169 (add_skeleton_AT_string): Likewise.
2170 (output_macinfo_op): Likewise.
2171 (index_string): Likewise.
2172 (output_index_string_offset): Likewise.
2173 (output_index_string): Likewise.
2174 (count_index_strings): Likewise.
2176 2018-05-16 Carl Love <cel@us.ibm.com>
2178 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2179 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2181 2018-05-16 Martin Jambor <mjambor@suse.cz>
2183 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2184 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2186 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2188 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2189 (fnma<mode>4): Likewise.
2190 (fms<mode>4): Likewise.
2191 (fnms<mode>4): Likewise.
2192 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2193 (aarch64_fnma<mode>4): Likewise.
2194 (aarch64_fms<mode>4): Likewise.
2195 (aarch64_fnms<mode>4): Likewise.
2196 (aarch64_fnmadd<mode>4): Likewise.
2198 2018-05-16 Jason Merrill <jason@redhat.com>
2200 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2202 2018-05-16 Richard Biener <rguenther@suse.de>
2204 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2205 (dump_stmt_cost): Declare.
2206 (add_stmt_cost): Dump cost we add.
2207 (add_stmt_costs): New function.
2208 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2210 (vect_analyze_stmt): Adjust prototype.
2211 (vectorizable_condition): Likewise.
2212 (vectorizable_live_operation): Likewise.
2213 (vectorizable_reduction): Likewise.
2214 (vectorizable_induction): Likewise.
2215 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2216 cost vector to pass to vectorizable_ and record afterwards.
2217 (vect_model_reduction_cost): Take cost vector argument and adjust.
2218 (vect_model_induction_cost): Likewise.
2219 (vectorizable_reduction): Likewise.
2220 (vectorizable_induction): Likewise.
2221 (vectorizable_live_operation): Likewise.
2222 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2223 SLP_TREE_NUMBER_OF_VEC_STMTS.
2224 (vect_analyze_slp_cost_1): Remove.
2225 (vect_analyze_slp_cost): Likewise.
2226 (vect_slp_analyze_node_operations): Take visited args and
2227 a target cost vector. Avoid processing already visited stmt sets.
2228 (vect_slp_analyze_operations): Use a local cost vector to gather
2229 costs and register those of non-discarded instances.
2230 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2231 (vect_schedule_slp_instance): Remove copying of
2232 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2234 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2235 adding cost. Record cost entry location.
2236 (vect_prologue_cost_for_slp_op): Function to compute cost of
2237 a constant or invariant generated for SLP vect in the prologue,
2238 split out from vect_analyze_slp_cost_1.
2239 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2240 (vect_model_promotion_demotion_cost): Likewise.
2241 (vect_model_store_cost): Likewise, make static.
2242 (vect_model_load_cost): Likewise.
2243 (vectorizable_bswap): Add cost vector arg and adjust.
2244 (vectorizable_call): Likewise.
2245 (vectorizable_simd_clone_call): Likewise.
2246 (vectorizable_conversion): Likewise.
2247 (vectorizable_assignment): Likewise.
2248 (vectorizable_shift): Likewise.
2249 (vectorizable_operation): Likewise.
2250 (vectorizable_store): Likewise.
2251 (vectorizable_load): Likewise.
2252 (vectorizable_condition): Likewise.
2253 (vectorizable_comparison): Likewise.
2254 (can_vectorize_live_stmts): Likewise.
2255 (vect_analyze_stmt): Likewise.
2256 (vect_transform_stmt): Adjust calls to vectorizable_*.
2257 * tree-vectorizer.c: Include gimple-pretty-print.h.
2258 (dump_stmt_cost): New function.
2260 2018-05-16 Richard Biener <rguenther@suse.de>
2262 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2263 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2264 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2265 (check_name): New callback.
2266 (dse_classify_store): Track cycles via a visited bitmap of PHI
2267 defs and simplify handling of in-loop and across loop dead stores
2268 and properly fail for loop-variant refs. Handle byte-tracking with
2269 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2272 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2274 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2275 (vect_get_mask_type_for_stmt): Likewise.
2276 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2278 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2279 to determine the statement's vector type and the vector type that
2280 should be used for calculating nunits. Deal with cases in which
2281 the type has to be deferred.
2282 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2283 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2284 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2285 (vect_determine_vf_for_stmt): New functions, split out from...
2286 (vect_determine_vectorization_factor): ...here.
2287 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2288 (vect_get_mask_type_for_stmt): New functions, split out from
2289 vect_determine_vectorization_factor.
2291 2018-05-16 Richard Biener <rguenther@suse.de>
2293 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2294 verify the [VEC_]COND_EXPR embedded comparison.
2296 2018-05-15 Martin Sebor <msebor@redhat.com>
2298 PR tree-optimization/85753
2299 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2300 RECORD_TYPE in addition to ARRAY_TYPE.
2302 2018-05-15 Martin Sebor <msebor@redhat.com>
2305 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2307 2018-05-15 Richard Biener <rguenther@suse.de>
2309 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2310 add by_clobber_p one. Change algorithm to collect all defs
2311 representing uses we need to walk and try reducing them to
2312 a single one before failing.
2313 (dse_dom_walker::dse_optimize_stmt): Adjust.
2315 2018-05-13 Mark Wielaard <mark@klomp.org>
2317 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2318 (size_of_loc_descr): Likewise.
2319 (output_loc_operands): Likewise.
2320 (output_loc_operands_raw): Likewise.
2321 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2322 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2323 (hash_loc_operands): Likewise.
2324 (compare_loc_operands): Likewise.
2326 2018-05-14 Mark Wielaard <mark@klomp.org>
2328 * dwarf2out.c (count_index_addrs): New function.
2329 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2331 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2333 PR tree-optimization/83648
2334 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2335 return value as malloc candidate.
2337 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2340 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2341 as true in call to suggest_attribute.
2343 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2345 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2346 -mreadonly-in-sdata.
2348 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2350 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2352 (aarch64_crypto_aesd_fused): Likewise.
2354 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2356 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2357 (movsi_aarch64): Likewise.
2358 (load_pairsi): Likewise.
2359 (load_pairdi): Likewise.
2360 (store_pairsi): Likewise.
2361 (store_pairdi): Likewise.
2362 (load_pairsf): Likewise.
2363 (load_pairdf): Likewise.
2364 (store_pairsf): Likewise.
2365 (store_pairdf): Likewise.
2366 (zero_extend): Likewise.
2367 (trunc): Swap alternatives.
2368 (fcvt_target): Add '?' to prefer w over r.
2370 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2373 * config/i386/i386.md: Disallow non-commutative arithmetics in
2374 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2375 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2376 in the peephole2 before it.
2378 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2380 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2381 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2382 (ix86_handle_option): Handle -mcldemote.
2383 * config.gcc: New header.
2384 * config/i386/cldemoteintrin.h: New file.
2385 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2386 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2388 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2389 OPTION_MASK_ISA_CLDEMOTE.
2390 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2391 (ix86_valid_target_attribute_inner_p): Ditto.
2392 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2393 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2394 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2395 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2396 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2398 * config/i386/i386.opt: Add -mcldemote.
2399 * config/i386/x86intrin.h: New header.
2400 * doc/invoke.texi: Add -mcldemote.
2402 2018-05-14 Richard Biener <rguenther@suse.de>
2404 * doc/match-and-simplify.texi: Adjust :s documentation.
2406 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2408 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2409 intended memcpy size.
2410 (REORDER_45): Likewise.
2412 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2414 * sort.cc: New file.
2415 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2416 * vec.c (qsort_chk): Use gcc_qsort.
2417 * Makefile.in (OBJS-libcommon): Add sort.o.
2418 (build/sort.o): New target. Use it...
2419 (BUILD_RTL): ... here, and...
2420 (build/gencfn-macros): ... here, and...
2421 (build/genmatch): ... here.
2423 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2424 Chung-Ju Wu <jasonwucj@gmail.com>
2426 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2427 * config/nds32/nds32-graywolf.md: New file.
2428 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2429 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2431 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2432 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2433 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2434 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2435 * config/nds32/pipelines.md: Include n15 settings.
2437 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2438 Chung-Ju Wu <jasonwucj@gmail.com>
2440 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2441 * config/nds32/nds32-n13.md: New file.
2442 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2443 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2445 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2446 * config/nds32/nds32.md (pipeline_model): Add n13.
2447 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2448 * config/nds32/pipelines.md: Include n13 settings.
2450 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2451 Chung-Ju Wu <jasonwucj@gmail.com>
2453 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2454 * config/nds32/nds32-n10.md: New file.
2455 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2456 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2458 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2459 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2460 * config/nds32/nds32.md (pipeline_model): Add n10.
2461 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2462 * config/nds32/pipelines.md: Include n10 settings.
2464 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2465 Kito Cheng <kito.cheng@gmail.com>
2466 Chung-Ju Wu <jasonwucj@gmail.com>
2468 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2469 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2470 Add enum values for DSP extension instructions.
2471 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2473 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2474 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2476 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2477 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2478 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2479 * config/nds32/nds32-intrinsic.md: Likewise.
2480 * config/nds32/nds32_intrinsic.h: Likewise.
2481 * config/nds32/nds32-md-auxiliary.c: Likewise.
2482 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2483 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2484 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2485 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2486 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2487 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2488 TYPE_DMAC in switch statement.
2489 * config/nds32/nds32.c: New checking and implementation for DSP
2490 extension instructions.
2491 * config/nds32/nds32.h: Likewise.
2492 * config/nds32/nds32.md: Likewise.
2493 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2494 * config/nds32/predicates.md: Implement new predicates for DSP
2497 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2499 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2500 Reformat alternatives and attributes so it is easier to identify
2501 which constraints/attributes go with which instruction.
2502 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2503 (mov<mode>_softfloat32, FMOVE64): Likewise.
2504 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2505 (mov<mode>_softfloat64, FMOVE64): Likewise.
2507 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2509 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2511 (Basic PowerPC Built-in Functions): The new name of the
2512 subsection previously known as "PowerPC Built-in Functions".
2513 (Basic PowerPC Built-in Functions Available on all Configurations):
2515 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2516 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2517 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2518 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2520 2018-05-11 Martin Jambor <mjambor@suse.cz>
2523 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2526 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2529 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2531 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2533 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2534 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2535 (ix86_handle_option): Handle -mwaitpkg.
2536 * config.gcc: New header.
2537 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2538 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2539 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2541 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2542 OPTION_MASK_ISA_WAITPKG.
2543 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2544 (ix86_option_override_internal): Add PTA_WAITPKG.
2545 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2546 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2547 IX86_BUILTIN_TPAUSE.
2548 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2549 __builtin_ia32_umwait and __builtin_ia32_tpause.
2550 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2551 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2552 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2553 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2554 UNSPECV_TPAUSE): New.
2555 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2556 * config/i386/i386.opt: Add -mwaitpkg.
2557 * config/i386/waitpkgintrin.h: New file.
2558 * config/i386/x86intrin.h: New header.
2559 * doc/invoke.texi: Add -mwaitpkg.
2561 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2564 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2566 (cortex-m0): Use armv6s-m isa.
2567 (cortex-m0plus): Likewise.
2568 (cortex-m1): Likewise.
2569 (cortex-m0.small-multiply): Likewise.
2570 (cortex-m0plus.small-multiply): Likewise.
2571 (cortex-m1.small-multiply): Likewise.
2573 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2574 Jakub Jelinek <jakub@redhat.com>
2576 PR tree-optimization/85692
2577 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2578 source permute as well.
2580 2018-05-11 Martin Liska <mliska@suse.cz>
2583 * doc/extend.texi: Document LLVM style format for no_sanitize
2586 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2588 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2589 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2590 (mode_supports_vsx_dform_quad): Likewise.
2591 (mode_supports_vmx_dform): Move these functions to be next to the
2592 other mode_supports functions.
2593 (mode_supports_dq_form): Likewise.
2594 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2595 mode_supports_dq_form.
2596 (reg_offset_addressing_ok_p): Likewise.
2597 (offsettable_ok_by_alignment): Likewise.
2598 (rs6000_legitimate_offset_address_p): Likewise.
2599 (legitimate_lo_sum_address_p): Likewise.
2600 (rs6000_legitimize_address): Likewise.
2601 (rs6000_legitimize_reload_address): Likewise.
2602 (rs6000_secondary_reload_inner): Likewise.
2603 (rs6000_preferred_reload_class): Likewise.
2604 (rs6000_output_move_128bit): Likewise.
2606 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2608 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2609 Generate SImode target register for null target.
2610 <case IX86_BUILTIN_XGETBV>: Ditto.
2611 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2612 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2614 2018-05-10 Carl Love <cel@us.ibm.com>
2616 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2617 dcbtt and dcbtstt if operands[2] is 0.
2619 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2622 * config/i386/sse.md (usadv64qi): New expander.
2624 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2626 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2627 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2628 -maltivec=be support.
2629 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2630 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2631 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2632 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2633 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2634 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2635 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2636 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2637 altivec_vsumsws): Adjust.
2638 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2639 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2640 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2642 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2643 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2644 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2645 (altivec_lve<VI_char>x): Delete expand.
2646 (*altivec_lve<VI_char>x_internal): Rename to...
2647 (altivec_lve<VI_char>x): ... this.
2648 (altivec_lvxl_<mode>): Delete expand.
2649 (*altivec_lvxl_<mode>_internal): Rename to ...
2650 (altivec_lvxl_<mode>): ... this.
2651 (altivec_stvxl_<mode>): Delete expand.
2652 (*altivec_stvxl_<mode>_internal): Rename to ...
2653 (altivec_stvxl_<mode>): ... this.
2654 (altivec_stve<VI_char>x): Delete expand.
2655 (*altivec_stve<VI_char>x_internal): Rename to ...
2656 (altivec_stve<VI_char>x): ... this.
2657 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2658 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2659 reduc_plus_scal_<mode>): Adjust.
2660 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2662 (rs6000_cpu_cpp_builtins): Adjust.
2663 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2664 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2665 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2666 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2667 -maltivec=be support.
2668 (rs6000_split_vec_extract_var): Adjust.
2669 (rs6000_split_v4si_init): Adjust.
2670 (swap_selector_for_mode): Delete.
2671 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2672 altivec_expand_stvex_be): Delete.
2673 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2674 -maltivec=be support.
2675 (rs6000_gimple_fold_builtin): Ditto.
2676 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2678 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2679 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2680 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2681 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2682 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2683 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2684 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2685 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2686 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2687 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2688 anonymous split): Adjust.
2689 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2690 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2692 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2694 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2695 when --with-gxx-include-dir is also specified.
2696 * configure: Regenerate.
2698 2018-05-09 Jim Wilson <jimw@sifive.com>
2701 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2702 * config/riscv/t-withmultilib: New.
2703 * config/riscv/withmultilib.h: New.
2704 * doc/install.texi: Document RISC-V --with-multilib-list support.
2706 2018-05-09 Richard Biener <rguenther@suse.de>
2708 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2710 (vect_bb_vectorization_profitable_p): Adjust. Compute
2711 actual scalar cost using the cost vector and the add_stmt_cost
2714 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2716 PR rtl-optimization/85645
2717 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2718 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2720 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2722 PR rtl-optimization/85645
2723 * shrink-wrap.c (spread_components): Return a boolean saying if
2724 anything was changed.
2725 (try_shrink_wrapping_separate): Iterate spread_components until
2726 nothing changes anymore.
2728 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2730 PR rtl-optimization/85645
2731 * regrename.c (build_def_use): Also kill the chains that include the
2732 destination of a REG_CFA_REGISTER note.
2734 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2736 PR rtl-optimization/85645
2737 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2738 insn that has a REG_CFA_REGISTER note.
2740 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2742 * cfgexpand.c (expand_clobber): New function.
2743 (expand_gimple_stmt_1): Use it.
2744 * tree-vect-stmts.c (vect_clobber_variable): New function,
2746 (vectorizable_simd_clone_call): ...here.
2747 (vectorizable_store): Emit a clobber either side of an
2748 IFN_STORE_LANES sequence.
2749 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2751 2018-05-09 Tom de Vries <tom@codesourcery.com>
2754 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2755 (define_insn "trap_if_false"): Add exit after trap.
2757 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2759 PR rtl-optimization/85638
2760 * bb-reorder.c: Include common/common-target.h.
2761 (create_forwarder_block): New function extracted from...
2762 (fix_up_crossing_landing_pad): ...here. Rename into...
2763 (dw2_fix_up_crossing_landing_pad): ...this.
2764 (sjlj_fix_up_crossing_landing_pad): New function.
2765 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2766 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2767 from both partitions and exit the loop after one iteration.
2769 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2772 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2774 (Basic PowerPC Built-in Functions): The new name of the
2775 subsection previously known as "PowerPC Built-in Functions".
2776 (Basic PowerPC Built-in Functions Available on all Configurations):
2778 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2780 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2782 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2784 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2787 2018-05-08 Jim Wilson <jimw@sifive.com>
2789 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2790 (LD_EMUL_SUFFIX): New.
2791 (LINK_SPEC): Use it.
2793 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2795 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2797 (Basic PowerPC Built-in Functions): The new name of the
2798 subsection previously known as "PowerPC Built-in Functions".
2799 (Basic PowerPC Built-in Functions Available on all Configurations):
2801 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2803 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2805 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2807 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2810 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2813 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2814 after cmpelim optimization.
2816 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2818 * config.gcc: Support "goldmont".
2819 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2820 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2822 * config/i386/i386.c (m_GOLDMONT): Define.
2823 (processor_target_table): Add "goldmont".
2824 (PTA_GOLDMONT): Define.
2825 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2826 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2827 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2828 (fold_builtin_cpu): Add "goldmont".
2829 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2830 (ix86_option_override_internal): Add "goldmont".
2831 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2832 (processor_type): Add PROCESSOR_GOLDMONT.
2833 * config/i386/i386.md: Add CPU "glm".
2834 * config/i386/glm.md: New file.
2835 * config/i386/x86-tune.def: Add m_GOLDMONT.
2836 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2838 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2841 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2843 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2844 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2845 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2848 * config/i386/i386.c (ix86_fold_builtin): Handle
2849 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2852 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2853 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2855 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2858 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2859 (check_arch): Likewise.
2860 (check_fpu): Return the result rather than printing it.
2861 (end arch): Fix operator precedence.
2862 (end cpu): Likewise.
2863 (END): Print the result from check_fpu.
2865 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2866 Alan Hayward <alan.hayward@arm.com>
2867 David Sherwood <david.sherwood@arm.com>
2869 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2870 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2871 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2872 (*fcmuo<mode>_and): New patterns.
2874 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2876 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2877 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2878 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2879 (cmp_op, sve_imm_con): New code attributes.
2880 (SVE_COND_INT_CMP, imm_con): Delete.
2881 (cmp_op): Remove above unspecs from int attribute.
2882 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2884 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2885 comparison-specific unspecs.
2886 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2887 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2888 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2889 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2890 (*vec_fcm<cmp_op><mode>): Rename to...
2891 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2892 (*vec_fcmuo<mode>): Rename to...
2893 (*fcmuo<mode>): ...this and adjust likewise.
2894 (*pred_fcm<cmp_op><mode>): New pattern.
2895 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2896 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2898 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2900 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2901 (aarch64_emit_sve_predicated_cond): New function.
2902 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2903 (aarch64_emit_unspec_cond_or): Replace with...
2904 (aarch64_emit_sve_or_conds): ...this new function. Use
2905 aarch64_emit_sve_ptrue_op for the individual comparisons and
2906 aarch64_emit_binop to OR them together.
2907 (aarch64_emit_inverted_unspec_cond): Replace with...
2908 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2909 aarch64_emit_sve_ptrue_op for the comparison and
2910 aarch64_emit_unop to invert the result.
2911 (aarch64_expand_sve_vec_cmp_float): Update after the above
2912 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2914 2018-05-07 Nathan Sidwell <nathan@acm.org>
2916 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2917 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2918 (Backwards Compatibility): Likewise.
2920 2018-05-07 Luis Machado <luis.machado@linaro.org>
2924 2018-05-07 Luis Machado <luis.machado@linaro.org>
2926 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2927 <prefetch_dynamic_strides>: New const bool field.
2928 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2929 prefetch_dynamic_strides.
2930 (exynosm1_prefetch_tune): Likewise.
2931 (thunderxt88_prefetch_tune): Likewise.
2932 (thunderx_prefetch_tune): Likewise.
2933 (thunderx2t99_prefetch_tune): Likewise.
2934 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2935 (aarch64_override_options_internal): Update to set
2936 PARAM_PREFETCH_DYNAMIC_STRIDES.
2937 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2938 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2939 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2940 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2941 prefetch-dynamic-strides setting.
2943 2018-05-07 Luis Machado <luis.machado@linaro.org>
2945 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2946 <minimum_stride>: New const int field.
2947 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2948 minimum_stride field.
2949 (exynosm1_prefetch_tune): Likewise.
2950 (thunderxt88_prefetch_tune): Likewise.
2951 (thunderx_prefetch_tune): Likewise.
2952 (thunderx2t99_prefetch_tune): Likewise.
2953 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2954 (aarch64_override_options_internal): Update to set
2955 PARAM_PREFETCH_MINIMUM_STRIDE.
2956 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2957 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2958 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2959 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2960 stride is constant and is below the minimum stride threshold.
2962 2018-05-07 Luis Machado <luis.machado@linaro.org>
2964 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2967 2018-05-07 Luis Machado <luis.machado@linaro.org>
2969 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2970 <prefetch_dynamic_strides>: New const bool field.
2971 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2972 prefetch_dynamic_strides.
2973 (exynosm1_prefetch_tune): Likewise.
2974 (thunderxt88_prefetch_tune): Likewise.
2975 (thunderx_prefetch_tune): Likewise.
2976 (thunderx2t99_prefetch_tune): Likewise.
2977 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2978 (aarch64_override_options_internal): Update to set
2979 PARAM_PREFETCH_DYNAMIC_STRIDES.
2980 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2981 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2982 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2983 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2984 prefetch-dynamic-strides setting.
2986 2018-05-07 Luis Machado <luis.machado@linaro.org>
2988 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2989 <minimum_stride>: New const int field.
2990 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2991 minimum_stride field.
2992 (exynosm1_prefetch_tune): Likewise.
2993 (thunderxt88_prefetch_tune): Likewise.
2994 (thunderx_prefetch_tune): Likewise.
2995 (thunderx2t99_prefetch_tune): Likewise.
2996 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2997 (aarch64_override_options_internal): Update to set
2998 PARAM_PREFETCH_MINIMUM_STRIDE.
2999 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3000 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3001 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3002 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3003 stride is constant and is below the minimum stride threshold.
3005 2018-05-06 Jakub Jelinek <jakub@redhat.com>
3008 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
3009 the type is addressable. Don't force op into register if it has
3012 2018-05-05 Roland McGrath <mcgrathr@google.com>
3015 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
3016 any section for which we don't know a specific type it should have,
3017 regardless of name. Previously this was done only for the exact
3018 names ".init_array", ".fini_array", and ".preinit_array".
3019 (default_elf_asm_named_section): Add comment about
3020 relationship with default_section_type_flags and SECTION_NOTYPE.
3021 (get_section): Don't consider it a type conflict if one side has
3022 SECTION_NOTYPE and the other doesn't, as long as neither has the
3023 SECTION_BSS et al used in the default_section_type_flags logic.
3025 2018-05-05 Tom de Vries <tom@codesourcery.com>
3028 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
3029 (workaround_barsyncs): New function.
3030 (nvptx_reorg): Use workaround_barsyncs.
3031 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
3032 (define_expand "nvptx_membar_cta"): New define_expand.
3033 (define_insn "*nvptx_membar_cta"): New insn.
3035 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3037 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3038 To improve optimization opportunities.
3039 * builtin-types.def: The new needed builtin types for the above.
3041 2018-05-04 Richard Biener <rguenther@suse.de>
3043 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3044 * gimple-ssa-store-merging.c
3045 (imm_store_chain_info::output_merged_store): Remove redundant create,
3046 release split_store vector contents on failure.
3047 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3048 scalar stmt vector on cache hit.
3050 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3052 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3054 * config.gcc (powerpc-xilinx-eabi*): Remove.
3055 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3057 (fusion_addis_mem_combo_load): Ditto.
3058 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3060 (rs6000_cpu_cpp_builtins): Ditto.
3061 * config/rs6000/rs6000-linux.c
3062 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3063 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3064 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3066 (rs6000_setup_reg_addr_masks): Ditto.
3067 (rs6000_init_hard_regno_mode_ok): Ditto.
3068 (rs6000_option_override_internal): Ditto.
3069 (legitimate_lo_sum_address_p): Ditto.
3070 (rs6000_legitimize_address): Ditto.
3071 (rs6000_legitimize_reload_address): Ditto.
3072 (rs6000_legitimate_address_p): Ditto.
3073 (abi_v4_pass_in_fpr): Ditto.
3074 (setup_incoming_varargs): Ditto.
3075 (rs6000_gimplify_va_arg): Ditto.
3076 (rs6000_split_multireg_move): Ditto.
3077 (rs6000_savres_strategy): Ditto.
3078 (rs6000_emit_prologue_components): Ditto.
3079 (rs6000_emit_epilogue_components): Ditto.
3080 (rs6000_emit_prologue): Ditto.
3081 (rs6000_emit_epilogue): Ditto.
3082 (rs6000_elf_file_end): Ditto.
3083 (rs6000_function_value): Ditto.
3084 (rs6000_libcall_value): Ditto.
3085 * config/rs6000/rs6000.h: Ditto.
3086 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3087 (TARGET_MINMAX): ... this. New.
3088 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3089 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3090 (*movsi_internal1_single): Delete.
3091 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3092 mfpu=, mxilinx-fpu): Delete.
3093 * config/rs6000/singlefp.h: Delete.
3094 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3095 * config/rs6000/t-rs6000: Ditto.
3096 * config/rs6000/t-xilinx: Delete.
3097 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3098 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3099 (VStype_simple): Delete.
3100 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3101 * config/rs6000/xfpu.h: Delete.
3102 * config/rs6000/xfpu.md: Delete.
3103 * config/rs6000/xilinx.h: Delete.
3104 * config/rs6000/xilinx.opt: Delete.
3105 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3106 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3108 2018-05-04 Tom de Vries <tom@codesourcery.com>
3111 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3114 2018-05-04 Richard Biener <rguenther@suse.de>
3117 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3118 (expand_complex_div_wide): Likewise.
3119 (expand_complex_operations_1): Likewise.
3120 (expand_complex_libcall): Preserve EH info of the original stmt.
3121 (tree_lower_complex): Handle removed blocks.
3122 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3123 on complex multiplication and division libcall builtins.
3125 2018-05-04 Richard Biener <rguenther@suse.de>
3128 * fold-const.c (negate_expr_p): Restrict negation of operand
3129 zero of a division to when we know that can happen without
3131 (fold_negate_expr_1): Likewise.
3133 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3136 * real.h (real_nextafter): Declare.
3137 * real.c (real_nextafter): New function.
3138 * fold-const-call.c (fold_const_nextafter): New function.
3139 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3140 CASE_CFN_NEXTTOWARD.
3141 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3142 even when arg1_mode is different from arg0_mode.
3144 2018-05-03 Nathan Sidwell <nathan@acm.org>
3146 * doc/extend.texi (Deprecated Features): Remove
3148 (Backwards Compatibility): Likewise.
3149 * doc/invoke.texi (C++ Language Options): Likewise.
3150 (C++ Dialect Options): Likewise.
3152 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3155 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3156 _mm512_mask_mullox_epi64): New intrinsics.
3158 2018-05-03 Tom de Vries <tom@codesourcery.com>
3161 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3162 dump files): Add offload-tree.
3164 2018-05-03 Richard Biener <rguenther@suse.de>
3166 PR tree-optimization/85615
3167 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3168 to loops not nested in BBs loop father to avoid creating multi-entry
3171 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3173 PR tree-optimization/70291
3174 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3175 arguments. Change return type to tree. Emit libcall as a new
3176 statement rather than replacing existing one when inplace_p is true.
3177 (expand_complex_multiplication_components): New function.
3178 (expand_complex_multiplication): Expand floating-point complex
3179 multiplication using the above.
3180 (expand_complex_division): Rename inner_type parameter to type.
3181 Update expand_complex_libcall call-site.
3182 (expand_complex_operations_1): Update expand_complex_multiplication
3183 and expand_complex_division call-sites.
3185 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3188 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3189 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3190 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3191 the highest significant bit of the shift count mask is clear. In
3192 check whether and[sq]i3 is needed verify that all significant bits
3193 of the shift count other than the highest are set.
3195 2018-05-02 Tom de Vries <tom@codesourcery.com>
3198 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3199 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3200 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3201 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3202 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3203 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3204 __builtin_goacc_parlevel_size.
3206 2018-05-02 Richard Biener <rguenther@suse.de>
3208 PR tree-optimization/85597
3209 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3210 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3213 2018-05-02 Tom de Vries <tom@codesourcery.com>
3216 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3217 dump files): Add ltrans-tree.
3219 2018-05-02 Tom de Vries <tom@codesourcery.com>
3222 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3223 dump files): Add wpa-ipa.
3225 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3227 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3228 powerpc*-*-linux*paired* target.
3229 * config/rs6000/750cl.h: Delete.
3230 * config/rs6000/paired.h: Delete.
3231 * config/rs6000/paired.md: Delete.
3232 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3234 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3235 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3236 comment. Remove paired float support.
3237 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3238 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3240 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3241 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3243 * config/rs6000/rs6000.c: Remove paired float support.
3244 (paired_expand_vector_init, paired_expand_vector_move,
3245 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3246 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3247 paired_expand_builtin, paired_expand_predicate_builtin,
3248 paired_init_builtins): Delete.
3249 * config/rs6000/rs6000.h: Remove paired float support.
3250 * config/rs6000/rs6000.md: Remove paired float support.
3251 (move_from_CR_ov_bit): Delete.
3252 * config/rs6000/rs6000.opt (mpaired): Delete.
3253 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3254 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3256 2018-05-02 Richard Biener <rguenther@suse.de>
3259 * gimplify.c (gimplify_save_expr): When in SSA form allow
3260 SAVE_EXPRs to compute to SSA vars.
3262 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3265 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3266 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3267 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3268 clobber operands[2], instead use a new pseudo. Formatting fixes.
3270 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3272 PR tree-optimization/85586
3273 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3274 exit early for statements in the same group if the accesses are
3277 2018-05-02 Tom de Vries <tom@codesourcery.com>
3280 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3283 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3285 PR tree-optimization/85143
3286 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3288 2018-05-01 Tom de Vries <tom@codesourcery.com>
3291 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3292 not found" error message.
3294 2018-05-01 Tom de Vries <tom@codesourcery.com>
3297 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3298 * vec.c (test_ordered_remove_if): New function.
3299 (vec_c_tests): Call test_ordered_remove_if.
3300 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3301 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3302 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3303 VEC_ORDERED_REMOVE_IF.
3305 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3307 PR tree-optimization/82665
3308 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3309 pointer subtraction where arguments come from a memchr call.
3311 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3313 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3314 --push-state --as-needed and --pop-state instead of --as-needed and
3315 --no-as-needed if ld supports it.
3316 * configure: Regenerated.
3319 * doc/install.texi2html: Replace _002d with - and _002a with * in
3320 generated html files using sed.
3322 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3325 * gcc-rich-location.c (blank_line_before_p): New function.
3326 (use_new_line): New function.
3327 (gcc_rich_location::add_fixit_insert_formatted): New function.
3328 * gcc-rich-location.h
3329 (gcc_rich_location::add_fixit_insert_formatted): New function.
3331 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3333 * selftest.c (assert_streq): Rename "expected" and "actual" to
3334 "val1" and "val2". Extend NULL-handling to cover both inputs
3335 symmetrically, while still requiring both to be non-NULL for a pass.
3336 * selftest.h (assert_streq): Rename "expected" and "actual" to
3338 (ASSERT_EQ): Likewise.
3339 (ASSERT_EQ_AT): Likewise.
3340 (ASSERT_KNOWN_EQ): Likewise.
3341 (ASSERT_KNOWN_EQ_AT): Likewise.
3342 (ASSERT_NE): Likewise.
3343 (ASSERT_MAYBE_NE): Likewise.
3344 (ASSERT_MAYBE_NE_AT): Likewise.
3345 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3346 the assertion to pass.
3347 (ASSERT_STREQ_AT): Likewise.
3349 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3351 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3352 interaction with -pie.
3354 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3356 * selftest.h: Fix alphabetization of per-source-file selftest
3359 2018-04-30 Jason Merrill <jason@redhat.com>
3361 PR c++/61982 - dead stores to destroyed objects.
3362 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3365 2018-04-30 Jason Merrill <jason@redhat.com>
3367 * tree.c (build_clobber): New.
3368 * tree.h: Declare it.
3369 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3371 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3373 * diagnostic-show-locus.c (layout::layout): Update for
3374 location_get_source_line returning a char_span.
3375 (struct char_span): Move to input.h.
3376 (struct correction): Update for fields in char_span becoming
3378 (struct source_line): Update for location_get_source_line
3379 returning a char_span.
3380 (layout::print_line): Likewise.
3381 * edit-context.c (edited_file::print_content): Likewise.
3382 (edited_file::print_diff_hunk): Likewise.
3383 (edited_file::print_run_of_changed_lines): Likewise.
3384 (edited_file::get_num_lines): Likewise.
3385 (edited_line::edited_line): Likewise.
3386 * final.c (asm_show_source): Likewise.
3387 * input.c (location_get_source_line): Convert return type
3388 from const char * to char_span, losing the final "line_len"
3390 (dump_location_info): Update for the above.
3391 (get_substring_ranges_for_loc): Likewise. Use a char_span
3392 when handling the literal within the line.
3393 (test_reading_source_line): Update for location_get_source_line
3394 returning a char_span.
3395 * input.h (class char_span): Move here from
3396 diagnostic-show-locus.c, converting from a struct to a class.
3397 Make data members private.
3398 (char_span::operator bool): New.
3399 (char_span::length): New.
3400 (char_span::get_buffer): New.
3401 (char_span::operator[]): New.
3402 (char_span::subspan): Make const.
3403 (char_span::xstrdup): New.
3404 (location_get_source_line): Convert return type from const char *
3405 to char_span, losing the final "line_size" param.
3407 2018-04-30 Jan Hubicka <jh@suse.cz>
3409 * lto-wrapper.c (ltrans_priorities): New static var.
3410 (cmp_priority): New.
3411 (run_gcc): Read priorities and if doing parallel build order
3412 the Makefile by them.
3414 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3416 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3418 2018-04-30 Richard Biener <rguenther@suse.de>
3420 * tree-cfg.c (verify_address): Remove base argument, add
3421 flag whether to check TREE_ADDRESSABLE and do that.
3422 (verify_expr): Remove.
3423 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3424 (verify_gimple_assign_single): Likewise.
3425 (verify_gimple_switch): Likewise.
3426 (verify_expr_location_1): Dereference tp once. Add (disabled)
3427 piece from verify_expr.
3428 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3430 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3432 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3434 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3436 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3437 (small_data_pattern): Likewise.
3438 (arc_rewrite_small_data): Likewise.
3439 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3440 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3441 (get_symbol_alignment): New function.
3442 (legitimate_small_data_address_p): Likewise.
3443 (legitimate_scaled_address): Update, call
3444 legitimate_small_data_address_p.
3445 (output_sdata): New static variable.
3446 (arc_print_operand): Update how we handle small data operands.
3447 (arc_print_operand_address): Likewise.
3448 (arc_legitimate_address_p): Update, use
3449 legitimate_small_data_address_p.
3450 (arc_rewrite_small_data_p): Remove.
3451 (arc_rewrite_small_data_1): Likewise.
3452 (arc_rewrite_small_data): Likewise.
3453 (small_data_pattern): Likewise.
3454 (compact_sda_memory_operand): Update to use
3455 legitimate_small_data_address_p and get_symbol_alignment.
3456 (prepare_move_operands): Don't rewite sdata pattern.
3457 (prepare_extend_operands): Remove.
3458 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3460 (zero_extendqisi2): Likewise.
3461 (zero_extendhisi2): Likewise.
3462 (extendqihi2): Likewise.
3463 (extendqisi2): Likewise.
3464 (extendhisi2): Likewise.
3468 * config/arc/constraints.md (Usd): Change it to memory constraint.
3470 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3472 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3473 as source of std instructions.
3474 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3475 allow 6-bit constants as source for store instructions.
3476 (movdi_insn): Update instruction pattern to allow 6-bit constants
3477 as source for store instructions.
3479 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3481 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3483 2018-04-30 Nathan Sidwell <nathan@acm.org>
3484 Sandra Loosemore <sandra@codesourcery.com>
3486 * dumpfile.c (dump_open): Allow '-' for stdout.
3487 * doc/invoke.texi (Developer Options): Document dump filename
3488 determination early. Document stdin/stdout selection.
3490 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3492 Microblaze Target: PIC data text relative
3494 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3495 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3497 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3498 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3499 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3500 New addressing mode for data-text relative position indepenedent code.
3501 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3502 'ADDRESS_SYMBOLIC_TXT_REL'.
3503 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3504 (microblaze_legitimate_pic_operand): Exclude function calls from
3505 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3506 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3508 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3509 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3510 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3511 for 'address + offset'.
3512 (microblaze_expand_prologue): Add new function prologue call for
3514 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3515 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3516 table in case of TARGET_PIC_DATA_TEXT_REL.
3517 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3518 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3519 Add new macros 'UNSPEC_TEXT',
3520 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3521 + exclude function calls from 'UNSPEC_PLT' in case of data text
3523 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3524 new target hook for generating address diff vector tables in case of
3526 * doc/tm.texi : Regenerate.
3527 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3528 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3529 of addr diff vector generation.
3530 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3531 target hook definition.
3532 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3533 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3534 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3535 option for data text relative.
3537 2018-04-30 Richard Biener <rguenther@suse.de>
3539 * tree-chrec.h (evolution_function_is_constant_p): Remove
3541 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3543 2018-04-30 Richard Biener <rguenther@suse.de>
3546 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3548 2018-04-30 Richard Biener <rguenther@suse.de>
3550 PR tree-optimization/28364
3551 PR tree-optimization/85275
3552 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3553 copying first exit test.
3555 2018-04-28 Mark Wielaard <mark@klomp.org>
3557 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3559 (dwarf_AT): Handle DW_AT_addr_base.
3560 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3562 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3565 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3566 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3567 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3568 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3570 2018-04-28 Richard Biener <rguenther@suse.de>
3572 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3573 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3574 to reflect use. Only add interesting stmts.
3576 2018-04-27 Martin Jambor <mjambor@suse.cz>
3579 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3580 the jump function allows for passing through aggregate values.
3582 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3584 * input.h (in_system_header_at): Convert from macro to inline
3586 (from_macro_expansion_at): Likewise.
3587 (from_macro_definition_at): Likewise.
3589 2018-04-27 Jeff Law <law@redhat.com>
3591 * config.gcc: Mark tile* targets as deprecated/obsolete.
3593 2018-04-27 Richard Biener <rguenther@suse.de>
3595 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3598 2018-04-27 Richard Biener <rguenther@suse.de>
3600 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3602 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3604 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3605 with Yd constraint. Set "preferred_for_speed" attribute from
3606 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3608 (*movdi_internal): Ditto.
3609 (movti_interunit splitters): Remove
3610 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3611 (movdi_interunit splitters): Ditto.
3612 * config/i386/constraints.md (Ye): Remove.
3613 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3615 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3618 * config/aarch64/constraints.md (Usg): Limit to 31.
3621 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3623 PR tree-optimization/85529
3624 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3625 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3626 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3627 zero extension or masking of the MSB bit.
3628 (optimize_range_tests): Add FIRST_BB argument, pass it through
3629 to optimize_range_tests_var_bound.
3630 (maybe_optimize_range_tests, reassociate_bb): Adjust
3631 optimize_range_tests callers.
3633 2018-04-26 Richard Biener <rguenther@suse.de>
3634 Jakub Jelinek <jakub@redhat.com>
3636 * cgraph.h (symbol_table): Just declare debug method here.
3637 * symtab.c (symbol_table::debug): Define.
3639 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3641 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3643 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3645 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3646 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3647 (*movdi_internal): Substitute Yi and Yj constraint with x
3648 and Ym and Yn constraint with y constraint. Update "isa"
3649 attribute and set "preferred_for_speed" attribute from
3650 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3651 (*movsi_internal): Ditto.
3652 (*movdf_internal): Ditto.
3653 (*movsf_internal): Ditto.
3654 (*zero_extendsidi2): Ditto.
3655 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3656 (sse2_loadld): Ditto.
3657 (*vec_extract<ssevecmodelower>_0): Ditto.
3658 (*vec_extractv4si_0_zext_sse4): Ditto.
3659 (vec_concatv2di): Ditto.
3660 (*vec_dup<mode>): Ditto.
3661 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3662 * config/i386/constraints.md (Yi): Remove.
3667 2018-04-26 Nathan Sidwell <nathan@acm.org>
3669 * dumpfile.c (dump_open): New.
3670 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3671 (dump_finish): Detect stdio/stderr by value not name.
3673 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3675 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3677 2018-04-26 Tom de Vries <tom@codesourcery.com>
3680 * config/nvptx/nvptx.c (verify_neutering_jumps)
3681 (verify_neutering_labels): New function
3682 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3684 2018-04-26 Tom de Vries <tom@codesourcery.com>
3687 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3688 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3691 2018-04-26 Richard Biener <rguenther@suse.de>
3692 Tom de Vries <tom@codesourcery.com>
3695 * lto-streamer-out.c (output_function): Fixup loops if required to match
3696 discovery done in the reader.
3698 2018-04-26 Richard Biener <rguenther@suse.de>
3700 PR tree-optimization/85116
3701 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3702 have a loop exit from the single latch predecessor. Remove
3703 case of header with just condition.
3704 (ch_base::copy_headers): Exclude infinite loops from any
3706 (pass_ch::execute): Record exits.
3708 2018-04-26 Richard Biener <rguenther@suse.de>
3710 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3711 prologue cost vector and pass it to vect_get_load_cost.
3712 (vect_get_peeling_costs_all_drs): Likewise.
3713 (vect_peeling_hash_get_lowest_cost): Likewise.
3714 (vect_enhance_data_refs_alignment): Likewise.
3716 2018-04-26 Richard Biener <rguenther@suse.de>
3719 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3720 checking of integer<->pointer conversions.
3721 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3722 sign-/zero-extending pointer types.
3723 (expand_omp_for_static_chunk): Likewise.
3725 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3726 Jean Lee <xiaoyur347@gmail.com>
3728 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3729 (TARGET_ASAN_SHADOW_OFFSET): Define.
3730 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3731 true for -fsanitize=address.
3733 2018-04-25 Mark Wielaard <mark@klomp.org>
3735 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3738 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3740 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3741 than "alu", remove explicit "memory" and "imm_disp" attributes.
3742 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3745 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3746 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3747 gen_lowpart_no_emit.
3749 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3752 * config/i386/i386.c (ix86_expand_builtin): Change memory
3753 operand to XI, extend p0 to Pmode.
3754 * config/i386/i386.md: Change unspec volatile and operand
3755 1 mode to XI, change operand 0 mode to P.
3757 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3759 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3760 GET_MODE_MASK before any checking.
3761 (nds32_can_use_bset_p): Likewise.
3762 (nds32_can_use_btgl_p): Likewise.
3764 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3766 * config/nds32/nds32-doubleword.md: New define_split pattern for
3767 illegal register number.
3769 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3771 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3773 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3775 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3777 2018-04-25 Richard Biener <rguenther@suse.de>
3779 * lto-streamer.h (LTO_major_version): Bump to 8.
3781 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3783 * BASE-VER: Set to 9.0.0.
3785 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3787 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3788 in __abskf2 and __powikf2.
3790 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3793 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3794 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3795 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3796 Use the above on operand 2. Reindent.
3797 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3799 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3802 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3804 * config/i386/i386.opt (mcet): Removed.
3805 * doc/install.texi: Remove -mcet documentation.
3806 * doc/invoke.texi: Likewise.
3808 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3811 * doc/install.texi: Remove -mcet from bootstrap-cet.
3813 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3816 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3817 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3821 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3822 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3823 containing a CONST_VECTOR.
3825 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3827 * doc/install.texi: Update newlib dependency for nvptx.
3829 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3832 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3833 instead of INTVAL when shifting x left.
3835 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3837 PR tree-optimization/85478
3838 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3839 vect_grouped_store_supported for single element vectors.
3841 2018-04-24 Richard Biener <rguenther@suse.de>
3844 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3845 load cost increase to the case of non-constant step.
3847 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3850 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3851 destination if any_malformed_asm.
3853 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3856 * expr.c (store_field): In the bitfield case, if the value comes from
3857 a function call and is returned in registers by means of a PARALLEL,
3858 do not change the mode of the temporary unless BLKmode and VOIDmode.
3860 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3862 PR rtl-optimization/85423
3863 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3864 dependencies to debug insns when the previous insn is non-debug.
3866 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3868 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3869 enums into a single definition.
3870 (fls): Fix predicates and printing.
3873 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3875 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3876 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3877 and short u6 immediate.
3878 (check_if_valid_sleep_operand): Remove.
3879 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3881 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3883 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3884 flag_always_save_lp condition.
3885 * config/nds32/nds32.opt (malways-save-lp): New option.
3887 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3889 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3890 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3891 * config/nds32/nds32.h
3892 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3893 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3895 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3897 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3898 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3900 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3901 Chung-Ju Wu <jasonwucj@gmail.com>
3903 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3905 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3906 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3908 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3910 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3912 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3914 * config/nds32/nds32-protos.h (nds32_data_alignment,
3915 nds32_local_alignment): Declare.
3916 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3917 nds32_local_alignment): New functions.
3918 (TARGET_CONSTANT_ALIGNMENT): Define.
3919 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3921 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3923 * config/nds32/nds32.c
3924 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3925 (TARGET_MODES_TIEABLE_P): Likewise.
3927 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3929 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3932 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3933 Chung-Ju Wu <jasonwucj@gmail.com>
3935 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3936 for unaligned access.
3937 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3939 * config/nds32/nds32-intrinsic.md: Likewise.
3940 * config/nds32/nds32_intrinsic.h: Likewise.
3941 * config/nds32/nds32.h (nds32_builtins): Likewise.
3942 * config/nds32/nds32.opt (munaligned-access): New option.
3943 * config/nds32/nds32.c (nds32_asm_file_start): Display
3944 flag_unaligned_access status.
3946 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3948 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3949 -mno-relax is present.
3950 * config/riscv/linux.h (LINK_SPEC): Ditto.
3952 2018-04-20 Martin Sebor <msebor@redhat.com>
3955 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3957 (gimple_fold_builtin_stxcpy_chk): Same.
3958 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3960 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3963 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3964 __powikf2 when long double is IEEE 128-bit.
3966 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3968 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3969 step to make sure stack always aligned.
3971 2018-04-20 Carl Love <cel@us.ibm.com>
3974 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3975 size check for arg0.
3977 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3978 Tom de Vries <tom@codesourcery.com>
3981 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3982 Emit insns for calls too.
3983 (nvptx_find_par): Always look for worker-level predecessor insn.
3984 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3986 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3987 (nvptx_process_pars): Propagate frames for calls.
3989 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3992 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3994 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3995 (ix86_handle_option): Don't handle OPT_mibt.
3996 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3998 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
4000 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
4001 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
4002 (ix86_target_macros): Define __CET__ with flag_cf_protection
4003 for -fcf-protection.
4004 * config/i386/i386.c (isa2_opts): Remove -mibt.
4005 * config/i386/i386.h (TARGET_IBT): Removed.
4006 (TARGET_IBT_P): Likewise.
4007 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
4008 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
4009 * config/i386/i386.opt (mcet): Update help message.
4012 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
4013 -mcet as an alias for -mshstk.
4015 2018-04-20 Richard Biener <rguenther@suse.de>
4018 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
4019 complexity by forcing a single use of the multiply operand.
4021 2018-04-20 Martin Jambor <mjambor@suse.cz>
4024 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
4025 recursion dependency to only apply to non-clones.
4027 2018-04-20 Martin Jambor <mjambor@suse.cz>
4030 * ipa-cp.c (create_specialized_node): Check that clones of
4031 self-recursive edges exist during IPA-CP.
4033 2018-04-19 Toon Moene <toon@moene.org>
4035 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4038 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4040 PR tree-optimization/85467
4041 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4042 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4043 VECTOR_CST element to type.
4045 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4048 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4049 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4050 (builtin_longjmp): Likewise.
4051 (save_stack_nonlocal): New pattern.
4052 (restore_stack_nonlocal): Likewise.
4054 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4057 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4058 Replace ASM_OUTPUT_LABEL with fprintf.
4060 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4063 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4064 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4065 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4066 define __IBT__ and __SHSTK__ for -fcf-protection.
4067 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4069 (ix86_trampoline_init): Likewise.
4070 (x86_output_mi_thunk): Likewise.
4071 (ix86_notrack_prefixed_insn_p): Likewise.
4072 (ix86_option_override_internal): Don't disallow -fcf-protection.
4073 * config/i386/i386.md (rdssp<mode>): Also enable for
4075 (incssp<mode>): Likewise.
4076 (nop_endbr): Likewise.
4077 * config/i386/i386.opt (mcet): Change help message to built-in
4081 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4082 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4083 enable CET built-in functions.
4085 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4087 * common/config/i386/i386-common.c
4088 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4089 OPTION_MASK_ISA_MOVDIRI_UNSET,
4090 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4091 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4092 * config.gcc (movdirintrin.h): New header.
4093 * config/i386/cpuid.h (bit_MOVDIRI,
4094 bit_MOVDIR64B): New bits.
4095 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4097 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4098 (VOID, PVOID, PCVOID)): New function types.
4099 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4100 __builtin_ia32_directstoreu_u64,
4101 __builtin_ia32_movdir64b): New builtins.
4102 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4103 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4105 (ix86_valid_target_attribute_inner_p): Ditto.
4106 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4107 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4108 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4109 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4110 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4111 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4112 (movdiri<mode>, movdir64b_<mode>): New.
4113 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4114 * config/i386/immintrin.h: Include movdirintrin.h.
4115 * config/i386/movdirintrin.h: New file.
4116 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4118 2018-04-19 Richard Biener <rguenther@suse.de>
4121 * cfg.c (clear_bb_flags): When loop state says we have
4122 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4124 2018-04-19 Richard Biener <rguenther@suse.de>
4126 PR tree-optimization/84737
4127 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4128 copying restrict info.
4129 (vect_setup_realignment): Use it.
4130 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4131 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4132 the first DR to all generated stores.
4133 (vectorizable_load): Likewise for loads.
4135 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4137 PR tree-optimization/85446
4138 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4139 the integral and pointer types to have the same precision.
4141 * doc/install.texi: Document --disable-cet being the default and
4144 2018-04-18 Martin Liska <mliska@suse.cz>
4146 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4149 2018-04-18 Martin Liska <mliska@suse.cz>
4152 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4155 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4156 arguments if they are comparable.
4158 2018-04-18 Martin Liska <mliska@suse.cz>
4161 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4164 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4167 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4170 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4171 ENDBR after calling __morestack.
4173 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4176 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4177 by using gcc_base_ver to generate a gcc_driver_version, and use
4178 it when generating GCC_DRIVER_NAME.
4179 * configure: Regenerate.
4181 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4184 * config.gcc: Obsolete powerpc*-*-*spe*.
4186 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4189 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4190 (stabstr_D): Change type of unum from unsigned int to
4191 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4194 2018-04-17 Jim Wilson <jimw@sifive.com>
4197 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4198 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4199 Set arg_pointer_offset after using pretend_args_size.
4201 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4203 PR rtl-optimization/85431
4204 * dse.c (record_store): Ignore zero width stores.
4207 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4208 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4209 __builtin_stack_restore rather than after it.
4210 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4211 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4212 argument instead of virtual_dynamic_stack_rtx.
4214 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4216 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4218 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4219 Add note to error message to explain internal mapping of overloaded
4220 built-in function name to non-overloaded built-in function name.
4221 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4224 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4227 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4228 where the inputs overlap with the output.
4230 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4232 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4233 (=v, v) alternative and explicit "memory" attribute.
4234 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4235 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4237 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4238 "sselog1" type instead of "sselog".
4239 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4240 "sselog". Remove explicit "memory" attribute.
4241 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4242 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4244 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4245 "sselog1" type instead of "sselog". Remove explicit "memory"
4247 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4248 use "sselog1" type instead of "sselog". Remove explicit "memory"
4250 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4251 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4253 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4254 "sselog1" type instead of "sselog". Remove explicit "memory"
4256 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4257 use "sselog1" type instead of "sselog". Remove explicit "memory"
4261 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4264 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4267 2018-04-17 Martin Jambor <mjambor@suse.cz>
4270 * ipa-cp.c (create_specialized_node): Call
4271 expand_all_artificial_thunks if necessary.
4273 2018-04-17 Martin Liska <mliska@suse.cz>
4276 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4277 in message, remote space in between '_G' and '('.
4279 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4282 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4283 avx512f_vmcmp<mode>3<round_saeonly_name>,
4284 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4285 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4286 avx512f_rndscale<mode><round_saeonly_name>,
4287 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4288 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4289 Use %<iptr>2 instead of %2 for -masm=intel.
4290 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4291 avx512f_vcvttss2usi<round_saeonly_name>,
4292 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4294 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4295 avx512f_vcvttsd2usi<round_saeonly_name>,
4296 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4297 Use %q1 instead of %1 for -masm=intel.
4298 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4299 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4300 of %3 for -masm=intel.
4301 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4303 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4305 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4307 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4308 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4310 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4312 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4313 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4314 %g1 and one with %0 and %1.
4315 (avx512er_vmrcp28<mode><round_saeonly_name>,
4316 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4318 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4319 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4320 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4321 of %0 and %{%4%} for -masm=intel.
4322 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4323 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4324 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4325 order of %0 and %{%5%}%{z%} for -masm=intel.
4327 2018-04-17 Jan Hubicka <jh@suse.cz>
4330 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4332 2018-04-17 Martin Liska <mliska@suse.cz>
4335 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4336 for target_clone error message. Make default implementation
4337 clone to be a local declaration.
4338 (separate_attrs): Add new argument and check for an empty
4340 (expand_target_clones): Handle it.
4341 (ipa_target_clone): Make redirection just for target_clones
4344 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4345 Tom de Vries <tom@codesourcery.com>
4348 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4349 tiled basic blocks without omp continue statements.
4351 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4354 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4355 vec_extract expression as having side effects to make sure it gets
4358 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4361 * config/i386/i386.c (get_builtin_code_for_version): Check
4364 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4367 * gcc/config.gcc: Support "skylake".
4368 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4370 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4371 (processor_target_table): Add "skylake".
4372 (ix86_option_override_internal): Add "skylake".
4373 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4374 PROCESSOR_CANNONLAKE.
4375 (get_builtin_code_for_version): Fix priority for
4376 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4377 PROCESSOR_SKYLAKE-AVX512.
4378 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4379 (processor_type): Add PROCESSOR_SKYLAKE.
4381 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4382 Jason Merrill <jason@redhat.com>
4385 * convert.c (convert_to_integer_1): Use direct recursion for
4386 enumeral types and types with a precision less than the number
4387 of bits in their mode.
4389 2018-04-16 Julia Koval <julia.koval@intel.com>
4392 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4393 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4395 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4398 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4399 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4400 and -mno-direct-move.
4402 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4405 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4406 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4407 Ensure negative shifts result in {0}.
4409 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4411 PR rtl-optimization/79916
4412 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4413 regs (if any) to define how to gnerate SD moves when LRA is in
4416 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4418 PR rtl-optimization/85393
4419 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4420 * except.c (expand_dw2_landing_pad_for_region): Make static.
4421 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4422 a label and unconditional jump to old_bb, rather than
4423 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4426 PR rtl-optimization/85376
4427 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4428 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4429 instead of a specific value.
4431 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4432 Bin Cheng <bin.cheng@arm.com>
4434 PR tree-optimization/82965
4435 PR tree-optimization/83991
4436 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4437 by_profile_only parameter.
4438 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4439 information if the loop was predicted to iterate too many times.
4440 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4442 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4445 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4448 2018-04-13 Martin Liska <mliska@suse.cz>
4449 Jakub Jelinek <jakub@redhat.com>
4452 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4453 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4454 * builtins.c (expand_builtin_memory_copy_args): Use
4455 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4456 handle dest_addr == pc_rtx.
4458 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4461 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4462 asked to not generate direct moves.
4463 (fix_trunc<mode>si2_stfiwx): Similar.
4464 (fix_trunc<mode>si2_internal): Similar.
4466 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4469 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4470 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4471 lookup if dest in some wider mode is known to be const0_rtx and
4472 if so, record permanent equivalence for it to be ZERO_EXTEND of
4473 the narrower mode destination.
4475 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4477 * lto-streamer-out.c (output_function): Revert 259346.
4478 * omp-expand.c (expand_oacc_for): Likewise.
4480 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4482 PR rtl-optimization/85354
4483 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4484 * sel-sched.c (sel_global_init): ... here.
4486 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4489 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4490 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4491 mode for PE-COFF targets.
4492 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4493 (i386_pe_asm_lto_end): Likewise.
4494 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4495 (TARGET_ASM_LTO_END): Likewise.
4496 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4497 (i386_pe_asm_lto_start): New function.
4498 (i386_pe_asm_lto_end): Likewise.
4500 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4501 Richard Biener <rguenther@suse.de>
4504 * lto-streamer-out.c (output_function): Fix CFG loop state before
4506 * omp-expand.c (expand_oacc_for): Handle calls to internal
4507 functions like regular functions.
4509 2018-04-12 Richard Biener <rguenther@suse.de>
4512 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4513 for the early LTO debug to properly generate references to it
4514 during DIE emission. Do not re-use that for the skeleton for
4516 (dwarf2out_early_finish): Likewise.
4518 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4521 * config/i386/sse.md
4522 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4523 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4524 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4525 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4526 and output is a reg, avoid creating invalid lowpart subreg, but
4527 instead split into a 512-bit move. Don't split if not AVX512VL,
4528 input is xmm16+ reg and output is a mem.
4529 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4530 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4531 xmm16+ reg and output is a mem.
4533 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4535 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4536 also for flag_dwarf2_cfi_asm.
4538 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4540 PR rtl-optimization/85342
4541 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4542 a bool scalar var inside of the loop instead. Don't try to update
4543 recog_data.operand after failed apply_change_group.
4545 2018-04-12 Tom de Vries <tom@codesourcery.com>
4548 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4549 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4550 array with flexible array member as array without given dimension.
4551 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4552 argument for undefined param to true.
4554 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4557 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4558 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4559 from PowerPC section.
4560 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4561 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4562 help text that is too long.
4563 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4564 help text that is too long.
4565 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4566 help text that is too long.
4568 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4570 * config/alpha/alpha.md (stack_probe_internal): Rename
4571 from "probe_stack". Update all callers.
4573 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4575 PR rtl-optimization/84566
4576 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4577 sched_macro_fuse_insns.
4579 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4582 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4583 (compute_block_dependences): ... from here.
4585 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4587 PR tree-optimization/85331
4588 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4589 from int to HOST_WIDE_INT.
4591 2018-04-11 Martin Jambor <mjambor@suse.cz>
4594 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4595 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4596 not the same as the source val.
4597 (cgraph_edge_brings_value_p): New parameter.
4598 (gather_edges_for_value): Pass destination value to
4599 cgraph_edge_brings_value_p.
4600 (perhaps_add_new_callers): Likewise.
4601 (get_info_about_necessary_edges): Likewise and exclude values brought
4602 only by self-recursive edges.
4603 (create_specialized_node): Redirect only clones of self-calling edges.
4604 (+self_recursive_pass_through_p): New function.
4605 (find_more_scalar_values_for_callers_subset): Use it.
4606 (find_aggregate_values_for_callers_subset): Likewise.
4607 (known_aggs_to_agg_replacement_list): Removed.
4608 (decide_whether_version_node): Re-calculate known constants for all
4609 remaining context clones.
4611 2018-04-11 Richard Biener <rguenther@suse.de>
4614 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4615 from early DWARF output.
4616 (dwarf2out_early_finish): Output line info unconditionally into
4617 early DWARF and add reference to it.
4619 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4622 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4623 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4624 other than V2DFmode using iptr mode attribute.
4625 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4627 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4629 PR rtl-optimization/84659
4630 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4632 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4635 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4637 (output_loc_list): Pass address of a dummy size variable even in the
4638 locview handling loop.
4639 (index_location_lists): Add comment on why skip_loc_list_entry can't
4642 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4645 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4648 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4651 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4652 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4653 and -mstring-compare-inline-limit.
4655 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4658 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4659 for stack clash protection in a register whenever we need it to be in
4662 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4664 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4665 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4667 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4670 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4672 (mlong-double-): Ditto.
4673 * config/rs6000/sysv4.opt (msdata=): Ditto.
4674 (mtls-size=): Ditto.
4676 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4678 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4679 erroneous entries for
4680 "vector int vec_ldl (int, long int *)", and
4681 "vector unsigned int vec_ldl (int, unsigned long int *)".
4682 Add comments and entries for
4683 "vector bool char vec_ldl (int, bool char *)",
4684 "vector bool short vec_ldl (int, bool short *)",
4685 "vector bool int vec_ldl (int, bool int *)",
4686 "vector bool long long vec_ldl (int, bool long long *)",
4687 "vector pixel vec_ldl (int, pixel *)",
4688 "vector long long vec_ldl (int, long long *)",
4689 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4690 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4691 type tree bool_long_long_type_node and correct definition of
4692 bool_V2DI_type_node to make reference to this new type tree.
4693 (rs6000_mangle_type): Replace erroneous reference to
4694 bool_long_type_node with bool_long_long_type_node.
4695 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4696 comments to emphasize sign distinctions for char and int types and
4697 replace RS6000_BTI_bool_long constant with
4698 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4699 use of RS6000_BTI_pixel.
4700 (bool_long_type_node): Remove this macro definition.
4701 (bool_long_long_type_node): New macro definition
4703 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4705 PR rtl-optimization/85300
4706 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4707 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4708 simplify_unary_operation fails.
4710 2018-04-10 Martin Liska <mliska@suse.cz>
4712 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4713 cgraph_edge and ipa_ref.
4715 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4719 * config/i386/sse.md
4720 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4721 computation of the VEC_MERGE selector from mask.
4722 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4723 Fix decoding of the VEC_MERGE selector into mask.
4725 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4727 PR tree-optimization/85286
4728 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4730 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4732 * final.c (final_1): Set insn_last_address as well as
4733 insn_current_address.
4735 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4738 * explow.c (emit_stack_probe): Call validize_mem on memory location
4739 before passing it to gen_probe_stack. Create address operand and
4740 legitimize it for the probe_stack_address case.
4742 2018-04-09 Jan Hubicka <jh@suse.cz>
4745 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4746 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4747 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4749 (free_lang_data): Rebuild type inheritance graph.
4751 2018-04-09 Martin Sebor <msebor@redhat.com>
4753 * invoke.texi (-finline-small-functions): Mention other optimization
4755 (-findirect-inlining, -fpartial-inlining): Same.
4756 (-finline-functions-called-once): Same.
4757 (-freorder-blocks-and-partition): Same.
4759 2018-04-09 Jan Hubicka <jh@suse.cz>
4762 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4763 jumps; choose last target that matches the criteria (i.e.
4764 no partition changes for non-crossing jumps).
4765 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4766 support for redirecting crossing jumps to non-crossing.
4768 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4770 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4771 also for naked functions.
4773 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4775 * config/arc/arc.md (add_shift): New pattern.
4776 (add_shift2): Likewise.
4777 (sub_shift): Likewise.
4778 (sub_shift_cmp0_noout): Likewise.
4779 (compare_si_ashiftsi): Likewise.
4780 (xbfu_cmp0_noout): New combine pattern.
4781 (xbfu_cmp0"): Likewise.
4782 (movsi_set_cc_insn): Place the predicable variant first.
4783 (commutative_binary_cmp0_noout): Remove clobber.
4784 (commutative_binary_cmp0): New pattern.
4785 (noncommutative_binary_cmp0): Likewise.
4786 (noncommutative_binary_cmp0_noout): Likewise.
4787 (noncommutative_binary_comparison_result_used): Removed.
4788 (rsub_cmp0): New pattern.
4789 (rsub_cmp0_noout): Likewise.
4790 (extzvsi): Changed, keep only meaningful variants.
4791 (SQH, SEZ): New iterators.
4792 (SQH_postfix): New mode attribute.
4793 (SEZ_prefix): New code attribute.
4794 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4795 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4796 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4798 (noncommutative_operator): Check the availability of barrel
4801 2018-04-09 Richard Biener <rguenther@suse.de>
4803 PR tree-optimization/85284
4804 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4805 Only use the niter constraining form of simple_iv when the exit
4808 2018-04-09 Tom de Vries <tom@codesourcery.com>
4811 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4812 (define_expand "*memory_barrier"): New define_expand.
4813 (define_insn "memory_barrier"): New insn.
4815 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4817 PR rtl-optimization/80463
4818 PR rtl-optimization/83972
4819 PR rtl-optimization/83480
4821 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4822 correct producer for the insn.
4823 (tidy_control_flow): Fixup seqnos in case of debug insns.
4825 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4827 PR rtl-optimization/83913
4829 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4830 different sched-times when merging exprs.
4832 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4834 PR rtl-optimization/83962
4836 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4837 tidy_fallthru_edge and tidy_control_flow.
4839 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4841 PR rtl-optimization/83530
4843 * sel-sched.c (force_next_insn): New global variable.
4844 (remove_insn_for_debug): When force_next_insn is true, also leave only
4845 next insn in the ready list.
4846 (sel_sched_region): When the region wasn't scheduled, make another pass
4847 over it with force_next_insn set to 1.
4849 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4851 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4853 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4854 for interrupt control.
4855 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4856 functions for interrupt control.
4857 * config/nds32/nds32-intrinsic.md: Likewise.
4858 * config/nds32/nds32_intrinsic.h: Likewise.
4859 * config/nds32/nds32.h (nds32_builtins): Likewise.
4861 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4863 * config/nds32/nds32.c (nds32_init_machine_status,
4864 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4865 strict_aligned_p field.
4866 (nds32_expand_to_rtl_hook): New function.
4867 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4868 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4870 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4871 Chung-Ju Wu <jasonwucj@gmail.com>
4873 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4874 * config/nds32/nds32-n7.md: New file.
4875 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4876 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4878 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4879 * config/nds32/nds32.md (pipeline_model): Add n7.
4880 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4881 * config/nds32/pipelines.md: Include n7 settings.
4883 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4884 Chung-Ju Wu <jasonwucj@gmail.com>
4886 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4887 * config/nds32/nds32-e8.md: New file.
4888 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4889 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4891 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4892 * config/nds32/nds32.md (pipeline_model): Add e8.
4893 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4894 * config/nds32/pipelines.md: Include e8 settings.
4896 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4897 Chung-Ju Wu <jasonwucj@gmail.com>
4899 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4900 * config/nds32/nds32-n8.md: New file.
4901 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4902 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4904 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4905 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4906 * config/nds32/nds32.md (pipeline_model): Add n8.
4907 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4908 * config/nds32/pipelines.md: Include n8 settings.
4910 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4911 Chung-Ju Wu <jasonwucj@gmail.com>
4913 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4914 * config/nds32/nds32-n9-2r1w.md: New file.
4915 * config/nds32/nds32-n9-3r2w.md: New file.
4916 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4917 nds32_register_ports): New or modify for cpu n9.
4918 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4920 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4921 * config/nds32/nds32-utils.c: New file.
4922 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4923 TARGET_MUL_SLOW): Define.
4924 * config/nds32/nds32.md (pipeline_model): New attribute.
4925 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4926 New options that support cpu n9.
4927 * config/nds32/pipelines.md: Include n9 settings.
4928 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4930 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4932 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4933 information if necessary.
4934 (output_cond_branch_compare_zero): Likewise.
4935 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4936 (nds32_target_alignment): Refine for alignment.
4937 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4938 (FUNCTION_BOUNDARY): Modify.
4939 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4941 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4943 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4945 * config/nds32/constants.md (unspec_volatile_element): Add values for
4946 TLB operation and data prefetch.
4947 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4948 functions for TLB operation and data prefetch.
4949 * config/nds32/nds32-intrinsic.md: Likewise.
4950 * config/nds32/nds32_intrinsic.h: Likewise.
4951 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4952 (nds32_print_operand): Likewise.
4953 * config/nds32/nds32.h (nds32_builtins): Likewise.
4955 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4956 Andrew Pinski <pinsika@gcc.gnu.org>
4959 * match.pd: Use constant_boolean_node of correct type instead of
4960 boolean_true_node or boolean_false_node for simplifying
4961 pointer comparisons to zero.
4963 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4965 PR tree-optimization/80021
4966 * tree.c (verify_type_variant): Make error call in verify_variant_match
4967 translatable and remove final full stop.
4969 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4971 * config/nds32/constants.md (unspec_volatile_element): Add
4972 UNSPEC_VOLATILE_EH_RETURN.
4973 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4974 nds32_output_stack_pop): Support dwarf exception handling process.
4975 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4976 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4977 exception handling process.
4978 (nds32_compute_stack_frame): Likewise.
4979 (nds32_return_addr_rtx): Likewise.
4980 (nds32_initial_elimination_offset): Likewise.
4981 (nds32_expand_prologue): Likewise.
4982 (nds32_expand_epilogue): Likewise.
4983 (nds32_dynamic_chain_address): New function.
4984 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4986 (DYNAMIC_CHAIN_ADDRESS): Define.
4987 (EH_RETURN_DATA_REGNO): Define.
4988 (EH_RETURN_STACKADJ_RTX): Define.
4989 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4990 patterns for dwarf exception handling.
4992 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4994 * config/nds32/nds32.h: Clean up obsolete macros.
4996 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4998 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4999 Add enum values for particular instructions.
5000 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5001 particular intrinsic functions.
5002 * config/nds32/nds32-intrinsic.md: Likewise.
5003 * config/nds32/nds32_intrinsic.h: Likewise.
5004 * config/nds32/nds32.h (nds32_builtins): Likewise.
5005 * config/nds32/nds32.md (type): Add pbsad and pbsada.
5006 (btst, ave): New patterns for particular instructions.
5008 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5010 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
5011 Add enum values for atomic load/store and memory sync.
5012 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
5014 * config/nds32/nds32-intrinsic.md: Likewise.
5015 * config/nds32/nds32_intrinsic.h: Likewise.
5016 * config/nds32/nds32.h (nds32_builtins): Likewise.
5018 2018-04-07 Jakub Jelinek <jakub@redhat.com>
5020 PR tree-optimization/85257
5021 * fold-const.c (native_encode_vector): If not all elts could fit
5022 and off is -1, return 0 rather than offset.
5023 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
5024 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
5025 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
5026 adjust buffer in native_interpret_expr call.
5028 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5030 * config/nds32/constants.md (unspec_volatile_element): Add cache
5031 control enum values.
5032 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
5033 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
5034 * config/nds32/nds32.c (nds32_cctl_names): New.
5035 (nds32_print_operand): Handle cache control register names.
5036 * config/nds32/nds32.h (nds32_builtins): New enum values.
5037 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5039 * config/nds32/nds32.md (type): Add mmu.
5040 * config/nds32/pipelines.md (simple_insn): Add mmu.
5042 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5044 * config/nds32/nds32.md (type): Remove call.
5045 * config/nds32/pipelines.md (simple_insn): Likewise.
5047 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5049 * config/nds32/constants.md (unspec_volatile_element): Add
5050 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5051 UNSPEC_VOLATILE_FMFCFG.
5052 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5053 description for fmfcfg and fmfcsr.
5054 (bdesc_1arg): Add fmtcsr.
5055 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5056 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5057 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5058 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5059 unspec_fmfcfg): New patterns.
5060 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5061 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5062 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5063 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5064 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5065 __nds32__fmfcfg): Define.
5067 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5069 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5070 intrinsic register names.
5071 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5072 intrinsic register enum values and macros.
5074 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5076 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5077 for load/store addressing form.
5078 (nds32_print_operand_address): Likewise.
5080 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5083 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5084 based on LABEL_REF. Remove useless assertion.
5085 (pic_address_needs_scratch): Fix formatting.
5086 (sparc_legitimize_pic_address): Minor tweaks.
5087 (sparc_delegitimize_address): Adjust assertion accordingly.
5088 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5089 into symbolic_operand.
5090 (movsi_high_pic_label_ref): Likewise.
5091 (movsi_lo_sum_pic_label_ref): Likewise.
5092 (movdi_pic_label_ref): Likewise.
5093 (movdi_high_pic_label_ref): Likewise.
5094 (movdi_lo_sum_pic_label_ref): Likewise.
5096 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5098 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5099 custom LIB_SPEC setup.
5101 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5102 Kito Cheng <kito.cheng@gmail.com>
5104 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5105 * config/riscv/freebsd.h: New.
5107 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5109 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5110 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5113 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5114 Kito Cheng <kito.cheng@gmail.com>
5116 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5117 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5118 * config/nds32/nds32-protos.h (nds32_output_call,
5119 nds32_output_return): Declare.
5120 * config/nds32/nds32.md: Refine all the call and return patterns.
5122 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5125 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5126 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5128 PR rtl-optimization/84872
5129 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5130 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5133 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5135 * expr.c (copy_blkmode_to_reg): Revert 254862.
5136 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5138 2018-04-06 Richard Biener <rguenther@suse.de>
5141 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5142 after seeing a component reference with an adjacent field. Treat
5143 refs to arrays at struct end of external decls similar to
5144 refs to unconstrained commons.
5146 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5149 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5150 look through SAVE_EXPRs with non-side-effects argument. Adjust
5152 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5155 2018-04-06 Richard Biener <rguenther@suse.de>
5158 * alias.c (find_base_term): New wrapper around find_base_term
5159 unwinding CSELIB_VAL_PTR changes.
5160 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5163 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5165 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5167 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5168 constant definitions.
5169 ("nop"): lr 0,0 -> nopr r0
5170 ("nop_lr0", "nop_lr1"): New insn definitions.
5172 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5174 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5175 NDS32_V3PUSH_AVAILABLE_P macro.
5177 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5178 Chung-Ju Wu <jasonwucj@gmail.com>
5180 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5181 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5182 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5183 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5184 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5185 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5186 * config/nds32/constraints.md: New constraints and checking for hard
5187 float configuration.
5188 * config/nds32/iterators.md: New mode iterator and attribute for hard
5189 float configuration.
5190 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5192 * config/nds32/nds32-fpu.md: New file.
5193 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5194 deal with hard float code generation.
5195 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5197 (abi_type, float_reg_number): New enum type.
5198 * config/nds32/nds32-predicates.c: New predicates for hard float.
5199 * config/nds32/nds32-protos.h: Declare functions for hard float.
5200 * config/nds32/nds32.c: Implementation for hard float configuration.
5201 * config/nds32/nds32.h: Definitions for hard float configuration.
5202 * config/nds32/nds32.md: Include hard float machine description and
5203 modify patterns for hard float configuration.
5204 * config/nds32/nds32.opt: New options for hard float configuration.
5205 * config/nds32/predicates.md: New predicates for hard float
5208 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5210 * common/config/nds32/nds32-common.c
5211 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5213 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5216 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5217 CONSTRUCTOR_ELT (ctor, ...)->value.
5219 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5222 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5224 2018-04-05 Tom de Vries <tom@codesourcery.com>
5227 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5230 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5231 Kito Cheng <kito.cheng@gmail.com>
5233 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5234 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5235 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5236 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5238 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5239 Kito Cheng <kito.cheng@gmail.com>
5241 * config/nds32/constraints.md (Ufe): New memory constraint.
5242 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5243 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5244 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5246 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5247 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5249 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5251 * config/nds32/nds32.md: Use optimize_size in the condition for
5252 alu-shift instructions.
5254 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5256 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5258 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5260 * config/nds32/nds32.md (negsi2): Refine pattern.
5262 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5263 Chung-Ju Wu <jasonwucj@gmail.com>
5265 * config/nds32/iterators.md (shift_rotate): New code iterator.
5266 (shift): New code attribute.
5267 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5268 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5269 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5270 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5271 bit-wise operations.
5272 (andsi3, *andsi3): Ditto.
5273 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5274 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5275 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5276 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5277 nds32_ior_operand, nds32_xor_operand): New predicates.
5279 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5281 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5282 (addsi3, subsi3): ... this.
5284 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5286 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5288 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5290 * config/nds32/nds32.md: Adjust indention.
5292 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5294 * config/nds32/nds32.md (feature): New attribute.
5296 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5298 * config/nds32/nds32.md (subtype): New attribute.
5300 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5303 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5304 expansion to perform a bitwise AND of the argument followed by a
5305 boolean negation of the result.
5307 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5309 PR rtl-optimization/84878
5310 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5311 the basic block. Assert the use reference is not artificial and that
5312 it has an associated insn.
5314 2018-04-04 Michael Matz <matz@suse.de>
5316 * builtins.c (compute_objsize): Pass correct operand
5317 to array_at_struct_end_p.
5319 2018-04-04 Richard Biener <rguenther@suse.de>
5322 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5323 from contexts for DINFO_LEVEL_TERSE and below.
5325 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5327 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5328 resiter_operand condition.
5329 * config/nds32/nds32.md (*move<mode>): Ditto.
5331 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5332 Monk Chiang <sh.chiang04@gmail.com>
5334 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5336 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5338 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5340 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5341 Kito Cheng <kito.cheng@gmail.com>
5343 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5344 nds32_cond_code_str, output_cond_branch,
5345 output_cond_branch_compare_zero, nds32_expand_cbranch,
5346 nds32_expand_cstore, nds32_expand_movcc,
5347 nds32_output_cbranchsi4_equality_zero,
5348 nds32_output_cbranchsi4_equality_reg,
5349 nds32_output_cbranchsi4_equality_reg_or_const_int,
5350 nds32_output_cbranchsi4_greater_less_zero: New functions.
5351 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5352 nds32_expand_cstore, nds32_expand_movcc,
5353 nds32_output_cbranchsi4_equality_zero,
5354 nds32_output_cbranchsi4_equality_reg,
5355 nds32_output_cbranchsi4_equality_reg_or_const_int,
5356 nds32_output_cbranchsi4_greater_less_zero): Declare.
5357 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5358 nds32_rimm11s_operand): New predicates.
5359 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5360 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5363 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5365 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5366 * config/nds32/nds32.md: Ditto.
5367 * config/nds32/pipelines.md: Ditto.
5369 2018-04-04 Richard Biener <rguenther@suse.de>
5371 PR tree-optimization/85168
5372 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5373 propagating abnormals.
5375 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5377 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5379 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5380 Kito Cheng <kito.cheng@gmail.com>
5382 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5383 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5384 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5385 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5386 * config/nds32/nds32.md (sibcall_internal): New.
5387 (sibcall_register): Remove.
5388 (sibcall_immediate): Remove.
5389 (sibcall_value_internal): New.
5390 (sibcall_value_register): Remove.
5391 (sibcall_value_immediate): Remove.
5392 * config/nds32/predicates.md (nds32_general_register_operand): New.
5393 (nds32_call_address_operand): New.
5395 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5397 PR rtl-optimization/85167
5398 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5399 bb_defs if *split_p, instead preinitialize it to NULL.
5401 PR tree-optimization/85156
5402 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5403 evaluating the argument multiple times.
5405 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5407 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5409 (_mm_cvtpd_ps): Likewise.
5410 (_mm_cvttpd_epi32): Likewise.
5411 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5412 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5413 vector, pixel, and bool following altivec.h include.
5415 2018-04-03 Martin Sebor <msebor@redhat.com>
5417 * doc/extend.texi (Common Function Attributes): Clarify.
5418 (const attribute): Likewise.
5419 (pure attribute): Likewise.
5421 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5424 * config/i386/i386.c (ix86_expand_vector_set): Use
5425 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5427 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5429 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5430 instructions when changing rounding bits to preserve precision bits
5431 in the x87 control word.
5433 2018-04-03 Martin Liska <mliska@suse.cz>
5435 PR tree-optimization/82491
5436 * rtl.h (strip_offset_and_add): Replace += suboffset with
5437 poly_uint64 () + suboffset.
5439 2018-03-29 Martin Liska <mliska@suse.cz>
5440 Martin Jambor <mjambor@suse.cz>
5443 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5444 param_type is not an integral or pointer type.
5446 2018-04-03 Richard Biener <rguenther@suse.de>
5448 * sese.h (recompute_all_dominators): Remove.
5450 2018-04-02 Martin Sebor <msebor@redhat.com>
5452 * doc/invoke.texi (-Wrestrict): Fix typos.
5454 2018-04-02 Jim Wilson <jimw@sifive.com>
5456 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5457 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5458 (<optab>di3, <optab>si3_extend): Likewise.
5459 (<optab>si3_mask, <optab>si3_mask_1): New.
5460 (<optab>di3_mask, <optab>di3_mask_1): New.
5461 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5462 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5463 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5465 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5467 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5470 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5472 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5473 (nds32_canonicalize_comparison): New function.
5475 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5476 Kito Cheng <kito.cheng@gmail.com>
5477 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5479 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5480 * config/nds32/constants.md (unspec_volatile_element): Add
5481 UNSPEC_VOLATILE_RELAX_GROUP.
5482 * config/nds32/nds32-relax-opt.c: New file.
5483 * config/nds32/nds32-predicates.c
5484 (nds32_symbol_load_store_p): New function.
5485 * config/nds32/nds32-protos.h
5486 (nds32_symbol_load_store_p): Declare function.
5487 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5488 * config/nds32/nds32.c
5489 (nds32_register_pass): New function to register pass.
5490 (nds32_register_passes): New function to register passes.
5491 * config/nds32/nds32.md (relax_group): New pattern.
5492 * config/nds32/nds32.opt (mrelax-hint): New option.
5493 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5495 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5497 * config/nds32/t-nds32: Modify files dependency.
5499 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5501 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5502 (PROFILE_HOOK): Define its implementation.
5504 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5506 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5507 type and 32-bit size.
5509 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5512 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5513 (V_128_256): New mode iterator.
5514 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5515 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5516 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5518 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5519 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5521 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5524 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5525 NaN inputs correctly.
5527 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5530 * config/rs6000/vsx.md (??r): New mode attribute.
5531 (*vsx_mov<mode>_64bit): Use it.
5532 (*vsx_mov<mode>_32bit): Likewise.
5534 2018-03-30 Martin Sebor <msebor@redhat.com>
5536 PR tree-optimization/84818
5537 * builtins.c (check_access): Use warning_n.
5539 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5542 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5544 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5547 2018-03-30 Julia Koval <julia.koval@intel.com>
5550 * x86-tune.def (movx, partial_reg_dependency): Enable for
5553 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5556 * lra-constraints.c (process_alt_operands): Move setting
5557 this_alternative_matches below.
5559 2018-03-29 Martin Liska <mliska@suse.cz>
5562 * doc/invoke.texi: Document how LTO works with debug info.
5563 Describe auto-load support of binutils. Mention 'x86-64'
5564 as valid option value of -march option.
5566 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5568 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5571 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5572 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5573 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5576 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5579 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5580 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5581 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5582 for RS6000_BTM_POWERPC64.
5583 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5584 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5585 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5590 2018-03-28 Carl Love <cel@us.ibm.com>
5593 2017-09-27 Carl Love <cel@us.ibm.com>
5595 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5596 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5597 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5600 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5602 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5603 instead of __vector bool.
5604 (_mm_max_pu8): Likewise.
5605 (_mm_min_pi16): Likewise.
5607 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5610 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5611 (DIVWEUO): Likewise.
5613 (DIVDEUO): Likewise.
5614 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5615 DIVWEUO and DIVDEUO.
5616 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5617 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5618 (div_extend): Likewise.
5619 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5621 (__builtin_divweuo): Likewise.
5622 (__builtin_divdeo): Likewise.
5623 (__builtin_divdeuo): Likewise.
5625 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5628 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5629 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5631 PR tree-optimization/82004
5632 * gimple-match-head.c (optimize_pow_to_exp): New function.
5633 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5634 Don't fold to exp if optimize_pow_to_exp is false.
5636 2018-03-28 Martin Liska <mliska@suse.cz>
5639 * calls.c (initialize_argument_information): Fix trailing space.
5640 * common.opt: Fix typo and provide better explanation for
5641 -fsanitize-coverage option.
5642 * config/i386/i386.opt: Fix typo.
5644 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5645 Martin Liska <mliska@suse.cz>
5648 * gimplify.c (asan_poison_variable): Don't do the check for
5649 gimplify_omp_ctxp here.
5650 (gimplify_decl_expr): Do it here.
5651 (gimplify_target_expr): Likewise.
5653 2018-03-28 Martin Liska <mliska@suse.cz>
5656 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5657 chkp_type_bounds_count if MPX is not enabled.
5659 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5661 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5663 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5666 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5667 function to create the function decl for complex long double
5668 multiply and divide for -mabi=ieeelongdouble.
5669 (init_float128_ieee): Call it.
5671 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5674 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5675 -fcf-protection=branch -mibt.
5676 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5678 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5681 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5683 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5686 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5687 extern array declarations.
5689 2018-03-27 Richard Biener <rguenther@suse.de>
5692 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5693 explicit single_use checks.
5695 2018-03-27 Richard Biener <rguenther@suse.de>
5697 PR tree-optimization/85082
5698 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5701 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5703 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5704 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5705 Turn on fasynchronous-unwind-tables and funwind-tables.
5707 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5710 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5711 (*bmi_blsr_<mode>_ccz): Ditto.
5713 2018-03-26 Tom de Vries <tom@codesourcery.com>
5715 PR tree-optimization/85063
5716 * omp-general.c (offloading_function_p): New function. Factor out
5718 * omp-offload.c (pass_omp_target_link::gate): ... here.
5719 * omp-general.h (offloading_function_p): Declare.
5720 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5721 with attribute omp declare target for offloading functions.
5723 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5725 PR tree-optimization/84005
5726 * tree-data-ref.h (get_base_for_alignment): Declare.
5727 * tree-data-ref.c (get_base_for_alignment_1): New function.
5728 (get_base_for_alignment): Likewise.
5729 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5730 get_base_for_alignment to find a suitable base object, instead
5731 of always using drb->base_address.
5733 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5736 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5737 known size by default.
5739 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5742 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5743 and non BLKmode operands.
5745 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5748 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5749 Clean up attributes.
5751 2018-03-23 Richard Biener <rguenther@suse.de>
5754 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5755 we are going to emit early debug for LTO.
5757 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5760 * function.c (match_asm_constraints_1): Don't optimize if input
5761 doesn't satisfy general_operand predicate for output's mode.
5764 * alias.c (write_dependence_p): Don't require for x_canonicalized
5765 non-VOIDmode if x has VOIDmode.
5768 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5769 just don't try to optimize it rather than assert it never happens.
5771 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5773 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5774 macro expansions for definition of ST_INTERNAL_<mode> and
5775 LD_INTERNAL_<mode> builtins.
5776 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5778 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5780 (altivec_expand_st_builtin): Likewise.
5781 (altivec_expand_builtin): Remove calls to deleted functions.
5782 (rs6000_address_for_altivec): Delete this function.
5783 * config/rs6000/vector.md: Remove expands for
5784 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5786 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5789 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5790 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5791 re-computing once computed.
5792 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5793 (arm_init_machine_status): Initialize
5794 machine->static_chain_stack_bytes.
5796 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5799 * doc/extend.texi: Add four new prototypes for vec_ld.
5800 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5801 definitions for more logical presentation.
5802 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5803 entries for V1TI variants of __builtin_altivec_ld builtin.
5804 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5805 handling of V1TI variant of LVX icode pattern.
5806 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5807 (rs6000_gimple_fold_builtin): Likewise.
5808 (altivec_init_builtins): Add code to define
5809 __builtin_altivec_lvx_v1ti function.
5811 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5814 * function.c (match_asm_constraints_1): Don't do the optimization
5815 if input isn't a REG, SUBREG, MEM or constant.
5817 2018-03-22 Tom de Vries <tom@codesourcery.com>
5819 PR tree-optimization/84956
5820 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5821 bb_has_abnormal_pred.
5823 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5826 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5827 DECL_INITIAL (decl) to decl at the end.
5828 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5831 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5833 * doc/extend.texi (__builtin_tgmath): Document when complex
5834 integer types are treated as _Complex _Float64.
5836 2018-03-21 Tom de Vries <tom@codesourcery.com>
5838 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5840 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5842 PR tree-optimization/84960
5843 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5844 if it is ENTRY block, move them into single succ of ENTRY in that case.
5846 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5848 PR tree-optimization/84811
5849 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5850 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5851 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5852 (known_subrange_p): Update accordingly. Cast each value involved
5853 in the size comparison, rather than casting the result of the
5856 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5858 PR tree-optimization/84982
5859 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5860 by flipping the least significant bit rather than all bits from
5861 bitpos to bitpos + bitsize - 1.
5863 2018-03-21 Nathan Sidwell <nathan@acm.org>
5865 * doc/extend.texi (Deprecated Features): Remove mention of
5866 long-deleted deprecations.
5868 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5871 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5872 * configure: Regenerate.
5874 2018-03-21 Tom de Vries <tom@codesourcery.com>
5876 PR tree-optimization/83126
5877 * tree-parloops.c (num_phis): New function.
5878 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5880 2018-03-21 Nathan Sidwell <nathan@acm.org>
5882 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5883 mention anon-struct/union members and trailing attributes.
5885 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5887 PR tree-optimization/84969
5888 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5889 builtin memset partitions if they set different rhs values.
5891 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5893 PR rtl-optimization/84989
5894 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5895 VEC_DUPLICATE with scalar result mode.
5897 2018-03-21 Martin Liska <mliska@suse.cz>
5900 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5901 not intended return statement.
5903 2018-03-21 Martin Liska <mliska@suse.cz>
5906 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5907 (chkp_find_bound_slots_1): Limit number of iterations.
5909 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5912 * Minor grammar fixes for x86 options.
5914 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5917 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5918 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5920 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5923 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5924 (altivec_lvx_<mode>_1op): Likewise.
5925 (altivec_stvx_<mode>_2op): Likewise.
5926 (altivec_stvx_<mode>_1op): Likewise.
5927 (altivec_lvx_<VM2:mode>): New define_expand.
5928 (altivec_stvx_<VM2:mode>): Likewise.
5929 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5930 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5931 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5932 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5933 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5934 (rs6000_gen_lvx): Likewise.
5935 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5936 (altivec_expand_stv_builtin): Likewise.
5937 (altivec_expand_builtin): Likewise.
5938 * config/rs6000/vector.md: Likewise.
5940 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5943 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5946 2018-03-20 Richard Biener <rguenther@suse.de>
5949 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5950 sign-conversions as zero, fall back to standard scalar_stmt
5953 2018-03-20 Martin Liska <mliska@suse.cz>
5956 * predict.c (rebuild_frequencies): Handle case when we have
5957 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5959 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5962 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5963 flag_section_anchors.
5964 * varasm.c (use_blocks_for_decl_p): Remove hack for
5965 dw2_force_const_mem.
5968 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5970 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5971 be created, use lowpart_subreg of operands[0] rather than operands[0]
5973 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5974 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5975 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5976 and n constraint instead of aarch64_shift_imm_di and Usd.
5977 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5978 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5980 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5983 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5984 to favor GPR over NEON registers.
5985 (<shift>di3_neon): Likewise.
5987 2018-03-20 Tom de Vries <tom@codesourcery.com>
5990 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5991 (nvptx_process_pars): Emit bar.sync asap and alap.
5993 2018-03-20 Tom de Vries <tom@codesourcery.com>
5996 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5997 seen_label if seen_label is already set.
5999 2018-03-20 Jakub Jelinek <jakub@redhat.com>
6002 * config/i386/i386.c (fold_builtin_cpu): For features above 31
6003 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
6004 Use 1U instead of 1. Formatting fixes.
6007 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
6008 instead of TREE_TYPE (s1) for the return value.
6010 2018-03-19 Jakub Jelinek <jakub@redhat.com>
6012 PR tree-optimization/84946
6013 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
6014 bitsize + bitsize in poly_uint64 rather than poly_int64.
6017 * dwarf2asm.c: Include fold-const.c.
6018 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
6019 of decl rather than decl itself.
6021 PR rtl-optimization/84643
6022 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
6024 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
6027 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
6028 calling assemble_variable.
6030 2018-03-19 Sudakshina Das <sudi.das@arm.com>
6033 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
6034 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6036 2018-03-19 Jim Wilson <jimw@sifive.com>
6039 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6040 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6041 (riscv_first_stack_step): Likewise.
6042 (riscv_option_override): Use STACK_BOUNDARY instead of
6044 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6046 (BIGGEST_ALIGNMENT): Set to 128.
6047 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6048 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6051 2018-03-19 Richard Biener <rguenther@suse.de>
6053 PR tree-optimization/84933
6054 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6055 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6057 2018-03-19 Richard Biener <rguenther@suse.de>
6059 PR tree-optimization/84859
6060 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6061 (cond_if_else_store_replacement): Perform sinking operation on
6062 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6063 Generalize what a BB with a single eligible store is.
6065 2018-03-19 Richard Biener <rguenther@suse.de>
6067 PR tree-optimization/84929
6068 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6069 chrec_is_positive against non-chrec arg.
6071 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6074 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6076 2018-03-18 Martin Liska <mliska@suse.cz>
6078 PR rtl-optimization/84635
6079 * regrename.c (build_def_use): Use matches_mode only when
6082 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6084 PR tree-optimization/84913
6085 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6086 vectorize chains of COND_EXPRs.
6088 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6090 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6092 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6094 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6096 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6098 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6100 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6101 Kito Cheng <kito.cheng@gmail.com>
6103 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6104 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6105 (nds32_adjust_reg_alloc_order): New function.
6106 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6108 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6110 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6111 nds32_print_operand, nds32_print_operand_address): Use
6112 HOST_WIDE_INT_PRINT_DEC instead.
6114 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6116 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6118 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6121 * config/i386/i386.c (initial_ix86_tune_features,
6122 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6124 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6125 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6126 rather than 1u << ix86_tune. Formatting fix.
6127 (ix86_option_override_internal): Change ix86_arch_mask from
6128 unsigned int to unsigned HOST_WIDE_INT, initialize to
6129 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6130 (ix86_function_specific_restore): Likewise.
6132 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6135 * postreload.c (reload_combine_recognize_pattern): Perform
6136 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6137 truncate_int_for_mode the result for the destination's mode.
6140 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6141 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6143 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6146 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6149 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6150 regno_allocno_class_array and sorted_pseudos.
6151 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6152 insns where regno is used.
6154 2018-03-16 Martin Liska <mliska@suse.cz>
6157 * multiple_target.c (create_dispatcher_calls): Redirect
6158 reference in the symbol table.
6160 2018-03-16 Martin Liska <mliska@suse.cz>
6163 * multiple_target.c (create_dispatcher_calls): Redirect also
6166 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6170 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6171 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6172 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6174 2018-03-16 Julia Koval <julia.koval@intel.com>
6176 * doc/invoke.texi (Skylake Server): Add CLWB.
6177 Cannonlake): Remove CLWB.
6179 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6181 PR tree-optimization/84841
6182 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6184 (FLOAT_ONE_CONST_TYPE): Define.
6185 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6186 (sort_by_operand_rank): Put entries with higher constant_type last
6187 rather than first to match comments.
6189 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6191 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6194 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6197 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6198 check and use address of target always.
6200 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6203 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6204 (indirect_thunk_bnd_needed): Likewise.
6205 (indirect_thunks_used): Likewise.
6206 (indirect_thunks_bnd_used): Likewise.
6207 (indirect_return_needed): New.
6208 (indirect_return_bnd_needed): Likewise.
6209 (output_indirect_thunk_function): Add a bool argument for
6211 (output_indirect_thunk_function): Don't generate alias for
6212 function return thunk.
6213 (ix86_code_end): Call output_indirect_thunk_function to generate
6214 function return thunks.
6215 (ix86_output_function_return): Set indirect_return_bnd_needed
6216 and indirect_return_needed instead of indirect_thunk_bnd_needed
6217 and indirect_thunk_needed.
6219 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6221 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6222 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6223 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6225 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6226 Paul Hua <paul.hua.gm@gmail.com>
6229 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6231 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6233 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6234 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6237 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6240 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6241 instead of GET_MODE_SIZE when comparing Units.
6243 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6246 * varasm.c (hash_section): Return an unchangeble hash value
6247 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6248 Return !aarch64_can_use_per_function_literal_pools_p ().
6250 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6253 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6254 rather than address of cmode as last argument to prepare_cmp_insn.
6256 2018-03-15 Julia Koval <julia.koval@intel.com>
6258 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6259 F_AVX512VNNI, F_AVX512BITALG): New.
6261 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6264 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6265 insn for floating-point loads and stores.
6267 2018-03-14 Carl Love <cel@us.ibm.com>
6269 * config/rs6000/rs6000-c.c: Add macro definitions for
6270 ALTIVEC_BUILTIN_VEC_PERMXOR.
6271 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6272 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6273 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6274 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6276 * config/doc/extend.texi: Add prototypes for vec_permxor.
6278 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6281 * diagnostic-show-locus.c (class layout_point): Convert m_line
6282 from int to linenum_type.
6283 (line_span::comparator): Use linenum "compare" function when
6284 comparing line numbers.
6285 (test_line_span): New function.
6286 (layout_range::contains_point): Convert param "row" from int to
6288 (layout_range::intersects_line_p): Likewise.
6289 (layout::will_show_line_p): Likewise.
6290 (layout::print_source_line): Likewise.
6291 (layout::should_print_annotation_line_p): Likewise.
6292 (layout::print_annotation_line): Likewise.
6293 (layout::print_leading_fixits): Likewise.
6294 (layout::annotation_line_showed_range_p): Likewise.
6295 (struct line_corrections): Likewise for field m_row.
6296 (line_corrections::line_corrections): Likewise for param "row".
6297 (layout::print_trailing_fixits): Likewise.
6298 (layout::get_state_at_point): Likewise.
6299 (layout::get_x_bound_for_row): Likewise.
6300 (layout::print_line): Likewise.
6301 (diagnostic_show_locus): Likewise for locals "last_line" and
6303 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6304 * input.c (selftest::test_linenum_comparisons): New function.
6305 (selftest::input_c_tests): Call it.
6306 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6307 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6308 * selftest.h (ASSERT_GT): New macro.
6309 (ASSERT_GT_AT): New macro.
6310 (ASSERT_LT): New macro.
6311 (ASSERT_LT_AT): New macro.
6313 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6315 PR rtl-optimization/84780
6316 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6318 2018-03-14 Martin Liska <mliska@suse.cz>
6320 * tree.c (record_node_allocation_statistics): Use
6321 get_stats_node_kind.
6322 (get_stats_node_kind): New function extracted from
6323 record_node_allocation_statistics.
6324 (free_node): Use get_stats_node_kind.
6326 2018-03-14 Richard Biener <rguenther@suse.de>
6328 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6329 that the value-set of ANTIC_IN doesn't grow.
6332 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6334 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6335 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6337 2018-03-14 Julia Koval <julia.koval@intel.com>
6339 * config.gcc (icelake-client, icelake-server): New.
6341 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6342 (initial_ix86_arch_features): Ditto.
6343 (PTA_SKYLAKE): Add SGX.
6344 (PTA_ICELAKE): Remove.
6345 (PTA_ICELAKE_CLIENT): New.
6346 (PTA_ICELAKE_SERVER): New.
6347 (ix86_option_override_internal): Split up icelake on icelake client and
6349 (get_builtin_code_for_version): Ditto.
6350 (fold_builtin_cpu): Ditto.
6351 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6352 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6353 * config/i386/i386.h (processor_type): Ditto.
6354 * doc/invoke.texi: Ditto.
6356 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6359 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6360 INTEGER_CST offset, add it together with bitpos / 8 and
6361 sign extend based on POINTER_SIZE.
6365 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6368 * config/i386/constraints.md (Yc): New register constraint.
6369 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6370 Use Yc constraint for alternative 2 of operand 0. Remove
6371 preferred_for_speed attribute.
6373 2018-03-14 Richard Biener <rguenther@suse.de>
6375 PR tree-optimization/84830
6376 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6377 with the old one to avoid oscillations.
6379 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6382 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6384 (assign_by_spills): Return a flag of reload assignment failure.
6385 Do not process the reload assignment failures. Do not spill other
6386 reload pseudos if they has the same reg class. Update n if
6388 (lra_assign): Add a return arg. Set up from the result of
6389 assign_by_spills call.
6390 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6391 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6392 usage_insns if it is not NULL.
6393 (spill_hard_reg_in_range): New function.
6394 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6395 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6396 function prototypes.
6397 (lra_assign): Change prototype.
6398 * lra.c (lra): Add code to deal with fails by splitting hard reg
6401 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6403 * config/riscv/riscv.opt (mrelax): New option.
6404 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6405 "norelax" when riscv_mrelax is disabled.
6406 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6408 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6411 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6412 reassociation for int modes.
6414 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6416 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6417 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6419 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6420 * config/aarch64/aarch64-sve.md
6421 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6422 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6423 (*extend<mode><Vwide>2): Rename to...
6424 (aarch64_sve_extend<mode><Vwide>2): ...this.
6425 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6426 renaming the old pattern to...
6427 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6429 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6430 define_expand, renaming the old pattern to...
6431 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6432 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6433 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6434 account when deciding which SVE instruction the optab should use.
6435 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6437 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6439 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6440 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6441 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6442 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6443 (tlsdesc_small_<mode>): Turn a define_expand and use
6444 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6445 (tlsdesc_small_advsimd_<mode>): ...this.
6446 (tlsdesc_small_sve_<mode>): New pattern.
6448 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6450 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6451 (UNSPEC_UMUL_HIGHPART): New constants.
6452 (MUL_HIGHPART): New int iteraor.
6453 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6454 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6456 (*<su>mul<mode>3_highpart): New define_insn.
6458 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6461 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6464 2018-03-13 Martin Liska <mliska@suse.cz>
6467 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6469 (sem_item_optimizer::~sem_item_optimizer): Release it.
6470 (sem_item_optimizer::merge_classes): Register variable aliases.
6471 (sem_item_optimizer::fixup_pt_set): New function.
6472 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6473 * ipa-icf.h: Declare new variables and functions.
6475 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6478 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6479 integer_pow2p@2 and test integer_pow2p in condition.
6480 (A < 0 ? C : 0): Similarly for @1.
6483 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6484 characters starting at p contain '\0' character, don't look beyond
6488 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6489 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6492 * reg-stack.c (change_stack): Change update_end var from int to
6493 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6494 also call set_block_for_insn on the newly added insns and rescan.
6497 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6498 on the last operand.
6501 * tree.c (stabilize_reference_1): Return save_expr (e) for
6502 STATEMENT_LIST even if it doesn't have side-effects.
6504 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6506 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6508 2018-03-12 Renlin Li <renlin.li@arm.com>
6510 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6511 aarch64_output_scalar_simd_mov_immediate.
6513 2018-03-12 Martin Sebor <msebor@redhat.com>
6515 PR tree-optimization/83456
6516 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6517 for perfectly overlapping calls to memcpy.
6518 (gimple_fold_builtin_memory_chk): Same.
6519 (gimple_fold_builtin_strcpy): Handle no-warning.
6520 (gimple_fold_builtin_stxcpy_chk): Same.
6521 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6523 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6525 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6526 parameter. Use it for SFmode.
6527 (rs6000_function_arg_advance_1): Adjust.
6528 (rs6000_function_arg): Adjust.
6529 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6531 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6533 PR rtl-optimization/84169
6534 PR rtl-optimization/84780
6535 * combine.c (can_combine_p): Check for a 2-insn combination whether
6536 the destination register is used between the two insns, too.
6538 2018-03-12 Richard Biener <rguenther@suse.de>
6540 PR tree-optimization/84803
6541 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6542 for refs DR analysis didn't process.
6544 2018-03-12 Richard Biener <rguenther@suse.de>
6546 PR tree-optimization/84777
6547 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6548 force-vectorize loops ignore whether we are optimizing for size.
6550 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6552 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6553 (TARGET_MD_ASM_ADJUST): Define.
6555 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6556 Kito Cheng <kito.cheng@gmail.com>
6557 Chung-Ju Wu <jasonwucj@gmail.com>
6559 * config/nds32/nds32.c (nds32_compute_stack_frame,
6560 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6561 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6562 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6563 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6564 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6565 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6566 * config/nds32/nds32.md (prologue, epilogue): Use macro
6567 NDS32_V3PUSH_AVAILABLE_P to do checking.
6569 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6572 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6573 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6574 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6575 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6576 addition of most attributes on !orig_type_die or the attribute not
6577 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6579 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6580 Chung-Ju Wu <jasonwucj@gmail.com>
6582 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6584 * config/nds32/nds32.opt (mvh): New option.
6586 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6587 Chung-Ju Wu <jasonwucj@gmail.com>
6589 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6591 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6592 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6595 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6596 Chung-Ju Wu <jasonwucj@gmail.com>
6598 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6600 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6601 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6603 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6604 Kito Cheng <kito.cheng@gmail.com>
6605 Chung-Ju Wu <jasonwucj@gmail.com>
6607 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6608 UNSPEC_FFMISM and UNSPEC_FLMISM.
6609 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6610 for ffb, ffmism and flmism.
6611 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6612 (unspec_ffmism): Ditto.
6613 (unspec_flmism): Ditto.
6614 (nds32_expand_builtin_impl): Check if string extension is available.
6615 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6616 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6618 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6621 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6624 * lra-assigns.c (assign_by_spills): Return a flag of reload
6625 assignment failure. Do not process the reload assignment
6626 failures. Do not spill other reload pseudos if they has the same
6628 (lra_assign): Add a return arg. Set up from the result of
6629 assign_by_spills call.
6630 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6631 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6632 usage_insns if it is not NULL.
6633 (spill_hard_reg_in_range): New function.
6634 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6635 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6636 function prototypes.
6637 (lra_assign): Change prototype.
6638 * lra.c (lra): Add code to deal with fails by splitting hard reg
6641 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6644 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6646 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6649 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6650 (dw_val_node): Add val_symbolic_view.
6651 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6652 (symview_upper_bound): New.
6653 (new_line_info_table): Initialize symviews_since_reset.
6654 (dwarf2out_source_line): Count symviews_since_reset and set
6655 symview_upper_bound.
6656 (dw_val_equal_p): Handle symview.
6657 (add_AT_symview): New.
6658 (print_dw_val): Handle symview.
6659 (attr_checksum, attr_checksum_ordered): Likewise.
6660 (same_dw_val_p, size_of_die): Likewise.
6661 (value_format, output_die): Likewise.
6662 (add_high_low_attributes): Use add_AT_symview for entry_view.
6663 (dwarf2out_finish): Reset symview_upper_bound, clear
6666 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6669 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6670 Add strict argument and use it.
6671 (rs6000_split_multireg_move): Update for new strict argument.
6672 (mem_operand_gpr): Disallow all non-offsettable addresses.
6673 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6675 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6678 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6679 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6680 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6683 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6684 decl, use remap_type if we want to use the type.
6686 2018-03-09 Martin Sebor <msebor@redhat.com>
6688 PR tree-optimization/84526
6689 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6691 (builtin_access::generic_overlap): Be prepared to handle non-array
6694 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6696 PR rtl-optimization/84682
6697 * lra-constraints.c (process_address_1): Check is_address flag
6698 for address constraints.
6699 (process_alt_operands): Likewise.
6700 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6701 preprocess_constraints.
6702 * recog.h (preprocess_constraints): Add oploc parameter.
6704 * recog.c (preprocess_constraints): Test address_operand for
6705 CT_ADDRESS constraints.
6707 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6710 * lra-assigns.c (assign_by_spills): Return a flag of reload
6711 assignment failure. Do not process the reload assignment
6712 failures. Do not spill other reload pseudos if they has the same
6714 (lra_assign): Add a return arg. Set up from the result of
6715 assign_by_spills call.
6716 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6717 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6718 usage_insns if it is not NULL.
6719 (spill_hard_reg_in_range): New function.
6720 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6721 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6722 function prototypes.
6723 (lra_assign): Change prototype.
6724 * lra.c (lra): Add code to deal with fails by splitting hard reg
6727 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6730 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6731 Accept complain bool parameter. Only emit errors if it is true.
6732 (arm_parse_cpu_option_name): Likewise.
6733 (arm_target_thumb_only): Adjust callers of the above.
6734 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6735 prototype to take a default true bool parameter.
6736 (arm_parse_arch_option_name): Likewise.
6738 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6739 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6743 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6744 * configure: Regenerate.
6745 * configure.ac ("linker --version-script option"): New.
6746 ("linker soname option"): New.
6748 2018-03-09 Richard Biener <rguenther@suse.de>
6750 PR tree-optimization/84775
6751 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6752 immediate uses of predicate stmts and mark them modified.
6755 PR tree-optimization/84178
6756 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6758 (version_loop_for_if_conversion): Delay update_ssa call.
6759 (tree_if_conversion): Delay update_ssa until after predicate
6762 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6765 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6766 when the function accesses prior frames.
6768 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6771 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6772 gen_llsym, otherwise call maybe_gen_llsym.
6775 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6776 has ',' character inside of it.
6778 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6781 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6782 as clobbering CC_REGNUM.
6784 2018-03-08 Richard Biener <rguenther@suse.de>
6787 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6788 (follow_copies_to_constant): Do not follow SSA names registered
6791 2018-03-08 Richard Biener <rguenther@suse.de>
6793 PR tree-optimization/84178
6794 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6796 (version_loop_for_if_conversion): Delay update_ssa call.
6797 (tree_if_conversion): Delay update_ssa until after predicate
6800 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6802 PR tree-optimization/84178
6803 * tree-if-conv.c (release_bb_predicate): Remove the
6804 the assertion that the stmts have NULL use_ops.
6805 Discard the statements, asserting that they haven't
6806 yet been added to a BB.
6808 2018-03-08 Richard Biener <rguenther@suse.de>
6810 PR tree-optimization/84746
6811 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6812 (phi_translate): Pass in destination ANTIC_OUT set.
6813 (phi_translate_1): Likewise. For a simplified result lookup
6814 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6815 (phi_translate_set): Adjust.
6816 (do_pre_regular_insertion): Likewise.
6817 (do_pre_partial_partial_insertion): Likewise.
6819 2018-03-08 Martin Liska <mliska@suse.cz>
6821 PR gcov-profile/84735
6822 * doc/gcov.texi: Document usage of profile files.
6823 * gcov-io.h: Document changes in the format.
6825 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6829 * dwarf2out.c (struct dw_line_info_table): Update comments for
6831 (FORCE_RESET_NEXT_VIEW): New.
6832 (FORCE_RESETTING_VIEW_P): New.
6833 (RESETTING_VIEW_P): Check for -1 too.
6834 (ZERO_VIEW_P): Likewise.
6835 (new_line_info_table): Force-reset next view.
6836 (dwarf2out_begin_function): Likewise.
6837 (dwarf2out_source_line): Simplify zero_view_p initialization.
6838 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6839 view directly. Omit view when omitting .loc at line 0.
6841 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6843 PR tree-optimization/84740
6844 * tree-switch-conversion.c (process_switch): Call build_constructors
6845 only if info.phi_count is non-zero.
6847 PR tree-optimization/84739
6848 * tree-tailcall.c (find_tail_calls): Check call arguments against
6849 DECL_ARGUMENTS (current_function_decl) rather than
6850 DECL_ARGUMENTS (func) when checking for tail recursion.
6852 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6854 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6855 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6856 Volker Reichelt's entry and add entries for people that perform
6857 GCC fuzzy testing and report numerous bugs.
6859 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6862 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6863 readonly data in sdata, if that is disabled.
6864 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6865 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6866 -mreadonly-in-sdata option.
6868 2018-03-07 Martin Sebor <msebor@redhat.com>
6870 PR tree-optimization/84468
6871 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6872 basic block when looking for nul assignment.
6874 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6877 * except.h (output_function_exception_table): Adjust prototype.
6878 * except.c (output_function_exception_table): Remove FNNAME parameter
6879 and add SECTION parameter. Ouput one part of the table at a time.
6880 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6881 the first part of the exception table and emit unwind directives.
6882 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6883 (i386_pe_seh_cold_init): Likewise.
6884 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6885 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6886 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6887 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6888 * config/i386/winnt.c: Include except.h.
6889 (struct seh_frame_state): Add reg_offset, after_prologue and
6890 in_cold_section fields.
6891 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6892 (i386_pe_seh_cold_init): New function.
6893 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6894 to seh->in_cold_section.
6895 (seh_emit_push): Record the offset of the push.
6896 (seh_emit_save): Record the offet of the save.
6897 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6898 Test seh->after_prologue to disregard the epilogue.
6899 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6900 (i386_pe_end_cold_function): New function.
6902 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6905 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6906 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6909 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6911 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6912 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6915 * multiple_target.c: Include tree-inline.h and intl.h.
6916 (expand_target_clones): Diagnose and fail if node->definition and
6917 !tree_versionable_function_p (node->decl).
6919 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6921 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6923 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6924 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6925 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6927 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6930 * combine.c (try_combine): Use reg_or_subregno instead of handling
6931 just paradoxical SUBREGs and REGs.
6933 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6935 * config/arc/arc.c (arc_finalize_pic): Remove function.
6936 (arc_must_save_register): We use single base PIC register, remove
6937 checks to save/restore the PIC register.
6938 (arc_expand_prologue): Likewise.
6939 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6941 (arc_verify_short): Likewise.
6942 (arc_attr_type): Likewise.
6943 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6944 (walk_stores): Likewise.
6945 (arc_address_cost): Make it static.
6946 (arc_verify_short): Likewise.
6947 (branch_dest): Likewise.
6948 (arc_attr_type): Likewise.
6949 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6950 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6951 (arc_final_prescan_insn): Remove inserting the nops due to
6952 hardware hazards. It is done in reorg step.
6953 (insn_length_variant_t): Remove.
6954 (insn_length_parameters_t): Likewise.
6955 (arc_insn_length_parameters): Likewise.
6956 (arc_get_insn_variants): Likewise.
6957 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6959 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6962 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6965 PR tree-optimization/84687
6966 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6968 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6970 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6972 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6973 Rename to ppc_speculation_barrier.
6974 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6975 __builtin_ppc_speculation_barrier.
6977 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6980 * combine.c (combine_simplify_rtx): Don't try to simplify if
6981 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6984 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6986 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6987 to 32 bytes when compiling for POWER9.
6989 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6992 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6993 regparm >= 3 with no arg reg available also for calls with
6994 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6997 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6999 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
7001 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
7004 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
7006 2018-03-05 Richard Biener <rguenther@suse.de>
7008 PR tree-optimization/84486
7009 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
7010 When inserting a __builtin_assume_aligned call set the LHS
7011 SSA name alignment info accordingly.
7013 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
7015 PR tree-optimization/84114
7016 * config/aarch64/aarch64.c (aarch64_reassociation_width)
7017 Avoid reassociation of FLOAT_MODE addition.
7019 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
7021 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
7022 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
7023 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
7024 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
7025 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
7026 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
7027 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
7029 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
7030 __builtin_ia32_wbinvd): New builtins.
7031 (SPECIAL_ARGS2): New.
7032 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
7033 (SPECIAL_ARGS2): New.
7034 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
7035 (ix86_valid_target_attribute_inner_p): Ditto.
7036 (ix86_init_mmx_sse_builtins): Add special_args2.
7037 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7038 TARGET_WBNOINVD_P): New.
7039 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7040 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7041 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7042 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7043 * config/i386/pconfigintrin.h: New file.
7044 * config/i386/wbnoinvdintrin.h: Ditto.
7045 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7046 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7048 2018-03-05 Richard Biener <rguenther@suse.de>
7050 PR tree-optimization/84670
7051 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7053 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7054 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7055 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7056 doesn't grow if all successors (recursively) were visited at least
7059 2018-03-05 Richard Biener <rguenther@suse.de>
7061 PR tree-optimization/84650
7062 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7063 if executed in the loop pipeline.
7065 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7067 * doc/configfiles.texi (Configuration Files): Move info about
7068 conditionalizing $target-protos.h to...
7069 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7070 differs from $target-protos.h.
7072 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7073 Chung-Ju Wu <jasonwucj@gmail.com>
7075 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7076 * config/nds32/nds32-multiple.md (setmemsi): Define.
7077 * config/nds32/nds32-memory-manipulation.c
7078 (nds32_gen_dup_4_byte_to_word_value): New.
7079 (emit_setmem_word_loop): New.
7080 (emit_setmem_byte_loop): New.
7081 (nds32_expand_setmem_loop): New.
7082 (nds32_expand_setmem_loop_v3m): New.
7083 (nds32_expand_setmem_unroll): New.
7084 (nds32_expand_setmem): New.
7086 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7087 Chung-Ju Wu <jasonwucj@gmail.com>
7089 * config/nds32/nds32-memory-manipulation.c
7090 (nds32_emit_load_store): New.
7091 (nds32_emit_post_inc_load_store): New.
7092 (nds32_emit_mem_move): New.
7093 (nds32_emit_mem_move_block): New.
7094 (nds32_expand_movmemsi_loop_unknown_size): New.
7095 (nds32_expand_movmemsi_loop_known_size): New.
7096 (nds32_expand_movmemsi_loop): New.
7097 (nds32_expand_movmemsi_unroll): New.
7098 (nds32_expand_movmemqi): Rename ...
7099 (nds32_expand_movmemsi): ... to this.
7100 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7101 (movmemsi): ... to this.
7102 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7103 (nds32_expand_movmemsi): ... to this.
7105 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7106 Monk Chiang <sh.chiang04@gmail.com>
7107 Chung-Ju Wu <jasonwucj@gmail.com>
7109 * config/nds32/nds32-protos.h
7110 (nds32_expand_load_multiple): New arguments.
7111 (nds32_expand_store_multiple): Ditto.
7112 (nds32_valid_multiple_load_store): Rename ...
7113 (nds32_valid_multiple_load_store_p): ... to this.
7114 * config/nds32/nds32-memory-manipulation.c
7115 (nds32_expand_load_multiple): Refine implementation.
7116 (nds32_expand_store_multiple): Ditto.
7117 * config/nds32/nds32-multiple.md
7118 (load_multiple): Update nds32_expand_load_multiple interface.
7119 (store_multiple): Update nds32_expand_store_multiple interface.
7120 * config/nds32/nds32-predicates.c
7121 (nds32_valid_multiple_load_store): Rename ...
7122 (nds32_valid_multiple_load_store_p): ... to this and refine
7124 * config/nds32/predicates.md
7125 (nds32_load_multiple_and_update_address_operation): New predicate.
7126 (nds32_store_multiple_and_update_address_operation): New predicate.
7128 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7129 Chung-Ju Wu <jasonwucj@gmail.com>
7131 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7132 (combo): New attribute.
7133 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7135 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7137 * config/nds32/nds32.opt: Change -mcmodel= default value.
7139 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7140 Monk Chiang <sh.chiang04@gmail.com>
7141 Chung-Ju Wu <jasonwucj@gmail.com>
7143 * config/nds32/constants.md (unspec_element): New enum.
7144 * config/nds32/constraints.md (Umw): New constraint.
7145 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7146 * config/nds32/nds32-intrinsic.md: Likewise.
7147 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7148 (nds32_valid_smw_lwm_base_p): New.
7149 (nds32_output_smw_single_word): New.
7150 (nds32_output_lmw_single_word): New.
7151 (nds32_expand_unaligned_load): New.
7152 (nds32_expand_unaligned_store): New.
7153 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7154 (nds32_output_smw_single_word): Declare.
7155 (nds32_output_lmw_single_word): Declare.
7156 (nds32_expand_unaligned_load): Declare.
7157 (nds32_expand_unaligned_store): Declare.
7158 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7159 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7160 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7161 NDS32_BUILTIN_UASTORE_DW.
7162 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7165 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7166 Kito Cheng <kito.cheng@gmail.com>
7167 Chung-Ju Wu <jasonwucj@gmail.com>
7169 * config/nds32/nds32-intrinsic.c
7170 (nds32_expand_builtin_null_ftype_reg): Delete.
7171 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7172 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7173 (nds32_read_argument): New.
7174 (nds32_legitimize_target): Ditto.
7175 (nds32_legitimize_argument): Ditto.
7176 (nds32_check_constant_argument): Ditto.
7177 (nds32_expand_unop_builtin): Ditto.
7178 (nds32_expand_unopimm_builtin): Ditto.
7179 (nds32_expand_binop_builtin): Ditto.
7180 (nds32_builtin_decl_impl): Ditto.
7181 (builtin_description): Ditto.
7182 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7183 (nds32_init_builtins_impl): Ditto.
7184 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7185 (nds32_builtin_decl): New.
7186 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7187 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7189 2018-03-02 Jeff Law <law@redhat.com>
7191 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7192 (redundant_insn, fill_simple_delay_slots): Likewise.
7193 (fill_slots_from_thread): Likewise.
7194 * resource.c (mark_referenced_resources): Likewise.
7195 (mark_set_resources, find_dead_or_set_registers): Likewise.
7197 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7199 * substring-locations.h (format_warning_va): Formatting fix for
7201 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7203 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7204 * substring-locations.c: Include intl.h.
7205 (format_warning_va): Turned into small wrapper around
7206 format_warning_n_va, renamed to ...
7207 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7208 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7210 (format_warning_at_substring_n): New function.
7211 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7212 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7213 format_warning_at_substring with just a shorter name instead of
7214 const function pointer.
7215 (fmtwarn_n): New function.
7216 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7217 appropriate, get rid of all the fmtstr temporaries, move conditionals
7218 with G_() wrapped string literals directly into fmtwarn arguments,
7219 cast dir.len to (int), formatting fixes.
7221 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7223 * doc/invoke.texi: Remove "Cilk Plus" references.
7225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7226 Richard Biener <rguenther@suse.de>
7229 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7230 for error or warning attributes if CALL_FROM_THUNK_P is set.
7233 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7236 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7237 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7240 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7241 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7243 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7244 instead of -1U in last predictors element's probability member.
7246 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7249 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7250 arguments if they are comparable.
7252 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7254 PR tree-optimization/84634
7255 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7256 masks and masked_loop_p with a single loop_masks, making sure it's
7257 null for bb vectorization.
7259 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7261 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7262 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7263 loop->force_vectorize to check whether there is no alias.
7265 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7268 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7270 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7272 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7273 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7274 instead of a loop around prev_real_insn.
7275 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7279 * config/i386/i386.c (ix86_print_operand): Use conditional
7280 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7283 2018-03-02 Richard Biener <rguenther@suse.de>
7285 PR tree-optimization/84427
7286 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7287 (bitmap_set_subtract_values): Rewrite to handle multiple
7290 (prune_clobbered_mems): Likewise.
7291 (phi_translate): Take edge instead of pred/phiblock.
7292 (phi_translate_1): Likewise.
7293 (phi_translate_set): Likewise. Insert all translated
7294 exprs for a value into the set, keeping possibly multiple
7295 expressions per value.
7296 (compute_antic_aux): Adjust for phi_translate changes.
7297 When intersecting union the expressions and prune those
7298 not in the final value set, keeping possibly multiple
7299 expressions per value. Do not use value-insertion
7300 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7301 all expressions. Add verification that the value-sets
7302 only shrink during iteration.
7303 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7304 (do_pre_regular_insertion): Likewise.
7305 (do_pre_partial_partial_insertion): Likewise.
7307 2018-03-02 Richard Biener <rguenther@suse.de>
7310 * config/darwin.c (saved_debug_info_level): New static global.
7311 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7312 (darwin_asm_lto_end): Restore debug info generation settings.
7314 2018-03-01 Martin Liska <mliska@suse.cz>
7317 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7320 2018-03-01 Richard Biener <rguenther@suse.de>
7323 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7324 type annotation with LTO when debug was disabled at compile-time.
7326 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7328 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7330 (mips_final_postscan_insn): Likewise.
7332 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7334 PR rtl-optimization/84528
7335 * alias.c (init_alias_target): Add commentary.
7336 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7337 a unique base value if the frame pointer is not eliminated
7338 to the stack pointer.
7340 2018-03-01 Tom de Vries <tom@codesourcery.com>
7342 PR rtl-optimization/83327
7343 * lra-int.h (hard_regs_spilled_into): Declare.
7344 * lra.c (hard_regs_spilled_into): Define.
7345 (init_reg_info): Init hard_regs_spilled_into.
7346 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7347 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7348 (process_bb_lives): Handle hard_regs_spilled_into.
7349 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7350 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7352 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7354 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7355 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7356 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7357 * config/rs6000/aix72.h: New file.
7359 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7361 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7362 instead of warning_at with conditional singular and plural messages
7366 * stor-layout.c (update_alignment_for_field): For
7367 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7368 && !DECL_PACKED (field), do the alignment update, just use
7369 only desired_align instead of MAX (type_align, desired_align)
7371 (place_field): Don't do known_align < desired_align handling
7372 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7373 is non-NULL, instead do it after rli->prev_field handling and
7374 only if not within a bitfield word. For DECL_PACKED (field)
7375 use type_align of BITS_PER_UNIT.
7377 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7379 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7380 superfluous parentheses and trailing spaces.
7382 2018-02-28 Richard Biener <rguenther@suse.de>
7384 PR tree-optimization/84584
7385 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7386 SCoPs with fake exit edge.
7388 2018-02-28 Martin Liska <mliska@suse.cz>
7391 * timevar.c (timer::print): Fix format to properly print 100%
7394 2018-02-28 Richard Biener <rguenther@suse.de>
7397 * genmatch.c (capture_info::walk_match): Do not mark
7398 captured expressions without operands as expr_p given
7399 they act more like predicates and should be subject to
7400 "lost tail" side-effect preserving.
7402 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7404 PR rtl-optimization/81611
7405 * auto-inc-dec.c (attempt_change): Move dead note from
7406 mem_insn if it's the next use of regno
7407 (find_address): Take address use of reg holding
7408 non-incremented value. Add parm to limit search to the named
7410 (merge_in_block): Attempt to use a mem insn that is the next
7411 use of the original regno.
7413 2018-02-27 Martin Sebor <msebor@redhat.com>
7416 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7417 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7419 2018-02-27 Martin Sebor <msebor@redhat.com>
7421 PR translation/84207
7422 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7423 n argument to unsigned HOST_WIDE_INT.
7424 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7425 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7426 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7429 2018-02-27 Richard Biener <rguenther@suse.de>
7431 PR tree-optimization/84512
7432 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7433 Do not use the estimate returned from record_stmt_cost for
7434 the scalar iteration cost but sum properly using add_stmt_cost.
7436 2018-02-27 Richard Biener <rguenther@suse.de>
7438 PR tree-optimization/84466
7439 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7440 Adjust last change to less strictly validate use operands.
7442 2018-02-27 Martin Liska <mliska@suse.cz>
7444 PR gcov-profile/84548
7445 * gcov.c (process_file): Allow partial overlap and consider it
7446 also as group functions.
7447 (output_lines): Properly calculate range of lines for a group.
7449 2018-02-27 Martin Liska <mliska@suse.cz>
7451 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7452 'ggc' suffixes. Change first column width.
7453 (timer::print): Fix formatting of the column.
7455 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7457 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7458 preserve inline entry blocks for the sake of debug inline
7459 entry point markers alone.
7460 (remove_unused_locals): Suggest in comments a better place to
7461 force the preservation of inline entry blocks that are
7462 otherwise unused, but do not preserve them.
7464 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7466 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7468 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7471 * config/i386/constraints.md (Bs): Replace
7472 ix86_indirect_branch_register with
7473 TARGET_INDIRECT_BRANCH_REGISTER.
7475 * config/i386/i386.md (indirect_jump): Likewise.
7476 (tablejump): Likewise.
7477 (*sibcall_memory): Likewise.
7478 (*sibcall_value_memory): Likewise.
7479 Peepholes of indirect call and jump via memory: Likewise.
7480 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7481 (*sibcall_value_GOT_32): Likewise.
7482 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7483 (GOT_memory_operand): Likewise.
7484 (call_insn_operand): Likewise.
7485 (sibcall_insn_operand): Likewise.
7486 (GOT32_symbol_operand): Likewise.
7487 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7489 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7491 PR rtl-optimization/83496
7492 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7493 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7494 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7495 redundant insn, if any.
7496 (relax_delay_slots): Likewise.
7497 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7499 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7501 PR tree-optimization/83965
7502 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7503 that grouped statements are part of a reduction chain. Return
7504 true if the statement is not marked as a reduction itself but
7506 (vect_recog_dot_prod_pattern): Don't check whether the statement
7507 is part of a group here.
7508 (vect_recog_sad_pattern): Likewise.
7509 (vect_recog_widen_sum_pattern): Likewise.
7511 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7514 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7516 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7519 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7521 (ix86_output_indirect_function_return): New prototype.
7522 (ix86_split_simple_return_pop_internal): Likewise.
7523 * config/i386/i386.c (indirect_return_via_cx): New.
7524 (indirect_return_via_cx_bnd): Likewise.
7525 (indirect_thunk_name): Handle return va CX_REG.
7526 (output_indirect_thunk_function): Create alias for
7527 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7528 (ix86_output_indirect_jmp): Remove the bool argument.
7529 (ix86_output_indirect_function_return): New function.
7530 (ix86_split_simple_return_pop_internal): Likewise.
7531 * config/i386/i386.md (*indirect_jump): Don't pass false
7532 to ix86_output_indirect_jmp.
7533 (*tablejump_1): Likewise.
7534 (simple_return_pop_internal): Change it to define_insn_and_split.
7535 Call ix86_split_simple_return_pop_internal to split it for
7537 (simple_return_indirect_internal): Call
7538 ix86_output_indirect_function_return instead of
7539 ix86_output_indirect_jmp.
7541 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7544 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7545 memset and value initialization afterwards.
7547 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7549 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7551 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7554 * common/config/aarch64/aarch64-common.c
7555 (aarch_option_optimization_table[]): Switch
7556 off fomit-frame-pointer
7558 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7559 Chung-Ju Wu <jasonwucj@gmail.com>
7561 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7563 (store_multiple): Ditto.
7565 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7567 * config.gcc: Add --with-cpu support for nds32 target.
7568 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7569 * config/nds32/nds32.opt: Add -mcpu= option.
7571 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7573 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7574 isel=yes): Warn for these deprecated options.
7576 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7578 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7579 ISA_2_5_MASKS_EMBEDDED.
7581 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7583 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7584 p->max as pointers rather than using iterative_hash_expr.
7586 2018-02-23 Carl Love <cel@us.ibm.com>
7588 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7589 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7591 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7592 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7593 P8V_BUILTIN_VEC_VUNSIGNED2.
7595 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7598 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7599 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7601 * lra-constraints.c (curr_insn_transform): Set up
7602 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7604 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7605 * lra-eliminations.c (spill_pseudos): Ditto.
7606 (process_insn_for_elimination): Ditto.
7607 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7608 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7609 LRA_NON_CLOBBERED_ALT.
7611 2018-02-22 Martin Sebor <msebor@redhat.com>
7613 PR tree-optimization/84480
7614 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7615 to maybe_diag_stxncpy_trunc. Call it.
7616 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7617 from gimple_fold_builtin_strcpy. Print inlining stack.
7618 (handle_builtin_stxncpy): Print inlining stack.
7619 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7621 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7624 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7625 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7626 and -fcheck-pointer-bounds are used together.
7627 (indirect_thunk_prefix): New enum.
7628 (indirect_thunk_need_prefix): New function.
7629 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7630 "_nt" instead of "_bnd" for NOTRACK prefix.
7631 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7632 (output_indirect_thunk_function): Likewise.
7634 (ix86_code_end): Update output_indirect_thunk_function calls.
7635 (ix86_output_indirect_branch_via_reg): Replace
7636 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7637 (ix86_output_indirect_branch_via_push): Likewise.
7638 (ix86_output_function_return): Likewise.
7639 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7640 incompatible with -fcf-protection=branch and
7641 -fcheck-pointer-bounds.
7643 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7646 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7647 Change gcc_assert call to output_operand_lossage.
7649 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7651 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7653 2018-02-22 DJ Delorie <dj@redhat.com>
7654 Sebastian Perta <sebastian.perta@renesas.com>
7655 Oleg Endo <olegendo@gcc.gnu.org>
7657 * config/rx/rx.c (rx_rtx_costs): New function.
7658 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7660 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7662 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7664 2018-02-22 Martin Liska <mliska@suse.cz>
7667 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7668 Add "native" as a possible value.
7670 2018-02-22 Martin Liska <mliska@suse.cz>
7673 * config/i386/i386.c (ix86_option_override_internal):
7674 Add "native" as a possible value for -march and -mtune.
7676 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7679 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7680 to all type variants.
7682 PR tree-optimization/84503
7683 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7684 width as info->bitpos + info->bitsize - start.
7685 (merged_store_group::merge_overlapping): Simplify width computation.
7686 (check_no_overlap): New function.
7687 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7688 start + width and last_order of the group, fail if check_no_overlap
7690 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7691 to group if check_no_overlap fails.
7693 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7695 * config/rs6000/altivec.md: Delete contraint arguments to
7696 define_expand, define_split, and define_peephole2, and in
7697 define_insn_and_split if always unused.
7698 * config/rs6000/darwin.md: Ditto.
7699 * config/rs6000/dfp.md: Ditto.
7700 * config/rs6000/rs6000.md: Ditto.
7701 * config/rs6000/sync.md: Ditto.
7702 * config/rs6000/vector.md: Ditto.
7703 * config/rs6000/vsx.md: Ditto.
7705 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7707 * config/rs6000/altivec.md: Write output control strings as braced
7708 blocks instead of double-quoted strings.
7709 * config/rs6000/darwin.md: Ditto.
7710 * config/rs6000/rs6000.md: Ditto.
7711 * config/rs6000/vector.md: Ditto.
7712 * config/rs6000/vsx.md: Ditto.
7714 2018-02-21 Jason Merrill <jason@redhat.com>
7716 PR c++/84314 - ICE with templates and fastcall attribute.
7717 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7719 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7721 * ipa-cp.c (determine_versionability): Fix comment typos.
7723 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7726 * ipa-cp.c (determine_versionability): Do not version functions caling
7729 2018-02-21 Martin Liska <mliska@suse.cz>
7732 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7733 Add "native" as a possible value.
7734 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7735 the macro when native cpu detection is available.
7737 2018-02-21 Martin Liska <mliska@suse.cz>
7740 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7741 Add "native" as a possible value.
7742 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7743 when native cpu detection is available.
7745 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7746 Martin Sebor <msebor@redhat.com>
7748 PR tree-optimization/84478
7749 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7751 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7752 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7753 support which is conservatively correct, for 2 only stay conservative
7754 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7755 argument to the 2 argument get_range_strlen, adjust 6 arg
7756 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7758 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7759 (gimple_fold_builtin_strlen): Pass true as last argument to
7762 2018-02-20 Martin Sebor <msebor@redhat.com>
7765 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7766 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7767 (builtin_memref::builtin_memref): Factor out parts into
7768 set_base_and_offset and call it.
7770 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7773 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7774 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7775 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7776 search at the associated MODE_INT.
7778 2018-02-20 Jeff Law <law@redhat.com>
7781 PR tree-optimization/81592
7783 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7784 for range data rather than using global data.
7785 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7786 range data rather than using global data.
7787 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7788 pass it to children as needed.
7789 (struct directive::fmtresult): Similarly.
7790 (struct directive::set_width): Similarly.
7791 (struct directive::set_precision): Similarly.
7792 (format_integer, format_directive, parse_directive): Similarly.
7793 (format_none): Accept unnamed vr_values parameter.
7794 (format_percent, format_floating, format_character): Similarly.
7795 (format_string, format_plain): Similarly.
7796 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7797 the EVRP range analyzer for range data rather than using global data.
7798 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7799 gimple-ssa-evrp-analyze.h
7800 (class sprintf_dom_walker): Add after_dom_children member function.
7801 Add evrp_range_analyzer member.
7802 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7803 range analyzer as needed.
7804 (sprintf_dom_walker::after_dom_children): New member function.
7805 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7807 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7808 (evrp_range_analyzer::pop_to_marker): Likewise.
7810 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7812 PR tree-optimization/84419
7813 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7814 with the required type if its current type is compatible but
7817 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7820 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7821 after vectorization.
7823 2018-02-20 Martin Liska <mliska@suse.cz>
7826 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7827 possible values if we don't have a hint.
7829 2018-02-20 Martin Liska <mliska@suse.cz>
7833 * final.c (shorten_branches): Build align_tab array with one
7835 * opts.c (finish_options): Add alignment option limit check.
7836 (MAX_CODE_ALIGN): Likewise.
7837 (MAX_CODE_ALIGN_VALUE): Likewise.
7838 * doc/invoke.texi: Document maximum allowed option value for
7839 all -falign-* options.
7841 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7844 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7845 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7846 * var-tracking.c (emit_note_insn_var_location): Remove all references
7847 to NOTE_INSN_CALL_ARG_LOCATION.
7848 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7849 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7850 Use copy_rtx_if_shared.
7851 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7852 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7853 (dwarf2out_var_location): Remove handling of
7854 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7856 * final.c (final_scan_insn): Remove all references to
7857 NOTE_INSN_CALL_ARG_LOCATION.
7858 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7859 before dumping final insns.
7860 * except.c (emit_note_eh_region_end): Remove all references to
7861 NOTE_INSN_CALL_ARG_LOCATION.
7862 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7863 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7864 * config/arc/arc.c (hwloop_optimize): Likewise.
7865 * config/arm/arm.c (create_fix_barrier): Likewise.
7866 * config/s390/s390.c (s390_chunkify_start): Likewise.
7867 * config/sh/sh.c (find_barrier): Likewise.
7868 * config/i386/i386.c (rest_of_insert_endbranch,
7869 ix86_seh_fixup_eh_fallthru): Likewise.
7870 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7871 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7872 * config/frv/frv.c (frv_function_prologue): Likewise.
7873 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7875 (note_outside_basic_block_p): Remove all references to
7876 NOTE_INSN_CALL_ARG_LOCATION.
7877 * gengtype.c (adjust_field_rtx_def): Likewise.
7878 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7880 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7881 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7884 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7887 PR tree-optimization/84452
7888 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7889 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7892 2018-02-19 Martin Liska <mliska@suse.cz>
7895 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7897 2018-02-19 Martin Liska <mliska@suse.cz>
7898 Richard Sandiford <richard.sandiford@linaro.org>
7900 PR tree-optimization/82491
7901 * gimple-fold.c (get_base_constructor): Make earlier bail out
7904 2018-02-19 Carl Love <cel@us.ibm.com>
7906 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7907 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7909 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7910 P8V_BUILTIN_VEC_NEG.
7912 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7914 * config/rl78/rl78.md (movdf): New define expand.
7916 2018-02-19 Martin Liska <mliska@suse.cz>
7919 * doc/invoke.texi: Fix typo.
7920 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7922 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7924 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7925 handle rs6000_single_float and rs6000_double_float specially for
7928 2018-02-16 Jeff Law <law@redhat.com>
7930 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7933 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7936 * ipa-inline.c (inline_small_functions): Fix a typo.
7938 2018-02-16 Nathan Sidwell <nathan@acm.org>
7940 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7942 2018-02-16 Carl Love <cel@us.ibm.com>
7944 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7945 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7946 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7947 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7948 expansion to P8V_BUILTIN_VEC_FLOAT2.
7950 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7952 PR rtl-optimization/70023
7953 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7954 src_regno into account.
7956 2018-02-16 Carl Love <cel@us.ibm.com>
7958 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7959 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7960 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7961 * config/rs6000/rs6000.c: Remove case statements for
7962 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7963 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7964 and P9V_BUILTIN_VEC_VINSERT4B.
7965 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7966 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7967 * config/rs6000/vsx.md:
7968 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7971 2018-02-16 Carl Love <cel@us.ibm.com>
7973 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7975 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7977 * config/rs6000/rs6000-c.c: Add the definitions for
7978 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7979 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7980 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7981 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7982 definition for insert4b and define insn *insert3b_internal.
7983 * doc/extend.texi: Add documentation for vec_extract4b.
7985 2018-02-16 Nathan Sidwell <nathan@acm.org>
7987 * doc/extend.texi (Backwards Compatibility): Mention friend
7988 injection. Note for-scope is deprecated.
7989 * doc/invoke.texi (-ffriend-injection): Deprecate.
7991 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7993 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7994 that moved to I2, also allow destinations that are a paradoxical
7995 subreg (instead of a normal reg).
7997 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
8000 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
8003 2018-02-16 Richard Biener <rguenther@suse.de>
8005 PR tree-optimization/84037
8006 PR tree-optimization/84016
8008 * config/i386/i386.c (ix86_builtin_vectorization_cost):
8009 Adjust vec_construct for the fact we need additional higher latency
8010 128bit inserts for AVX256 and AVX512 vector builds.
8011 (ix86_add_stmt_cost): Scale vector construction cost for
8014 2018-02-16 Richard Biener <rguenther@suse.de>
8016 PR tree-optimization/84417
8017 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
8018 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
8019 (non_rewritable_lvalue_p): Likewise, use poly-ints.
8021 2018-02-16 Martin Liska <mliska@suse.cz>
8024 * internal-fn.def (ASAN_CHECK): Set proper flags.
8025 (ASAN_MARK): Likewise.
8027 2018-02-16 Julia Koval <julia.koval@intel.com>
8029 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
8030 from PTA_CANNONLAKE.
8032 2018-02-16 Jakub Jelinek <jakub@redhat.com>
8035 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8036 Use ++iter rather than iter++ for std::list iterators.
8037 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8038 defer deleting them until all nodes in the forest are processed. Do
8039 free even leaf nodes. Change to_process into auto_vec.
8042 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8043 * vec.h (vec_default_construct): Use memset instead of placement new
8044 if BROKEN_VALUE_INITIALIZATION is defined.
8045 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8046 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8049 PR rtl-optimization/83723
8050 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8051 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8052 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8054 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8056 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8058 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8060 PR rtl-optimization/81443
8061 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8062 from inner REGs to paradoxical SUBREGs.
8064 2018-02-16 Richard Biener <rguenther@suse.de>
8066 PR tree-optimization/84399
8067 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8068 For operands we can analyze at their definition make sure we can
8069 analyze them at each use as well.
8071 2018-02-16 Richard Biener <rguenther@suse.de>
8073 PR tree-optimization/84190
8074 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8075 volatile accesses if the decl isn't volatile.
8077 2018-02-15 Jason Merrill <jason@redhat.com>
8079 PR c++/84314 - ICE with templates and fastcall attribute.
8080 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8081 TYPE_CANONICAL on an existing type.
8083 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8085 PR tree-optimization/84383
8086 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8087 dstoff nor call operand_equal_p if dstbase is NULL.
8089 PR tree-optimization/84334
8090 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8091 also a CONSTANT_CLASS_P, punt.
8093 2018-02-14 Jim Wilson <jimw@sifive.com>
8095 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8096 first SMALL_OPERAND check. New local min_second_step. Move assert
8097 to where locals are set. Add TARGET_RVC support.
8098 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8100 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8102 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8104 2018-02-14 Martin Sebor <msebor@redhat.com>
8106 PR tree-optimization/83698
8107 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8108 arrays constrain the offset range to their bounds.
8109 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8110 (builtin_access::overlap): Avoid setting the size of overlap if it's
8112 (maybe_diag_overlap): Also consider arrays when deciding what values
8113 of offsets to include in diagnostics.
8115 2018-02-14 Martin Sebor <msebor@redhat.com>
8118 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8119 that correspond to the kind of a declaration.
8121 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8124 * config/pa/pa.md: Load address of PIC label using the linkage table
8125 if the label is nonlocal.
8127 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8129 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8130 warning message if user requests -maltivec=be.
8131 * doc/invoke.texi: Document deprecation of -maltivec=be.
8133 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8136 * config/rs6000/rs6000-c.c: Update definitions for
8137 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8138 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8140 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8143 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8144 add _get_ssp intrinsics. Remove argument from
8145 __builtin_ia32_rdssp[d|q].
8146 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8147 * config/i386/i386-builtin.def: Remove argument from
8148 __builtin_ia32_rdssp[d|q].
8149 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8150 ix86_expand_special_args_builtin for _rdssp[d|q].
8151 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8152 Clear register before usage.
8153 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8154 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8156 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8158 PR tree-optimization/84357
8159 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8160 operand 1 of an ARRAY_REF too.
8162 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8165 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8166 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8168 (set_of_reg): New struct.
8169 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8170 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8171 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8173 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8174 Split into bitclr, bitset, bitinvert patterns if appropriate.
8175 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8176 use rx_fuse_in_memory_bitop.
8177 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8178 to named insn, correct maximum insn length.
8180 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8183 * machmode.def: Define a complex mode for PARTIAL_INT.
8184 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8186 * doc/rtl.texi: Document CSPImode.
8187 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8189 (msp430_hard_regno_nregs_with_padding): Likewise.
8191 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8194 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8196 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8198 PR rtl-optimization/84169
8199 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8200 we generated a parallel as new i3 and we split that to new i2 and i3
8201 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8202 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8203 those to i2, not i1. Partially rewrite this scan code.
8205 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8208 * stor-layout.c (place_field): For variable length fields, adjust
8209 offset_align afterwards not just based on the field's alignment,
8210 but also on the size.
8213 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8214 of exps and logs in the use_exp2 case.
8216 2018-02-13 Jeff Law <law@redhat.com>
8218 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8221 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8224 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8228 * common.opt (gas-loc-support, gas-locview-support): New.
8229 (ginline-points, ginternal-reset-location-views): New.
8230 * doc/invoke.texi: Document them. Use @itemx where intended.
8231 (gvariable-location-views): Adjust.
8232 * target.def (reset_location_view): New.
8233 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8234 (TARGET_RESET_LOCATION_VIEW): New.
8235 * doc/tm.texi: Rebuilt.
8236 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8237 (dwarf2out_default_as_locview_support): New.
8238 (output_asm_line_debug_info): Use option variables.
8239 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8240 (output_loc_list): Likewise.
8241 (add_high_low_attributes): Check option variables.
8242 Don't output entry view attribute in strict mode.
8243 (gen_inlined_subroutine_die): Check option variables.
8244 (dwarf2out_inline_entry): Likewise.
8245 (init_sections_and_labels): Likewise.
8246 (dwarf2out_early_finish): Likewise.
8247 (maybe_reset_location_view): New, from...
8248 (dwarf2out_var_location): ... here. Call it.
8249 * debug.h (dwarf2out_default_as_loc_support): Declare.
8250 (dwarf2out_default_as_locview_support): Declare.
8251 * hooks.c (hook_int_rtx_insn_0): New.
8252 * hooks.h (hook_int_rtx_insn_0): Declare.
8253 * toplev.c (process_options): Take -gas-loc-support and
8254 -gas-locview-support from dwarf2out. Enable
8255 -gvariable-location-views by default only with locview
8256 assembler support. Enable -ginternal-reset-location-views by
8257 default only if the target defines the corresponding hook.
8258 Enable -ginline-points by default if location views are
8259 enabled; force it disabled if statement frontiers are
8261 * tree-inline.c (expand_call_inline): Check option variables.
8262 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8264 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8266 PR tree-optimization/84321
8267 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8268 handling. Also check whether the anti-range contains any values
8269 that satisfy the mask; switch to a VR_RANGE if not.
8271 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8274 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8276 2018-02-13 Martin Jambor <mjambor@suse.cz>
8279 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8280 of call statements, also set location of a load to a temporary.
8282 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8284 * config/rl78/rl78.c (add_vector_labels): New function.
8285 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8286 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8287 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8288 which checks that no arguments are passed.
8289 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8290 * doc/extend.texi: Documentation for the new attribute.
8292 2018-02-13 Andreas Schwab <schwab@suse.de>
8294 * config/riscv/linux.h (CPP_SPEC): Define.
8296 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8299 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8300 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8301 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8302 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8303 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8304 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8305 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8306 temporarily for AES and PCLMUL builtins.
8308 PR tree-optimization/84339
8309 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8310 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8314 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8315 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8316 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8318 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8320 * omp-simd-clone.h: New file.
8321 * omp-simd-clone.c: Include omp-simd-clone.h.
8322 (expand_simd_clones): No longer static.
8323 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8324 cgraph.h and omp-simd-clone.h.
8325 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8326 (vect_recog_widen_shift_pattern): Formatting fix.
8327 (vect_pattern_recog_1): Don't check optab for calls.
8330 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8331 operands[2] into a REG before using gen_lowpart on it.
8333 2018-02-12 Jeff Law <law@redhat.com>
8336 * config/sh/sh.c (find_barrier): Consider a sibling call
8339 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8340 successfully back substituting a reg.
8342 2018-02-12 Richard Biener <rguenther@suse.de>
8344 PR tree-optimization/84037
8345 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8346 parameter, move visited init to caller.
8347 (vect_slp_analyze_operations): Separate cost from validity
8348 check, initialize visited once for all instances.
8349 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8351 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8353 (vect_model_promotion_demotion_cost): Likewise.
8354 (vectorizable_bswap): Guard cost modeling with !slp_node
8355 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8357 (vectorizable_call): Likewise.
8358 (vectorizable_conversion): Likewise.
8359 (vectorizable_assignment): Likewise.
8360 (vectorizable_shift): Likewise.
8361 (vectorizable_operation): Likewise.
8362 (vectorizable_store): Likewise.
8363 (vectorizable_load): Likewise.
8364 (vectorizable_condition): Likewise.
8365 (vectorizable_comparison): Likewise.
8367 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8370 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8371 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8372 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8374 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8377 * params.def (inline-min-speedup): Increase from 8 to 15.
8378 (max-inline-insns-auto): Decrease from 40 to 30.
8379 * ipa-split.c (consider_split): Add some buffer for function to
8380 be considered inlining candidate.
8381 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8384 2018-02-12 Richard Biener <rguenther@suse.de>
8386 PR tree-optimization/84037
8387 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8388 matched stmts if we cannot swap the non-matched ones.
8390 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8392 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8393 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8394 _mm_maskz_scalef_round_ss): New intrinsics.
8395 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8396 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8397 __builtin_ia32_scalefss_round): Remove.
8398 (__builtin_ia32_scalefsd_mask_round,
8399 __builtin_ia32_scalefss_mask_round): New intrinsics.
8400 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8401 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8402 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8403 "<round_constraint>")): Changed to ...
8404 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8405 "<round_scalar_constraint>")): ... this.
8406 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8407 %0, %1, %2<round_op3>}"): Changed to ...
8408 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8409 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8410 %2<round_scalar_mask_op3>}"): ... this.
8411 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8413 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8415 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8416 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8417 (_mm_maskz_sqrt_round_ss): New intrinsics.
8418 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8419 (__builtin_ia32_sqrtsd_mask_round)
8420 (__builtin_ia32_sqrtss_mask_round): New builtins.
8421 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8422 (__builtin_ia32_sqrtss_round): Remove.
8423 (__builtin_ia32_sqrtsd_mask_round)
8424 (__builtin_ia32_sqrtss_mask_round): New builtins.
8425 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8426 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8427 ((match_operand:VF_128 1 "vector_operand"
8428 "xBm,<round_constraint>")): Changed to ...
8429 ((match_operand:VF_128 1 "vector_operand"
8430 "xBm,<round_scalar_constraint>")): ... this.
8431 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8432 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8433 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8434 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8435 %<iptr>1<round_scalar_mask_op3>}): ... this.
8436 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8437 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8439 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8442 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8443 Cast vec_cmpeq result to correct type.
8444 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8445 Cast vec_cmpgt result to correct type.
8447 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8449 * final.c (final_scan_insn_1): Renamed from...
8450 (final_scan_insn): ... this. New wrapper, to recover
8451 seen from the outermost call in recursive ones.
8452 * config/sparc/sparc.c (output_return): Drop seen from call.
8453 (output_sibcall): Likewise.
8454 * config/visium/visium.c (output_branch): Likewise.
8456 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8458 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8461 2018-02-10 Alan Modra <amodra@gmail.com>
8464 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8465 Specify LR as an input.
8467 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8470 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8471 remove_member_access_dummy_vars): New functions.
8472 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8473 lower_omp_1, execute_lower_omp): Use them.
8475 PR rtl-optimization/84308
8476 * shrink-wrap.c (spread_components): Release todo vector.
8478 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8480 PR rtl-optimization/57193
8481 * ira-color.c (struct allocno_color_data): Add member
8482 conflict_allocno_hard_prefs.
8483 (update_conflict_allocno_hard_prefs): New.
8484 (bucket_allocno_compare_func): Add a preference based on
8485 conflict_allocno_hard_prefs.
8486 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8487 (color_allocnos): Remove a dead code. Initiate
8488 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8490 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8493 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8494 constraint from =wa to wa. Avoid a subreg on the output operand,
8495 instead use a pseudo and subreg it in a move.
8496 (p9_xxbrd_<mode>): Changed to ...
8497 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8498 (p9_xxbrd_v2df): New expander.
8499 (p9_xxbrw_<mode>): Changed to ...
8500 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8501 (p9_xxbrw_v4sf): New expander.
8503 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8505 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8507 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8510 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8511 multiply in 32-bit mode.
8512 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8513 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8516 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8518 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8519 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8520 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8521 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8523 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8526 * dwarf2out.c (is_trivial_indirect_ref): New function.
8527 (dwarf2out_late_global_decl): Do not generate a location
8528 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8529 and that are not defined in the current unit.
8531 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8533 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8534 instead of a libcall for UNORDERED.
8536 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8539 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8540 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8542 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8545 * config/s390/s390.c (s390_set_current_function): Invoke
8546 s390_indirect_branch_settings also if fndecl didn't change.
8548 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8550 * config/rs6000/rs6000.md (blockage): Set length to zero.
8552 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8554 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8556 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8559 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8560 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8564 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8565 PARALLEL incoming that failed vt_get_decl_and_offset check.
8568 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8569 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8571 (get_variable_section): For decls in named .bss* sections pass true as
8572 second argument to bss_initializer_p.
8574 2018-02-09 Marek Polacek <polacek@redhat.com>
8575 Jakub Jelinek <jakub@redhat.com>
8578 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8579 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8580 Sync some changes from cxx_fold_indirect_ref.
8582 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8584 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8586 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8587 (BLOCK_INLINE_ENTRY_LABEL): New.
8588 (dwarf2out_var_location): Disregard inline entry markers.
8589 (inline_entry_data): New struct.
8590 (inline_entry_data_hasher): New hashtable type.
8591 (inline_entry_data_hasher::hash): New.
8592 (inline_entry_data_hasher::equal): New.
8593 (inline_entry_data_table): New variable.
8594 (add_high_low_attributes): Add DW_AT_entry_pc and
8595 DW_AT_GNU_entry_view attributes if a pending entry is found
8596 in inline_entry_data_table. Add old entry_pc attribute only
8597 if debug nonbinding markers are disabled.
8598 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8599 markers are enabled.
8600 (block_within_block_p, dwarf2out_inline_entry): New.
8601 (dwarf2out_finish): Check that no entries remained in
8602 inline_entry_data_table.
8603 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8604 (final_scan_insn, notice_source_line): Likewise.
8605 (rest_of_clean_state): Skip inline entry markers.
8606 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8608 * gimple.c (gimple_build_debug_inline_entry): New.
8609 * gimple.h (enum gimple_debug_subcode): Add
8610 GIMPLE_DEBUG_INLINE_ENTRY.
8611 (gimple_build_debug_inline_entry): Declare.
8612 (gimple_debug_inline_entry_p): New.
8613 (gimple_debug_nonbind_marker_p): Adjust.
8614 * insn-notes.def (INLINE_ENTRY): New.
8615 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8616 inline entry marker notes.
8617 (print_insn): Likewise.
8618 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8619 (INSN_DEBUG_MARKER_KIND): Likewise.
8620 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8621 * tree-inline.c (expand_call_inline): Build and insert
8622 debug_inline_entry stmt.
8623 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8624 inline entry blocks early, if nonbind markers are enabled.
8625 (dump_scope_block): Dump fragment info.
8626 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8627 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8628 (gimple_build_debug_inline_entry): New.
8629 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8630 Enable/disable inline entry points too.
8631 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8632 (DEBUG_INSN): Describe inline entry markers.
8634 * common.opt (gvariable-location-views): New.
8635 (gvariable-location-views=incompat5): New.
8636 * config.in: Rebuilt.
8637 * configure: Rebuilt.
8638 * configure.ac: Test assembler for view support.
8639 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8640 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8641 * dwarf2out.c (var_loc_view): New typedef.
8642 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8643 (dwarf2out_locviews_in_attribute): New.
8644 (dwarf2out_locviews_in_loclist): New.
8645 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8646 (enum dw_line_info_opcode): Add LI_adv_address.
8647 (struct dw_line_info_table): Add view.
8648 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8649 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8650 (zero_view_p): New variable.
8651 (ZERO_VIEW_P): New macro.
8652 (output_asm_line_debug_info): New.
8653 (struct var_loc_node): Add view.
8654 (add_AT_view_list, AT_loc_list): New.
8655 (add_var_loc_to_decl): Add view param. Test it against last.
8656 (new_loc_list): Add view params. Record them.
8657 (AT_loc_list_ptr): Handle loc and view lists.
8658 (view_list_to_loc_list_val_node): New.
8659 (print_dw_val): Handle dw_val_class_view_list.
8660 (size_of_die): Likewise.
8661 (value_format): Likewise.
8662 (loc_list_has_views): New.
8663 (gen_llsym): Set vl_symbol too.
8664 (maybe_gen_llsym, skip_loc_list_entry): New.
8665 (dwarf2out_maybe_output_loclist_view_pair): New.
8666 (output_loc_list): Output view list or entries too.
8667 (output_view_list_offset): New.
8668 (output_die): Handle dw_val_class_view_list.
8669 (output_dwarf_version): New.
8670 (output_compilation_unit_header): Use it.
8671 (output_skeleton_debug_sections): Likewise.
8672 (output_rnglists, output_line_info): Likewise.
8673 (output_pubnames, output_aranges): Update version comments.
8674 (output_one_line_info_table): Output view numbers in asm comments.
8675 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8676 Call maybe_gen_llsym.
8677 (loc_list_from_tree_1): Adjust.
8678 (add_AT_location_description): Create view list attribute if
8679 needed, check it's absent otherwise.
8680 (convert_cfa_to_fb_loc_list): Adjust.
8681 (maybe_emit_file): Call output_asm_line_debug_info for test.
8682 (dwarf2out_var_location): Reset views as needed. Precompute
8683 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8684 attribute. Set view.
8685 (new_line_info_table): Reset next view.
8686 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8687 (dwarf2out_source_line): Likewise. Output view resets and labels to
8688 the assembler, or select appropriate line info opcodes.
8689 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8690 (optimize_string_length): Catch it. Adjust.
8691 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8692 dw_val_class_view_list, and remove it if no longer needed.
8693 (hash_loc_list): Hash view numbers.
8694 (loc_list_hasher::equal): Compare them.
8695 (optimize_location_lists): Check whether a view list symbol is
8696 needed, and whether the locview attribute is present, and
8697 whether they match. Remove the locview attribute if no longer
8699 (index_location_lists): Call skip_loc_list_entry for test.
8700 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8701 Use output_dwarf_version.
8702 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8703 (struct dw_val_node): Add val_view_list.
8704 * final.c (SEEN_NEXT_VIEW): New.
8705 (set_next_view_needed): New.
8706 (clear_next_view_needed): New.
8707 (maybe_output_next_view): New.
8708 (final_start_function): Rename to...
8709 (final_start_function_1): ... this. Take pointer to FIRST,
8710 add SEEN parameter. Emit param bindings in the initial view.
8711 (final_start_function): Reintroduce SEEN-less interface.
8712 (final): Rename to...
8713 (final_1): ... this. Take SEEN parameter. Output final pending
8714 next view at the end.
8715 (final): Reintroduce seen-less interface.
8716 (final_scan_insn): Output pending next view before switching
8717 sections or ending a block. Mark the next view as needed when
8718 outputting variable locations. Notify debug backend of section
8719 changes, and of location view changes.
8720 (rest_of_handle_final): Adjust.
8721 * toplev.c (process_options): Autodetect value for debug variable
8722 location views option. Warn on incompat5 without -gdwarf-5.
8723 * doc/invoke.texi (gvariable-location-views): New.
8724 (gvariable-location-views=incompat5): New.
8725 (gno-variable-location-views): New.
8727 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8729 PR tree-optimization/84136
8730 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8731 that the result of find_edge is non-NULL.
8733 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8736 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8737 storing integer register in SImode. Fix cost of 256 and 512
8738 byte aligned SSE register store.
8740 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8742 * config/i386/i386.c (ix86_multiplication_cost): Fix
8743 multiplication cost for TARGET_AVX512DQ.
8745 2018-02-08 Marek Polacek <polacek@redhat.com>
8747 PR tree-optimization/84238
8748 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8751 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8753 PR tree-optimization/84265
8754 * tree-vect-stmts.c (vectorizable_store): Don't treat
8755 VMAT_CONTIGUOUS accesses as grouped.
8756 (vectorizable_load): Likewise.
8758 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8760 PR tree-optimization/81635
8761 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8762 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8763 (test_round_for_mask): New functions.
8764 (wide_int_cc_tests): Call test_round_for_mask.
8765 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8766 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8767 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8768 range returned by get_range_info.
8770 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8773 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8774 * symtab.c: Include builtins.h
8775 (symtab_node::output_to_lto_symbol_table_p): Move here
8776 from lto-streamer-out.c:output_symbol_p.
8777 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8778 (output_symbol_p): Move all logic to symtab.c
8779 (produce_symtab): Update.
8781 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8783 * config/s390/s390-opts.h (enum indirect_branch): Define.
8784 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8785 (s390_indirect_branch_via_thunk)
8786 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8787 (enum s390_indirect_branch_type): Define.
8788 * config/s390/s390.c (struct s390_frame_layout, struct
8789 machine_function): Remove.
8790 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8791 (indirect_branch_table_label_no, indirect_branch_table_name):
8793 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8794 (enum s390_indirect_branch_option): Define.
8795 (s390_return_addr_from_memory): New function.
8796 (s390_handle_string_attribute): New function.
8797 (s390_attribute_table): Add new attribute handler.
8798 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8799 (s390_indirect_branch_via_thunk): New function.
8800 (s390_indirect_branch_via_inline_thunk): New function.
8801 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8802 sibling call optimization for non z10 compiles.
8803 (s390_emit_call): Force indirect branch target to be a single
8804 register. Add r1 clobber for non-z10 compiles.
8805 (s390_emit_epilogue): Emit return jump via return_use expander.
8806 (s390_reorg): Handle JUMP_INSNs as execute targets.
8807 (s390_option_override_internal): Perform validity checks for the
8808 new command line options.
8809 (s390_indirect_branch_attrvalue): New function.
8810 (s390_indirect_branch_settings): New function.
8811 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8812 (s390_output_indirect_thunk_function): New function.
8813 (s390_code_end): Implement target hook.
8814 (s390_case_values_threshold): Implement target hook.
8815 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8817 * config/s390/s390.h (struct s390_frame_layout)
8818 (struct machine_function): Move here from s390.c.
8819 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8820 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8821 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8822 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8823 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8824 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8825 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8826 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8827 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8828 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8829 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8830 (mnemonic attribute): Add values which aren't recognized
8832 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8833 pattern for branch conversion. Fix mnemonic attribute.
8834 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8835 indirect branch via thunk if requested.
8836 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8837 ("*indirect_jump"): Disable for branch conversion using out of
8839 ("indirect_jump_via_thunk<mode>_z10")
8840 ("indirect_jump_via_thunk<mode>")
8841 ("indirect_jump_via_inlinethunk<mode>_z10")
8842 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8843 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8844 ("casesi_jump_via_inlinethunk<mode>_z10")
8845 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8846 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8847 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8848 ("*indirect2_jump"): Disable for branch conversion.
8849 ("casesi_jump"): Turn into expander and expand patterns for branch
8851 ("return_use"): New expander.
8852 ("*return"): Emit return via thunk and rename it to ...
8853 ("*return<mode>"): ... this one.
8854 * config/s390/s390.opt: Add new options and and enum for the
8857 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8859 * lra-constraints.c (match_reload): Unconditionally use
8860 gen_lowpart_SUBREG, rather than selecting between that
8861 and equivalent gen_rtx_SUBREG code.
8863 2018-02-08 Richard Biener <rguenther@suse.de>
8865 PR tree-optimization/84233
8866 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8867 changed flag instead of boguously re-using phi_inserted.
8869 2018-02-08 Martin Jambor <mjambor@suse.cz>
8871 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8872 static local variables.
8874 2018-02-08 Richard Biener <rguenther@suse.de>
8876 PR tree-optimization/84278
8877 * tree-vect-stmts.c (vectorizable_store): When looking for
8878 smaller vector types to perform grouped strided loads/stores
8879 make sure the mode is supported by the target.
8880 (vectorizable_load): Likewise.
8882 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8884 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8885 Increase LDP/STP opportunities by adding adjacent callee-saves.
8887 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8889 PR rtl-optimization/84068
8890 PR rtl-optimization/83459
8891 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8893 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8895 PR tree-optimization/84224
8896 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8897 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8900 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8903 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8904 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8905 count, remove one USE.
8907 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8909 * doc/install.texi (Configuration): Document the
8910 --with-long-double-format={ibm,ieee} PowerPC configuration
8914 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8915 Convert from define_expand to be define_insn_and_split. Rework
8916 float/double/_Float128 conversions to QI/HI/SImode to work with
8917 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8918 conversions to QI/HImode types did a store and then a load to
8919 truncate the value. For conversions to VSX registers, don't split
8920 the insn, instead emit the code directly. Use the code iterator
8921 any_fix to combine signed and unsigned conversions.
8922 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8923 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8924 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8925 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8926 (fix_<mode>di2_hw): Likewise.
8927 (fixuns_<mode>di2_hw): Likewise.
8928 (fix_<mode>si2_hw): Likewise.
8929 (fixuns_<mode>si2_hw): Likewise.
8930 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8931 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8932 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8933 fix<uns>_trunc<SFDF:mode>si2_p8.
8934 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8936 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8937 (fix<uns>_<mode>_mem): Likewise.
8938 (fctiw<u>z_<mode>_mem): Likewise.
8939 (fix<uns>_<mode>_mem): Likewise.
8940 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8941 the register allocator from doing a direct move to the GPRs to do
8942 a store, and instead use the ISA 3.0 store byte/half-word from
8943 vector register instruction. For IEEE 128-bit floating point,
8944 also optimize stores of 32-bit ints.
8945 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8947 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8949 * genextract.c (push_pathstr_operand): New function to support
8951 (walk_rtx): Call push_pathstr_operand.
8952 (print_path): Support [a-zA-Z].
8954 2018-02-07 Richard Biener <rguenther@suse.de>
8956 PR tree-optimization/84037
8957 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8958 (cse_and_gimplify_to_preheader): Declare.
8959 (vect_get_place_in_interleaving_chain): Likewise.
8960 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8962 (_loop_vec_info::~_loop_vec_info): Delete it.
8963 (cse_and_gimplify_to_preheader): New function.
8964 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8965 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8966 (vectorizable_load): Likewise. For grouped stores always base
8967 the IV on the first element.
8968 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8969 condition before gimplifying.
8971 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8973 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8974 *DIV_EXPR and *MOD_EXPR.
8976 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8979 * config/i386/i386.c (ix86_option_override_internal): Mask out
8980 the CF_SET bit when checking -fcf-protection.
8982 2018-02-07 Tom de Vries <tom@codesourcery.com>
8985 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8988 2018-02-07 Richard Biener <rguenther@suse.de>
8990 PR tree-optimization/84204
8991 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8994 PR tree-optimization/84205
8995 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8996 special-case isl_ast_op_zdiv_r.
8998 PR tree-optimization/84223
8999 * graphite-scop-detection.c (gather_bbs::before_dom_children):
9000 Only add conditions from within the region.
9001 (gather_bbs::after_dom_children): Adjust.
9003 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
9006 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
9007 * config/avr/avr.md: Only post-reload split REG-REG moves if
9008 either register is GENERAL_REG_P.
9010 2018-02-07 Jakub Jelinek <jakub@redhat.com>
9012 PR tree-optimization/84235
9013 * tree-ssa-scopedtables.c
9014 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
9015 if the subtraction is performed in floating point type where NaNs are
9016 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
9017 build 1. Formatting fix.
9019 2018-02-06 Jakub Jelinek <jakub@redhat.com>
9022 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
9023 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
9024 and skip it regardless of bb boundaries. Use CALL_P macro,
9025 don't test INSN_P (insn) together with CALL_P or JUMP_P check
9026 unnecessarily, formatting fix.
9028 2018-02-06 Michael Collison <michael.collison@arm.com>
9030 * config/arm/thumb2.md:
9031 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
9032 (*thumb_mov_notscc): Ditto.
9034 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9037 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9038 unsigned_fix, not "s".
9040 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9042 * configure.ac (gcc_fn_eh_frame_ro): New function.
9043 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9044 correct .eh_frame permissions.
9045 * configure: Regenerate.
9047 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9049 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9052 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9054 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9055 Display warning message for -mno-speculate-indirect-jumps.
9057 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9059 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9061 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9063 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9065 PR tree-optimization/84225
9066 * tree-eh.c (find_trapping_overflow): Only call
9067 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9069 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9072 * config/i386/i386.c: Reimplement the check of possible options
9073 -mibt/-mshstk conbination. Change error messages.
9074 * doc/invoke.texi: Fix a typo: remove extra '='.
9076 2018-02-06 Marek Polacek <polacek@redhat.com>
9078 PR tree-optimization/84228
9079 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9081 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9084 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9085 emitted arch directives.
9086 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9087 __ARM_FEATURE_COPROC before changing architectures.
9089 2018-02-06 Richard Biener <rguenther@suse.de>
9091 * config/i386/i386.c (print_reg): Fix typo.
9092 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9094 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9096 * configure: Regenerate.
9098 2018-02-05 Martin Sebor <msebor@redhat.com>
9100 PR tree-optimization/83369
9101 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9104 2018-02-05 Martin Liska <mliska@suse.cz>
9106 * doc/invoke.texi: Cherry-pick upstream r323995.
9108 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9110 * ira.c (ira_init_register_move_cost): Adjust comment.
9112 2018-02-05 Martin Liska <mliska@suse.cz>
9114 PR gcov-profile/84137
9115 * doc/gcov.texi: Fix typo in documentation.
9117 2018-02-05 Martin Liska <mliska@suse.cz>
9119 PR gcov-profile/83879
9120 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9121 using dlopen functionality.
9123 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9125 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9126 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9127 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9128 _mm_maskz_range_round_ss): New intrinsics.
9129 (__builtin_ia32_rangesd128_round)
9130 (__builtin_ia32_rangess128_round): Remove.
9131 (__builtin_ia32_rangesd128_mask_round,
9132 __builtin_ia32_rangess128_mask_round): New builtins.
9133 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9134 __builtin_ia32_rangess128_round): Remove.
9135 (__builtin_ia32_rangesd128_mask_round,
9136 __builtin_ia32_rangess128_mask_round): New builtins.
9137 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9138 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9139 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9140 "<round_saeonly_constraint>")): Changed to ...
9141 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9142 "<round_saeonly_scalar_constraint>")): ... this.
9143 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9144 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9145 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9146 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9147 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9149 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9151 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9153 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9154 Remove all values except native, 8540 and 8548.
9156 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9158 * config/i386/i386.c (ix86_output_function_return): Pass
9159 INVALID_REGNUM, instead of -1, as invalid register number to
9160 indirect_thunk_name and output_indirect_thunk.
9162 2018-02-02 Julia Koval <julia.koval@intel.com>
9164 * config.gcc: Add -march=icelake.
9165 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9166 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9167 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9168 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9169 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9170 (processor_target_table): Add icelake.
9171 (ix86_option_override_internal): Handle new PTAs.
9172 (get_builtin_code_for_version): Handle icelake.
9173 (M_INTEL_COREI7_ICELAKE): New.
9174 (fold_builtin_cpu): Handle icelake.
9175 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9176 * doc/invoke.texi: Add -march=icelake.
9178 2018-02-02 Julia Koval <julia.koval@intel.com>
9180 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9181 to wide_int_bitmask.
9182 * wide-int-bitmask.h: New.
9184 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9187 * config/i386/i386.md: Replace Pmode with word_mode in
9188 builtin_setjmp_setup and builtin_longjmp to support x32.
9190 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9194 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9196 (rs6000_supported_cpu_names): New static variable.
9197 (linux_cpu_translation_table): Likewise.
9198 (elf_platform) <cpu>: Define new static variable and use it.
9199 Translate kernel AT_PLATFORM name to canonical name if needed.
9200 Error if platform name is unknown.
9202 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9205 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9207 2018-02-01 Jeff Law <law@redhat.com>
9210 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9211 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9212 the scratch if RELEASE_VIA_POP is false.
9213 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9214 If we have to save a temporary register, decrement SIZE appropriately.
9215 Pass new arguments to release_scratch_register_on_entry.
9216 (ix86_adjust_stack_and_probe): Likewise.
9217 (ix86_emit_probe_stack_range): Pass new arguments to
9218 release_scratch_register_on_entry.
9220 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9222 PR rtl-optimization/84157
9223 * combine.c (change_zero_ext): Use REG_P predicate in
9224 front of HARD_REGISTER_P predicate.
9226 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9228 * config/avr/avr.c (avr_option_override): Move disabling of
9229 -fdelete-null-pointer-checks to...
9230 * common/config/avr/avr-common.c (avr_option_optimization_table):
9233 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9235 PR tree-optimization/81635
9236 * tree-data-ref.c (split_constant_offset_1): For types that
9237 wrap on overflow, try to use range info to prove that wrapping
9240 2018-02-01 Renlin Li <renlin.li@arm.com>
9243 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9245 (aarch64_register_move_cost): Likewise.
9246 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9248 (REG_CLASS_NAMES): Likewise.
9249 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9250 TAILCALL_ADDR_REGS. Remove IP registers.
9251 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9253 2018-02-01 Richard Biener <rguenther@suse.de>
9255 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9256 for specifying RPO order and allow NULL for that.
9257 * domwalk.c (dom_walker::dom_walker): Likewise.
9258 (dom_walker::walk): Handle NULL RPO order.
9259 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9261 (rewrite_update_dom_walker): Likewise.
9262 (mark_def_dom_walker): Likewise.
9264 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9266 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9267 (aarch64_maybe_expand_sve_subreg_move): Declare.
9268 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9269 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9271 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9272 that are semantically a reverse operation.
9273 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9274 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9275 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9277 (aarch64_can_change_mode_class): For big-endian, forbid changes
9278 between two SVE modes if they have different element sizes.
9280 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9282 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9283 the TImode handling for big-endian targets.
9285 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9287 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9288 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9290 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9291 Remove BSWAP handing for big-endian targets and use the form of
9292 LD1RQ appropariate for the mode.
9294 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9296 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9297 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9300 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9303 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9304 check for operands that need to go through aarch64_sve_reload_be.
9306 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9308 PR tree-optimization/81661
9309 PR tree-optimization/84117
9310 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9311 * tree-eh.c: Include gimplify.h.
9312 (find_trapping_overflow, replace_trapping_overflow,
9313 rewrite_to_non_trapping_overflow): New functions.
9314 * tree-vect-loop.c: Include tree-eh.h.
9315 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9316 * tree-data-ref.c: Include tree-eh.h.
9317 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9319 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9321 PR rtl-optimization/84123
9322 * combine.c (change_zero_ext): Check if hard register satisfies
9323 can_change_dest_mode before calling gen_lowpart_SUBREG.
9325 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9328 * ira.c (ira_init_register_move_cost): Remove assert.
9330 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9332 PR rtl-optimization/84071
9333 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9334 * doc/tm.texi: Regenerate.
9336 2018-01-31 Richard Biener <rguenther@suse.de>
9338 PR tree-optimization/84132
9339 * tree-data-ref.c (analyze_miv_subscript): Properly
9340 check whether evolution_function_is_affine_multivariate_p
9341 before calling gcd_of_steps_may_divide_p.
9343 2018-01-31 Julia Koval <julia.koval@intel.com>
9346 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9347 * config/i386/i386.md (rdpid_rex64) New.
9348 (rdpid): Make 32bit only.
9350 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9353 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9354 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9356 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9359 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9361 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9363 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9365 PR rtl-optimization/84071
9366 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9367 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9369 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9371 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9372 (arc_attribute_table): Add 'aux' attribute.
9373 (arc_in_small_data_p): Consider aux like variables.
9374 (arc_is_aux_reg_p): New function.
9375 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9376 (arc_get_aux_arg): New function.
9377 (prepare_move_operands): Handle aux-register access.
9378 (arc_handle_aux_attribute): New function.
9379 * doc/extend.texi (ARC Variable attributes): Add subsection.
9381 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9383 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9384 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9385 (arc_attribute_table): Add 'uncached' attribute.
9386 (arc_print_operand): Print '.di' flag for uncached memory
9388 (arc_in_small_data_p): Do not consider for small data the uncached
9390 (arc_is_uncached_mem_p): New function.
9391 * config/arc/predicates.md (compact_store_memory_operand): Check
9392 for uncached memory accesses.
9393 (nonvol_nonimm_operand): Likewise.
9394 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9396 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9399 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9400 falign-loops=): Add Optimization flag.
9402 2018-01-30 Jeff Law <law@redhat.com>
9405 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9406 INT_REGISTERS_SAVED. Check it prior to calling
9407 get_scratch_register_on_entry.
9408 (ix86_adjust_stack_and_probe): Similarly.
9409 (ix86_emit_probe_stack_range): Similarly.
9410 (ix86_expand_prologue): Corresponding changes.
9412 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9415 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9416 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9418 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9421 * lra-constraints.c (curr_insn_transform): Process AND in the
9424 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9426 PR rtl-optimization/83986
9427 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9428 dependence against last_pending_memory_flush in addition to
9431 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9433 PR tree-optimization/81611
9434 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9437 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9440 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9443 2018-01-30 Richard Biener <rguenther@suse.de>
9444 Jakub Jelinek <jakub@redhat.com>
9446 PR tree-optimization/84111
9447 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9448 inner loops added during recursion, as they don't have up-to-date
9451 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9454 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9455 (can_inline_edge_by_limits_p): ... here.
9456 (can_early_inline_edge_p, check_callers,
9457 update_caller_keys, update_callee_keys, recursive_inlining,
9458 add_new_edges_to_heap, speculation_useful_p,
9459 inline_small_functions,
9460 inline_small_functions, flatten_function,
9461 inline_to_all_callers_1): Update.
9463 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9465 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9468 2018-01-30 Richard Biener <rguenther@suse.de>
9470 PR tree-optimization/83008
9471 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9472 invariant and constant vector uses in stmts when they need
9475 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9478 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9479 * configure: Regenerate.
9481 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9483 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9485 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9486 Use gen_rtx_REG rather than gen_lowpart.
9488 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9490 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9491 rather than 0 when creating partial subregs.
9493 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9495 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9498 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9501 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9502 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9503 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9504 flags. This restores the settings used before the 2017-07-24.
9505 Turning off pre increment/decrement/modify allows IVOPTS to
9506 optimize DF/SF loops where the index is an int.
9508 2018-01-29 Richard Biener <rguenther@suse.de>
9509 Kelvin Nilsen <kelvin@gcc.gnu.org>
9512 * tree-vect-stmts.c (vectorizable_call): Don't call
9513 targetm.vectorize_builtin_md_vectorized_function if callee is
9516 2018-01-22 Carl Love <cel@us.ibm.com>
9518 * doc/extend.tex: Fix typo in second arg in
9519 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9521 2018-01-29 Richard Biener <rguenther@suse.de>
9523 PR tree-optimization/84086
9524 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9525 (flush_ssaname_freelist): When SSA names were released reset
9526 the SCEV hash table.
9528 2018-01-29 Richard Biener <rguenther@suse.de>
9530 PR tree-optimization/84057
9531 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9532 removed paths when removing edges.
9534 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9536 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9537 -mfunction-return=@var{choice}.
9539 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9542 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9544 (layout::print_source_line): Likewise.
9545 (test_get_line_width_without_trailing_whitespace): Add test cases.
9547 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9550 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9553 2018-01-26 Jim Wilson <jimw@sifive.com>
9555 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9557 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9560 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9562 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9563 and CMP + SUB-immediate -> SUBS.
9565 2018-01-26 Martin Sebor <msebor@redhat.com>
9567 PR tree-optimization/83896
9568 * tree-ssa-strlen.c (get_string_len): Rename...
9569 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9570 Avoid assuming length is constant.
9571 (handle_char_store): Use HOST_WIDE_INT for string length.
9573 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9576 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9577 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9579 2018-01-26 Richard Biener <rguenther@suse.de>
9581 PR rtl-optimization/84003
9582 * dse.c (record_store): Only record redundant stores when
9583 the earlier store aliases at least all accesses the later one does.
9585 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9587 PR rtl-optimization/83985
9588 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9589 REG_CFA_RESTORE insns.
9590 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9591 REG_CFA_RESTORE insns here.
9594 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9595 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9597 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9599 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9600 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9601 (arc_init): Likewise.
9602 (arc_override_options): Likewise.
9603 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9605 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9607 * config/arc/arc.h (TARGET_DBNZ): Define.
9608 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9609 properly set the tune attribute.
9610 (dbnz): Use TARGET_DBNZ guard.
9611 * config/arc/arc.opt (mtune): Add core3 option.
9613 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9615 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9616 recognize new pic like addresses.
9617 (arc_delegitimize_address): Clean up.
9619 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9621 * config/arc/arc-arches.def: Option mrf16 valid for all
9623 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9624 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9625 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9626 * config/arc/arc-tables.opt: Regenerate.
9627 * config/arc/arc.c (arc_conditional_register_usage): Handle
9628 reduced register file case.
9629 (arc_file_start): Set must have build attributes.
9630 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9632 * config/arc/arc.opt (mrf16): Add new option.
9633 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9634 * config/arc/genmultilib.awk: Handle new mrf16 option.
9635 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9636 * config/arc/t-multilib: Regenerate.
9637 * doc/invoke.texi (ARC Options): Document mrf16 option.
9639 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9641 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9642 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9643 (arc_attribute_table): Add 'secure_call' attribute.
9644 (arc_print_operand): Print secure call operand.
9645 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9647 (arc_is_secure_call_p): New function. * config/arc/arc.md
9648 (call_i): Add support for sjli instruction.
9649 (call_value_i): Likewise.
9650 * config/arc/constraints.md (Csc): New constraint.
9652 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9653 John Eric Martin <John.Martin@emmicro-us.com>
9655 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9656 * config/arc/arc.c (_arc_jli_section): New struct.
9657 (arc_jli_section): New type.
9658 (rc_jli_sections): New static variable.
9659 (arc_handle_jli_attribute): New function.
9660 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9661 (arc_file_end): New function.
9662 (TARGET_ASM_FILE_END): Define.
9663 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9664 (arc_add_jli_section): New function.
9665 (jli_call_scan): Likewise.
9666 (arc_reorg): Call jli_call_scan.
9667 (arc_output_addsi): Remove 'S' from printing asm operand.
9668 (arc_is_jli_call_p): New function.
9669 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9671 (movhi_insn): Likewise.
9672 (movsi_insn): Likewise.
9673 (movsi_set_cc_insn): Likewise.
9674 (loadqi_update): Likewise.
9675 (load_zeroextendqisi_update): Likewise.
9676 (load_signextendqisi_update): Likewise.
9677 (loadhi_update): Likewise.
9678 (load_zeroextendhisi_update): Likewise.
9679 (load_signextendhisi_update): Likewise.
9680 (loadsi_update): Likewise.
9681 (loadsf_update): Likewise.
9682 (movsicc_insn): Likewise.
9683 (bset_insn): Likewise.
9684 (bxor_insn): Likewise.
9685 (bclr_insn): Likewise.
9686 (bmsk_insn): Likewise.
9687 (bicsi3_insn): Likewise.
9688 (cmpsi_cc_c_insn): Likewise.
9689 (movsi_ne): Likewise.
9690 (movsi_cond_exec): Likewise.
9691 (clrsbsi2): Likewise.
9702 (call_i): Remove 'S' asm letter, add jli instruction.
9703 (call_value_i): Likewise.
9704 * config/arc/arc.op (mjli-always): New option.
9705 * config/arc/constraints.md (Cji): New constraint.
9706 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9708 (subsf3_fpx): Likewise.
9709 (mulsf3_fpx): Likewise.
9710 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9712 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9714 * doc/invoke.texi (ARC): Document mjli-always option.
9716 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9718 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9719 avoid addition with 0 and use incw and decw where possible.
9721 2018-01-26 Richard Biener <rguenther@suse.de>
9723 PR tree-optimization/81082
9724 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9725 association if it requires casting to unsigned.
9726 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9727 from fold_plusminus_mult_expr to catch important cases late when
9728 range info is available.
9730 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9732 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9733 * configure.ac (hidden_linkonce): New test.
9734 * configure: Regenerate.
9735 * config.in: Regenerate.
9737 2018-01-26 Julia Koval <julia.koval@intel.com>
9739 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9740 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9741 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9742 _mm_mask_bitshuffle_epi64_mask): Fix type.
9743 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9744 USI_FTYPE_V4DI_V4DI_USI): Remove.
9745 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9746 __builtin_ia32_vpshufbitqmb256_mask,
9747 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9748 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9749 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9751 2018-01-26 Alan Modra <amodra@gmail.com>
9754 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9755 UNSPEC_VBPERMQ. Sort other unspecs.
9757 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9759 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9761 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9764 * predict.c (drop_profile): Do not push/pop cfun; update also
9766 (handle_missing_profiles): Fix logic looking for zero profiles.
9768 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9771 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9772 on functions with #pragma omp declare simd or functions with simd
9774 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9775 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9776 Remove trailing \n from warning_at calls.
9778 2018-01-25 Tom de Vries <tom@codesourcery.com>
9781 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9782 for neutered workers.
9784 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9787 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9788 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9790 2018-01-24 Jeff Law <law@redhat.com>
9793 * i386.c (get_probe_interval): Move to earlier point.
9794 (ix86_compute_frame_layout): If -fstack-clash-protection and
9795 the frame is larger than the probe interval, then use pushes
9796 to save registers rather than reg->mem moves.
9797 (ix86_expand_prologue): Remove conditional for int_registers_saved
9800 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9803 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9804 min/max for never referenced object.
9806 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9809 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9811 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9812 attributes from DECL_ATTRIBUTES (decl) without affecting
9813 DECL_ATTRIBUTES (current_function_decl).
9814 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9815 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9817 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9819 PR tree-optimization/83979
9820 * fold-const.c (fold_comparison): Use constant_boolean_node
9821 instead of boolean_{true,false}_node.
9823 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9825 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9828 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9830 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9831 Simplify the clause that sets the length attribute.
9832 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9833 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9834 clause that sets the length attribute.
9835 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9837 2018-01-24 Tom de Vries <tom@codesourcery.com>
9840 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9841 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9842 Add strict parameter.
9843 (prevent_branch_around_nothing): Insert dummy insn between branch to
9844 label and label with no ptx insn inbetween.
9845 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9847 2018-01-24 Tom de Vries <tom@codesourcery.com>
9850 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9851 for neutered threads in warp.
9852 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9854 2018-01-24 Richard Biener <rguenther@suse.de>
9856 PR tree-optimization/83176
9857 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9860 2018-01-24 Richard Biener <rguenther@suse.de>
9862 PR tree-optimization/82819
9863 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9864 code generating pluses that are no-ops in the target precision.
9866 2018-01-24 Richard Biener <rguenther@suse.de>
9869 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9871 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9873 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9874 to merge probabilities.
9875 * predict.c (probably_never_executed): Also mark as cold functions
9876 with global 0 profile and guessed local profile.
9877 * profile-count.c (profile_probability::combine_with_count): New
9879 * profile-count.h (profile_probability::operator*,
9880 profile_probability::operator*=, profile_probability::operator/,
9881 profile_probability::operator/=): Reduce precision to adjusted
9882 and set value to guessed on contradictory divisions.
9883 (profile_probability::combine_with_freq): Remove.
9884 (profile_probability::combine_wiht_count): Declare.
9885 (profile_count::force_nonzero):: Set to adjusted.
9886 (profile_count::probability_in):: Set quality to adjusted.
9887 * tree-ssa-tail-merge.c (replace_block_by): Use
9890 2018-01-23 Andrew Waterman <andrew@sifive.com>
9891 Jim Wilson <jimw@sifive.com>
9893 * config/riscv/riscv.c (riscv_stack_boundary): New.
9894 (riscv_option_override): Set riscv_stack_boundary. Handle
9895 riscv_preferred_stack_boundary_arg.
9896 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9897 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9898 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9899 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9900 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9901 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9903 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9906 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9907 of struct ix86_frame.
9908 (ix86_expand_epilogue): Likewise. Add a local variable for
9909 the reg_save_offset field in struct ix86_frame.
9911 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9913 PR tree-optimization/82604
9914 * tree-loop-distribution.c (enum partition_kind): New enum item
9915 PKIND_PARTIAL_MEMSET.
9916 (partition_builtin_p): Support above new enum item.
9917 (generate_code_for_partition): Ditto.
9918 (compute_access_range): Differentiate cases that equality can be
9919 proven at all loops, the innermost loops or no loops.
9920 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9921 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9922 (finalize_partitions, distribute_loop): Don't fuse partition of
9923 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9924 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9927 2018-01-23 Martin Liska <mliska@suse.cz>
9929 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9930 order to ignore the predictor.
9931 (PRED_POLYMORPHIC_CALL): Likewise.
9932 (PRED_RECURSIVE_CALL): Likewise.
9934 2018-01-23 Martin Liska <mliska@suse.cz>
9936 * tree-profile.c (tree_profiling): Print function header to
9937 aware reader which function we are working on.
9938 * value-prof.c (gimple_find_values_to_profile): Do not print
9939 not interesting value histograms.
9941 2018-01-23 Martin Liska <mliska@suse.cz>
9943 * profile-count.h (enum profile_quality): Add
9944 profile_uninitialized as the first value. Do not number values
9945 as they are zero based.
9946 (profile_count::verify): Update sanity check.
9947 (profile_probability::verify): Likewise.
9949 2018-01-23 Nathan Sidwell <nathan@acm.org>
9951 * doc/invoke.texi (ffor-scope): Deprecate.
9953 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9955 PR tree-optimization/83510
9956 * domwalk.c (set_all_edges_as_executable): New function.
9957 (dom_walker::dom_walker): Convert bool param
9958 "skip_unreachable_blocks" to enum reachability. Move setup of
9959 edge flags to set_all_edges_as_executable and only do it when
9960 reachability is REACHABLE_BLOCKS.
9961 * domwalk.h (enum dom_walker::reachability): New enum.
9962 (dom_walker::dom_walker): Convert bool param
9963 "skip_unreachable_blocks" to enum reachability.
9964 (set_all_edges_as_executable): New decl.
9965 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9966 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9968 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9969 but converting true to REACHABLE_BLOCKS.
9970 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9972 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9973 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9974 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9976 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9977 if check_all_array_refs will be called.
9979 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9981 * tree.c (selftest::test_location_wrappers): Add more test
9984 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9986 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9987 (selftest::test_bit_in_range): Likewise.
9989 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9992 * doc/sourcebuild.texi (vect_float): Say that the selector
9993 only describes the situation when -funsafe-math-optimizations is on.
9994 (vect_float_strict): Document.
9996 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9998 PR tree-optimization/83965
9999 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
10000 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
10001 instead of checking only for a reduction.
10002 (vect_recog_widen_sum_pattern): Likewise.
10004 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10006 * predict.c (probably_never_executed): Only use precise profile info.
10007 (compute_function_frequency): Skip after inlining hack since we now
10008 have quality checking.
10010 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
10012 * profile-count.h (profile_probability::very_unlikely,
10013 profile_probability::unlikely, profile_probability::even): Set
10014 precision to guessed.
10016 2018-01-23 Richard Biener <rguenther@suse.de>
10018 PR tree-optimization/83963
10019 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
10020 Properly terminate dominator walk when crossing the exit edge not
10021 when visiting its source block.
10023 2018-01-23 Jakub Jelinek <jakub@redhat.com>
10026 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
10027 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
10029 2018-01-22 Jakub Jelinek <jakub@redhat.com>
10031 PR tree-optimization/83957
10032 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
10033 semicolon after for body surrounded by braces.
10035 PR tree-optimization/83081
10036 * profile-count.h (profile_probability::split): New method.
10037 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10038 Use profile_probability::split.
10039 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10040 when splitting a single conditional jump into 2.
10042 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10044 PR tree-optimization/69452
10045 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10048 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10050 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10051 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10052 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10054 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10056 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10057 * config/rl78/rl78.md (movdi): New define_expand.
10058 * config/rl78/rl78.c (rl78_split_movdi): New function.
10060 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10063 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10065 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10066 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10067 128-bit to produce an UNSPEC move to get the double word with the
10068 signbit and then a shift directly to do signbit.
10069 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10070 implementation with a new version that just does either a direct
10071 move or a regular move. Move memory interface to separate insns.
10072 Move insns so they are next to the expander.
10073 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10074 with signbit move. Split big and little endian case.
10075 (signbit<mode>2_dm_mem_le): Likewise.
10076 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10077 (signbit<mode>2_dm2): Likewise.
10079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10081 * config/rl78/rl78.md (anddi3): New define_expand.
10083 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10085 * config/rl78/rl78.md (umindi3): New define_expand.
10087 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10089 * config/rl78/rl78.md (smindi3): New define_expand.
10091 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10093 * config/rl78/rl78.md (smaxdi3): New define_expand.
10095 2018-01-22 Carl Love <cel@us.ibm.com>
10097 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10098 LVX_V1TI): Add macro expansion.
10099 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10100 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10101 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10102 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10103 Change check to determine if the instruction is a byte reversing
10104 entry. Fix typo in comment.
10105 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10106 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10107 Add def_builtin calls for new builtins.
10108 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10109 Add define_insn expansion.
10111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10113 * config/rl78/rl78.md (umaxdi3): New define_expand.
10115 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10117 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10118 for non-QImode registers.
10120 2018-01-22 Richard Biener <rguenther@suse.de>
10122 PR tree-optimization/83963
10123 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10124 including the loop exit block.
10125 (scop_detection::merge_sese): Likewise.
10126 (scop_detection::add_scop): Do it here instead.
10128 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10130 * doc/sourcebuild.texi (arm_softfloat): Document.
10132 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10135 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10136 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10137 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10139 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10140 David Edelsohn <dje.gcc@gmail.com>
10143 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10144 Change "crset eq" to "crset 2".
10145 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10146 (*call_indirect_aix<mode>_nospec): Likewise.
10147 (*call_value_indirect_aix<mode>_nospec): Likewise.
10148 (*call_indirect_elfv2<mode>_nospec): Likewise.
10149 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10150 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10151 change assembly output from . to $.
10152 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10153 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10154 (*tablejump<mode>_internal1_nospec): Likewise.
10156 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10159 * config/sh/sh_optimize_sett_clrt.cc:
10160 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10162 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10164 PR tree-optimization/83940
10165 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10166 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10167 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10168 use it to pass back the definition type.
10169 (vect_check_store_rhs): Likewise rhs_dt_out.
10170 (vect_build_gather_load_calls): Add a mask_dt argument and use
10171 it instead of a call to vect_is_simple_use.
10172 (vectorizable_store): Update calls to vect_check_load_store_mask
10173 and vect_check_store_rhs. Use the dt returned by the latter instead
10174 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10175 instead of calls to vect_is_simple_use. Pass the scalar rather
10176 than the vector operand to vect_is_simple_use when handling
10177 second and subsequent copies of an rhs value.
10178 (vectorizable_load): Update calls to vect_check_load_store_mask
10179 and vect_build_gather_load_calls. Use the cached mask_dt and
10180 gs_info.offset_dt instead of calls to vect_is_simple_use.
10182 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10184 PR middle-end/83945
10185 * tree-emutls.c: Include gimplify.h.
10186 (lower_emutls_2): New function.
10187 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10188 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10189 it before further processing.
10192 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10193 UINTVAL (trueop1) instead of INTVAL (op1).
10195 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10199 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10200 INCOMING_FRAME_SP_OFFSET if not defined.
10201 (scan_trace): Add ENTRY argument. If true and
10202 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10203 emit a note to adjust the CFA offset.
10204 (create_cfi_notes): Adjust scan_trace callers.
10205 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10206 INCOMING_FRAME_SP_OFFSET in the CIE.
10207 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10208 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10210 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10211 * doc/tm.texi: Regenerated.
10213 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10215 PR rtl-optimization/83147
10216 * lra-constraints.c (remove_inheritance_pseudos): Use
10217 lra_substitute_pseudo_within_insn.
10219 2018-01-19 Tom de Vries <tom@codesourcery.com>
10220 Cesar Philippidis <cesar@codesourcery.com>
10223 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10225 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10228 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10229 spaces for function labels.
10231 2018-01-19 Martin Liska <mliska@suse.cz>
10233 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10234 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10235 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10236 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10237 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10238 (PRED_CONST_RETURN): Change from 69 to 65.
10239 (PRED_NULL_RETURN): Change from 91 to 71.
10240 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10241 (PRED_LOOP_GUARD): Change from 66 to 73.
10243 2018-01-19 Martin Liska <mliska@suse.cz>
10245 * predict.c (predict_insn_def): Add new assert.
10246 (struct branch_predictor): Change type to signed integer.
10247 (test_prediction_value_range): Amend test to cover
10248 PROB_UNINITIALIZED.
10249 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10250 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10251 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10252 (PRED_LOOP_IV_COMPARE): Likewise.
10253 * predict.h (PROB_UNINITIALIZED): Define new constant.
10255 2018-01-19 Martin Liska <mliska@suse.cz>
10257 * predict.c (dump_prediction): Add new format for
10258 analyze_brprob.py script which is enabled with -details
10260 * profile-count.h (precise_p): New function.
10262 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10264 PR tree-optimization/83922
10265 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10266 there are no statements that need masking.
10267 (vect_active_double_reduction_p): New function.
10268 (vect_analyze_loop_operations): Use it when handling phis that
10269 are not in the loop header.
10271 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10273 PR tree-optimization/83914
10274 * tree-vect-loop.c (vectorizable_induction): Don't convert
10275 init_expr or apply the peeling adjustment for inductions
10276 that are nested within the vectorized loop.
10278 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10280 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10283 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10287 * function.h (gimplify_parameters): Add gimple_seq * argument.
10288 * function.c: Include gimple.h and options.h.
10289 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10290 for the added local temporaries if needed.
10291 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10292 if there are any parameter cleanups, wrap whole body into a
10293 try/finally with the cleanups.
10295 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10298 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10299 Use GET_MODE_CLASS for scalar floating point.
10301 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10305 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10306 Fix call of call_cgraph_insertion_hooks.
10308 2018-01-18 Martin Sebor <msebor@redhat.com>
10310 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10312 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10315 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10318 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10321 * common.opt: (-ffile-prefix-map): New option.
10322 * opts.c (common_handle_option): Defer it.
10323 * opts-global.c (handle_common_deferred_options): Handle it.
10324 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10325 * file-prefix-map.h: New file.
10326 (remap_debug_filename, add_debug_prefix_map): ...here.
10327 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10328 * final.c (debug_prefix_map, add_debug_prefix_map
10329 remap_debug_filename): Move to...
10330 * file-prefix-map.c: New file.
10331 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10332 generalize, get rid of alloca(), use strrchr() instead of strchr().
10333 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10334 Implement in terms of add_prefix_map().
10335 (remap_macro_filename, remap_debug_filename): Implement in term of
10337 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10338 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10339 * dbxout.c: Include file-prefix-map.h.
10340 * varasm.c: Likewise.
10341 * vmsdbgout.c: Likewise.
10342 * xcoffout.c: Likewise.
10343 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10344 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10345 * doc/invoke.texi (-ffile-prefix-map): Document.
10346 (-fdebug-prefix-map): Update description.
10348 2018-01-18 Martin Liska <mliska@suse.cz>
10350 * config/i386/i386.c (indirect_thunk_name): Document that also
10352 (output_indirect_thunk): Document why both instructions
10353 (pause and lfence) are generated.
10355 2018-01-18 Richard Biener <rguenther@suse.de>
10357 PR tree-optimization/83887
10358 * graphite-scop-detection.c
10359 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10360 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10361 (scop_detection::merge_sese): Re-implement with a flood-fill
10362 algorithm that properly finds a SESE region if it exists.
10364 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10367 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10368 pointer_diff optimizations use view_convert instead of convert.
10370 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10372 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10373 Generate different code for -mno-speculate-indirect-jumps.
10374 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10375 (*call_indirect_aix<mode>): Disable for
10376 -mno-speculate-indirect-jumps.
10377 (*call_indirect_aix<mode>_nospec): New define_insn.
10378 (*call_value_indirect_aix<mode>): Disable for
10379 -mno-speculate-indirect-jumps.
10380 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10381 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10382 -mno-speculate-indirect-jumps.
10383 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10385 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10387 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10388 long double type, set the flags for noting the default long double
10389 type, even if we don't pass or return a long double type.
10391 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10394 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10397 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10399 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10400 support for merge[hl].
10401 (fold_mergehl_helper): New helper function.
10402 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10403 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10404 (altivec_vmrglw_direct): Add xxmrglw insn.
10406 2018-01-17 Andrew Waterman <andrew@sifive.com>
10408 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10409 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10411 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10414 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10415 call the lto_location_cache before reading the
10416 DECL_SOURCE_LOCATION of the types.
10418 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10419 Richard Sandiford <richard.sandiford@linaro.org>
10421 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10422 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10423 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10424 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10425 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10427 * config/aarch64/constraints.md (aarch64_movti_operand):
10429 * config/aarch64/predicates.md (Uti): Add new constraint.
10431 2018-01-17 Carl Love <cel@us.ibm.com>
10433 * config/rs6000/vsx.md (define_expand xl_len_r,
10434 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10435 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10437 (define_expand, define_insn): Move the shift left from the
10438 define_insn to the define_expand for lxvl and stxvl instructions.
10439 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10440 and XL_LEN_R definitions to PURE.
10442 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10444 * config/i386/i386.c (indirect_thunk_name): Declare regno
10445 as unsigned int. Compare regno with INVALID_REGNUM.
10446 (output_indirect_thunk): Ditto.
10447 (output_indirect_thunk_function): Ditto.
10448 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10449 in the call to output_indirect_thunk_function.
10451 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10453 PR middle-end/83884
10454 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10455 rather than the size of inner_type to determine the stack slot size
10456 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10458 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10461 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10464 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10466 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10467 endian Linux systems to optionally enable multilibs for selecting
10468 the long double type if the user configured an explicit type.
10469 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10470 have no long double multilibs if not defined.
10471 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10472 warn if the user used -mabi={ieee,ibm}longdouble and we built
10473 multilibs for long double.
10474 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10475 appropriate multilib option.
10476 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10478 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10479 for building long double multilibs.
10480 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10482 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10484 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10487 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10489 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10492 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10495 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10498 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10500 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10501 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10503 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10505 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10506 different rtl trees depending on TARGET_64BIT.
10507 (rs6000_gen_lvx): Likewise.
10509 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10511 * config/visium/visium.md (nop): Tweak comment.
10512 (hazard_nop): Likewise.
10514 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10516 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10517 -mspeculate-indirect-jumps.
10518 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10519 for -mno-speculate-indirect-jumps.
10520 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10521 (*call_value_indirect_elfv2<mode>): Disable for
10522 -mno-speculate-indirect-jumps.
10523 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10524 (indirect_jump): Emit different RTL for
10525 -mno-speculate-indirect-jumps.
10526 (*indirect_jump<mode>): Disable for
10527 -mno-speculate-indirect-jumps.
10528 (*indirect_jump<mode>_nospec): New define_insn.
10529 (tablejump): Emit different RTL for
10530 -mno-speculate-indirect-jumps.
10531 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10532 (tablejumpsi_nospec): New define_expand.
10533 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10534 (tablejumpdi_nospec): New define_expand.
10535 (*tablejump<mode>_internal1): Disable for
10536 -mno-speculate-indirect-jumps.
10537 (*tablejump<mode>_internal1_nospec): New define_insn.
10538 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10541 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10543 * caller-save.c (insert_save): Drop unnecessary parameter. All
10546 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10547 Richard Biener <rguenth@suse.de>
10550 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10551 return early, inline manually is_gimple_sizepos. Make sure if we
10552 call gimplify_expr we don't end up with a gimple constant.
10553 * tree.c (variably_modified_type_p): Don't return true for
10554 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10555 * gimplify.h (is_gimple_sizepos): Remove.
10557 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10559 PR tree-optimization/83857
10560 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10561 vectorizable_live_operation for pure SLP statements.
10562 (vectorizable_live_operation): Handle PHIs.
10564 2018-01-16 Richard Biener <rguenther@suse.de>
10566 PR tree-optimization/83867
10567 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10568 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10570 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10573 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10574 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10575 If off is not INTEGER_CST, issue a may not be aligned warning
10576 rather than isn't aligned. Use isn%'t rather than isn't.
10577 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10579 <case MULT_EXPR>: Improve the case when bottom and one of the
10580 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10581 operand, in that case check if the other operand is multiple of
10582 bottom divided by the INTEGER_CST operand.
10584 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10587 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10588 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10589 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10590 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10591 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10592 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10593 (pa_function_arg_size): New function.
10595 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10597 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10598 in a separate statement.
10600 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10602 PR tree-optimization/83847
10603 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10604 group gathers and scatters.
10606 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10608 PR rtl-optimization/86620
10609 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10611 PR rtl-optimization/83213
10612 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10613 to last if both are JUMP_INSNs.
10615 PR tree-optimization/83843
10616 * gimple-ssa-store-merging.c
10617 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10618 store_immediate_info for bswap/nop orig_stores.
10620 2018-01-15 Andrew Waterman <andrew@sifive.com>
10622 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10624 <UDIV>: Increase cost if !TARGET_DIV.
10626 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10628 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10629 (define_attr "cr_logical_3op"): New.
10630 (cceq_ior_compare): Adjust.
10631 (cceq_ior_compare_complement): Adjust.
10632 (*cceq_rev_compare): Adjust.
10633 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10634 (is_cracked_insn): Adjust.
10635 (insn_must_be_first_in_group): Adjust.
10636 * config/rs6000/40x.md: Adjust.
10637 * config/rs6000/440.md: Adjust.
10638 * config/rs6000/476.md: Adjust.
10639 * config/rs6000/601.md: Adjust.
10640 * config/rs6000/603.md: Adjust.
10641 * config/rs6000/6xx.md: Adjust.
10642 * config/rs6000/7450.md: Adjust.
10643 * config/rs6000/7xx.md: Adjust.
10644 * config/rs6000/8540.md: Adjust.
10645 * config/rs6000/cell.md: Adjust.
10646 * config/rs6000/e300c2c3.md: Adjust.
10647 * config/rs6000/e500mc.md: Adjust.
10648 * config/rs6000/e500mc64.md: Adjust.
10649 * config/rs6000/e5500.md: Adjust.
10650 * config/rs6000/e6500.md: Adjust.
10651 * config/rs6000/mpc.md: Adjust.
10652 * config/rs6000/power4.md: Adjust.
10653 * config/rs6000/power5.md: Adjust.
10654 * config/rs6000/power6.md: Adjust.
10655 * config/rs6000/power7.md: Adjust.
10656 * config/rs6000/power8.md: Adjust.
10657 * config/rs6000/power9.md: Adjust.
10658 * config/rs6000/rs64.md: Adjust.
10659 * config/rs6000/titan.md: Adjust.
10661 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10663 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10664 ix86_indirect_branch_register logic.
10666 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10668 * config/i386/constraints.md (Bs): Update
10669 ix86_indirect_branch_register check. Don't check
10670 ix86_indirect_branch_register with GOT_memory_operand.
10672 * config/i386/predicates.md (GOT_memory_operand): Don't check
10673 ix86_indirect_branch_register here.
10674 (GOT32_symbol_operand): Likewise.
10676 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10678 * config/i386/predicates.md (constant_call_address_operand):
10679 Rewrite ix86_indirect_branch_register logic.
10680 (sibcall_insn_operand): Likewise.
10682 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10684 * config/i386/constraints.md (Bs): Replace
10685 ix86_indirect_branch_thunk_register with
10686 ix86_indirect_branch_register.
10688 * config/i386/i386.md (indirect_jump): Likewise.
10689 (tablejump): Likewise.
10690 (*sibcall_memory): Likewise.
10691 (*sibcall_value_memory): Likewise.
10692 Peepholes of indirect call and jump via memory: Likewise.
10693 * config/i386/i386.opt: Likewise.
10694 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10695 (GOT_memory_operand): Likewise.
10696 (call_insn_operand): Likewise.
10697 (sibcall_insn_operand): Likewise.
10698 (GOT32_symbol_operand): Likewise.
10700 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10702 PR middle-end/83837
10703 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10704 type rather than type addr's type points to.
10705 (expand_omp_atomic_mutex): Likewise.
10706 (expand_omp_atomic): Likewise.
10708 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10711 * config/i386/i386.c (output_indirect_thunk_function): Use
10712 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10713 for __x86_return_thunk.
10715 2018-01-15 Richard Biener <rguenther@suse.de>
10717 PR middle-end/83850
10718 * expmed.c (extract_bit_field_1): Fix typo.
10720 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10723 * config/arm/iterators.md (VF): New mode iterator.
10724 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10725 Remove integer-related logic from pattern.
10726 (neon_vabd<mode>_3): Likewise.
10728 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10730 PR middle-end/82694
10731 * common.opt (fstrict-overflow): No longer an alias.
10732 (fwrapv-pointer): New option.
10733 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10734 also for pointer types based on flag_wrapv_pointer.
10735 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10736 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10737 opts->x_flag_wrapv got set.
10738 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10739 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10740 POINTER_TYPE_OVERFLOW_UNDEFINED.
10741 * match.pd: Likewise in address comparison pattern.
10742 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10744 2018-01-15 Richard Biener <rguenther@suse.de>
10747 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10748 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10749 Reset type names to their identifier if their TYPE_DECL doesn't
10750 have linkage (and thus is used for ODR and devirt).
10751 (save_debug_info_for_decl): Remove.
10752 (save_debug_info_for_type): Likewise.
10753 (add_tree_to_fld_list): Adjust.
10754 * tree-pretty-print.c (dump_generic_node): Make dumping of
10755 type names more robust.
10757 2018-01-15 Richard Biener <rguenther@suse.de>
10759 * BASE-VER: Bump to 8.0.1.
10761 2018-01-14 Martin Sebor <msebor@redhat.com>
10764 * builtins.c (check_access): Avoid warning when the no-warning bit
10767 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10769 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10770 * ira-color (allocno_hard_regs_compare): Likewise.
10772 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10775 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10776 Use .pushsection/.popsection.
10778 2018-01-14 Martin Sebor <msebor@redhat.com>
10781 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10783 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10785 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10786 entry from extra_headers.
10787 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10788 extra_headers, make the list bitwise identical to the i?86-*-* one.
10790 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10792 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10793 -mcmodel=large with -mindirect-branch=thunk,
10794 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10795 -mfunction-return=thunk-extern.
10796 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10797 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10798 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10800 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10802 * config/i386/i386.c (print_reg): Print the name of the full
10803 integer register without '%'.
10804 (ix86_print_operand): Handle 'V'.
10805 * doc/extend.texi: Document 'V' modifier.
10807 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10809 * config/i386/constraints.md (Bs): Disallow memory operand for
10810 -mindirect-branch-register.
10812 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10813 (GOT_memory_operand): Likewise.
10814 (call_insn_operand): Likewise.
10815 (sibcall_insn_operand): Likewise.
10816 (GOT32_symbol_operand): Likewise.
10817 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10818 for -mindirect-branch-register.
10819 (tablejump): Likewise.
10820 (*sibcall_memory): Likewise.
10821 (*sibcall_value_memory): Likewise.
10822 Disallow peepholes of indirect call and jump via memory for
10823 -mindirect-branch-register.
10824 (*call_pop): Replace m with Bw.
10825 (*call_value_pop): Likewise.
10826 (*sibcall_pop_memory): Replace m with Bs.
10827 * config/i386/i386.opt (mindirect-branch-register): New option.
10828 * doc/invoke.texi: Document -mindirect-branch-register option.
10830 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10832 * config/i386/i386-protos.h (ix86_output_function_return): New.
10833 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10834 set function_return_type.
10835 (indirect_thunk_name): Add ret_p to indicate thunk for function
10837 (output_indirect_thunk_function): Pass false to
10838 indirect_thunk_name.
10839 (ix86_output_indirect_branch_via_reg): Likewise.
10840 (ix86_output_indirect_branch_via_push): Likewise.
10841 (output_indirect_thunk_function): Create alias for function
10842 return thunk if regno < 0.
10843 (ix86_output_function_return): New function.
10844 (ix86_handle_fndecl_attribute): Handle function_return.
10845 (ix86_attribute_table): Add function_return.
10846 * config/i386/i386.h (machine_function): Add
10847 function_return_type.
10848 * config/i386/i386.md (simple_return_internal): Use
10849 ix86_output_function_return.
10850 (simple_return_internal_long): Likewise.
10851 * config/i386/i386.opt (mfunction-return=): New option.
10852 (indirect_branch): Mention -mfunction-return=.
10853 * doc/extend.texi: Document function_return function attribute.
10854 * doc/invoke.texi: Document -mfunction-return= option.
10856 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10858 * config/i386/i386-opts.h (indirect_branch): New.
10859 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10860 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10861 with local indirect jump when converting indirect call and jump.
10862 (ix86_set_indirect_branch_type): New.
10863 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10864 (indirectlabelno): New.
10865 (indirect_thunk_needed): Likewise.
10866 (indirect_thunk_bnd_needed): Likewise.
10867 (indirect_thunks_used): Likewise.
10868 (indirect_thunks_bnd_used): Likewise.
10869 (INDIRECT_LABEL): Likewise.
10870 (indirect_thunk_name): Likewise.
10871 (output_indirect_thunk): Likewise.
10872 (output_indirect_thunk_function): Likewise.
10873 (ix86_output_indirect_branch_via_reg): Likewise.
10874 (ix86_output_indirect_branch_via_push): Likewise.
10875 (ix86_output_indirect_branch): Likewise.
10876 (ix86_output_indirect_jmp): Likewise.
10877 (ix86_code_end): Call output_indirect_thunk_function if needed.
10878 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10880 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10881 (ix86_attribute_table): Add indirect_branch.
10882 * config/i386/i386.h (machine_function): Add indirect_branch_type
10883 and has_local_indirect_jump.
10884 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10886 (tablejump): Likewise.
10887 (*indirect_jump): Use ix86_output_indirect_jmp.
10888 (*tablejump_1): Likewise.
10889 (simple_return_indirect_internal): Likewise.
10890 * config/i386/i386.opt (mindirect-branch=): New option.
10891 (indirect_branch): New.
10894 (thunk-inline): Likewise.
10895 (thunk-extern): Likewise.
10896 * doc/extend.texi: Document indirect_branch function attribute.
10897 * doc/invoke.texi: Document -mindirect-branch= option.
10899 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10902 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10904 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10906 * ipa-inline.c (want_inline_small_function_p): Return false if
10907 inlining has already failed with CIF_FINAL_ERROR.
10908 (update_caller_keys): Call want_inline_small_function_p before
10910 (update_callee_keys): Likewise.
10912 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10914 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10916 (rs6000_quadword_masked_address_p): Likewise.
10917 (quad_aligned_load_p): Likewise.
10918 (quad_aligned_store_p): Likewise.
10919 (const_load_sequence_p): Add comment to describe the outer-most loop.
10920 (mimic_memory_attributes_and_flags): New function.
10921 (rs6000_gen_stvx): Likewise.
10922 (replace_swapped_aligned_store): Likewise.
10923 (rs6000_gen_lvx): Likewise.
10924 (replace_swapped_aligned_load): Likewise.
10925 (replace_swapped_load_constant): Capitalize argument name in
10926 comment describing this function.
10927 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10928 and stores that access quad-word aligned addresses and replace
10929 with stvx or lvx instructions when appropriate.
10930 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10931 New function prototype.
10932 (rs6000_quadword_masked_address_p): Likewise.
10933 (rs6000_gen_lvx): Likewise.
10934 (rs6000_gen_stvx): Likewise.
10935 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10936 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10937 when memory address is aligned.
10938 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10939 this split to select lvx instruction when memory address is aligned.
10940 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10941 instruction when memory address is aligned.
10942 (*vsx_le_perm_load_v16qi): Likewise.
10943 (four unnamed splitters): Modify to select the stvx instruction
10944 when memory is aligned.
10946 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10948 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10949 which appears in the queue multiple times.
10951 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10952 Alan Hayward <alan.hayward@arm.com>
10953 David Sherwood <david.sherwood@arm.com>
10955 * tree-vectorizer.h (vec_lower_bound): New structure.
10956 (_loop_vec_info): Add check_nonzero and lower_bounds.
10957 (LOOP_VINFO_CHECK_NONZERO): New macro.
10958 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10959 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10960 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10961 fields. Make seg_len the distance travelled, not including the
10963 (dr_direction_indicator): Declare.
10964 (dr_zero_step_indicator): Likewise.
10965 (dr_known_forward_stride_p): Likewise.
10966 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10968 (runtime_alias_check_p): Allow runtime alias checks with
10970 (operator ==): Compare access_size and align.
10971 (prune_runtime_alias_test_list): Rework for new distinction between
10972 the access_size and seg_len.
10973 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10975 (get_segment_min_max): New function.
10976 (create_intersect_range_checks): Use it.
10977 (dr_step_indicator): New function.
10978 (dr_direction_indicator): Likewise.
10979 (dr_zero_step_indicator): Likewise.
10980 (dr_known_forward_stride_p): Likewise.
10981 * tree-loop-distribution.c (data_ref_segment_size): Return
10982 DR_STEP * (niters - 1).
10983 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10985 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10986 (vect_preserves_scalar_order_p): New function, split out from...
10987 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10988 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10989 (vect_vfa_access_size): New function.
10990 (vect_vfa_align): Likewise.
10991 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10992 (dump_lower_bound): New function.
10993 (vect_check_lower_bound): Likewise.
10994 (vect_small_gap_p): Likewise.
10995 (vectorizable_with_step_bound_p): Likewise.
10996 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10997 depencies if the vectorization factor is 1. Convert the checks
10998 for nonzero steps into checks on the bounds of DR_STEP. Try using
10999 a bunds check for variable steps if the minimum required step is
11000 relatively small. Update calls to the dr_with_seg_len
11001 constructor and to vect_compile_time_alias.
11002 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
11004 (vect_loop_versioning): Call it.
11005 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
11007 (vect_estimate_min_profitable_iters): Account for any bounds checks.
11009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11010 Alan Hayward <alan.hayward@arm.com>
11011 David Sherwood <david.sherwood@arm.com>
11013 * doc/sourcebuild.texi (vect_scatter_store): Document.
11014 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
11016 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
11018 * genopinit.c (main): Add supports_vec_scatter_store and
11019 supports_vec_scatter_store_cached to target_optabs.
11020 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
11021 IFN_MASK_SCATTER_STORE.
11022 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
11024 * internal-fn.h (internal_store_fn_p): Declare.
11025 (internal_fn_stored_value_index): Likewise.
11026 * internal-fn.c (scatter_store_direct): New macro.
11027 (expand_scatter_store_optab_fn): New function.
11028 (direct_scatter_store_optab_supported_p): New macro.
11029 (internal_store_fn_p): New function.
11030 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
11031 IFN_MASK_SCATTER_STORE.
11032 (internal_fn_mask_index): Likewise.
11033 (internal_fn_stored_value_index): New function.
11034 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
11035 for scatter stores.
11036 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11037 * optabs-query.c (supports_vec_scatter_store_p): New function.
11038 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11039 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11040 true for scatter stores.
11041 (vect_gather_scatter_fn_p): Handle scatter stores too.
11042 (vect_check_gather_scatter): Consider using scatter stores if
11043 supports_vec_scatter_store_p.
11044 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11045 scatter stores too.
11046 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11047 internal_fn_stored_value_index.
11048 (check_load_store_masking): Handle scatter stores too.
11049 (vect_get_store_rhs): Make public.
11050 (vectorizable_call): Use internal_store_fn_p.
11051 (vectorizable_store): Handle scatter store internal functions.
11052 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11053 when deciding whether the end of the group has been reached.
11054 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11055 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11056 (mask_scatter_store<mode>): New insns.
11058 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11059 Alan Hayward <alan.hayward@arm.com>
11060 David Sherwood <david.sherwood@arm.com>
11062 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11063 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11064 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11066 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11067 Use vect_truncate_gather_scatter_offset if we can't treat the
11068 operation as a normal gather load or scatter store.
11069 (get_group_load_store_type): Take the gather_scatter_info
11070 as argument. Try using a gather load or scatter store for
11071 single-element groups.
11072 (get_load_store_type): Update calls to get_group_load_store_type
11073 and vect_use_strided_gather_scatters_p.
11075 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11076 Alan Hayward <alan.hayward@arm.com>
11077 David Sherwood <david.sherwood@arm.com>
11079 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11080 optional tree argument.
11081 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11083 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11084 but continue to use the current value as a fallback.
11085 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11086 to compare the updates.
11087 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11088 (get_load_store_type): Use it when handling a strided access.
11089 (vect_get_strided_load_store_ops): New function.
11090 (vect_get_data_ptr_increment): Likewise.
11091 (vectorizable_load): Handle strided gather loads. Always pass
11092 a step to vect_create_data_ref_ptr and bump_vector_ptr.
11094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11095 Alan Hayward <alan.hayward@arm.com>
11096 David Sherwood <david.sherwood@arm.com>
11098 * doc/md.texi (gather_load@var{m}): Document.
11099 (mask_gather_load@var{m}): Likewise.
11100 * genopinit.c (main): Add supports_vec_gather_load and
11101 supports_vec_gather_load_cached to target_optabs.
11102 * optabs-tree.c (init_tree_optimization_optabs): Use
11103 ggc_cleared_alloc to allocate target_optabs.
11104 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11105 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11107 * internal-fn.h (internal_load_fn_p): Declare.
11108 (internal_gather_scatter_fn_p): Likewise.
11109 (internal_fn_mask_index): Likewise.
11110 (internal_gather_scatter_fn_supported_p): Likewise.
11111 * internal-fn.c (gather_load_direct): New macro.
11112 (expand_gather_load_optab_fn): New function.
11113 (direct_gather_load_optab_supported_p): New macro.
11114 (direct_internal_fn_optab): New function.
11115 (internal_load_fn_p): Likewise.
11116 (internal_gather_scatter_fn_p): Likewise.
11117 (internal_fn_mask_index): Likewise.
11118 (internal_gather_scatter_fn_supported_p): Likewise.
11119 * optabs-query.c (supports_at_least_one_mode_p): New function.
11120 (supports_vec_gather_load_p): Likewise.
11121 * optabs-query.h (supports_vec_gather_load_p): Declare.
11122 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11123 and memory_type field.
11124 (NUM_PATTERNS): Bump to 15.
11125 * tree-vect-data-refs.c: Include internal-fn.h.
11126 (vect_gather_scatter_fn_p): New function.
11127 (vect_describe_gather_scatter_call): Likewise.
11128 (vect_check_gather_scatter): Try using internal functions for
11129 gather loads. Recognize existing calls to a gather load function.
11130 (vect_analyze_data_refs): Consider using gather loads if
11131 supports_vec_gather_load_p.
11132 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11133 (vect_get_gather_scatter_offset_type): Likewise.
11134 (vect_convert_mask_for_vectype): Likewise.
11135 (vect_add_conversion_to_patterm): Likewise.
11136 (vect_try_gather_scatter_pattern): Likewise.
11137 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11138 (vect_vect_recog_func_ptrs): Add it.
11139 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11140 internal_fn_mask_index and internal_gather_scatter_fn_p.
11141 (check_load_store_masking): Take the gather_scatter_info as an
11142 argument and handle gather loads.
11143 (vect_get_gather_scatter_ops): New function.
11144 (vectorizable_call): Check internal_load_fn_p.
11145 (vectorizable_load): Likewise. Handle gather load internal
11147 (vectorizable_store): Update call to check_load_store_masking.
11148 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11149 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11150 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11151 (aarch64_gather_scale_operand_d): New predicates.
11152 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11153 (mask_gather_load<mode>): New insns.
11155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11156 Alan Hayward <alan.hayward@arm.com>
11157 David Sherwood <david.sherwood@arm.com>
11159 * optabs.def (fold_left_plus_optab): New optab.
11160 * doc/md.texi (fold_left_plus_@var{m}): Document.
11161 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11162 * internal-fn.c (fold_left_direct): Define.
11163 (expand_fold_left_optab_fn): Likewise.
11164 (direct_fold_left_optab_supported_p): Likewise.
11165 * fold-const-call.c (fold_const_fold_left): New function.
11166 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11167 * tree-parloops.c (valid_reduction_p): New function.
11168 (gather_scalar_reductions): Use it.
11169 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11170 (vect_finish_replace_stmt): Declare.
11171 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11172 (needs_fold_left_reduction_p): New function, split out from...
11173 (vect_is_simple_reduction): ...here. Accept reductions that
11174 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11175 (vect_force_simple_reduction): Also store the reduction type in
11176 the assignment's STMT_VINFO_REDUC_TYPE.
11177 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11178 (merge_with_identity): New function.
11179 (vect_expand_fold_left): Likewise.
11180 (vectorize_fold_left_reduction): Likewise.
11181 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11182 scalar phi in place for it. Check for target support and reject
11183 cases that would reassociate the operation. Defer the transform
11184 phase to vectorize_fold_left_reduction.
11185 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11186 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11187 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11191 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11192 call to ifc_temp_var.
11194 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11195 Alan Hayward <alan.hayward@arm.com>
11196 David Sherwood <david.sherwood@arm.com>
11198 * target.def (legitimize_address_displacement): Take the original
11199 offset as a poly_int.
11200 * targhooks.h (default_legitimize_address_displacement): Update
11202 * targhooks.c (default_legitimize_address_displacement): Likewise.
11203 * doc/tm.texi: Regenerate.
11204 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11205 as an argument, moving assert of ad->disp == ad->disp_term to...
11206 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11207 Try calling targetm.legitimize_address_displacement before expanding
11208 the address rather than afterwards, and adjust for the new interface.
11209 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11210 Match the new hook interface. Handle SVE addresses.
11211 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11212 new hook interface.
11214 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11216 * Makefile.in (OBJS): Add early-remat.o.
11217 * target.def (select_early_remat_modes): New hook.
11218 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11219 * doc/tm.texi: Regenerate.
11220 * targhooks.h (default_select_early_remat_modes): Declare.
11221 * targhooks.c (default_select_early_remat_modes): New function.
11222 * timevar.def (TV_EARLY_REMAT): New timevar.
11223 * passes.def (pass_early_remat): New pass.
11224 * tree-pass.h (make_pass_early_remat): Declare.
11225 * early-remat.c: New file.
11226 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11228 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
11230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11231 Alan Hayward <alan.hayward@arm.com>
11232 David Sherwood <david.sherwood@arm.com>
11234 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11235 vfm1 with a bound_epilog parameter.
11236 (vect_do_peeling): Update calls accordingly, and move the prologue
11237 call earlier in the function. Treat the base bound_epilog as 0 for
11238 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11239 this base when peeling for gaps.
11240 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11241 with fully-masked loops.
11242 (vect_estimate_min_profitable_iters): Handle the single peeled
11243 iteration in that case.
11245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11246 Alan Hayward <alan.hayward@arm.com>
11247 David Sherwood <david.sherwood@arm.com>
11249 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11250 single-element interleaving even if the size is not a power of 2.
11251 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11252 accesses for single-element interleaving if the group size is
11255 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11256 Alan Hayward <alan.hayward@arm.com>
11257 David Sherwood <david.sherwood@arm.com>
11259 * doc/md.texi (fold_extract_last_@var{m}): Document.
11260 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11261 * optabs.def (fold_extract_last_optab): New optab.
11262 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11263 * internal-fn.c (fold_extract_direct): New macro.
11264 (expand_fold_extract_optab_fn): Likewise.
11265 (direct_fold_extract_optab_supported_p): Likewise.
11266 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11267 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11268 EXTRACT_LAST_REDUCTION.
11269 (get_initial_def_for_reduction): Do not create an initial vector
11270 for EXTRACT_LAST_REDUCTION reductions.
11271 (vectorizable_reduction): Leave the scalar phi in place for
11272 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11273 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11274 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11275 transform phase to vectorizable_condition.
11276 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11278 (vect_finish_stmt_generation): ...here.
11279 (vect_finish_replace_stmt): New function.
11280 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11281 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11283 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11286 Alan Hayward <alan.hayward@arm.com>
11287 David Sherwood <david.sherwood@arm.com>
11289 * doc/md.texi (extract_last_@var{m}): Document.
11290 * optabs.def (extract_last_optab): New optab.
11291 * internal-fn.def (EXTRACT_LAST): New internal function.
11292 * internal-fn.c (cond_unary_direct): New macro.
11293 (expand_cond_unary_optab_fn): Likewise.
11294 (direct_cond_unary_optab_supported_p): Likewise.
11295 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11296 loops using EXTRACT_LAST.
11297 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11298 (extract_last_<mode>): ...this optab.
11299 (vec_extract<mode><Vel>): Update accordingly.
11301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11302 Alan Hayward <alan.hayward@arm.com>
11303 David Sherwood <david.sherwood@arm.com>
11305 * target.def (empty_mask_is_expensive): New hook.
11306 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11307 * doc/tm.texi: Regenerate.
11308 * targhooks.h (default_empty_mask_is_expensive): Declare.
11309 * targhooks.c (default_empty_mask_is_expensive): New function.
11310 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11311 if the target says that empty masks are expensive.
11312 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11314 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11317 Alan Hayward <alan.hayward@arm.com>
11318 David Sherwood <david.sherwood@arm.com>
11320 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11321 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11322 (vect_use_loop_mask_for_alignment_p): New function.
11323 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11324 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11325 niters_skip argument. Make sure that the first niters_skip elements
11326 of the first iteration are inactive.
11327 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11328 Update call to vect_set_loop_masks_directly.
11329 (get_misalign_in_elems): New function, split out from...
11330 (vect_gen_prolog_loop_niters): ...here.
11331 (vect_update_init_of_dr): Take a code argument that specifies whether
11332 the adjustment should be added or subtracted.
11333 (vect_update_init_of_drs): Likewise.
11334 (vect_prepare_for_masked_peels): New function.
11335 (vect_do_peeling): Skip prologue peeling if we're using a mask
11336 instead. Update call to vect_update_inits_of_drs.
11337 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11339 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11340 alignment. Do not include the number of peeled iterations in
11341 the minimum threshold in that case.
11342 (vectorizable_induction): Adjust the start value down by
11343 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11344 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11345 Take the number of skipped iterations into account when calculating
11347 * tree-vect-stmts.c (vect_gen_while_not): New function.
11349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11350 Alan Hayward <alan.hayward@arm.com>
11351 David Sherwood <david.sherwood@arm.com>
11353 * doc/sourcebuild.texi (vect_fully_masked): Document.
11354 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11355 default value to 0.
11356 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11358 (vect_analyze_loop_2): ...here. Don't check the vectorization
11359 factor against the number of loop iterations if the loop is
11362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11363 Alan Hayward <alan.hayward@arm.com>
11364 David Sherwood <david.sherwood@arm.com>
11366 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11367 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11368 (dump_groups): Update accordingly.
11369 (iv_use::mem_type): New member variable.
11370 (address_p): New function.
11371 (record_use): Add a mem_type argument and initialize the new
11373 (record_group_use): Add a mem_type argument. Use address_p.
11374 Remove obsolete null checks of base_object. Update call to record_use.
11375 (find_interesting_uses_op): Update call to record_group_use.
11376 (find_interesting_uses_cond): Likewise.
11377 (find_interesting_uses_address): Likewise.
11378 (get_mem_type_for_internal_fn): New function.
11379 (find_address_like_use): Likewise.
11380 (find_interesting_uses_stmt): Try find_address_like_use before
11381 calling find_interesting_uses_op.
11382 (addr_offset_valid_p): Use the iv mem_type field as the type
11383 of the addressed memory.
11384 (add_autoinc_candidates): Likewise.
11385 (get_address_cost): Likewise.
11386 (split_small_address_groups_p): Use address_p.
11387 (split_address_groups): Likewise.
11388 (add_iv_candidate_for_use): Likewise.
11389 (autoinc_possible_for_pair): Likewise.
11390 (rewrite_groups): Likewise.
11391 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11392 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11393 (get_alias_ptr_type_for_ptr_address): New function.
11394 (rewrite_use_address): Rewrite address uses in calls that were
11395 identified by find_address_like_use.
11397 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11398 Alan Hayward <alan.hayward@arm.com>
11399 David Sherwood <david.sherwood@arm.com>
11401 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11403 * gimple-expr.h (is_gimple_addressable: Likewise.
11404 * gimple-expr.c (is_gimple_address): Likewise.
11405 * internal-fn.c (expand_call_mem_ref): New function.
11406 (expand_mask_load_optab_fn): Use it.
11407 (expand_mask_store_optab_fn): Likewise.
11409 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11410 Alan Hayward <alan.hayward@arm.com>
11411 David Sherwood <david.sherwood@arm.com>
11413 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11414 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11415 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11416 (cond_umax@var{mode}): Document.
11417 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11418 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11419 (cond_umin_optab, cond_umax_optab): New optabs.
11420 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11421 (COND_IOR, COND_XOR): New internal functions.
11422 * internal-fn.h (get_conditional_internal_fn): Declare.
11423 * internal-fn.c (cond_binary_direct): New macro.
11424 (expand_cond_binary_optab_fn): Likewise.
11425 (direct_cond_binary_optab_supported_p): Likewise.
11426 (get_conditional_internal_fn): New function.
11427 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11428 Cope with reduction statements that are vectorized as calls rather
11430 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11431 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11432 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11433 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11434 (UNSPEC_COND_EOR): New unspecs.
11435 (optab): Add mappings for them.
11436 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11437 (sve_int_op, sve_fp_op): New int attributes.
11439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11440 Alan Hayward <alan.hayward@arm.com>
11441 David Sherwood <david.sherwood@arm.com>
11443 * optabs.def (while_ult_optab): New optab.
11444 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11445 * internal-fn.def (WHILE_ULT): New internal function.
11446 * internal-fn.h (direct_internal_fn_supported_p): New override
11447 that takes two types as argument.
11448 * internal-fn.c (while_direct): New macro.
11449 (expand_while_optab_fn): New function.
11450 (convert_optab_supported_p): Likewise.
11451 (direct_while_optab_supported_p): New macro.
11452 * wide-int.h (wi::udiv_ceil): New function.
11453 * tree-vectorizer.h (rgroup_masks): New structure.
11454 (vec_loop_masks): New typedef.
11455 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11456 and fully_masked_p.
11457 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11458 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11459 (vect_max_vf): New function.
11460 (slpeel_make_loop_iterate_ntimes): Delete.
11461 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11462 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11463 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11464 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11465 internal-fn.h, stor-layout.h and optabs-query.h.
11466 (vect_set_loop_mask): New function.
11467 (add_preheader_seq): Likewise.
11468 (add_header_seq): Likewise.
11469 (interleave_supported_p): Likewise.
11470 (vect_maybe_permute_loop_masks): Likewise.
11471 (vect_set_loop_masks_directly): Likewise.
11472 (vect_set_loop_condition_masked): Likewise.
11473 (vect_set_loop_condition_unmasked): New function, split out from
11474 slpeel_make_loop_iterate_ntimes.
11475 (slpeel_make_loop_iterate_ntimes): Rename to..
11476 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11477 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11478 (vect_do_peeling): Update call accordingly.
11479 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11481 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11482 mask_compare_type, can_fully_mask_p and fully_masked_p.
11483 (release_vec_loop_masks): New function.
11484 (_loop_vec_info): Use it to free the loop masks.
11485 (can_produce_all_loop_masks_p): New function.
11486 (vect_get_max_nscalars_per_iter): Likewise.
11487 (vect_verify_full_masking): Likewise.
11488 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11489 retries, and free the mask rgroups before retrying. Check loop-wide
11490 reasons for disallowing fully-masked loops. Make the final decision
11491 about whether use a fully-masked loop or not.
11492 (vect_estimate_min_profitable_iters): Do not assume that peeling
11493 for the number of iterations will be needed for fully-masked loops.
11494 (vectorizable_reduction): Disable fully-masked loops.
11495 (vectorizable_live_operation): Likewise.
11496 (vect_halve_mask_nunits): New function.
11497 (vect_double_mask_nunits): Likewise.
11498 (vect_record_loop_mask): Likewise.
11499 (vect_get_loop_mask): Likewise.
11500 (vect_transform_loop): Handle the case in which the final loop
11501 iteration might handle a partial vector. Call vect_set_loop_condition
11502 instead of slpeel_make_loop_iterate_ntimes.
11503 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11504 (check_load_store_masking): New function.
11505 (prepare_load_store_mask): Likewise.
11506 (vectorizable_store): Handle fully-masked loops.
11507 (vectorizable_load): Likewise.
11508 (supportable_widening_operation): Use vect_halve_mask_nunits for
11510 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11511 (vect_gen_while): New function.
11512 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11513 (aarch64_uqdec<mode>): New insn.
11515 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11516 Alan Hayward <alan.hayward@arm.com>
11517 David Sherwood <david.sherwood@arm.com>
11519 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11520 (reduc_xor_scal_optab): New optabs.
11521 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11522 (reduc_xor_scal_@var{m}): Document.
11523 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11524 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11525 internal functions.
11526 * fold-const-call.c (fold_const_call): Handle them.
11527 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11528 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11529 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11530 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11531 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11532 (UNSPEC_XORV): New unspecs.
11533 (optab): Add entries for them.
11534 (BITWISEV): New int iterator.
11535 (bit_reduc_op): New int attributes.
11537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11538 Alan Hayward <alan.hayward@arm.com>
11539 David Sherwood <david.sherwood@arm.com>
11541 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11542 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11543 * optabs.def (vec_shl_insert_optab): New optab.
11544 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11545 (duplicate_and_interleave): Likewise.
11546 * tree-vect-loop.c: Include internal-fn.h.
11547 (neutral_op_for_slp_reduction): New function, split out from
11548 get_initial_defs_for_reduction.
11549 (get_initial_def_for_reduction): Handle option 2 for variable-length
11550 vectors by loading the neutral value into a vector and then shifting
11551 the initial value into element 0.
11552 (get_initial_defs_for_reduction): Replace the code argument with
11553 the neutral value calculated by neutral_op_for_slp_reduction.
11554 Use gimple_build_vector for constant-length vectors.
11555 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11556 but the first group_size elements have a neutral value.
11557 Use duplicate_and_interleave otherwise.
11558 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11559 Update call to get_initial_defs_for_reduction. Handle SLP
11560 reductions for variable-length vectors by creating one vector
11561 result for each scalar result, with the elements associated
11562 with other scalar results stubbed out with the neutral value.
11563 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11564 Require IFN_VEC_SHL_INSERT for double reductions on
11565 variable-length vectors, or SLP reductions that have
11566 a neutral value. Require can_duplicate_and_interleave_p
11567 support for variable-length unchained SLP reductions if there
11568 is no neutral value, such as for MIN/MAX reductions. Also require
11569 the number of vector elements to be a multiple of the number of
11570 SLP statements when doing variable-length unchained SLP reductions.
11571 Update call to vect_create_epilog_for_reduction.
11572 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11573 and remove initial values.
11574 (duplicate_and_interleave): Make public.
11575 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11576 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11579 Alan Hayward <alan.hayward@arm.com>
11580 David Sherwood <david.sherwood@arm.com>
11582 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11583 (can_duplicate_and_interleave_p): New function.
11584 (vect_get_and_check_slp_defs): Take the vector of statements
11585 rather than just the current one. Remove excess parentheses.
11586 Restriction rejectinon of vect_constant_def and vect_external_def
11587 for variable-length vectors to boolean types, or types for which
11588 can_duplicate_and_interleave_p is false.
11589 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11590 (duplicate_and_interleave): New function.
11591 (vect_get_constant_vectors): Use gimple_build_vector for
11592 constant-length vectors and suitable variable-length constant
11593 vectors. Use duplicate_and_interleave for other variable-length
11594 vectors. Don't defer the update when inserting new statements.
11596 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11597 Alan Hayward <alan.hayward@arm.com>
11598 David Sherwood <david.sherwood@arm.com>
11600 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11601 min_profitable_iters doesn't go negative.
11603 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11604 Alan Hayward <alan.hayward@arm.com>
11605 David Sherwood <david.sherwood@arm.com>
11607 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11608 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11609 * optabs.def (vec_mask_load_lanes_optab): New optab.
11610 (vec_mask_store_lanes_optab): Likewise.
11611 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11612 (MASK_STORE_LANES): Likewise.
11613 * internal-fn.c (mask_load_lanes_direct): New macro.
11614 (mask_store_lanes_direct): Likewise.
11615 (expand_mask_load_optab_fn): Handle masked operations.
11616 (expand_mask_load_lanes_optab_fn): New macro.
11617 (expand_mask_store_optab_fn): Handle masked operations.
11618 (expand_mask_store_lanes_optab_fn): New macro.
11619 (direct_mask_load_lanes_optab_supported_p): Likewise.
11620 (direct_mask_store_lanes_optab_supported_p): Likewise.
11621 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11623 (vect_load_lanes_supported): Likewise.
11624 * tree-vect-data-refs.c (strip_conversion): New function.
11625 (can_group_stmts_p): Likewise.
11626 (vect_analyze_data_ref_accesses): Use it instead of checking
11627 for a pair of assignments.
11628 (vect_store_lanes_supported): Take a masked_p parameter.
11629 (vect_load_lanes_supported): Likewise.
11630 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11631 vect_store_lanes_supported and vect_load_lanes_supported.
11632 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11633 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11634 parameter. Don't allow gaps for masked accesses.
11635 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11636 and vect_load_lanes_supported.
11637 (get_load_store_type): Take a masked_p parameter and update
11638 call to get_group_load_store_type.
11639 (vectorizable_store): Update call to get_load_store_type.
11640 Handle IFN_MASK_STORE_LANES.
11641 (vectorizable_load): Update call to get_load_store_type.
11642 Handle IFN_MASK_LOAD_LANES.
11644 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11645 Alan Hayward <alan.hayward@arm.com>
11646 David Sherwood <david.sherwood@arm.com>
11648 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11650 * config/aarch64/aarch64-protos.h
11651 (aarch64_sve_struct_memory_operand_p): Declare.
11652 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11653 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11654 (VPRED, vpred): Handle SVE structure modes.
11655 * config/aarch64/constraints.md (Utx): New constraint.
11656 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11657 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11658 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11659 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11660 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11661 structure modes. Split into pieces after RA.
11662 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11663 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11665 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11666 SVE structure modes.
11667 (aarch64_classify_address): Likewise.
11668 (sizetochar): Move earlier in file.
11669 (aarch64_print_operand): Handle SVE register lists.
11670 (aarch64_array_mode): New function.
11671 (aarch64_sve_struct_memory_operand_p): Likewise.
11672 (TARGET_ARRAY_MODE): Redefine.
11674 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11675 Alan Hayward <alan.hayward@arm.com>
11676 David Sherwood <david.sherwood@arm.com>
11678 * target.def (array_mode): New target hook.
11679 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11680 * doc/tm.texi: Regenerate.
11681 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11682 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11683 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11684 targetm.array_mode.
11685 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11688 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11689 Alan Hayward <alan.hayward@arm.com>
11690 David Sherwood <david.sherwood@arm.com>
11692 * fold-const.c (fold_binary_loc): Check the argument types
11693 rather than the result type when testing for a vector operation.
11695 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11697 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11698 * doc/tm.texi: Regenerate.
11700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11701 Alan Hayward <alan.hayward@arm.com>
11702 David Sherwood <david.sherwood@arm.com>
11704 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11705 (sve): Document new AArch64 extension.
11706 * doc/md.texi (w): Extend the description of the AArch64
11707 constraint to include SVE vectors.
11708 (Upl, Upa): Document new AArch64 predicate constraints.
11709 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11711 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11712 (msve-vector-bits=): New option.
11713 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11714 SVE when these are disabled.
11715 (sve): New extension.
11716 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11717 modes. Adjust their number of units based on aarch64_sve_vg.
11718 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11719 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11720 aarch64_addr_query_type.
11721 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11722 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11723 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11724 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11725 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11726 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11727 (aarch64_simd_imm_zero_p): Delete.
11728 (aarch64_check_zero_based_sve_index_immediate): Declare.
11729 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11730 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11731 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11732 (aarch64_sve_float_mul_immediate_p): Likewise.
11733 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11734 rather than an rtx.
11735 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11736 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11737 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11738 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11739 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11740 (aarch64_regmode_natural_size): Likewise.
11741 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11742 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11744 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11745 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11746 for VG and the SVE predicate registers.
11747 (V_ALIASES): Add a "z"-prefixed alias.
11748 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11749 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11750 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11751 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11752 (REG_CLASS_NAMES): Add entries for them.
11753 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11754 and the predicate registers.
11755 (aarch64_sve_vg): Declare.
11756 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11757 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11758 (REGMODE_NATURAL_SIZE): Define.
11759 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11761 * config/aarch64/aarch64.c: Include cfgrtl.h.
11762 (simd_immediate_info): Add a constructor for series vectors,
11763 and an associated step field.
11764 (aarch64_sve_vg): New variable.
11765 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11766 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11767 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11768 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11769 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11770 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11771 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11772 (aarch64_get_mask_mode): New functions.
11773 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11774 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11775 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11776 predicate modes and predicate registers. Explicitly restrict
11777 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11778 to store a vector mode if it is recognized by
11779 aarch64_classify_vector_mode.
11780 (aarch64_regmode_natural_size): New function.
11781 (aarch64_hard_regno_caller_save_mode): Return the original mode
11783 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11784 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11785 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11786 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11788 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11789 does not overlap dest if the function is frame-related. Handle
11791 (aarch64_split_add_offset): New function.
11792 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11793 them aarch64_add_offset.
11794 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11795 and update call to aarch64_sub_sp.
11796 (aarch64_add_cfa_expression): New function.
11797 (aarch64_expand_prologue): Pass extra temporary registers to the
11798 functions above. Handle the case in which we need to emit new
11799 DW_CFA_expressions for registers that were originally saved
11800 relative to the stack pointer, but now have to be expressed
11801 relative to the frame pointer.
11802 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11804 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11805 IP0 and IP1 values for SVE frames.
11806 (aarch64_expand_vec_series): New function.
11807 (aarch64_expand_sve_widened_duplicate): Likewise.
11808 (aarch64_expand_sve_const_vector): Likewise.
11809 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11810 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11811 into the register, rather than emitting a SET directly.
11812 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11813 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11814 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11815 (offset_9bit_signed_scaled_p): New functions.
11816 (aarch64_replicate_bitmask_imm): New function.
11817 (aarch64_bitmask_imm): Use it.
11818 (aarch64_cannot_force_const_mem): Reject expressions involving
11819 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11820 (aarch64_classify_index): Handle SVE indices, by requiring
11821 a plain register index with a scale that matches the element size.
11822 (aarch64_classify_address): Handle SVE addresses. Assert that
11823 the mode of the address is VOIDmode or an integer mode.
11824 Update call to aarch64_classify_symbol.
11825 (aarch64_classify_symbolic_expression): Update call to
11826 aarch64_classify_symbol.
11827 (aarch64_const_vec_all_in_range_p): New function.
11828 (aarch64_print_vector_float_operand): Likewise.
11829 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11830 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11831 and the FP immediates 1.0 and 0.5.
11832 (aarch64_print_address_internal): Handle SVE addresses.
11833 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11834 (aarch64_regno_regclass): Handle predicate registers.
11835 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11837 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11838 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11839 (aarch64_convert_sve_vector_bits): New function.
11840 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11841 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11842 rather than an rtx.
11843 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11844 Handle SVE vector and predicate modes. Accept VL-based constants
11845 that need only one temporary register, and VL offsets that require
11846 no temporary registers.
11847 (aarch64_conditional_register_usage): Mark the predicate registers
11848 as fixed if SVE isn't available.
11849 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11850 Return true for SVE vector and predicate modes.
11851 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11852 rather than an unsigned int. Handle SVE modes.
11853 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11855 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11857 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11858 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11859 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11860 (aarch64_sve_float_mul_immediate_p): New functions.
11861 (aarch64_sve_valid_immediate): New function.
11862 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11863 Explicitly reject structure modes. Check for INDEX constants.
11864 Handle PTRUE and PFALSE constants.
11865 (aarch64_check_zero_based_sve_index_immediate): New function.
11866 (aarch64_simd_imm_zero_p): Delete.
11867 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11868 vector modes. Accept constants in the range of CNT[BHWD].
11869 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11870 ask for an Advanced SIMD mode.
11871 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11872 (aarch64_simd_vector_alignment): Handle SVE predicates.
11873 (aarch64_vectorize_preferred_vector_alignment): New function.
11874 (aarch64_simd_vector_alignment_reachable): Use it instead of
11876 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11877 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11879 (MAX_VECT_LEN): Delete.
11880 (expand_vec_perm_d): Add a vec_flags field.
11881 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11882 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11883 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11885 (aarch64_evpc_rev): Rename to...
11886 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11887 (aarch64_evpc_rev_global): New function.
11888 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11889 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11891 (aarch64_evpc_sve_tbl): New function.
11892 (aarch64_expand_vec_perm_const_1): Update after rename of
11893 aarch64_evpc_rev. Handle SVE permutes too, trying
11894 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11895 than aarch64_evpc_tbl.
11896 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11897 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11898 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11899 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11900 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11901 (aarch64_expand_sve_vcond): New functions.
11902 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11903 of aarch64_vector_mode_p.
11904 (aarch64_dwarf_poly_indeterminate_value): New function.
11905 (aarch64_compute_pressure_classes): Likewise.
11906 (aarch64_can_change_mode_class): Likewise.
11907 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11908 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11909 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11910 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11911 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11912 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11913 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11914 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11916 (Dn, Dl, Dr): Accept const as well as const_vector.
11917 (Dz): Likewise. Compare against CONST0_RTX.
11918 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11919 of "vector" where appropriate.
11920 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11921 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11922 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11923 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11924 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11925 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11926 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11927 (v_int_equiv): Extend to SVE modes.
11928 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11930 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11931 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11932 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11933 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11934 (SVE_COND_FP_CMP): New int iterators.
11935 (perm_hilo): Handle the new unpack unspecs.
11936 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11938 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11939 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11940 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11941 (aarch64_equality_operator, aarch64_constant_vector_operand)
11942 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11943 (aarch64_sve_nonimmediate_operand): Likewise.
11944 (aarch64_sve_general_operand): Likewise.
11945 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11946 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11947 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11948 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11949 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11950 (aarch64_sve_float_arith_immediate): Likewise.
11951 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11952 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11953 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11954 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11955 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11956 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11957 (aarch64_sve_float_arith_operand): Likewise.
11958 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11959 (aarch64_sve_float_mul_operand): Likewise.
11960 (aarch64_sve_vec_perm_operand): Likewise.
11961 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11962 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11963 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11964 as well as const_vector.
11965 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11966 in file. Use CONST0_RTX and CONSTM1_RTX.
11967 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11968 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11969 Use aarch64_simd_imm_zero.
11970 * config/aarch64/aarch64-sve.md: New file.
11971 * config/aarch64/aarch64.md: Include it.
11972 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11973 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11974 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11975 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11976 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11977 (sve): New attribute.
11978 (enabled): Disable instructions with the sve attribute unless
11980 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11981 aarch64_expand_mov_immediate.
11982 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11983 CNT[BHSD] immediates.
11984 (movti): Split CONST_POLY_INT moves into two halves.
11985 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11986 Split additions that need a temporary here if the destination
11987 is the stack pointer.
11988 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11989 (*add<mode>3_poly_1): New instruction.
11990 (set_clobber_cc): New expander.
11992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11994 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11995 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11996 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11997 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11998 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11999 Change innermode from fixed_mode_size to machine_mode.
12000 (simplify_subreg): Update call accordingly. Handle a constant-sized
12001 subreg of a variable-length CONST_VECTOR.
12003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12004 Alan Hayward <alan.hayward@arm.com>
12005 David Sherwood <david.sherwood@arm.com>
12007 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
12008 (add_offset_to_base): New function, split out from...
12009 (create_mem_ref): ...here. When handling a scale other than 1,
12010 check first whether the address is valid without the offset.
12011 Add it into the base if so, leaving the index and scale as-is.
12013 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12016 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
12017 fold_for_warn before checking if arg2 is INTEGER_CST.
12019 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
12021 * config/rs6000/predicates.md (load_multiple_operation): Delete.
12022 (store_multiple_operation): Delete.
12023 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
12024 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
12025 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
12026 guarded by TARGET_STRING.
12027 (rs6000_output_load_multiple): Delete.
12028 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
12029 OPTION_MASK_STRING / TARGET_STRING handling.
12030 (print_operand) <'N', 'O'>: Add comment that these are unused now.
12031 (const rs6000_opt_masks) <"string">: Change mask to 0.
12032 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
12033 (MASK_STRING): Delete.
12034 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12036 (load_multiple): Delete.
12043 (store_multiple): Delete.
12050 (movmemsi_8reg): Delete.
12051 (corresponding unnamed define_insn): Delete.
12052 (movmemsi_6reg): Delete.
12053 (corresponding unnamed define_insn): Delete.
12054 (movmemsi_4reg): Delete.
12055 (corresponding unnamed define_insn): Delete.
12056 (movmemsi_2reg): Delete.
12057 (corresponding unnamed define_insn): Delete.
12058 (movmemsi_1reg): Delete.
12059 (corresponding unnamed define_insn): Delete.
12060 * config/rs6000/rs6000.opt (mno-string): New.
12061 (mstring): Replace by deprecation warning stub.
12062 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12064 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12066 * regrename.c (regrename_do_replace): If replacing the same
12067 reg multiple times, try to reuse last created gen_raw_REG.
12070 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12071 main to workaround a bug in GDB.
12073 2018-01-12 Tom de Vries <tom@codesourcery.com>
12076 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12078 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12080 PR rtl-optimization/80481
12081 * ira-color.c (get_cap_member): New function.
12082 (allocnos_conflict_by_live_ranges_p): Use it.
12083 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12084 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12086 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12089 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12090 (*saddl_se_1): Ditto.
12091 (*ssubsi_1): Ditto.
12092 (*ssubl_se_1): Ditto.
12094 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12096 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12097 rather than wi::to_widest for DR_INITs.
12098 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12099 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12100 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12102 (vect_analyze_group_access_1): Note that here.
12104 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12106 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12107 polynomial type sizes.
12109 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12111 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12112 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12113 (gimple_add_tmp_var): Likewise.
12115 2018-01-12 Martin Liska <mliska@suse.cz>
12117 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12118 (gimple_alloc_sizes): Likewise.
12119 (dump_gimple_statistics): Use PRIu64 in printf format.
12120 * gimple.h: Change uint64_t to int.
12122 2018-01-12 Martin Liska <mliska@suse.cz>
12124 * tree-core.h: Use uint64_t instead of int.
12125 * tree.c (tree_node_counts): Likewise.
12126 (tree_node_sizes): Likewise.
12127 (dump_tree_statistics): Use PRIu64 in printf format.
12129 2018-01-12 Martin Liska <mliska@suse.cz>
12131 * Makefile.in: As qsort_chk is implemented in vec.c, add
12132 vec.o to linkage of gencfn-macros.
12133 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12134 passing the info to record_node_allocation_statistics.
12135 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12137 * ggc-common.c (struct ggc_usage): Add operator== and use
12138 it in operator< and compare function.
12139 * mem-stats.h (struct mem_usage): Likewise.
12140 * vec.c (struct vec_usage): Remove operator< and compare
12141 function. Can be simply inherited.
12143 2018-01-12 Martin Jambor <mjambor@suse.cz>
12146 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12147 * tree-ssa-math-opts.c: Include domwalk.h.
12148 (convert_mult_to_fma_1): New function.
12149 (fma_transformation_info): New type.
12150 (fma_deferring_state): Likewise.
12151 (cancel_fma_deferring): New function.
12152 (result_of_phi): Likewise.
12153 (last_fma_candidate_feeds_initial_phi): Likewise.
12154 (convert_mult_to_fma): Added deferring logic, split actual
12155 transformation to convert_mult_to_fma_1.
12156 (math_opts_dom_walker): New type.
12157 (math_opts_dom_walker::after_dom_children): New method, body moved
12158 here from pass_optimize_widening_mul::execute, added deferring logic
12160 (pass_optimize_widening_mul::execute): Moved most of code to
12161 math_opts_dom_walker::after_dom_children.
12162 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12163 * config/i386/i386.c (ix86_option_override_internal): Added
12164 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12166 2018-01-12 Richard Biener <rguenther@suse.de>
12169 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12170 inline instance vars.
12172 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12175 * config/rx/rx.c (rx_is_restricted_memory_address):
12176 Handle SUBREG case.
12178 2018-01-12 Richard Biener <rguenther@suse.de>
12180 PR tree-optimization/80846
12181 * target.def (split_reduction): New target hook.
12182 * targhooks.c (default_split_reduction): New function.
12183 * targhooks.h (default_split_reduction): Declare.
12184 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12185 target requests first reduce vectors by combining low and high
12187 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12188 (get_vectype_for_scalar_type_and_size): Export.
12189 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12190 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12191 * doc/tm.texi: Regenerate.
12192 * config/i386/i386.c (ix86_split_reduction): Implement
12193 TARGET_VECTORIZE_SPLIT_REDUCTION.
12195 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12198 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12199 in PIC mode except for TARGET_VXWORKS_RTP.
12200 * config/sparc/sparc.c: Include cfgrtl.h.
12201 (TARGET_INIT_PIC_REG): Define.
12202 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12203 (sparc_pic_register_p): New predicate.
12204 (sparc_legitimate_address_p): Use it.
12205 (sparc_legitimize_pic_address): Likewise.
12206 (sparc_delegitimize_address): Likewise.
12207 (sparc_mode_dependent_address_p): Likewise.
12208 (gen_load_pcrel_sym): Remove 4th parameter.
12209 (load_got_register): Adjust call to above. Remove obsolete stuff.
12210 (sparc_expand_prologue): Do not call load_got_register here.
12211 (sparc_flat_expand_prologue): Likewise.
12212 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12213 (sparc_use_pseudo_pic_reg): New function.
12214 (sparc_init_pic_reg): Likewise.
12215 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12216 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12218 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12220 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12221 Add item for branch_cost.
12223 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12225 PR rtl-optimization/83565
12226 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12227 not extend the result to a larger mode for rotate operations.
12228 (num_sign_bit_copies1): Likewise.
12230 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12233 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12235 Use values-Xc.o for -pedantic.
12236 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12238 2018-01-12 Martin Liska <mliska@suse.cz>
12241 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12243 (possible_polymorphic_call_targets): Use it.
12244 (ipa_devirt): Likewise.
12246 2018-01-12 Martin Liska <mliska@suse.cz>
12248 * profile-count.h (enum profile_quality): Use 0 as invalid
12249 enum value of profile_quality.
12251 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12253 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12254 -mext-string options.
12256 2018-01-12 Richard Biener <rguenther@suse.de>
12258 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12259 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12260 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12262 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12264 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12266 * configure.ac (--with-long-double-format): Add support for the
12267 configuration option to change the default long double format on
12269 * config.gcc (powerpc*-linux*-*): Likewise.
12270 * configure: Regenerate.
12271 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12272 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12273 used without modification.
12275 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12277 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12278 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12279 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12280 MISC_BUILTIN_SPEC_BARRIER.
12281 (rs6000_init_builtins): Likewise.
12282 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12284 (speculation_barrier): New define_insn.
12285 * doc/extend.texi: Document __builtin_speculation_barrier.
12287 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12290 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12291 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12292 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12294 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12295 integral modes instead of "ss" and "sd".
12296 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12297 vectors with 32-bit and 64-bit elements.
12298 (vecdupssescalarmodesuffix): New mode attribute.
12299 (vec_dup<mode>): Use it.
12301 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12304 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12305 frame if argument is passed on stack.
12307 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12310 * ree.c (combine_reaching_defs): Optimize also
12311 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12312 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12314 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12316 PR middle-end/83189
12317 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12319 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12321 PR middle-end/83718
12322 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12323 after they are computed.
12325 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12327 PR tree-optimization/83695
12328 * gimple-loop-linterchange.cc
12329 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12330 reset cached scev information after interchange.
12331 (pass_linterchange::execute): Remove call to scev_reset_htab.
12333 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12335 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12336 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12337 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12338 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12339 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12340 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12341 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12342 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12343 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12344 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12345 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12346 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12347 (V_lane_reg): Likewise.
12348 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12350 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12351 (vfmal_lane_low<mode>_intrinsic,
12352 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12353 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12354 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12355 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12356 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12357 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12359 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12361 * config/arm/arm-cpus.in (fp16fml): New feature.
12362 (ALL_SIMD): Add fp16fml.
12363 (armv8.2-a): Add fp16fml as an option.
12364 (armv8.3-a): Likewise.
12365 (armv8.4-a): Add fp16fml as part of fp16.
12366 * config/arm/arm.h (TARGET_FP16FML): Define.
12367 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12369 * config/arm/arm-modes.def (V2HF): Define.
12370 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12371 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12372 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12373 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12374 vfmsl_low, vfmsl_high): New set of builtins.
12375 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12376 (vfml_op): New code attribute.
12377 (VFMLHALVES): New int iterator.
12378 (VFML, VFMLSEL): New mode attributes.
12379 (V_reg): Define mapping for V2HF.
12380 (V_hi, V_lo): New mode attributes.
12381 (VF_constraint): Likewise.
12382 (vfml_half, vfml_half_selector): New int attributes.
12383 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12385 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12386 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12388 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12389 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12390 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12391 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12393 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12394 Document new effective target and option set.
12396 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12398 * config/arm/arm-cpus.in (armv8_4): New feature.
12399 (ARMv8_4a): New fgroup.
12400 (armv8.4-a): New arch.
12401 * config/arm/arm-tables.opt: Regenerate.
12402 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12403 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12404 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12405 Add matching rules for -march=armv8.4-a and extensions.
12406 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12408 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12411 * config/rx/rx.md (BW): New mode attribute.
12412 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12414 2018-01-11 Richard Biener <rguenther@suse.de>
12416 PR tree-optimization/83435
12417 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12418 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12419 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12421 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12422 Alan Hayward <alan.hayward@arm.com>
12423 David Sherwood <david.sherwood@arm.com>
12425 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12427 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12428 (aarch64_print_address_internal): Use it to check for a zero offset.
12430 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12431 Alan Hayward <alan.hayward@arm.com>
12432 David Sherwood <david.sherwood@arm.com>
12434 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12435 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12436 Return a poly_int64 rather than a HOST_WIDE_INT.
12437 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12438 rather than a HOST_WIDE_INT.
12439 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12440 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12441 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12442 final_offset from HOST_WIDE_INT to poly_int64.
12443 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12444 to_constant when getting the number of units in an Advanced SIMD
12446 (aarch64_builtin_vectorized_function): Check for a constant number
12448 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12450 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12451 attribute instead of GET_MODE_NUNITS.
12452 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12453 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12454 GET_MODE_SIZE for fixed-size registers.
12455 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12456 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12457 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12458 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12459 (aarch64_print_operand, aarch64_print_address_internal)
12460 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12461 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12462 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12463 Handle polynomial GET_MODE_SIZE.
12464 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12465 wider than SImode without modification.
12466 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12467 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12468 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12469 passing and returning SVE modes.
12470 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12471 rather than GEN_INT.
12472 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12473 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12474 (aarch64_allocate_and_probe_stack_space): Likewise.
12475 (aarch64_layout_frame): Cope with polynomial offsets.
12476 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12477 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12478 polynomial offsets.
12479 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12480 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12481 poly_int64 rather than a HOST_WIDE_INT.
12482 (aarch64_get_separate_components, aarch64_process_components)
12483 (aarch64_expand_prologue, aarch64_expand_epilogue)
12484 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12485 (aarch64_anchor_offset): New function, split out from...
12486 (aarch64_legitimize_address): ...here.
12487 (aarch64_builtin_vectorization_cost): Handle polynomial
12488 TYPE_VECTOR_SUBPARTS.
12489 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12491 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12492 number of elements from the PARALLEL rather than the mode.
12493 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12494 rather than GET_MODE_BITSIZE.
12495 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12496 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12497 (aarch64_expand_vec_perm_const_1): Handle polynomial
12498 d->perm.length () and d->perm elements.
12499 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12500 Apply to_constant to d->perm elements.
12501 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12502 polynomial CONST_VECTOR_NUNITS.
12503 (aarch64_move_pointer): Take amount as a poly_int64 rather
12505 (aarch64_progress_pointer): Avoid temporary variable.
12506 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12507 the mode attribute instead of GET_MODE.
12509 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12510 Alan Hayward <alan.hayward@arm.com>
12511 David Sherwood <david.sherwood@arm.com>
12513 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12514 x exists before using it.
12515 (aarch64_add_constant_internal): Rename to...
12516 (aarch64_add_offset_1): ...this. Replace regnum with separate
12517 src and dest rtxes. Handle the case in which they're different,
12518 including when the offset is zero. Replace scratchreg with an rtx.
12519 Use 2 additions if there is no spare register into which we can
12520 move a 16-bit constant.
12521 (aarch64_add_constant): Delete.
12522 (aarch64_add_offset): Replace reg with separate src and dest
12523 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12524 Use aarch64_add_offset_1.
12525 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12526 an rtx rather than an int. Take the delta as a poly_int64
12527 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12528 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12529 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12530 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12531 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12532 and aarch64_add_sp.
12533 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12534 aarch64_add_constant.
12536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12538 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12539 Use scalar_float_mode.
12541 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12543 * config/aarch64/aarch64-simd.md
12544 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12545 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12546 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12547 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12548 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12549 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12550 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12551 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12552 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12553 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12555 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12558 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12559 targ_options->x_arm_arch_string is non NULL.
12561 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12563 * config/aarch64/aarch64.h
12564 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12566 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12569 * expmed.c (emit_store_flag_force): Swap if const op0
12570 and change VOIDmode to mode of op0.
12572 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12574 PR rtl-optimization/83761
12575 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12576 than bytes to mode_for_size.
12578 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12580 PR middle-end/83189
12581 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12582 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12585 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12587 PR middle-end/83575
12588 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12589 when in layout mode.
12590 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12591 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12594 2018-01-10 Michael Collison <michael.collison@arm.com>
12596 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12597 * config/aarch64/aarch64-option-extension.def: Add
12598 AARCH64_OPT_EXTENSION of 'fp16fml'.
12599 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12600 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12601 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12602 * config/aarch64/constraints.md (Ui7): New constraint.
12603 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12604 (VFMLA_SEL_W): Ditto.
12607 (VFMLA16_LOW): New int iterator.
12608 (VFMLA16_HIGH): Ditto.
12609 (UNSPEC_FMLAL): New unspec.
12610 (UNSPEC_FMLSL): Ditto.
12611 (UNSPEC_FMLAL2): Ditto.
12612 (UNSPEC_FMLSL2): Ditto.
12613 (f16mac): New code attribute.
12614 * config/aarch64/aarch64-simd-builtins.def
12615 (aarch64_fmlal_lowv2sf): Ditto.
12616 (aarch64_fmlsl_lowv2sf): Ditto.
12617 (aarch64_fmlalq_lowv4sf): Ditto.
12618 (aarch64_fmlslq_lowv4sf): Ditto.
12619 (aarch64_fmlal_highv2sf): Ditto.
12620 (aarch64_fmlsl_highv2sf): Ditto.
12621 (aarch64_fmlalq_highv4sf): Ditto.
12622 (aarch64_fmlslq_highv4sf): Ditto.
12623 (aarch64_fmlal_lane_lowv2sf): Ditto.
12624 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12625 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12626 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12627 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12628 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12629 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12630 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12631 (aarch64_fmlal_lane_highv2sf): Ditto.
12632 (aarch64_fmlsl_lane_highv2sf): Ditto.
12633 (aarch64_fmlal_laneq_highv2sf): Ditto.
12634 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12635 (aarch64_fmlalq_lane_highv4sf): Ditto.
12636 (aarch64_fmlsl_lane_highv4sf): Ditto.
12637 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12638 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12639 * config/aarch64/aarch64-simd.md:
12640 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12641 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12642 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12643 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12644 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12645 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12646 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12647 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12648 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12649 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12650 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12651 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12652 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12653 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12654 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12655 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12656 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12657 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12658 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12659 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12660 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12661 (vfmlsl_low_u32): Ditto.
12662 (vfmlalq_low_u32): Ditto.
12663 (vfmlslq_low_u32): Ditto.
12664 (vfmlal_high_u32): Ditto.
12665 (vfmlsl_high_u32): Ditto.
12666 (vfmlalq_high_u32): Ditto.
12667 (vfmlslq_high_u32): Ditto.
12668 (vfmlal_lane_low_u32): Ditto.
12669 (vfmlsl_lane_low_u32): Ditto.
12670 (vfmlal_laneq_low_u32): Ditto.
12671 (vfmlsl_laneq_low_u32): Ditto.
12672 (vfmlalq_lane_low_u32): Ditto.
12673 (vfmlslq_lane_low_u32): Ditto.
12674 (vfmlalq_laneq_low_u32): Ditto.
12675 (vfmlslq_laneq_low_u32): Ditto.
12676 (vfmlal_lane_high_u32): Ditto.
12677 (vfmlsl_lane_high_u32): Ditto.
12678 (vfmlal_laneq_high_u32): Ditto.
12679 (vfmlsl_laneq_high_u32): Ditto.
12680 (vfmlalq_lane_high_u32): Ditto.
12681 (vfmlslq_lane_high_u32): Ditto.
12682 (vfmlalq_laneq_high_u32): Ditto.
12683 (vfmlslq_laneq_high_u32): Ditto.
12684 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12685 (AARCH64_FL_FOR_ARCH8_4): New.
12686 (AARCH64_ISA_F16FML): New ISA flag.
12687 (TARGET_F16FML): New feature flag for fp16fml.
12688 (doc/invoke.texi): Document new fp16fml option.
12690 2018-01-10 Michael Collison <michael.collison@arm.com>
12692 * config/aarch64/aarch64-builtins.c:
12693 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12694 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12695 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12696 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12697 (AARCH64_ISA_SHA3): New ISA flag.
12698 (TARGET_SHA3): New feature flag for sha3.
12699 * config/aarch64/iterators.md (sha512_op): New int attribute.
12700 (CRYPTO_SHA512): New int iterator.
12701 (UNSPEC_SHA512H): New unspec.
12702 (UNSPEC_SHA512H2): Ditto.
12703 (UNSPEC_SHA512SU0): Ditto.
12704 (UNSPEC_SHA512SU1): Ditto.
12705 * config/aarch64/aarch64-simd-builtins.def
12706 (aarch64_crypto_sha512hqv2di): New builtin.
12707 (aarch64_crypto_sha512h2qv2di): Ditto.
12708 (aarch64_crypto_sha512su0qv2di): Ditto.
12709 (aarch64_crypto_sha512su1qv2di): Ditto.
12710 (aarch64_eor3qv8hi): Ditto.
12711 (aarch64_rax1qv2di): Ditto.
12712 (aarch64_xarqv2di): Ditto.
12713 (aarch64_bcaxqv8hi): Ditto.
12714 * config/aarch64/aarch64-simd.md:
12715 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12716 (aarch64_crypto_sha512su0qv2di): Ditto.
12717 (aarch64_crypto_sha512su1qv2di): Ditto.
12718 (aarch64_eor3qv8hi): Ditto.
12719 (aarch64_rax1qv2di): Ditto.
12720 (aarch64_xarqv2di): Ditto.
12721 (aarch64_bcaxqv8hi): Ditto.
12722 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12723 (vsha512h2q_u64): Ditto.
12724 (vsha512su0q_u64): Ditto.
12725 (vsha512su1q_u64): Ditto.
12726 (veor3q_u16): Ditto.
12727 (vrax1q_u64): Ditto.
12728 (vxarq_u64): Ditto.
12729 (vbcaxq_u16): Ditto.
12730 * config/arm/types.md (crypto_sha512): New type attribute.
12731 (crypto_sha3): Ditto.
12732 (doc/invoke.texi): Document new sha3 option.
12734 2018-01-10 Michael Collison <michael.collison@arm.com>
12736 * config/aarch64/aarch64-builtins.c:
12737 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12738 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12739 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12740 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12741 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12742 (AARCH64_ISA_SM4): New ISA flag.
12743 (TARGET_SM4): New feature flag for sm4.
12744 * config/aarch64/aarch64-simd-builtins.def
12745 (aarch64_sm3ss1qv4si): Ditto.
12746 (aarch64_sm3tt1aq4si): Ditto.
12747 (aarch64_sm3tt1bq4si): Ditto.
12748 (aarch64_sm3tt2aq4si): Ditto.
12749 (aarch64_sm3tt2bq4si): Ditto.
12750 (aarch64_sm3partw1qv4si): Ditto.
12751 (aarch64_sm3partw2qv4si): Ditto.
12752 (aarch64_sm4eqv4si): Ditto.
12753 (aarch64_sm4ekeyqv4si): Ditto.
12754 * config/aarch64/aarch64-simd.md:
12755 (aarch64_sm3ss1qv4si): Ditto.
12756 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12757 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12758 (aarch64_sm4eqv4si): Ditto.
12759 (aarch64_sm4ekeyqv4si): Ditto.
12760 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12761 (sm3part_op): Ditto.
12762 (CRYPTO_SM3TT): Ditto.
12763 (CRYPTO_SM3PART): Ditto.
12764 (UNSPEC_SM3SS1): New unspec.
12765 (UNSPEC_SM3TT1A): Ditto.
12766 (UNSPEC_SM3TT1B): Ditto.
12767 (UNSPEC_SM3TT2A): Ditto.
12768 (UNSPEC_SM3TT2B): Ditto.
12769 (UNSPEC_SM3PARTW1): Ditto.
12770 (UNSPEC_SM3PARTW2): Ditto.
12771 (UNSPEC_SM4E): Ditto.
12772 (UNSPEC_SM4EKEY): Ditto.
12773 * config/aarch64/constraints.md (Ui2): New constraint.
12774 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12775 * config/arm/types.md (crypto_sm3): New type attribute.
12776 (crypto_sm4): Ditto.
12777 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12778 (vsm3tt1aq_u32): Ditto.
12779 (vsm3tt1bq_u32): Ditto.
12780 (vsm3tt2aq_u32): Ditto.
12781 (vsm3tt2bq_u32): Ditto.
12782 (vsm3partw1q_u32): Ditto.
12783 (vsm3partw2q_u32): Ditto.
12784 (vsm4eq_u32): Ditto.
12785 (vsm4ekeyq_u32): Ditto.
12786 (doc/invoke.texi): Document new sm4 option.
12788 2018-01-10 Michael Collison <michael.collison@arm.com>
12790 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12791 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12792 (AARCH64_FL_FOR_ARCH8_4): New.
12793 (AARCH64_FL_V8_4): New flag.
12794 (doc/invoke.texi): Document new armv8.4-a option.
12796 2018-01-10 Michael Collison <michael.collison@arm.com>
12798 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12799 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12800 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12801 * config/aarch64/aarch64-option-extension.def: Add
12802 AARCH64_OPT_EXTENSION of 'sha2'.
12803 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12804 (crypto): Disable sha2 and aes if crypto disabled.
12805 (crypto): Enable aes and sha2 if enabled.
12806 (simd): Disable sha2 and aes if simd disabled.
12807 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12809 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12810 (TARGET_SHA2): New feature flag for sha2.
12811 (TARGET_AES): New feature flag for aes.
12812 * config/aarch64/aarch64-simd.md:
12813 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12814 conditional on TARGET_AES.
12815 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12816 (aarch64_crypto_sha1hsi): Make pattern conditional
12818 (aarch64_crypto_sha1hv4si): Ditto.
12819 (aarch64_be_crypto_sha1hv4si): Ditto.
12820 (aarch64_crypto_sha1su1v4si): Ditto.
12821 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12822 (aarch64_crypto_sha1su0v4si): Ditto.
12823 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12824 (aarch64_crypto_sha256su0v4si): Ditto.
12825 (aarch64_crypto_sha256su1v4si): Ditto.
12826 (doc/invoke.texi): Document new aes and sha2 options.
12828 2018-01-10 Martin Sebor <msebor@redhat.com>
12830 PR tree-optimization/83781
12831 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12834 2018-01-11 Martin Sebor <msebor@gmail.com>
12835 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12837 PR tree-optimization/83501
12838 PR tree-optimization/81703
12840 * tree-ssa-strlen.c (get_string_cst): Rename...
12841 (get_string_len): ...to this. Handle global constants.
12842 (handle_char_store): Adjust.
12844 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12845 Jim Wilson <jimw@sifive.com>
12847 * config/riscv/riscv-protos.h (riscv_output_return): New.
12848 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12849 (riscv_attribute_table, riscv_output_return),
12850 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12851 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12852 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12854 (riscv_expand_prologue): Add early return for naked function.
12855 (riscv_expand_epilogue): Likewise.
12856 (riscv_function_ok_for_sibcall): Return false for naked function.
12857 (riscv_set_current_function): New.
12858 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12859 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12860 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12861 * doc/extend.texi (RISC-V Function Attributes): New.
12863 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12865 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12866 check for 128-bit long double before checking TCmode.
12867 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12868 128-bit long doubles before checking TFmode or TCmode.
12869 (FLOAT128_IBM_P): Likewise.
12871 2018-01-10 Martin Sebor <msebor@redhat.com>
12873 PR tree-optimization/83671
12874 * builtins.c (c_strlen): Unconditionally return zero for the empty
12876 Use -Warray-bounds for warnings.
12877 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12878 for non-constant array indices with COMPONENT_REF, arrays of
12879 arrays, and pointers to arrays.
12880 (gimple_fold_builtin_strlen): Determine and set length range for
12881 non-constant character arrays.
12883 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12885 PR middle-end/81897
12886 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12889 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12891 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12893 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12896 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12897 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12898 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12899 indexed_or_indirect_operand predicate.
12900 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12901 (*vsx_le_perm_load_v8hi): Likewise.
12902 (*vsx_le_perm_load_v16qi): Likewise.
12903 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12904 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12905 (*vsx_le_perm_store_v8hi): Likewise.
12906 (*vsx_le_perm_store_v16qi): Likewise.
12907 (eight unnamed splitters): Likewise.
12909 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12911 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12912 * config/rs6000/emmintrin.h: Likewise.
12913 * config/rs6000/mmintrin.h: Likewise.
12914 * config/rs6000/xmmintrin.h: Likewise.
12916 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12919 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12921 * tree.c (tree_nop_conversion): Return true for location wrapper
12923 (maybe_wrap_with_location): New function.
12924 (selftest::check_strip_nops): New function.
12925 (selftest::test_location_wrappers): New function.
12926 (selftest::tree_c_tests): Call it.
12927 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12928 (maybe_wrap_with_location): New decl.
12929 (EXPR_LOCATION_WRAPPER_P): New macro.
12930 (location_wrapper_p): New inline function.
12931 (tree_strip_any_location_wrapper): New inline function.
12933 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12936 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12937 stack_realign_offset for the largest alignment of stack slot
12939 (ix86_find_max_used_stack_alignment): New function.
12940 (ix86_finalize_stack_frame_flags): Use it. Set
12941 max_used_stack_alignment if we don't realign stack.
12942 * config/i386/i386.h (machine_function): Add
12943 max_used_stack_alignment.
12945 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12947 * config/arm/arm.opt (-mbranch-cost): New option.
12948 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12951 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12954 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12955 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12957 2018-01-10 Richard Biener <rguenther@suse.de>
12960 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12961 early out so it also covers the case where we have a non-NULL
12964 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12966 PR tree-optimization/83753
12967 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12968 for non-strided grouped accesses if the number of elements is 1.
12970 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12973 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12974 * i386.h (TARGET_USE_GATHER): Define.
12975 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12977 2018-01-10 Martin Liska <mliska@suse.cz>
12980 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12981 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12983 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12984 CLEANUP_NO_PARTITIONING is not set.
12986 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12988 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12989 for vectors, as a partial revert of r254296.
12990 * rtl.h (const_vec_p): Delete.
12991 (const_vec_duplicate_p): Don't test for vector CONSTs.
12992 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12993 * expmed.c (make_tree): Likewise.
12996 * common.md (E, F): Use CONSTANT_P instead of checking for
12998 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12999 checking for CONST_VECTOR.
13001 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13003 PR middle-end/83575
13004 * predict.c (force_edge_cold): Handle in more sane way edges
13005 with no prediction.
13007 2018-01-09 Carl Love <cel@us.ibm.com>
13009 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
13011 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
13012 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
13013 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
13014 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
13015 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
13016 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
13017 * config/rs6000/rs6000-protos.h: Add extern defition for
13018 rs6000_generate_float2_double_code.
13019 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
13021 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
13022 (float2_v2df): Add define_expand.
13024 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
13027 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
13028 op_mode in the force_to_mode call.
13030 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13032 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
13033 instead of checking each element individually.
13034 (aarch64_evpc_uzp): Likewise.
13035 (aarch64_evpc_zip): Likewise.
13036 (aarch64_evpc_ext): Likewise.
13037 (aarch64_evpc_rev): Likewise.
13038 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13039 instead of checking each element individually. Return true without
13041 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13042 whether all selected elements come from the same input, instead of
13043 checking each element individually. Remove calls to gen_rtx_REG,
13044 start_sequence and end_sequence and instead assert that no rtl is
13047 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13049 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13050 order of HIGH and CONST checks.
13052 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13054 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13055 if the destination isn't an SSA_NAME.
13057 2018-01-09 Richard Biener <rguenther@suse.de>
13059 PR tree-optimization/83668
13060 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13062 (canonicalize_loop_form): ... here, renamed from ...
13063 (canonicalize_loop_closed_ssa_form): ... this and amended to
13064 swap successor edges for loop exit blocks to make us use
13065 the RPO order we need for initial schedule generation.
13067 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13069 PR tree-optimization/64811
13070 * match.pd: When optimizing comparisons with Inf, avoid
13071 introducing or losing exceptions from comparisons with NaN.
13073 2018-01-09 Martin Liska <mliska@suse.cz>
13076 * asan.c (shadow_mem_size): Add gcc_assert.
13078 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13080 Don't save registers in main().
13083 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13084 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13085 * config/avr/avr.c (avr_set_current_function): Don't error if
13086 naked, OS_task or OS_main are specified at the same time.
13087 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13089 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13091 * common/config/avr/avr-common.c (avr_option_optimization_table):
13092 Switch on -mmain-is-OS_task for optimizing compilations.
13094 2018-01-09 Richard Biener <rguenther@suse.de>
13096 PR tree-optimization/83572
13097 * graphite.c: Include cfganal.h.
13098 (graphite_transform_loops): Connect infinite loops to exit
13099 and remove fake edges at the end.
13101 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13103 * ipa-inline.c (edge_badness): Revert accidental checkin.
13105 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13108 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13109 symbols; not inline clones.
13111 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13114 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13115 hard registers. Formatting fixes.
13117 PR preprocessor/83722
13118 * gcc.c (try_generate_repro): Pass
13119 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13120 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13123 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13124 Kito Cheng <kito.cheng@gmail.com>
13126 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13127 (riscv_leaf_function_p): Delete.
13128 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13130 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13132 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13134 (do_ifelse): New function.
13135 (do_isel): New function.
13136 (do_sub3): New function.
13137 (do_add3): New function.
13138 (do_load_mask_compare): New function.
13139 (do_overlap_load_compare): New function.
13140 (expand_compare_loop): New function.
13141 (expand_block_compare): Call expand_compare_loop() when appropriate.
13142 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13143 option description.
13144 (-mblock-compare-inline-loop-limit): New option.
13146 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13149 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13150 Reverse order of second and third operands in first alternative.
13151 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13152 of first and second elements in UNSPEC_VPERMR vector.
13153 (altivec_expand_vec_perm_le): Likewise.
13155 2018-01-08 Jeff Law <law@redhat.com>
13157 PR rtl-optimizatin/81308
13158 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13159 (process_switch): If group_case_labels makes a change, then set
13161 (pass_convert_switch::execute): If a switch is converted, then
13162 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13164 PR rtl-optimization/81308
13165 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13168 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13170 PR target/83663 - Revert r255946
13171 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13172 generation for cases where splatting a value is not useful.
13173 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13174 across a vec_duplicate and a paradoxical subreg forming a vector
13175 mode to a vec_concat.
13177 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13179 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13180 -march=armv8.3-a variants.
13181 * config/arm/t-multilib: Likewise.
13182 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13184 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13186 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13188 (cceq_ior_compare_complement): Give it a name so I can use it, and
13189 change boolean_or_operator predicate to boolean_operator so it can
13190 be used to generate a crand.
13191 (eqne): New code iterator.
13192 (bd/bd_neg): New code_attrs.
13193 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13194 a single define_insn.
13195 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13196 decrement (bdnzt/bdnzf/bdzt/bdzf).
13197 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13198 with the new names of the branch decrement patterns, and added the
13199 names of the branch decrement conditional patterns.
13201 2018-01-08 Richard Biener <rguenther@suse.de>
13203 PR tree-optimization/83563
13204 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13207 2018-01-08 Richard Biener <rguenther@suse.de>
13209 PR middle-end/83713
13210 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13212 2018-01-08 Richard Biener <rguenther@suse.de>
13214 PR tree-optimization/83685
13215 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13216 references to abnormals.
13218 2018-01-08 Richard Biener <rguenther@suse.de>
13221 * dwarf2out.c (output_indirect_strings): Handle empty
13222 skeleton_debug_str_hash.
13223 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13225 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13227 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13228 (emit_store_direct): Likewise.
13229 (arc_trampoline_adjust_address): Likewise.
13230 (arc_asm_trampoline_template): New function.
13231 (arc_initialize_trampoline): Use asm_trampoline_template.
13232 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13233 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13234 * config/arc/arc.md (flush_icache): Delete pattern.
13236 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13238 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13239 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13242 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13245 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13246 by not USED_FOR_TARGET.
13247 (make_pass_resolve_sw_modes): Likewise.
13249 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13251 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13254 2018-01-08 Richard Biener <rguenther@suse.de>
13256 PR middle-end/83580
13257 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13259 2018-01-08 Richard Biener <rguenther@suse.de>
13261 PR middle-end/83517
13262 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13264 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13266 PR middle-end/81897
13267 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13268 basic blocks with a small number of successors.
13269 (convert_control_dep_chain_into_preds): Improve handling of
13271 (dump_predicates): Split apart into...
13272 (dump_pred_chain): ...here...
13273 (dump_pred_info): ...and here.
13274 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13275 (can_chain_union_be_invalidated_p): Improve check for invalidation
13277 (uninit_uses_cannot_happen): Avoid unnecessary if
13278 convert_control_dep_chain_into_preds yielded nothing.
13280 2018-01-06 Martin Sebor <msebor@redhat.com>
13282 PR tree-optimization/83640
13283 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13284 subtracting negative offset from size.
13285 (builtin_access::overlap): Adjust offset bounds of the access to fall
13286 within the size of the object if possible.
13288 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13290 PR rtl-optimization/83699
13291 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13292 extract_bit_field_as_subreg to cases in which the extracted
13293 value is also a vector.
13295 * lra-constraints.c (process_alt_operands): Test for the equivalence
13296 substitutions when detecting a possible reload cycle.
13298 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13301 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13302 by default if flag_selective_schedling{,2}. Formatting fixes.
13304 PR rtl-optimization/83682
13305 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13306 if it has non-VECTOR_MODE element mode.
13307 (vec_duplicate_p): Likewise.
13309 PR middle-end/83694
13310 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13311 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13313 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13316 * config/i386/i386-builtin.def
13317 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13318 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13319 Require also OPTION_MASK_ISA_AVX512F in addition to
13320 OPTION_MASK_ISA_GFNI.
13321 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13322 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13323 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13324 to OPTION_MASK_ISA_GFNI.
13325 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13326 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13327 OPTION_MASK_ISA_AVX512BW.
13328 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13329 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13330 addition to OPTION_MASK_ISA_GFNI.
13331 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13332 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13333 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13334 to OPTION_MASK_ISA_GFNI.
13335 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13336 a requirement for all ISAs rather than any of them with a few
13338 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13340 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13341 bitmasks to be enabled with 3 exceptions, instead of requiring any
13342 enabled ISA with lots of exceptions.
13343 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13344 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13345 Change avx512bw in isa attribute to avx512f.
13346 * config/i386/sgxintrin.h: Add license boilerplate.
13347 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13348 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13349 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13350 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13352 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13353 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13354 temporarily sse2 rather than sse if not enabled already.
13357 * config/i386/sse.md (VI248_VLBW): Rename to ...
13358 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13359 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13360 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13361 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13362 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13363 mode iterator instead of VI248_VLBW.
13365 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13367 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13368 (record_modified): Skip clobbers; add debug output.
13369 (param_change_prob): Use sreal frequencies.
13371 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13373 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13374 punt for user-aligned variables.
13376 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13378 * tree-chrec.c (chrec_contains_symbols): Return true for
13381 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13384 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13385 of (x|y) == x for BICS pattern.
13387 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13389 PR tree-optimization/83605
13390 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13391 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13394 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13396 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13397 * config/epiphany/rtems.h: New file.
13399 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13400 Uros Bizjak <ubizjak@gmail.com>
13403 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13404 QIreg_operand instead of register_operand predicate.
13405 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13406 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13407 comments instead of -fmitigate[-_]rop.
13409 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13412 * cgraphunit.c (symbol_table::compile): Switch to text_section
13413 before calling assembly_start debug hook.
13414 * run-rtl-passes.c (run_rtl_passes): Likewise.
13417 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13419 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13420 range_int_cst_p rather than !symbolic_range_p before calling
13421 extract_range_from_multiplicative_op_1.
13423 2018-01-04 Jeff Law <law@redhat.com>
13425 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13426 redundant test in assertion.
13428 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13430 * doc/rtl.texi: Document machine_mode wrapper classes.
13432 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13434 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13435 using tree_to_uhwi.
13437 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13439 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13440 the VEC_PERM_EXPR fold to fail.
13442 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13445 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13446 to switched_sections.
13448 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13451 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13452 test for d.testing.
13454 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13457 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13458 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13460 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13463 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13464 is BLKmode and bitpos not zero or mode change is needed.
13466 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13469 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13472 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13475 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13476 instead of MULT rtx. Update all corresponding splitters.
13477 (*saddl_se): Ditto.
13478 (*ssub<modesuffix>): Ditto.
13479 (*ssubl_se): Ditto.
13480 (*cmp_sadd_di): Update split patterns.
13481 (*cmp_sadd_si): Ditto.
13482 (*cmp_sadd_sidi): Ditto.
13483 (*cmp_ssub_di): Ditto.
13484 (*cmp_ssub_si): Ditto.
13485 (*cmp_ssub_sidi): Ditto.
13486 * config/alpha/predicates.md (const23_operand): New predicate.
13487 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13488 Look for ASHIFT, not MULT inner operand.
13489 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13491 2018-01-04 Martin Liska <mliska@suse.cz>
13493 PR gcov-profile/83669
13494 * gcov.c (output_intermediate_file): Add version to intermediate
13496 * doc/gcov.texi: Document new field 'version' in intermediate
13497 file format. Fix location of '-k' option of gcov command.
13499 2018-01-04 Martin Liska <mliska@suse.cz>
13502 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13504 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13506 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13508 2018-01-03 Martin Sebor <msebor@redhat.com>
13510 PR tree-optimization/83655
13511 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13512 checking calls with invalid arguments.
13514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13516 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13517 (vectorizable_mask_load_store): Delete.
13518 (vectorizable_call): Return false for masked loads and stores.
13519 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13520 instead of gimple_assign_rhs1.
13521 (vectorizable_load): Handle IFN_MASK_LOAD.
13522 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13526 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13528 (vectorizable_mask_load_store): ...here.
13529 (vectorizable_load): ...and here.
13531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13533 * tree-vect-stmts.c (vect_build_all_ones_mask)
13534 (vect_build_zero_merge_argument): New functions, split out from...
13535 (vectorizable_load): ...here.
13537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13539 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13541 (vectorizable_mask_load_store): ...here.
13542 (vectorizable_store): ...and here.
13544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13546 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13548 (vectorizable_mask_load_store): ...here.
13550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13552 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13553 (vect_model_store_cost): Take a vec_load_store_type instead of a
13555 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13556 (vect_model_store_cost): Take a vec_load_store_type instead of a
13558 (vectorizable_mask_load_store): Update accordingly.
13559 (vectorizable_store): Likewise.
13560 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13564 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13565 IFN_MASK_LOAD calls here rather than...
13566 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13569 Alan Hayward <alan.hayward@arm.com>
13570 David Sherwood <david.sherwood@arm.com>
13572 * expmed.c (extract_bit_field_1): For vector extracts,
13573 fall back to extract_bit_field_as_subreg if vec_extract
13576 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13577 Alan Hayward <alan.hayward@arm.com>
13578 David Sherwood <david.sherwood@arm.com>
13580 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13581 they are variable or constant sized.
13582 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13583 slots for constant-sized data.
13585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13586 Alan Hayward <alan.hayward@arm.com>
13587 David Sherwood <david.sherwood@arm.com>
13589 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13590 handling COND_EXPRs with boolean comparisons, try to find a better
13591 basis for the mask type than the boolean itself.
13593 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13595 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13596 is calculated and how it can be overridden.
13597 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13598 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13600 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13604 Alan Hayward <alan.hayward@arm.com>
13605 David Sherwood <david.sherwood@arm.com>
13607 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13608 Remove the mode argument.
13609 (aarch64_simd_valid_immediate): Remove the mode and inverse
13611 * config/aarch64/iterators.md (bitsize): New iterator.
13612 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13613 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13614 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13615 aarch64_simd_valid_immediate.
13616 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13617 (aarch64_reg_or_bic_imm): Likewise.
13618 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13619 with an insn_type enum and msl with a modifier_type enum.
13620 Replace element_width with a scalar_mode. Change the shift
13621 to unsigned int. Add constructors for scalar_float_mode and
13622 scalar_int_mode elements.
13623 (aarch64_vect_float_const_representable_p): Delete.
13624 (aarch64_can_const_movi_rtx_p)
13625 (aarch64_simd_scalar_immediate_valid_for_move)
13626 (aarch64_simd_make_constant): Update call to
13627 aarch64_simd_valid_immediate.
13628 (aarch64_advsimd_valid_immediate_hs): New function.
13629 (aarch64_advsimd_valid_immediate): Likewise.
13630 (aarch64_simd_valid_immediate): Remove mode and inverse
13631 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13632 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13633 and aarch64_float_const_representable_p on the result.
13634 (aarch64_output_simd_mov_immediate): Remove mode argument.
13635 Update call to aarch64_simd_valid_immediate and use of
13636 simd_immediate_info.
13637 (aarch64_output_scalar_simd_mov_immediate): Update call
13640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13641 Alan Hayward <alan.hayward@arm.com>
13642 David Sherwood <david.sherwood@arm.com>
13644 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13645 (mode_nunits): Likewise CONST_MODE_NUNITS.
13646 * machmode.def (ADJUST_NUNITS): Document.
13647 * genmodes.c (mode_data::need_nunits_adj): New field.
13648 (blank_mode): Update accordingly.
13649 (adj_nunits): New variable.
13650 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13652 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13653 listed in adj_nunits.
13654 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13655 listed in adj_nunits. Don't emit case statements for such modes.
13656 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13657 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13658 nothing if adj_nunits is nonnull.
13659 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13660 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13661 (emit_mode_fbit): Update use of print_maybe_const_decl.
13662 (emit_move_size): Likewise. Treat the array as non-const
13664 (emit_mode_adjustments): Handle adj_nunits.
13666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13668 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13669 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13670 (VECTOR_MODES): Use it.
13671 (make_vector_modes): Take the prefix as an argument.
13673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13674 Alan Hayward <alan.hayward@arm.com>
13675 David Sherwood <david.sherwood@arm.com>
13677 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13678 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13679 for MODE_VECTOR_BOOL.
13680 * machmode.def (VECTOR_BOOL_MODE): Document.
13681 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13682 (make_vector_bool_mode): New function.
13683 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13685 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13686 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13688 * stor-layout.c (int_mode_for_mode): Likewise.
13689 * tree.c (build_vector_type_for_mode): Likewise.
13690 * varasm.c (output_constant_pool_2): Likewise.
13691 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13692 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13693 for MODE_VECTOR_BOOL.
13694 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13695 of mode class checks.
13696 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13697 instead of a list of mode class checks.
13698 (expand_vector_scalar_condition): Likewise.
13699 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13702 Alan Hayward <alan.hayward@arm.com>
13703 David Sherwood <david.sherwood@arm.com>
13705 * machmode.h (mode_size): Change from unsigned short to
13707 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13708 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13709 or if measurement_type is not polynomial.
13710 (fixed_size_mode::includes_p): Check for constant-sized modes.
13711 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13712 return a poly_uint16 rather than an unsigned short.
13713 (emit_mode_size): Change the type of mode_size from unsigned short
13714 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13715 (emit_mode_adjustments): Cope with polynomial vector sizes.
13716 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13718 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13720 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13721 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13722 * caller-save.c (setup_save_areas): Likewise.
13723 (replace_reg_with_saved_mem): Likewise.
13724 * calls.c (emit_library_call_value_1): Likewise.
13725 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13726 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13727 (gen_lowpart_for_combine): Likewise.
13728 * convert.c (convert_to_integer_1): Likewise.
13729 * cse.c (equiv_constant, cse_insn): Likewise.
13730 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13731 (cselib_subst_to_values): Likewise.
13732 * dce.c (word_dce_process_block): Likewise.
13733 * df-problems.c (df_word_lr_mark_ref): Likewise.
13734 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13735 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13736 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13737 (rtl_for_decl_location): Likewise.
13738 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13739 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13740 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13741 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13742 (expand_expr_real_1): Likewise.
13743 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13744 (pad_below): Likewise.
13745 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13746 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13747 * ira.c (get_subreg_tracking_sizes): Likewise.
13748 * ira-build.c (ira_create_allocno_objects): Likewise.
13749 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13750 (ira_sort_regnos_for_alter_reg): Likewise.
13751 * ira-costs.c (record_operand_costs): Likewise.
13752 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13753 (resolve_simple_move): Likewise.
13754 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13755 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13756 (lra_constraints): Likewise.
13757 (CONST_POOL_OK_P): Reject variable-sized modes.
13758 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13759 (add_pseudo_to_slot, lra_spill): Likewise.
13760 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13761 * optabs-query.c (get_best_extraction_insn): Likewise.
13762 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13763 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13764 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13765 * recog.c (offsettable_address_addr_space_p): Likewise.
13766 * regcprop.c (maybe_mode_change): Likewise.
13767 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13768 * regrename.c (build_def_use): Likewise.
13769 * regstat.c (dump_reg_info): Likewise.
13770 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13771 (find_reloads, find_reloads_subreg_address): Likewise.
13772 * reload1.c (eliminate_regs_1): Likewise.
13773 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13774 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13775 (simplify_binary_operation_1, simplify_subreg): Likewise.
13776 * targhooks.c (default_function_arg_padding): Likewise.
13777 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13778 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13779 (verify_gimple_assign_ternary): Likewise.
13780 * tree-inline.c (estimate_move_cost): Likewise.
13781 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13782 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13783 (get_address_cost_ainc): Likewise.
13784 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13785 (vect_supportable_dr_alignment): Likewise.
13786 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13787 (vectorizable_reduction): Likewise.
13788 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13789 (vectorizable_operation, vectorizable_load): Likewise.
13790 * tree.c (build_same_sized_truth_vector_type): Likewise.
13791 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13792 * var-tracking.c (emit_note_insn_var_location): Likewise.
13793 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13794 (ADDR_VEC_ALIGN): Likewise.
13796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13797 Alan Hayward <alan.hayward@arm.com>
13798 David Sherwood <david.sherwood@arm.com>
13800 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13802 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13803 or if measurement_type is polynomial.
13804 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13805 * combine.c (make_extraction): Likewise.
13806 * dse.c (find_shift_sequence): Likewise.
13807 * dwarf2out.c (mem_loc_descriptor): Likewise.
13808 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13809 (extract_bit_field, extract_low_bits): Likewise.
13810 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13811 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13812 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13813 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13814 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13815 * reload.c (find_reloads): Likewise.
13816 * reload1.c (alter_reg): Likewise.
13817 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13818 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13819 * tree-if-conv.c (predicate_mem_writes): Likewise.
13820 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13821 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13822 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13823 * valtrack.c (dead_debug_insert_temp): Likewise.
13824 * varasm.c (mergeable_constant_section): Likewise.
13825 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13827 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13828 Alan Hayward <alan.hayward@arm.com>
13829 David Sherwood <david.sherwood@arm.com>
13831 * expr.c (expand_assignment): Cope with polynomial mode sizes
13832 when assigning to a CONCAT.
13834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13835 Alan Hayward <alan.hayward@arm.com>
13836 David Sherwood <david.sherwood@arm.com>
13838 * machmode.h (mode_precision): Change from unsigned short to
13840 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13842 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13843 or if measurement_type is not polynomial.
13844 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13845 in which the mode is already known to be a scalar_int_mode.
13846 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13847 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13849 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13850 for GET_MODE_PRECISION.
13851 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13852 for GET_MODE_PRECISION.
13853 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13855 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13856 (expand_field_assignment, make_extraction): Likewise.
13857 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13858 (get_last_value): Likewise.
13859 * convert.c (convert_to_integer_1): Likewise.
13860 * cse.c (cse_insn): Likewise.
13861 * expr.c (expand_expr_real_1): Likewise.
13862 * lra-constraints.c (simplify_operand_subreg): Likewise.
13863 * optabs-query.c (can_atomic_load_p): Likewise.
13864 * optabs.c (expand_atomic_load): Likewise.
13865 (expand_atomic_store): Likewise.
13866 * ree.c (combine_reaching_defs): Likewise.
13867 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13868 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13869 * tree.h (type_has_mode_precision_p): Likewise.
13870 * ubsan.c (instrument_si_overflow): Likewise.
13872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13873 Alan Hayward <alan.hayward@arm.com>
13874 David Sherwood <david.sherwood@arm.com>
13876 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13877 polynomial numbers of units.
13878 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13879 (valid_vector_subparts_p): New function.
13880 (build_vector_type): Remove temporary shim and take the number
13881 of units as a poly_uint64 rather than an int.
13882 (build_opaque_vector_type): Take the number of units as a
13883 poly_uint64 rather than an int.
13884 * tree.c (build_vector_from_ctor): Handle polynomial
13885 TYPE_VECTOR_SUBPARTS.
13886 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13887 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13888 (build_vector_from_val): If the number of units is variable,
13889 use build_vec_duplicate_cst for constant operands and
13890 VEC_DUPLICATE_EXPR otherwise.
13891 (make_vector_type): Remove temporary is_constant ().
13892 (build_vector_type, build_opaque_vector_type): Take the number of
13893 units as a poly_uint64 rather than an int.
13894 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13896 * cfgexpand.c (expand_debug_expr): Likewise.
13897 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13898 (store_constructor, expand_expr_real_1): Likewise.
13899 (const_scalar_mask_from_tree): Likewise.
13900 * fold-const-call.c (fold_const_reduction): Likewise.
13901 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13902 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13903 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13904 (fold_relational_const): Likewise.
13905 (native_interpret_vector): Likewise. Change the size from an
13906 int to an unsigned int.
13907 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13908 TYPE_VECTOR_SUBPARTS.
13909 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13910 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13911 duplicating a non-constant operand into a variable-length vector.
13912 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13913 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13914 * ipa-icf.c (sem_variable::equals): Likewise.
13915 * match.pd: Likewise.
13916 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13917 * print-tree.c (print_node): Likewise.
13918 * stor-layout.c (layout_type): Likewise.
13919 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13920 * tree-cfg.c (verify_gimple_comparison): Likewise.
13921 (verify_gimple_assign_binary): Likewise.
13922 (verify_gimple_assign_ternary): Likewise.
13923 (verify_gimple_assign_single): Likewise.
13924 * tree-pretty-print.c (dump_generic_node): Likewise.
13925 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13926 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13927 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13928 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13929 (vect_shift_permute_load_chain): Likewise.
13930 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13931 (expand_vector_condition, optimize_vector_constructor): Likewise.
13932 (lower_vec_perm, get_compute_type): Likewise.
13933 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13934 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13935 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13936 (vect_recog_mask_conversion_pattern): Likewise.
13937 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13938 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13939 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13940 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13941 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13942 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13943 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13944 (supportable_widening_operation): Likewise.
13945 (supportable_narrowing_operation): Likewise.
13946 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13948 * varasm.c (output_constant): Likewise.
13950 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13951 Alan Hayward <alan.hayward@arm.com>
13952 David Sherwood <david.sherwood@arm.com>
13954 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13955 so that both the length == 3 and length != 3 cases set up their
13956 own permute vectors. Add comments explaining why we know the
13957 number of elements is constant.
13958 (vect_permute_load_chain): Likewise.
13960 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13961 Alan Hayward <alan.hayward@arm.com>
13962 David Sherwood <david.sherwood@arm.com>
13964 * machmode.h (mode_nunits): Change from unsigned char to
13966 (ONLY_FIXED_SIZE_MODES): New macro.
13967 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13968 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13969 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13971 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13972 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13973 or if measurement_type is not polynomial.
13974 * genmodes.c (ZERO_COEFFS): New macro.
13975 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13977 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13978 Use ZERO_COEFFS when emitting initializers.
13979 * data-streamer.h (bp_pack_poly_value): New function.
13980 (bp_unpack_poly_value): Likewise.
13981 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13982 for GET_MODE_NUNITS.
13983 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13984 for GET_MODE_NUNITS.
13985 * tree.c (make_vector_type): Remove temporary shim and make
13986 the real function take the number of units as a poly_uint64
13987 rather than an int.
13988 (build_vector_type_for_mode): Handle polynomial nunits.
13989 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13990 * emit-rtl.c (const_vec_series_p_1): Likewise.
13991 (gen_rtx_CONST_VECTOR): Likewise.
13992 * fold-const.c (test_vec_duplicate_folding): Likewise.
13993 * genrecog.c (validate_pattern): Likewise.
13994 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13995 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13996 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13997 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13998 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13999 * rtlanal.c (subreg_get_info): Likewise.
14000 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14001 (vect_grouped_load_supported): Likewise.
14002 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
14003 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
14004 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
14005 (simplify_const_unary_operation, simplify_binary_operation_1)
14006 (simplify_const_binary_operation, simplify_ternary_operation)
14007 (test_vector_ops_duplicate, test_vector_ops): Likewise.
14008 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
14009 instead of CONST_VECTOR_NUNITS.
14010 * varasm.c (output_constant_pool_2): Likewise.
14011 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
14012 explicit-encoded elements in the XVEC for variable-length vectors.
14014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14016 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
14018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14019 Alan Hayward <alan.hayward@arm.com>
14020 David Sherwood <david.sherwood@arm.com>
14022 * coretypes.h (fixed_size_mode): Declare.
14023 (fixed_size_mode_pod): New typedef.
14024 * builtins.h (target_builtins::x_apply_args_mode)
14025 (target_builtins::x_apply_result_mode): Change type to
14026 fixed_size_mode_pod.
14027 * builtins.c (apply_args_size, apply_result_size, result_vector)
14028 (expand_builtin_apply_args_1, expand_builtin_apply)
14029 (expand_builtin_return): Update accordingly.
14031 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14033 * cse.c (hash_rtx_cb): Hash only the encoded elements.
14034 * cselib.c (cselib_hash_rtx): Likewise.
14035 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14036 CONST_VECTOR encoding.
14038 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14039 Jeff Law <law@redhat.com>
14042 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14043 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14044 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14045 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14048 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14049 explicitly probe *sp in a noreturn function if there were any callee
14050 register saves or frame pointer is needed.
14052 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14055 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14056 BLKmode for ternary, binary or unary expressions.
14059 * var-tracking.c (delete_vta_debug_insn): New inline function.
14060 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14061 insns from get_insns () to NULL instead of each bb separately.
14062 Use delete_vta_debug_insn. No longer static.
14063 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14064 delete_vta_debug_insns callers.
14065 * rtl.h (delete_vta_debug_insns): Declare.
14066 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14067 instead of variable_tracking_main.
14069 2018-01-03 Martin Sebor <msebor@redhat.com>
14071 PR tree-optimization/83603
14072 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14073 arguments past the endof the argument list in functions declared
14074 without a prototype.
14075 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14076 Avoid checking when arguments are null.
14078 2018-01-03 Martin Sebor <msebor@redhat.com>
14081 * doc/extend.texi (attribute const): Fix a typo.
14082 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14083 issuing -Wsuggest-attribute for void functions.
14085 2018-01-03 Martin Sebor <msebor@redhat.com>
14087 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14088 offset_int::from instead of wide_int::to_shwi.
14089 (maybe_diag_overlap): Remove assertion.
14090 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14091 * gimple-ssa-sprintf.c (format_directive): Same.
14092 (parse_directive): Same.
14093 (sprintf_dom_walker::compute_format_length): Same.
14094 (try_substitute_return_value): Same.
14096 2018-01-03 Jeff Law <law@redhat.com>
14098 PR middle-end/83654
14099 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14100 non-constant residual for zero at runtime and avoid probing in
14101 that case. Reorganize code for trailing problem to mirror handling
14104 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14106 PR tree-optimization/83501
14107 * tree-ssa-strlen.c (get_string_cst): New.
14108 (handle_char_store): Call get_string_cst.
14110 2018-01-03 Martin Liska <mliska@suse.cz>
14112 PR tree-optimization/83593
14113 * tree-ssa-strlen.c: Include tree-cfg.h.
14114 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14115 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14116 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14118 (strlen_dom_walker::before_dom_children): Call
14119 gimple_purge_dead_eh_edges. Dump tranformation with details
14121 (strlen_dom_walker::before_dom_children): Update call by adding
14122 new argument cleanup_eh.
14123 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14125 2018-01-03 Martin Liska <mliska@suse.cz>
14128 * cif-code.def (VARIADIC_THUNK): New enum value.
14129 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14132 2018-01-03 Jan Beulich <jbeulich@suse.com>
14134 * sse.md (mov<mode>_internal): Tighten condition for when to use
14135 vmovdqu<ssescalarsize> for TI and OI modes.
14137 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14139 Update copyright years.
14141 2018-01-03 Martin Liska <mliska@suse.cz>
14144 * ipa-visibility.c (function_and_variable_visibility): Skip
14145 functions with noipa attribure.
14147 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14149 * gcc.c (process_command): Update copyright notice dates.
14150 * gcov-dump.c (print_version): Ditto.
14151 * gcov.c (print_version): Ditto.
14152 * gcov-tool.c (print_version): Ditto.
14153 * gengtype.c (create_file): Ditto.
14154 * doc/cpp.texi: Bump @copying's copyright year.
14155 * doc/cppinternals.texi: Ditto.
14156 * doc/gcc.texi: Ditto.
14157 * doc/gccint.texi: Ditto.
14158 * doc/gcov.texi: Ditto.
14159 * doc/install.texi: Ditto.
14160 * doc/invoke.texi: Ditto.
14162 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14164 * vector-builder.h (vector_builder::m_full_nelts): Change from
14165 unsigned int to poly_uint64.
14166 (vector_builder::full_nelts): Update prototype accordingly.
14167 (vector_builder::new_vector): Likewise.
14168 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14169 (vector_builder::operator ==): Likewise.
14170 (vector_builder::finalize): Likewise.
14171 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14172 Take the number of elements as a poly_uint64 rather than an
14174 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14175 from unsigned int to poly_uint64.
14176 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14177 (vec_perm_indices::new_vector): Likewise.
14178 (vec_perm_indices::length): Likewise.
14179 (vec_perm_indices::nelts_per_input): Likewise.
14180 (vec_perm_indices::input_nelts): Likewise.
14181 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14182 number of elements per input as a poly_uint64 rather than an
14183 unsigned int. Use the original encoding for variable-length
14184 vectors, rather than clamping each individual element.
14185 For the second and subsequent elements in each pattern,
14186 clamp the step and base before clamping their sum.
14187 (vec_perm_indices::series_p): Handle polynomial element counts.
14188 (vec_perm_indices::all_in_range_p): Likewise.
14189 (vec_perm_indices_to_tree): Likewise.
14190 (vec_perm_indices_to_rtx): Likewise.
14191 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14192 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14193 (tree_vector_builder::new_binary_operation): Handle polynomial
14194 element counts. Return false if we need to know the number
14195 of elements at compile time.
14196 * fold-const.c (fold_vec_perm): Punt if the number of elements
14197 isn't known at compile time.
14199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14201 * vec-perm-indices.h (vec_perm_builder): Change element type
14202 from HOST_WIDE_INT to poly_int64.
14203 (vec_perm_indices::element_type): Update accordingly.
14204 (vec_perm_indices::clamp): Handle polynomial element_types.
14205 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14206 (vec_perm_indices::all_in_range_p): Likewise.
14207 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14209 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14210 polynomial vec_perm_indices element types.
14211 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14212 * fold-const.c (fold_vec_perm): Likewise.
14213 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14214 * tree-vect-generic.c (lower_vec_perm): Likewise.
14215 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14216 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14217 element type to HOST_WIDE_INT.
14219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14220 Alan Hayward <alan.hayward@arm.com>
14221 David Sherwood <david.sherwood@arm.com>
14223 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14224 rather than an int. Use plus_constant.
14225 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14226 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14229 Alan Hayward <alan.hayward@arm.com>
14230 David Sherwood <david.sherwood@arm.com>
14232 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14233 a HOST_WIDE_INT to a poly_int64.
14235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14236 Alan Hayward <alan.hayward@arm.com>
14237 David Sherwood <david.sherwood@arm.com>
14239 * calls.c (load_register_parameters): Cope with polynomial
14240 mode sizes. Require a constant size for BLKmode parameters
14241 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14242 forces a parameter to be padded at the lsb end in order to
14243 fill a complete number of words, require the parameter size
14244 to be ordered wrt UNITS_PER_WORD.
14246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14247 Alan Hayward <alan.hayward@arm.com>
14248 David Sherwood <david.sherwood@arm.com>
14250 * reload1.c (spill_stack_slot_width): Change element type
14251 from unsigned int to poly_uint64_pod.
14252 (alter_reg): Treat mode sizes as polynomial.
14254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14255 Alan Hayward <alan.hayward@arm.com>
14256 David Sherwood <david.sherwood@arm.com>
14258 * reload.c (complex_word_subreg_p): New function.
14259 (reload_inner_reg_of_subreg, push_reload): Use it.
14261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14262 Alan Hayward <alan.hayward@arm.com>
14263 David Sherwood <david.sherwood@arm.com>
14265 * lra-constraints.c (process_alt_operands): Reject matched
14266 operands whose sizes aren't ordered.
14267 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14274 that the mode size is in the set {1, 2, 4, 8, 16}.
14276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14277 Alan Hayward <alan.hayward@arm.com>
14278 David Sherwood <david.sherwood@arm.com>
14280 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14281 Use plus_constant instead of gen_rtx_PLUS.
14283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14284 Alan Hayward <alan.hayward@arm.com>
14285 David Sherwood <david.sherwood@arm.com>
14287 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14288 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14289 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14290 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14291 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14292 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14293 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14294 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14295 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14296 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14298 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14299 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14300 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14301 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14302 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14303 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14304 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14305 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14306 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14307 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14309 * expr.c (emit_move_resolve_push): Treat the input and result
14310 of PUSH_ROUNDING as a poly_int64.
14311 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14312 (emit_push_insn): Likewise.
14313 * lra-eliminations.c (mark_not_eliminable): Likewise.
14314 * recog.c (push_operand): Likewise.
14315 * reload1.c (elimination_effects): Likewise.
14316 * rtlanal.c (nonzero_bits1): Likewise.
14317 * calls.c (store_one_arg): Likewise. Require the padding to be
14318 known at compile time.
14320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14321 Alan Hayward <alan.hayward@arm.com>
14322 David Sherwood <david.sherwood@arm.com>
14324 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14325 Use plus_constant instead of gen_rtx_PLUS.
14327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14328 Alan Hayward <alan.hayward@arm.com>
14329 David Sherwood <david.sherwood@arm.com>
14331 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14332 rather than an int.
14334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14335 Alan Hayward <alan.hayward@arm.com>
14336 David Sherwood <david.sherwood@arm.com>
14338 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14339 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14340 via stack temporaries. Treat the mode size as polynomial too.
14342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14343 Alan Hayward <alan.hayward@arm.com>
14344 David Sherwood <david.sherwood@arm.com>
14346 * expr.c (expand_expr_real_2): When handling conversions involving
14347 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14348 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14349 as a poly_uint64 too.
14351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14352 Alan Hayward <alan.hayward@arm.com>
14353 David Sherwood <david.sherwood@arm.com>
14355 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14358 Alan Hayward <alan.hayward@arm.com>
14359 David Sherwood <david.sherwood@arm.com>
14361 * combine.c (can_change_dest_mode): Handle polynomial
14362 REGMODE_NATURAL_SIZE.
14363 * expmed.c (store_bit_field_1): Likewise.
14364 * expr.c (store_constructor): Likewise.
14365 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14366 and polynomial REGMODE_NATURAL_SIZE.
14367 (gen_lowpart_common): Likewise.
14368 * reginfo.c (record_subregs_of_mode): Likewise.
14369 * rtlanal.c (read_modify_subreg_p): Likewise.
14371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14372 Alan Hayward <alan.hayward@arm.com>
14373 David Sherwood <david.sherwood@arm.com>
14375 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14376 numbers of elements.
14378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14379 Alan Hayward <alan.hayward@arm.com>
14380 David Sherwood <david.sherwood@arm.com>
14382 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14389 in a POINTER_PLUS_EXPR.
14391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14392 Alan Hayward <alan.hayward@arm.com>
14393 David Sherwood <david.sherwood@arm.com>
14395 * omp-simd-clone.c (simd_clone_subparts): New function.
14396 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14397 (ipa_simd_modify_function_body): Likewise.
14399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14400 Alan Hayward <alan.hayward@arm.com>
14401 David Sherwood <david.sherwood@arm.com>
14403 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14404 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14405 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14406 (expand_vector_condition, vector_element): Likewise.
14407 (subparts_gt): New function.
14408 (get_compute_type): Use subparts_gt.
14409 (count_type_subparts): Delete.
14410 (expand_vector_operations_1): Use subparts_gt instead of
14411 count_type_subparts.
14413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14414 Alan Hayward <alan.hayward@arm.com>
14415 David Sherwood <david.sherwood@arm.com>
14417 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14418 (vect_compile_time_alias): ...this new function. Do the calculation
14419 on poly_ints rather than trees.
14420 (vect_prune_runtime_alias_test_list): Update call accordingly.
14422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14423 Alan Hayward <alan.hayward@arm.com>
14424 David Sherwood <david.sherwood@arm.com>
14426 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14428 (vect_schedule_slp_instance): Likewise.
14430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14431 Alan Hayward <alan.hayward@arm.com>
14432 David Sherwood <david.sherwood@arm.com>
14434 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14435 constant and extern definitions for variable-length vectors.
14436 (vect_get_constant_vectors): Note that the number of units
14437 is known to be constant.
14439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14440 Alan Hayward <alan.hayward@arm.com>
14441 David Sherwood <david.sherwood@arm.com>
14443 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14444 of units as polynomial. Choose between WIDE and NARROW based
14447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14448 Alan Hayward <alan.hayward@arm.com>
14449 David Sherwood <david.sherwood@arm.com>
14451 * tree-vect-stmts.c (simd_clone_subparts): New function.
14452 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14455 Alan Hayward <alan.hayward@arm.com>
14456 David Sherwood <david.sherwood@arm.com>
14458 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14459 vectors as polynomial. Use build_index_vector for
14460 IFN_GOMP_SIMD_LANE.
14462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14463 Alan Hayward <alan.hayward@arm.com>
14464 David Sherwood <david.sherwood@arm.com>
14466 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14467 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14468 for variable-length vectors.
14469 (vectorizable_mask_load_store): Treat the number of units as
14470 polynomial, asserting that it is constant if the condition has
14471 already been enforced.
14472 (vectorizable_store, vectorizable_load): Likewise.
14474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14475 Alan Hayward <alan.hayward@arm.com>
14476 David Sherwood <david.sherwood@arm.com>
14478 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14479 of units as polynomial. Punt if we can't tell at compile time
14480 which vector contains the final result.
14482 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14483 Alan Hayward <alan.hayward@arm.com>
14484 David Sherwood <david.sherwood@arm.com>
14486 * tree-vect-loop.c (vectorizable_induction): Treat the number
14487 of units as polynomial. Punt on SLP inductions. Use an integer
14488 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14489 cast of such a series for variable-length floating-point
14492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14493 Alan Hayward <alan.hayward@arm.com>
14494 David Sherwood <david.sherwood@arm.com>
14496 * tree.h (build_index_vector): Declare.
14497 * tree.c (build_index_vector): New function.
14498 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14499 of units as polynomial, forcibly converting it to a constant if
14500 vectorizable_reduction has already enforced the condition.
14501 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14502 to create a {1,2,3,...} vector.
14503 (vectorizable_reduction): Treat the number of units as polynomial.
14504 Choose vectype_in based on the largest scalar element size rather
14505 than the smallest number of units. Enforce the restrictions
14508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14509 Alan Hayward <alan.hayward@arm.com>
14510 David Sherwood <david.sherwood@arm.com>
14512 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14513 number of units as polynomial.
14515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14516 Alan Hayward <alan.hayward@arm.com>
14517 David Sherwood <david.sherwood@arm.com>
14519 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14520 * target.def (autovectorize_vector_sizes): Return the vector sizes
14521 by pointer, using vector_sizes rather than a bitmask.
14522 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14523 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14524 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14526 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14527 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14528 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14529 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14530 * omp-general.c (omp_max_vf): Likewise.
14531 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14532 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14533 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14534 * tree-vect-slp.c (vect_slp_bb): Likewise.
14535 * doc/tm.texi: Regenerate.
14536 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14538 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14539 the vector size as a poly_uint64 rather than an unsigned int.
14540 (current_vector_size): Change from an unsigned int to a poly_uint64.
14541 (get_vectype_for_scalar_type): Update accordingly.
14542 * tree.h (build_truth_vector_type): Take the size and number of
14543 units as a poly_uint64 rather than an unsigned int.
14544 (build_vector_type): Add a temporary overload that takes
14545 the number of units as a poly_uint64 rather than an unsigned int.
14546 * tree.c (make_vector_type): Likewise.
14547 (build_truth_vector_type): Take the number of units as a poly_uint64
14548 rather than an unsigned int.
14550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14551 Alan Hayward <alan.hayward@arm.com>
14552 David Sherwood <david.sherwood@arm.com>
14554 * target.def (get_mask_mode): Take the number of units and length
14555 as poly_uint64s rather than unsigned ints.
14556 * targhooks.h (default_get_mask_mode): Update accordingly.
14557 * targhooks.c (default_get_mask_mode): Likewise.
14558 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14559 * doc/tm.texi: Regenerate.
14561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14562 Alan Hayward <alan.hayward@arm.com>
14563 David Sherwood <david.sherwood@arm.com>
14565 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14566 * omp-general.c (omp_max_vf): Likewise.
14567 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14568 (expand_omp_simd): Handle polynomial safelen.
14569 * omp-low.c (omplow_simd_context): Add a default constructor.
14570 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14571 (lower_rec_simd_input_clauses): Update accordingly.
14572 (lower_rec_input_clauses): Likewise.
14574 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14575 Alan Hayward <alan.hayward@arm.com>
14576 David Sherwood <david.sherwood@arm.com>
14578 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14579 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14580 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14581 (vect_analyze_slp_cost): Likewise.
14582 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14583 (vect_model_load_cost): Likewise.
14585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14586 Alan Hayward <alan.hayward@arm.com>
14587 David Sherwood <david.sherwood@arm.com>
14589 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14590 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14591 from an unsigned int * to a poly_uint64_pod *.
14592 (calculate_unrolling_factor): New function.
14593 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14596 Alan Hayward <alan.hayward@arm.com>
14597 David Sherwood <david.sherwood@arm.com>
14599 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14600 from an unsigned int to a poly_uint64.
14601 (_loop_vec_info::slp_unrolling_factor): Likewise.
14602 (_loop_vec_info::vectorization_factor): Change from an int
14604 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14605 (vect_get_num_vectors): New function.
14606 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14607 (vect_get_num_copies): Use vect_get_num_vectors.
14608 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14609 to an unsigned int *.
14610 (vect_analyze_data_refs): Change min_vf from an int * to a
14612 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14613 than an unsigned HOST_WIDE_INT.
14614 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14615 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14616 to an unsigned int *.
14617 (vect_analyze_data_ref_dependences): Likewise.
14618 (vect_compute_data_ref_alignment): Handle polynomial vf.
14619 (vect_enhance_data_refs_alignment): Likewise.
14620 (vect_prune_runtime_alias_test_list): Likewise.
14621 (vect_shift_permute_load_chain): Likewise.
14622 (vect_supportable_dr_alignment): Likewise.
14623 (dependence_distance_ge_vf): Take the vectorization factor as a
14624 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14625 (vect_analyze_data_refs): Change min_vf from an int * to a
14627 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14628 vfm1 as a poly_uint64 rather than an int. Make the same change
14629 for the returned bound_scalar.
14630 (vect_gen_vector_loop_niters): Handle polynomial vf.
14631 (vect_do_peeling): Likewise. Update call to
14632 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14633 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14635 * tree-vect-loop.c (vect_determine_vectorization_factor)
14636 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14637 (vect_get_known_peeling_cost): Likewise.
14638 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14639 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14640 (vect_transform_loop): Likewise. Use the lowest possible VF when
14641 updating the upper bounds of the loop.
14642 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14643 rather than an int.
14644 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14645 polynomial unroll factors.
14646 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14647 (vect_make_slp_decision): Likewise.
14648 (vect_supported_load_permutation_p): Likewise, and polynomial
14650 (vect_analyze_slp_cost): Handle polynomial vf.
14651 (vect_slp_analyze_node_operations): Likewise.
14652 (vect_slp_analyze_bb_1): Likewise.
14653 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14654 than an unsigned HOST_WIDE_INT.
14655 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14656 (vectorizable_load): Handle polynomial vf.
14657 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14659 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14662 Alan Hayward <alan.hayward@arm.com>
14663 David Sherwood <david.sherwood@arm.com>
14665 * match.pd: Handle bit operations involving three constants
14666 and try to fold one pair.
14668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14670 * tree-vect-loop-manip.c: Include gimple-fold.h.
14671 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14672 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14673 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14674 Add a path that uses a step of VF instead of 1, but disable it
14676 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14677 and niters_no_overflow parameters. Update calls to
14678 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14679 Create a new SSA name if the latter choses to use a ste other
14680 than zero, and return it via niters_vector_mult_vf_var.
14681 * tree-vect-loop.c (vect_transform_loop): Update calls to
14682 vect_do_peeling, vect_gen_vector_loop_niters and
14683 slpeel_make_loop_iterate_ntimes.
14684 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14685 (vect_gen_vector_loop_niters): Update declarations after above changes.
14687 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14689 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14690 128-bit round to integer instructions.
14691 (ceil<mode>2): Likewise.
14692 (btrunc<mode>2): Likewise.
14693 (round<mode>2): Likewise.
14695 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14697 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14698 unaligned VSX load/store on P8/P9.
14699 (expand_block_clear): Allow the use of unaligned VSX
14700 load/store on P8/P9.
14702 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14704 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14706 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14707 swap associated with both a load and a store.
14709 2018-01-02 Andrew Waterman <andrew@sifive.com>
14711 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14712 * config/riscv/riscv.md (clear_cache): Use it.
14714 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14716 * web.c: Remove out-of-date comment.
14718 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14720 * expr.c (fixup_args_size_notes): Check that any existing
14721 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14722 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14723 (emit_single_push_insn): ...here.
14725 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14727 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14728 (const_vector_encoded_nelts): New function.
14729 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14730 (const_vector_int_elt, const_vector_elt): Declare.
14731 * emit-rtl.c (const_vector_int_elt_1): New function.
14732 (const_vector_elt): Likewise.
14733 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14734 of CONST_VECTOR_ELT.
14736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14738 * expr.c: Include rtx-vector-builder.h.
14739 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14740 directly on the tree encoding.
14741 (const_vector_from_tree): Likewise.
14742 * optabs.c: Include rtx-vector-builder.h.
14743 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14744 sequence of "u" values.
14745 * vec-perm-indices.c: Include rtx-vector-builder.h.
14746 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14747 directly on the vec_perm_indices encoding.
14749 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14751 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14752 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14753 * rtx-vector-builder.h: New file.
14754 * rtx-vector-builder.c: Likewise.
14755 * rtl.h (rtx_def::u2): Add a const_vector field.
14756 (CONST_VECTOR_NPATTERNS): New macro.
14757 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14758 (CONST_VECTOR_DUPLICATE_P): Likewise.
14759 (CONST_VECTOR_STEPPED_P): Likewise.
14760 (CONST_VECTOR_ENCODED_ELT): Likewise.
14761 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14762 (unwrap_const_vec_duplicate): Likewise.
14763 (const_vec_series_p): Check for a non-duplicated vector encoding.
14764 Say that the function only returns true for integer vectors.
14765 * emit-rtl.c: Include rtx-vector-builder.h.
14766 (gen_const_vec_duplicate_1): Delete.
14767 (gen_const_vector): Call gen_const_vec_duplicate instead of
14768 gen_const_vec_duplicate_1.
14769 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14770 (gen_const_vec_duplicate): Use rtx_vector_builder.
14771 (gen_const_vec_series): Likewise.
14772 (gen_rtx_CONST_VECTOR): Likewise.
14773 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14774 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14775 Build a new vector rather than modifying a CONST_VECTOR in-place.
14776 (handle_special_swappables): Update call accordingly.
14777 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14778 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14779 Build a new vector rather than modifying a CONST_VECTOR in-place.
14780 (handle_special_swappables): Update call accordingly.
14782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14784 * simplify-rtx.c (simplify_const_binary_operation): Use
14785 CONST_VECTOR_ELT instead of XVECEXP.
14787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14789 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14790 the selector elements to be different from the data elements
14791 if the selector is a VECTOR_CST.
14792 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14793 ssizetype for the selector.
14795 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14797 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14798 before testing each element individually.
14799 * tree-vect-generic.c (lower_vec_perm): Likewise.
14801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14803 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14804 * selftest-run-tests.c (selftest::run_tests): Call it.
14805 * vector-builder.h (vector_builder::operator ==): New function.
14806 (vector_builder::operator !=): Likewise.
14807 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14808 (vec_perm_indices::all_from_input_p): New function.
14809 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14810 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14811 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14812 instead of reading the VECTOR_CST directly. Detect whether both
14813 vector inputs are the same before constructing the vec_perm_indices,
14814 and update the number of inputs argument accordingly. Use the
14815 utility functions added above. Only construct sel2 if we need to.
14817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14819 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14820 the broadcast of the low byte.
14821 (expand_mult_highpart): Use an explicit encoding for the permutes.
14822 * optabs-query.c (can_mult_highpart_p): Likewise.
14823 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14824 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14825 (vectorizable_bswap): Likewise.
14826 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14827 explicit encoding for the power-of-2 permutes.
14828 (vect_permute_store_chain): Likewise.
14829 (vect_grouped_load_supported): Likewise.
14830 (vect_permute_load_chain): Likewise.
14832 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14834 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14835 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14836 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14837 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14838 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14839 (vect_gen_perm_mask_any): Likewise.
14841 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14843 * int-vector-builder.h: New file.
14844 * vec-perm-indices.h: Include int-vector-builder.h.
14845 (vec_perm_indices): Redefine as an int_vector_builder.
14846 (auto_vec_perm_indices): Delete.
14847 (vec_perm_builder): Redefine as a stand-alone class.
14848 (vec_perm_indices::vec_perm_indices): New function.
14849 (vec_perm_indices::clamp): Likewise.
14850 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14851 (vec_perm_indices::new_vector): New function.
14852 (vec_perm_indices::new_expanded_vector): Update for new
14853 vec_perm_indices class.
14854 (vec_perm_indices::rotate_inputs): New function.
14855 (vec_perm_indices::all_in_range_p): Operate directly on the
14856 encoded form, without computing elided elements.
14857 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14858 encoding. Update for new vec_perm_indices class.
14859 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14860 the given vec_perm_builder.
14861 (expand_vec_perm_var): Update vec_perm_builder constructor.
14862 (expand_mult_highpart): Use vec_perm_builder instead of
14863 auto_vec_perm_indices.
14864 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14865 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14866 or double series encoding as appropriate.
14867 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14868 vec_perm_indices instead of auto_vec_perm_indices.
14869 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14870 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14871 (vect_permute_store_chain): Likewise.
14872 (vect_grouped_load_supported): Likewise.
14873 (vect_permute_load_chain): Likewise.
14874 (vect_shift_permute_load_chain): Likewise.
14875 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14876 (vect_transform_slp_perm_load): Likewise.
14877 (vect_schedule_slp_instance): Likewise.
14878 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14879 (vectorizable_mask_load_store): Likewise.
14880 (vectorizable_bswap): Likewise.
14881 (vectorizable_store): Likewise.
14882 (vectorizable_load): Likewise.
14883 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14884 vec_perm_indices instead of auto_vec_perm_indices. Use
14885 tree_to_vec_perm_builder to read the vector from a tree.
14886 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14887 vec_perm_builder instead of a vec_perm_indices.
14888 (have_whole_vector_shift): Use vec_perm_builder and
14889 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14890 truncation to calc_vec_perm_mask_for_shift.
14891 (vect_create_epilog_for_reduction): Likewise.
14892 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14893 from auto_vec_perm_indices to vec_perm_indices.
14894 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14895 instead of changing individual elements.
14896 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14897 the vector in d.perm.
14898 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14899 from auto_vec_perm_indices to vec_perm_indices.
14900 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14901 instead of changing individual elements.
14902 (arm_vectorize_vec_perm_const): Use new_vector to install
14903 the vector in d.perm.
14904 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14905 Update vec_perm_builder constructor.
14906 (rs6000_expand_interleave): Likewise.
14907 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14908 (rs6000_expand_interleave): Likewise.
14910 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14912 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14913 to qimode could truncate the indices.
14914 * optabs.c (expand_vec_perm_var): Likewise.
14916 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14918 * Makefile.in (OBJS): Add vec-perm-indices.o.
14919 * vec-perm-indices.h: New file.
14920 * vec-perm-indices.c: Likewise.
14921 * target.h (vec_perm_indices): Replace with a forward class
14923 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14924 * optabs.h: Include vec-perm-indices.h.
14925 (expand_vec_perm): Delete.
14926 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14927 (expand_vec_perm_const): Declare.
14928 * target.def (vec_perm_const_ok): Replace with...
14929 (vec_perm_const): ...this new hook.
14930 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14931 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14932 * doc/tm.texi: Regenerate.
14933 * optabs.def (vec_perm_const): Delete.
14934 * doc/md.texi (vec_perm_const): Likewise.
14935 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14936 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14937 expand_vec_perm for constant permutation vectors. Assert that
14938 the mode of variable permutation vectors is the integer equivalent
14939 of the mode that is being permuted.
14940 * optabs-query.h (selector_fits_mode_p): Declare.
14941 * optabs-query.c: Include vec-perm-indices.h.
14942 (selector_fits_mode_p): New function.
14943 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14944 is defined, instead of checking whether the vec_perm_const_optab
14945 exists. Use targetm.vectorize.vec_perm_const instead of
14946 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14947 fit in the vector mode before using a variable permute.
14948 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14949 vec_perm_indices instead of an rtx.
14950 (expand_vec_perm): Replace with...
14951 (expand_vec_perm_const): ...this new function. Take the selector
14952 as a vec_perm_indices rather than an rtx. Also take the mode of
14953 the selector. Update call to shift_amt_for_vec_perm_mask.
14954 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14955 Use vec_perm_indices::new_expanded_vector to expand the original
14956 selector into bytes. Check whether the indices fit in the vector
14957 mode before using a variable permute.
14958 (expand_vec_perm_var): Make global.
14959 (expand_mult_highpart): Use expand_vec_perm_const.
14960 * fold-const.c: Includes vec-perm-indices.h.
14961 * tree-ssa-forwprop.c: Likewise.
14962 * tree-vect-data-refs.c: Likewise.
14963 * tree-vect-generic.c: Likewise.
14964 * tree-vect-loop.c: Likewise.
14965 * tree-vect-slp.c: Likewise.
14966 * tree-vect-stmts.c: Likewise.
14967 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14969 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14970 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14971 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14972 (aarch64_vectorize_vec_perm_const): ...this new function.
14973 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14974 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14975 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14976 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14977 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14978 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14979 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14981 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14982 check for NEON modes.
14983 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14984 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14985 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14986 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14988 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14989 the old VEC_PERM_CONST conditions.
14990 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14991 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14992 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14993 (ia64_vectorize_vec_perm_const_ok): Merge into...
14994 (ia64_vectorize_vec_perm_const): ...this new function.
14995 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14996 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14997 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14998 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14999 * config/mips/mips.c (mips_expand_vec_perm_const)
15000 (mips_vectorize_vec_perm_const_ok): Merge into...
15001 (mips_vectorize_vec_perm_const): ...this new function.
15002 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
15003 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
15004 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
15005 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
15006 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
15007 (rs6000_expand_vec_perm_const): Delete.
15008 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
15010 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15011 (altivec_expand_vec_perm_const_le): Take each operand individually.
15012 Operate on constant selectors rather than rtxes.
15013 (altivec_expand_vec_perm_const): Likewise. Update call to
15014 altivec_expand_vec_perm_const_le.
15015 (rs6000_expand_vec_perm_const): Delete.
15016 (rs6000_vectorize_vec_perm_const_ok): Delete.
15017 (rs6000_vectorize_vec_perm_const): New function.
15018 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15019 an element count and rtx array.
15020 (rs6000_expand_extract_even): Update call accordingly.
15021 (rs6000_expand_interleave): Likewise.
15022 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
15023 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
15024 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
15025 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
15026 (rs6000_expand_vec_perm_const): Delete.
15027 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
15028 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15029 (altivec_expand_vec_perm_const_le): Take each operand individually.
15030 Operate on constant selectors rather than rtxes.
15031 (altivec_expand_vec_perm_const): Likewise. Update call to
15032 altivec_expand_vec_perm_const_le.
15033 (rs6000_expand_vec_perm_const): Delete.
15034 (rs6000_vectorize_vec_perm_const_ok): Delete.
15035 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15036 reference to the SPE evmerge intructions.
15037 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15038 an element count and rtx array.
15039 (rs6000_expand_extract_even): Update call accordingly.
15040 (rs6000_expand_interleave): Likewise.
15041 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15042 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15044 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15048 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15049 vector mode and that that mode matches the mode of the data
15051 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15052 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15053 directly using expand_vec_perm_1 when forcing selectors into
15055 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15057 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15059 * optabs-query.h (can_vec_perm_p): Delete.
15060 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15061 * optabs-query.c (can_vec_perm_p): Split into...
15062 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15063 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15064 particular selector is valid.
15065 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15066 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15067 (vect_grouped_load_supported): Likewise.
15068 (vect_shift_permute_load_chain): Likewise.
15069 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15070 (vect_transform_slp_perm_load): Likewise.
15071 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15072 (vectorizable_bswap): Likewise.
15073 (vect_gen_perm_mask_checked): Likewise.
15074 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15075 implementations of variable permutation vectors into account
15076 when deciding which selector to use.
15077 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15078 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15079 with a false third argument.
15080 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15081 to test whether the constant selector is valid and can_vec_perm_var_p
15082 to test whether a variable selector is valid.
15084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15086 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15087 * optabs-query.c (can_vec_perm_p): Likewise.
15088 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15089 instead of vec_perm_indices.
15090 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15091 (vect_gen_perm_mask_checked): Likewise,
15092 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15093 (vect_gen_perm_mask_checked): Likewise,
15095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15097 * optabs-query.h (qimode_for_vec_perm): Declare.
15098 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15099 (qimode_for_vec_perm): ...this new function.
15100 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15102 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15104 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15105 does not have a conditional at the top.
15107 2018-01-02 Richard Biener <rguenther@suse.de>
15109 * ipa-inline.c (big_speedup_p): Fix expression.
15111 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15114 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15117 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15121 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15122 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15123 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15124 cond_taken_branch_cost 3->4.
15126 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15128 PR tree-optimization/83581
15129 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15130 TODO_cleanup_cfg if any changes have been made.
15132 PR middle-end/83608
15133 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15134 convert_modes if target mode has the right side, but different mode
15137 PR middle-end/83609
15138 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15139 last argument when extracting from CONCAT. If either from_real or
15140 from_imag is NULL, use expansion through memory. If result is not
15141 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15142 the parts directly to inner mode, if even that fails, use expansion
15145 PR middle-end/83623
15146 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15147 check for bswap in mode rather than HImode and use that in expand_unop
15150 Copyright (C) 2018 Free Software Foundation, Inc.
15152 Copying and distribution of this file, with or without modification,
15153 are permitted in any medium without royalty provided the copyright
15154 notice and this notice are preserved.