1 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
6 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
8 PR tree-optimization/86034
9 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10 the unsigned bitfield type in a bit insertion sequence if it does not
11 have a larger precision than the bitfield size.
12 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
14 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
16 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
18 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
20 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
21 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
22 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
23 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
25 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
27 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
28 Disable -fdelete-null-pointer-checks for ELF toolchain.
30 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
31 Kito Cheng <kito.cheng@gmail.com>
33 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
34 (nds32le-*-*, nds32be-*-*): Integrate checking process.
35 (nds32*-*-*): Add glibc and uclibc conditions.
36 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
37 (TARGET_EXCEPT_UNWIND_INFO): Define.
38 * config/nds32/elf.h: New file.
39 * config/nds32/linux.h: New file.
40 * config/nds32/nds32-elf.opt: New file.
41 * config/nds32/nds32-linux.opt: New file.
42 * config/nds32/nds32-fp-as-gp.c
43 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
44 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
46 (nds32_asm_file_end): Ditto.
47 (nds32_print_operand): Ditto.
48 (nds32_insert_attributes): Ditto.
49 (nds32_init_libfuncs): New function.
50 (TARGET_HAVE_TLS): Define.
51 (TARGET_INIT_LIBFUNCS): Define.
52 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
54 (TARGET_ELF): Apply different mcmodel setting.
55 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
56 been migrated into elf.h and linux.h files.
57 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
58 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
59 (mcmodel): The content has been migrated into nds32-elf.opt and
60 nds32-linux.opt files.
61 * config/nds32/t-elf: New file.
62 * config/nds32/t-linux: New file.
64 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
65 Shiva Chen <shiva0217@gmail.com>
67 * config/nds32/constants.md (unspec_volatile_element): Add
68 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
69 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
71 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
72 (make_pass_nds32_fp_as_gp): Declare.
73 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
75 (nds32_asm_function_end_prologue): Remove unused asm output.
76 (nds32_asm_function_begin_epilogue): Remove unused asm output.
77 (nds32_asm_file_start): Output necessary fp_as_gp information.
78 (nds32_option_override): Adjust register usage.
79 (nds32_expand_prologue): Consider fp_as_gp situation.
80 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
81 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
84 (simple_return): Ditto.
85 (omit_fp_begin): Output special directive for fp_as_gp.
86 (omit_fp_end): Output special directive for fp_as_gp.
87 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
88 mforbid-fp-as-gp): New options.
90 2018-06-01 Mark Wielaard <mark@klomp.org>
92 * dwarf2out.c (dwarf2out_finish): Remove generation of
95 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
97 * gimple-ssa-store-merging.c: Include gimple-fold.h.
98 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
99 (struct merged_store_group): Add bit_insertion field.
100 (dump_char_array): Use standard hexadecimal format.
101 (merged_store_group::merged_store_group): Set bit_insertion to false.
102 (merged_store_group::apply_stores): Use optimal buffer size. Deal
103 with BIT_INSERT_EXPR stores. Move up code updating the mask and
104 also print the mask in the dump file.
105 (pass_store_merging::gate): Minor tweak.
106 (imm_store_chain_info::coalesce_immediate): Fix wrong association
107 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
108 stores with INTEGER_CST stores.
109 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
110 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
111 and use it throughout. Generate bit insertion sequences if need be.
112 (pass_store_merging::process_store): Remove redundant condition.
113 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
115 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
117 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
118 the 128-bit floating point types. Fix function comment.
120 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
122 * config/aarch64/aarch64-simd.md
123 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
125 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
128 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
130 PR tree-optimization/85989
131 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
133 (backprop::intersect_uses): Check it when deciding whether this
134 is a backedge reference.
135 (backprop::process_block): Add each phi to m_visited_phis
136 after visiting it, then clear it at the end.
138 2018-06-01 Richard Biener <rguenther@suse.de>
140 * tree-vectorizer.h (vect_dr_stmt): New function.
141 (vect_get_load_cost): Adjust.
142 (vect_get_store_cost): Likewise.
143 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
144 Use vect_dr_stmt instead of DR_SMTT.
145 (vect_record_base_alignments): Likewise.
146 (vect_calculate_target_alignment): Likewise.
147 (vect_compute_data_ref_alignment): Likewise and make static.
148 (vect_update_misalignment_for_peel): Likewise.
149 (vect_verify_datarefs_alignment): Likewise.
150 (vector_alignment_reachable_p): Likewise.
151 (vect_get_data_access_cost): Likewise. Pass down
152 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
153 (vect_get_peeling_costs_all_drs): Likewise.
154 (vect_peeling_hash_get_lowest_cost): Likewise.
155 (vect_enhance_data_refs_alignment): Likewise.
156 (vect_find_same_alignment_drs): Likewise.
157 (vect_analyze_data_refs_alignment): Likewise.
158 (vect_analyze_group_access_1): Likewise.
159 (vect_analyze_group_access): Likewise.
160 (vect_analyze_data_ref_access): Likewise.
161 (vect_analyze_data_ref_accesses): Likewise.
162 (vect_vfa_segment_size): Likewise.
163 (vect_small_gap_p): Likewise.
164 (vectorizable_with_step_bound_p): Likewise.
165 (vect_prune_runtime_alias_test_list): Likewise.
166 (vect_analyze_data_refs): Likewise.
167 (vect_supportable_dr_alignment): Likewise.
168 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
169 (vect_gen_prolog_loop_niters): Likewise.
170 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
171 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
173 (vect_recog_mask_conversion_pattern): Likewise.
174 (vect_try_gather_scatter_pattern): Likewise.
175 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
176 to vect_get_store_cost.
177 (vect_get_store_cost): Get stmt_info instead of DR.
178 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
179 (vect_get_load_cost): Get stmt_info instead of DR.
181 2018-06-01 Richard Biener <rguenther@suse.de>
184 * gimple-fold.c (var_decl_component_p): Also allow offsetted
185 vars wrapped in MEM_REFs.
187 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
189 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
190 Fix subreg tests so that we only return a choice between
191 GENERAL_REGS and FP_REGS if the original classes included both.
193 2018-06-01 Richard Biener <rguenther@suse.de>
196 * tree-ssa-structalias.c (get_function_part_constraint):
197 Handle NULL fi->decl.
198 (find_func_aliases_for_call): Properly handle indirect
200 (find_func_clobbers): Likewise.
201 (ipa_pta_execute): Likewise.
202 (create_variable_info_for): For functions that are ifunc_resolver
203 resolve to a varinfo that contains the result of the resolver
205 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
208 2018-05-31 Michael Collison <michael.collison@arm.com>
210 * config/aarch64/aarch64.md:
211 (*fix_to_zero_extenddfdi2): New pattern.
212 * gcc.target/aarch64/fix_extend1.c: New testcase.
214 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
218 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
219 and BUILT_IN_STRNCMP_EQ.
220 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
222 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
223 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
224 (gimple_fold_builtin): Likewise.
225 * tree-ssa-strlen.c (compute_string_length): New function.
226 (determine_min_obsize): New function.
227 (handle_builtin_string_cmp): New function to handle calls to
228 string compare functions.
229 (strlen_optimize_stmt): Add handling to builtin string compare
231 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
232 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
233 * tree.c (build_common_builtin_nodes): Add new defines of
234 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
236 2018-05-31 Jakub Jelinek <jakub@redhat.com>
239 * bb-reorder.c (pass_partition_blocks::gate): Return false for
240 functions with naked attribute.
242 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
244 * config/i386/sse.md (avx_vec_concat<mode>):
245 Substitute concat_tg_mode mode attribute with xtg_mode.
246 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
247 (concat_tg_mode): Remove mode attribute.
249 2018-05-31 Martin Sebor <msebor@redhat.com>
252 * calls.c (alloc_max_size): Correct a logic error/typo.
253 Treat excessive arguments as infinite. Warn for invalid arguments.
254 * doc/invoke.texi (-Walloc-size-larger-than): Update.
256 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
259 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
260 and movx for Haswell.
262 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
263 Cesar Philippidis <cesar@codesourcery.com>
266 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
267 when emitting error on private/firstprivate reductions.
268 * omp-low.c (lower_omp_target): Avoid reference-type processing
269 on pointers for firstprivate clause.
271 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
273 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
276 * config/aarch64/aarch64-simd.md
277 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
278 (aarch64_ld1_x3_<mode>): Likewise
279 (aarch64_st1x2<VALLDIF:mode>): Likewise
280 (aarch64_st1_x2_<mode>): Likewise
281 (aarch64_st1x3<VALLDIF:mode>): Likewise
282 (aarch64_st1_x3_<mode>): Likewise
283 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
284 (vld1_s8_x3): Likewise.
285 (vld1_u16_x3): Likewise.
286 (vld1_s16_x3): Likewise.
287 (vld1_u32_x3): Likewise.
288 (vld1_s32_x3): Likewise.
289 (vld1_u64_x3): Likewise.
290 (vld1_s64_x3): Likewise.
291 (vld1_f16_x3): Likewise.
292 (vld1_f32_x3): Likewise.
293 (vld1_f64_x3): Likewise.
294 (vld1_p8_x3): Likewise.
295 (vld1_p16_x3): Likewise.
296 (vld1_p64_x3): Likewise.
297 (vld1q_u8_x3): Likewise.
298 (vld1q_s8_x3): Likewise.
299 (vld1q_u16_x3): Likewise.
300 (vld1q_s16_x3): Likewise.
301 (vld1q_u32_x3): Likewise.
302 (vld1q_s32_x3): Likewise.
303 (vld1q_u64_x3): Likewise.
304 (vld1q_s64_x3): Likewise.
305 (vld1q_f16_x3): Likewise.
306 (vld1q_f32_x3): Likewise.
307 (vld1q_f64_x3): Likewise.
308 (vld1q_p8_x3): Likewise.
309 (vld1q_p16_x3): Likewise.
310 (vld1q_p64_x3): Likewise.
311 (vst1_s64_x2): Likewise.
312 (vst1_u64_x2): Likewise.
313 (vst1_f64_x2): Likewise.
314 (vst1_s8_x2): Likewise.
315 (vst1_p8_x2): Likewise.
316 (vst1_s16_x2): Likewise.
317 (vst1_p16_x2): Likewise.
318 (vst1_s32_x2): Likewise.
319 (vst1_u8_x2): Likewise.
320 (vst1_u16_x2): Likewise.
321 (vst1_u32_x2): Likewise.
322 (vst1_f16_x2): Likewise.
323 (vst1_f32_x2): Likewise.
324 (vst1_p64_x2): Likewise.
325 (vst1q_s8_x2): Likewise.
326 (vst1q_p8_x2): Likewise.
327 (vst1q_s16_x2): Likewise.
328 (vst1q_p16_x2): Likewise.
329 (vst1q_s32_x2): Likewise.
330 (vst1q_s64_x2): Likewise.
331 (vst1q_u8_x2): Likewise.
332 (vst1q_u16_x2): Likewise.
333 (vst1q_u32_x2): Likewise.
334 (vst1q_u64_x2): Likewise.
335 (vst1q_f16_x2): Likewise.
336 (vst1q_f32_x2): Likewise.
337 (vst1q_f64_x2): Likewise.
338 (vst1q_p64_x2): Likewise.
339 (vst1_s64_x3): Likewise.
340 (vst1_u64_x3): Likewise.
341 (vst1_f64_x3): Likewise.
342 (vst1_s8_x3): Likewise.
343 (vst1_p8_x3): Likewise.
344 (vst1_s16_x3): Likewise.
345 (vst1_p16_x3): Likewise.
346 (vst1_s32_x3): Likewise.
347 (vst1_u8_x3): Likewise.
348 (vst1_u16_x3): Likewise.
349 (vst1_u32_x3): Likewise.
350 (vst1_f16_x3): Likewise.
351 (vst1_f32_x3): Likewise.
352 (vst1_p64_x3): Likewise.
353 (vst1q_s8_x3): Likewise.
354 (vst1q_p8_x3): Likewise.
355 (vst1q_s16_x3): Likewise.
356 (vst1q_p16_x3): Likewise.
357 (vst1q_s32_x3): Likewise.
358 (vst1q_s64_x3): Likewise.
359 (vst1q_u8_x3): Likewise.
360 (vst1q_u16_x3): Likewise.
361 (vst1q_u32_x3): Likewise.
362 (vst1q_u64_x3): Likewise.
363 (vst1q_f16_x3): Likewise.
364 (vst1q_f32_x3): Likewise.
365 (vst1q_f64_x3): Likewise.
366 (vst1q_p64_x3): Likewise.
368 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
370 * config/msp430/msp430.c (msp430_output_labelref): Prepend
371 user_label_prefix to name.
373 * tree-core.h: Update comment about the format of NAME string
374 passed to handler in attribute_spec.
376 * config/msp430/msp430.md: Remove erroneous subreg expression from
377 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
378 zero_extend{q,h}isi2.
380 2018-05-30 Borislav Petkov <bp@suse.de>
382 * doc/extend.texi: Document some architecture specific
383 constraints and sort entries.
385 2018-05-30 Martin Sebor <msebor@redhat.com>
388 * builtins.c (expand_builtin_stpcpy_1): New function.
389 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
390 only if the former succeeds.
392 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
394 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
397 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
399 * doc/invoke.texi (-flinker-output): Document
401 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
403 * passes.c (ipa_write_summaries): Only modify statements if body
405 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
406 incrementally linking.
407 (ipa_passes): Likewise.
408 * lto-cgraph.c (lto_output_node): When incrementally linking do not
409 pass down resolution info.
410 * common.opt (flag_incremental_link): Update info.
411 * gcc.c (plugin specs): Turn flinker-output=* to
412 -plugin-opt=-linker-output-known
413 * toplev.c (compile_file): Also cut compilation when doing incremental
415 * flag-types. (enum lto_partition_model): Add
416 LTO_LINKER_OUTPUT_NOLTOREL.
417 (invoke.texi): Add -flinker-output docs.
418 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
419 link same way as WPA; do not stream in dead initializers.
421 * dwarf2out.c (dwarf2out_die_ref_for_decl,
422 darf2out_register_external_decl): Support incremental link.
424 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
426 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
428 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
430 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
431 it down to simple_object_copy_lto_debug_sections.
432 (run_gcc): Determine incremental LTO link time and configure
433 lto1 into non-wpa mode, disable renaming of debug sections.
435 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
437 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
438 descriptions of various incorrectly documented functions.
440 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
443 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
444 address check not strict.
446 2018-05-30 Richard Biener <rguenther@suse.de>
448 PR tree-optimization/85964
449 * tracer.c (better_p): Drop initialized count check, we only
450 call the function with initialized counts now.
451 (find_best_successor): Do find a best edge if one
452 has uninitialized count.
453 (find_best_predecessor): Likewise. Do BB frequency check only
454 if count is initialized.
456 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
458 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
459 (aarch64_ldrstr_offset_compare): New.
460 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
461 load/store orderings.
462 (aarch64_gen_adjusted_ldpstp): Likewise.
464 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
466 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
467 Check for subset of GENERAL_REGS and FP_REGS.
468 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
471 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
473 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
474 and wi::to_poly_offset. Add the current offset and then check
475 whether the sum fits, rather than using an unchecked addition of
476 a checked term. Check for a shwi rather than a uhwi.
477 * expr.c (get_bit_range): Use tree_to_poly_uint64.
478 (store_constructor): Use poly_int_tree_p.
479 (expand_expr_real_1): Likewise.
480 * function.c (assign_temp): Likewise.
481 * fold-const.c (const_binop): Use poly_int_tree_p and
483 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
485 * ipa-icf-gimple.c (func_checker::compare_operand): Use
486 to_poly_offset for MEM offsets.
487 * ipa-icf.c (sem_variable::equals): Likewise.
488 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
489 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
490 wi::to_poly_offset for BIT_FIELD_REF offsets.
491 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
493 * var-tracking.c (emit_note_insn_var_location): Use
496 2018-05-29 Jim Wilson <jimw@sifive.com>
498 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
500 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
503 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
504 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
506 (sse4_1_round<mode>2): Use nonimmediate_operand
507 for operand 1 predicate.
509 2018-05-29 Martin Sebor <msebor@redhat.com>
510 Richard Biener <rguenther@suse.de>
513 * calls.c (get_size_range): Call determine_value_range instead
515 * tree-vrp.h (determine_value_range): Declared new function.
516 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
518 2018-05-29 Richard Biener <rguenther@suse.de>
520 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
521 sure to use non-pattern stmts for get_earlier_stmt arguments.
522 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
523 called on pattern stmts.
524 (get_later_stmt): Likewise.
526 2018-05-29 Martin Liska <mliska@suse.cz>
528 PR gcov-profile/85759
529 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
532 2018-05-29 Jakub Jelinek <jakub@redhat.com>
534 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
536 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
540 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
541 VEC_PACK_FLOAT_EXPR): New tree codes.
542 * tree-pretty-print.c (op_code_prio): Handle
543 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
544 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
545 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
546 * tree-inline.c (estimate_operator_cost): Likewise.
547 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
548 * fold-const.c (const_binop): Likewise.
549 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
550 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
551 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
552 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
553 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
554 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
555 * expr.c (expand_expr_real_2): Likewise.
556 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
557 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
558 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
560 * optabs.c (expand_widen_pattern_expr): For
561 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
562 sign from result type rather than operand's type.
563 (expand_binop_directly): For vec_packu_float_optab and
564 vec_packs_float_optab allow result type to be different from operand's
566 * optabs-tree.c (optab_for_tree_code): Handle
567 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
568 VEC_PACK_FLOAT_EXPR. Formatting fixes.
569 * tree-vect-generic.c (expand_vector_operations_1): Handle
570 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
572 * tree-vect-stmts.c (supportable_widening_operation): Handle
574 (supportable_narrowing_operation): Handle FLOAT_EXPR.
575 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
576 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
577 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
578 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
580 (vec_pack<floatprefix>_float_<mode>): New expander.
581 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
583 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
584 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
585 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
586 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
587 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
589 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
590 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
591 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
592 VEC_PACK_FLOAT_EXPR): Document.
594 2018-05-29 Richard Biener <rguenther@suse.de>
596 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
598 (stmt_vec_info_vec): Make pointer.
599 (init_stmt_vec_info_vec): Remove.
600 (free_stmt_vec_info_vec): Likewise.
601 (set_stmt_vec_info_vec): New function.
602 (free_stmt_vec_infos): Likewise.
603 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
604 (set_vinfo_for_stmt): Likewise.
605 (get_earlier_stmt): Likewise.
606 (get_later_stmt): Likewise.
607 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
608 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
609 (vec_info::~vec_info): Free stmt_vec_infos.
610 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
611 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
612 (pass_slp_vectorize::execute): Likewise.
613 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
614 (free_stmt_vec_info_vec): Likewise.
615 (set_stmt_vec_info_vec): New function.
616 (free_stmt_vec_infos): Likewise.
617 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
618 the global stmt_vec_info_vec.
619 * tree-parloops.c (gather_scalar_reductions): Use
620 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
623 2018-05-29 Richard Biener <rguenther@suse.de>
625 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
627 2018-05-29 Martin Liska <mliska@suse.cz>
628 David Malcolm <dmalcolm@redhat.com>
630 * vec.c (test_reverse): New.
631 (vec_c_tests): Add new test.
632 * vec.h (vl_ptr>::reverse): New function.
634 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
636 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
638 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
641 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
643 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
645 2018-05-28 Richard Biener <rguenther@suse.de>
647 PR tree-optimization/85933
648 * tree-vect-data-refs.c (vect_record_base_alignments): Only
649 look at stmts marked as vectorizable.
651 2018-05-28 Richard Biener <rguenther@suse.de>
653 PR tree-optimization/85934
654 * tree-vect-generic.c (expand_vector_operations_1): Hoist
655 vector boolean check before scalar optimization.
657 2018-05-28 Jakub Jelinek <jakub@redhat.com>
659 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
662 2018-05-28 Mark Wielaard <mark@klomp.org>
664 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
665 if it is an expression containing a minus sign.
667 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
669 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
671 2018-05-27 Paul Koning <ni1d@arrl.net>
673 * config/pdp11/pdp11.md (truncsihi2): Remove.
675 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
676 Chung-Ju Wu <jasonwucj@gmail.com>
678 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
680 (unaligned_store_dw): Ditto.
681 * config/nds32/nds32-memory-manipulation.c
682 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
683 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
684 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
685 (emit_setmem_word_loop): Rename to ...
686 (emit_setmem_doubleword_loop): ... this.
687 (nds32_gen_dup_4_byte_to_word_value): New function.
688 (nds32_gen_dup_8_byte_to_double_word_value): New function.
689 (nds32_expand_setmem_loop): Refine implementation.
690 (nds32_expand_setmem_loop_v3m): Ditto.
691 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
694 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
696 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
698 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
700 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
701 (nds32_init_machine_status): Initialize machine->attr_naked_p and
702 machine->attr_no_prologue_p.
703 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
704 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
705 (nds32_expand_epilogue): Consider attr_naked_p.
706 (nds32_expand_epilogue_v3pop): Likewise.
707 (nds32_can_use_return_insn): Likewise.
708 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
709 attr_no_prologue_p fields.
710 * config/nds32/nds32.opt (mret-in-naked-func): New option.
712 2018-05-27 Jakub Jelinek <jakub@redhat.com>
715 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
718 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
720 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
722 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
724 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
726 (*<floatsuffix>floatv2div2sf2): Rename to ...
727 (*float<floatunssuffix>v2div2sf2): ... this.
728 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
729 (float<floatunssuffix>v2div2sf2_mask): ... this.
730 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
731 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
732 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
734 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
736 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
738 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
740 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
742 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
744 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
745 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
746 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
747 gen_ufix_truncv8dfv8si2.
748 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
749 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
750 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
751 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
752 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
753 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
754 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
755 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
757 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
761 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
763 2018-05-25 Jim Wilson <jimw@sifive.com>
765 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
766 * config/riscv/riscv.c (struct machine_function): Add
767 interrupt_handler_p and attribute_checked_p fields.
768 (riscv_attribute_table): Add interrupt.
769 (riscv_interrupt_type_p): New.
770 (riscv_save_reg_p): Save extra regs for interrupt handler.
771 (riscv_use_save_libcall): Return false for interrupt handler.
772 (riscv_first_stack_step): Add forward declaration.
773 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
774 for interrupt handler with large frame. Use it for saved reg list.
775 (riscv_expand_prologue): Move flag_stack_usage_info support to
776 eliminate duplication.
777 (riscv_expand_epilogue): Generate mret for interrupt handler.
778 (riscv_epilogue_uses): New.
779 (riscv_can_use_return_insn): Return false for interrupt handler.
780 (riscv_function_ok_for_sibcall): Likewise.
781 (riscv_set_current_function): Add interrupt handler support.
782 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
783 * config/riscv/riscv.md (UNSPECV_MRET): New.
785 (riscv_frflags, riscv_fsflags): Use tab after opcode.
787 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
789 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
791 PR tree-optimization/85712
792 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
793 this candidate has already been replaced in-situ by a copy.
795 2018-05-25 Jason Merrill <jason@redhat.com>
797 PR c++/80485 - inline function non-zero address.
798 * symtab.c (nonzero_address): Check DECL_COMDAT.
800 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
803 * config/alpha/alpha.md (ashlsi3): New insn pattern.
804 (*ashlsi_se): Rename from *ashldi_se. Define as sign
805 extension of SImode operation. Use const123_operand predicate.
807 (*saddl_se_1): Ditto.
809 (*ssubl_se_1): Ditto.
810 * config/alpha/predicates.md (const123_operand): New predicate.
811 * config/alpha/constraints.md (P): Use IN_RANGE.
813 2018-05-25 Richard Biener <rguenther@suse.de>
815 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
817 (ref_maybe_used_by_stmt_p): Likewise.
818 (stmt_may_clobber_ref_p): Likewise.
819 (stmt_may_clobber_ref_p_1): Likewise.
820 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
822 (ref_maybe_used_by_stmt_p): Likewise.
823 (stmt_may_clobber_ref_p): Likewise.
824 (stmt_may_clobber_ref_p_1): Likewise.
825 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
826 the alias oracle to disambiguate DRs with stmts DR analysis
828 (vect_analyze_data_refs): Do not give up on not analyzable
829 DRs for BB vectorization. Remove code truncating the dataref
832 2018-05-25 Jakub Jelinek <jakub@redhat.com>
835 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
836 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
837 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
839 2018-05-25 Richard Biener <rguenther@suse.de>
841 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
842 function, combining stmt data ref gathering and fatal analysis
844 (vect_analyze_data_refs): Remove now redudnant code and simplify.
845 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
846 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
847 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
848 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
850 2018-05-25 Bin Cheng <bin.cheng@arm.com>
852 PR tree-optimization/85720
853 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
854 SCC if all partitions are builtins.
855 (version_loop_by_alias_check): New parameter. Generate cancelable
856 runtime alias check if all partitions are builtins.
857 (distribute_loop): Update call to above function.
859 2018-05-25 Bin Cheng <bin.cheng@arm.com>
861 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
862 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
863 (parm_default_def_partition_arg): Ditto.
864 (set_parm_default_def_partition): Ditto.
865 (get_parm_default_def_partitions): Ditto and make it static.
866 (get_undefined_value_partitions): Ditto and make it static.
867 (remove_ssa_form): Refactor call to init_var_map here.
868 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
869 computation for loop region.
870 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
871 (register_default_def): Delete.
872 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
873 (parm_default_def_partition_arg): Ditto.
874 (set_parm_default_def_partition): Ditto.
875 (get_parm_default_def_partitions): Ditto and make it static.
876 (get_undefined_value_partitions): Ditto and make it static.
877 (coalesce_with_default, coalesce_with_default): Update comment.
878 (create_coalesce_list_for_region): New func factored out from
879 create_outofssa_var_map.
880 (populate_coalesce_list_for_outofssa): New func factored out from
881 create_outofssa_var_map and coalesce_ssa_name.
882 (create_outofssa_var_map): Delete.
883 (coalesce_ssa_name): Refactor to support live range computation.
884 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
885 (get_parm_default_def_partitions): Delete.
886 (get_undefined_value_partitions): Ditto.
887 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
888 computation for loop region.
889 (new_tree_live_info, loe_visit_block): Ditto.
890 (live_worklist, set_var_live_on_entry): Ditto.
891 (calculate_live_on_exit, verify_live_on_entry): Ditto.
892 * tree-ssa-live.h (struct _var_map): New fields.
893 (init_var_map): Change decl.
894 (region_contains_p): New.
896 2018-05-25 Bin Cheng <bin.cheng@arm.com>
898 * tree-ssa-live.h (live_merge_and_clear): Delete.
900 2018-05-25 Richard Biener <rguenther@suse.de>
903 * tree-dump.c (dequeue_and_dump): Remove access to removed
904 operand 2 of a SWITCH_EXPR.
906 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
908 * doc/sourcebuild.texi (vect_double_cond_arith): Include
909 multiplication and division.
910 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
911 (cond_udiv@var{m}, cond_umod@var{m}): Document.
912 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
913 (cond_udiv_optab, cond_umod_optab): New optabs.
914 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
915 (IFN_COND_RDIV): New internal functions.
916 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
917 TRUNC_MOD_EXPR and RDIV_EXPR.
918 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
919 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
921 (SVE_INT_BINARY): Include mult.
922 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
923 (optab, sve_int_op): Handle mult.
924 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
926 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
927 for SVE_INT_BINARY_SD.
929 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
931 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
932 (optab, sve_int_op): Handle div and udiv.
933 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
934 for SVE_INT_BINARY_SD.
935 (*<optab><mode>3): New insn for the same.
937 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
939 * tree-vect-patterns.c: Include predict.h.
940 (vect_recog_divmod_pattern): Restrict check for division support
941 to when optimizing for size.
943 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
945 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
946 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
947 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
948 (gimple_match_op::set_op): Likewise.
949 (gimple_resimplify4): Declare.
950 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
951 (expr::gen_transform): Likewise.
952 (decision_tree::gen): Generate a simplification routine for 4 operands.
953 * gimple-match-head.c (gimple_simplify): Add an overload for
954 4 operands. In the top-level function, handle up to 4 call
955 arguments and call gimple_resimplify4.
956 (gimple_resimplify4): New function.
957 (build_call_internal): Pass a fourth operand.
958 (maybe_push_to_seq): Likewise.
959 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
960 Fold VEC_COND_EXPRs of an operation and a default value into
961 an IFN_COND_* function if possible.
962 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
964 (SVE_COND_FP_BINARY): Include them.
965 (optab, sve_fp_op): Handle them.
966 (SVE_INT_BINARY_REV): New code iterator.
967 (SVE_COND_FP_BINARY_REV): New int iterator.
968 (commutative): New int attribute.
969 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
971 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
973 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
974 (*cond_<optab><mode>): New patterns for reversed operands.
976 2018-05-25 Richard Biener <rguenther@suse.de>
978 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
979 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
980 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
981 (STMT_VINFO_GROUPED_ACCESS): Adjust.
982 * tree-vect-data-refs.c (everywhere): Adjust users.
983 * tree-vect-loop.c (everywhere): Likewise.
984 * tree-vect-slp.c (everywhere): Likewise.
985 * tree-vect-stmts.c (everywhere): Likewise.
986 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
988 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
990 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
992 (gcc_cv_as_section_exclude): ... this.
993 Try Solaris as #exclude syntax.
994 * configure: Regenerate.
995 * config.in: Regenerate.
996 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
998 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
999 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
1001 * varasm.c (default_elf_asm_named_section): Don't check if
1002 HAVE_GAS_SECTION_EXCLUDE is defined.
1004 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1006 * doc/md.texi: Update the documentation of the cond_* optabs
1007 to mention the new final operand. Fix GET_MODE_NUNITS call.
1008 Describe the scalar case too.
1009 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
1010 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
1012 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
1013 (get_conditional_internal_fn): Update comment.
1014 * tree-vect-loop.c (vectorizable_reduction): Pass the original
1015 accumulator value as a final argument to conditional functions.
1016 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
1017 a define_expand and add an "else" operand. Assert for now that
1018 the else operand is equal to operand 2. Use SVE_INT_BINARY and
1019 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
1020 (*cond_<optab><mode>): New patterns.
1021 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
1022 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
1023 (UNSPEC_COND_EOR): Delete.
1024 (optab): Remove associated mappings.
1025 (SVE_INT_BINARY): New code iterator.
1026 (sve_int_op): Remove int attribute and add "minus" to the code
1028 (SVE_COND_INT_OP): Delete.
1029 (SVE_COND_FP_OP): Rename to...
1030 (SVE_COND_FP_BINARY): ...this.
1032 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
1034 * optabs.c (can_reuse_operands_p): New function.
1035 (maybe_legitimize_operands): Try to reuse the results for
1038 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1040 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
1041 Add {q} suffix to insn mnemonic.
1043 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1045 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
1046 (msp430_warn_func_return): New.
1048 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
1050 * fold-const.c (tree_nonzero_bits): New function.
1051 * fold-const.h (tree_nonzero_bits): Likewise.
1052 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
1053 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
1055 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
1059 * varasm.c (assemble_alias): Check ifunc_resolver only on
1062 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
1065 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
1066 when memory input operand is handled.
1068 2018-05-24 Luis Machado <luis.machado@linaro.org>
1070 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
1072 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
1074 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1076 * match.pd: Delay FMA folds until after vectorization.
1078 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
1081 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
1082 address check not strict.
1084 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
1086 * gimple-match.h (gimple_match_op): New class.
1087 (mprts_hook): Replace parameters with a gimple_match_op *.
1088 (maybe_build_generic_op): Likewise.
1089 (gimple_simplified_result_is_gimple_val): Replace parameters with
1090 a const gimple_match_op *.
1091 (gimple_simplify): Replace code_helper * and tree * parameters with
1092 a gimple_match_op * parameter.
1093 (gimple_resimplify1): Replace code_helper *, tree and tree *
1094 parameters with a gimple_match_op * parameter.
1095 (gimple_resimplify2): Likewise.
1096 (gimple_resimplify3): Likewise.
1097 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
1098 parameters with a gimple_match_op * parameter.
1099 * gimple-match-head.c (gimple_simplify): Change prototypes of
1100 auto-generated functions to take a gimple_match_op * instead of
1101 separate code_helper * and tree * parameters. Make the same
1102 change in the top-level overload and update calls to the
1103 gimple_resimplify routines. Update calls to the auto-generated
1104 functions and to maybe_push_res_to_seq in the publicly-facing
1105 operation-specific gimple_simplify overloads.
1106 (gimple_match_op::MAX_NUM_OPS): Define.
1107 (gimple_resimplify1): Replace rcode and ops with a single res_op
1108 parameter. Update call to gimple_simplify.
1109 (gimple_resimplify2): Likewise.
1110 (gimple_resimplify3): Likewise.
1111 (mprts_hook): Replace parameters with a gimple_match_op *.
1112 (maybe_build_generic_op): Likewise.
1113 (build_call_internal): Replace type, nargs and ops with
1114 a gimple_match_op *.
1115 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
1116 with a single gimple_match_op *. Update calls to mprts_hook,
1117 build_call_internal and gimple_simplified_result_is_gimple_val.
1118 Factor out code that is common to the tree_code and combined_fn cases.
1119 * genmatch.c (expr::gen_transform): Replace tem_code and
1120 tem_ops with a gimple_match_op called tem_op. Update calls
1121 to the gimple_resimplify functions and maybe_push_res_to_seq.
1122 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
1123 res_ops. Update call to the gimple_resimplify functions.
1124 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
1125 (decision_tree::gen): Make the functions take a gimple_match_op *
1126 called res_op instead of separate res_code and res_ops parameters.
1127 Update call accordingly.
1128 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
1129 and ops with a single res_op parameter. Update calls to
1130 maybe_build_generic_op and maybe_push_res_to_seq.
1131 (fold_stmt_1): Update calls to gimple_simplify and
1132 replace_stmt_with_simplification.
1133 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
1134 and gimple_simplified_result_is_gimple_val.
1135 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
1137 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
1138 with a gimple_match_op *.
1139 (vn_nary_build_or_lookup): Likewise. Update call to
1140 vn_nary_build_or_lookup_1.
1141 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
1142 gimple_match_op *. Update calls to the gimple_resimplify routines
1143 and to gimple_simplified_result_is_gimple_val.
1144 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
1145 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
1146 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
1147 (visit_nary_op): Likewise.
1148 (visit_reference_op_load): Likewise.
1150 2018-05-23 Luis Machado <luis.machado@linaro.org>
1152 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
1153 modifier for printing the step amount.
1155 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
1158 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
1161 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
1163 * doc/sourcebuild.texi (Endianness): New subsubsection.
1165 2018-05-23 Luis Machado <luis.machado@linaro.org>
1167 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1168 <prefetch_dynamic_strides>: New const bool field.
1169 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1170 prefetch_dynamic_strides.
1171 (exynosm1_prefetch_tune): Likewise.
1172 (thunderxt88_prefetch_tune): Likewise.
1173 (thunderx_prefetch_tune): Likewise.
1174 (thunderx2t99_prefetch_tune): Likewise.
1175 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
1177 (aarch64_override_options_internal): Update to set
1178 PARAM_PREFETCH_DYNAMIC_STRIDES.
1179 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
1180 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
1181 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
1182 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
1183 prefetch-dynamic-strides setting.
1185 2018-05-23 Luis Machado <luis.machado@linaro.org>
1187 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
1188 <minimum_stride>: New const int field.
1189 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
1190 minimum_stride field defaulting to -1.
1191 (exynosm1_prefetch_tune): Likewise.
1192 (thunderxt88_prefetch_tune): Likewise.
1193 (thunderx_prefetch_tune): Likewise.
1194 (thunderx2t99_prefetch_tune): Likewise.
1195 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
1196 <default_opt_level>: Set to 3.
1197 (aarch64_override_options_internal): Update to set
1198 PARAM_PREFETCH_MINIMUM_STRIDE.
1199 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
1200 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
1201 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
1202 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
1203 stride is constant and is below the minimum stride threshold.
1205 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1207 * config/arm/arm-cpus.in (mode26): Delete.
1208 (armv4): Delete mode26 reference.
1209 * config/arm/arm.c (arm_configure_build_target): Delete use of
1212 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1214 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
1216 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
1217 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
1219 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
1220 (floatunsdidf2): Ditto.
1222 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
1224 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
1225 (fixuns_trunc<mode>si2_avx512f): Ditto.
1226 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
1227 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
1228 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
1230 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1232 PR rtl-optimization/79985
1233 * df-scan.c (df_insn_refs_collect): Remove special case for
1234 global registers and asm statements.
1236 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
1238 * extend.texi (Global Register Variables): Rewrite the bullet list.
1239 Note that the register is available for allocation. Note that access
1240 via inline asm must use constraints. Add note about async-signal
1241 handlers. Remove paragraph about automagic register selection.
1243 2018-05-23 Richard Biener <rguenther@suse.de>
1245 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
1246 of fixed offset from memset VN.
1248 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
1250 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
1252 (alloc_cand_and_find_basis): Initialize first_interp field.
1253 (slsr_process_mul): Modify first_interp field.
1254 (slsr_process_add): Likewise.
1255 (slsr_process_cast): Modify first_interp field for each new
1257 (slsr_process_copy): Likewise.
1258 (dump_candidate): Dump first_interp field.
1259 (replace_mult_candidate): Process all interpretations, not just
1261 (replace_rhs_if_not_dup): Likewise.
1262 (replace_one_candidate): Likewise.
1264 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
1266 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
1268 (aarch64_needs_frame_chain): New function.
1269 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
1271 2018-05-23 Sudakshina Das <sudi.das@arm.com>
1274 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
1275 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
1276 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
1277 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
1278 as true for strict-align.
1279 (aarch64_can_inline_p): Perform checks even when callee has no
1280 attributes to check for strict alignment.
1281 * doc/extend.texi (AArch64 Function Attributes): Document
1283 * doc/invoke.texi: (AArch64 Options): Likewise.
1285 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
1287 PR tree-optimization/85853
1288 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
1289 the handling of the root of the node to...
1290 (vect_slp_analyze_node_operations_1): ...this new function,
1291 and run the whole thing with the child nodes' def types
1292 set according to their SLP node's def type.
1294 2018-05-23 Richard Biener <rguenther@suse.de>
1297 * tree-data-ref.c (create_runtime_alias_checks): Defer
1298 and ignore overflow warnings.
1300 2018-05-23 Yury Gribov <tetra2005@gmail.com>
1302 PR tree-optimization/85822
1303 * tree-vrp.c (is_masked_range_test): Fix handling of negative
1306 2018-05-23 Richard Biener <rguenther@suse.de>
1308 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1309 memset constants via native_interpret_expr.
1311 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
1314 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
1316 (cgraph_node::create_alias): Likewise.
1317 (cgraph_node::get_availability): Check ifunc_resolver instead
1318 of looking up ifunc attribute.
1319 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
1320 * varasm.c (do_assemble_alias): Likewise.
1321 (assemble_alias): Likewise.
1322 (default_binds_local_p_3): Likewise.
1323 * cgraph.h (cgraph_node): Add ifunc_resolver.
1324 (cgraph_node::only_called_directly_or_aliased_p): Return false
1326 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
1328 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
1329 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
1330 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
1331 instead of looking up ifunc attribute.
1333 2018-05-22 Luis Machado <luis.machado@linaro.org>
1335 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
1337 2018-05-22 Martin Sebor <msebor@redhat.com>
1340 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
1341 only when expasion succeeds.
1342 (expand_builtin_strcmp): Same.
1343 (expand_builtin_strncmp): Same.
1345 2018-05-22 Martin Sebor <msebor@redhat.com>
1347 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
1349 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1350 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1352 * config/aarch64/aarch64-ldpstp.md: Replace uses of
1353 aarch64_mem_pair_operand with memory_operand and delete operand swapping
1355 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1356 Add check for legitimate_address.
1357 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
1358 (aarch64_swap_ldrstr_operands): New.
1359 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
1362 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
1363 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1365 * config/aarch64/aarch64.md: New patterns to generate stp
1367 (store_pair_sw, store_pair_dw): New patterns to generate stp for
1368 single words and double words.
1369 (load_pair_sw, load_pair_dw): Likewise.
1370 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
1372 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
1374 * config/aarch64/aarch64-ldpstp.md: Modify peephole
1375 for different mode ldpstp and add peephole for merged zero stores.
1377 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
1379 (aarch64_gen_store_pair): Rename calls to match new patterns.
1380 (aarch64_gen_load_pair): Rename calls to match new patterns.
1381 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
1382 (load_pair<DREG:mode><DREG2:mode>): ... This.
1383 (store_pair<mode>): Rename to...
1384 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
1385 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
1387 (V_INT_EQUIV): Handle SImode.
1388 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
1391 2018-05-22 Martin Sebor <msebor@redhat.com>
1394 * calls.c (maybe_warn_nonstring_arg): Use string length to set
1395 or ajust the presumed bound on an operation to avoid unnecessary
1398 2018-05-22 Martin Sebor <msebor@redhat.com>
1400 PR tree-optimization/85826
1401 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
1402 assuming that a DECL necesarily has a constant size.
1404 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
1407 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
1409 2018-05-22 Richard Biener <rguenther@suse.de>
1411 PR tree-optimization/85834
1412 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
1413 non-constant and non-zero memset arguments.
1415 2018-05-22 Martin Liska <mliska@suse.cz>
1418 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
1420 2018-05-22 Richard Biener <rguenther@suse.de>
1422 PR tree-optimization/85863
1423 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
1424 comparisons when vectype is specified.
1425 (vectorizable_condition): Do not specify vectype for
1426 vect_is_simple_cond when SLP vectorizing.
1428 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
1431 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
1432 define __ibm128 as long double.
1433 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
1434 as a distinct type when IEEE 128-bit support is enabled.
1435 (init_float128_ieee): Fix up conversions between IFmode and IEEE
1436 128-bit types to use the correct functions.
1437 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
1438 convert between 128-bit floating point types that have different
1439 modes but the same representation, instead of using gen_lowpart to
1441 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
1443 (IFKF_reg): New attributes to give the register constraints for
1445 (extend<mode>tf2_internal): New insns to mark an explicit
1446 conversion between 128-bit floating point types that have a
1447 different mode but share the same representation.
1449 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
1451 PR tree-optimization/85814
1452 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
1453 a null return from get_strinfo when unsharing the next
1454 strinfo in the chain.
1456 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1459 * varasm.c (weak_finish): Clean up weak_decls.
1461 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1463 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
1464 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
1465 UNSPEC_UADALP values.
1466 * config/aarch64/iterators.md (ABAL): New int iterator.
1469 (sur): Add mappings for the above.
1470 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
1472 (aarch64_<sur>abal<mode>_4): Likewise.
1473 (aarch64_<sur>adalp<mode>_3): Likewise.
1474 (<sur>sadv16qi): New define_expand.
1476 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
1478 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
1479 (*movdf_internal): Ditto.
1480 (*rcpsf2_sse): Ditto.
1481 (*rsqrtsf2_sse): Ditto.
1482 (*sqrt<mode>2_sse): Ditto.
1484 2018-05-21 Tamar Christina <tamar.christina@arm.com>
1486 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
1488 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
1489 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
1490 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
1491 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
1493 * config/aarch64/arm_neon.h: Likewise.
1494 * config/aarch64/iterators.md (VQ_I): New.
1496 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
1498 * config.gcc: Add arc/t-multilib-linux to tmake_file for
1500 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
1503 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1505 * config/nds32/constraints.md (S): New constraint.
1506 * config/nds32/nds32.md (call_internal): Use constraint S.
1507 (call_value_internal): Likewise.
1508 (sibcall_internal): Likewise.
1509 (sibcall_value_internal): Likewise.
1511 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1512 Chung-Ju Wu <jasonwucj@gmail.com>
1514 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
1517 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
1518 Chung-Ju Wu <jasonwucj@gmail.com>
1520 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
1521 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
1522 (nds32_rtx_costs_impl): Simplify.
1523 (nds32_address_cost_impl): Simplify.
1524 (nds32_init_rtx_costs): New function.
1525 (nds32_rtx_costs_speed_prefer): Likewise.
1526 (nds32_rtx_costs_size_prefer): Likewise.
1527 (nds32_address_cost_speed_prefer): Likewise.
1528 (nds32_address_cost_speed_fwprop): Likewise.
1529 (nds32_address_cost_size_prefer): Likewise.
1530 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
1531 * config/nds32/nds32.c (nds32_option_override): Use
1532 nds32_init_rtx_costs function.
1534 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
1536 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
1537 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
1538 (TARGET_PIPELINE_N8): Likewise.
1539 (TARGET_PIPELINE_N10): Likewise.
1540 (TARGET_PIPELINE_N13): Likewise.
1541 (TARGET_PIPELINE_GRAYWOLF): Likewise.
1543 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
1545 * config/nds32/nds32-fpu.md: Update copyright year.
1547 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1549 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
1551 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1553 * config/nds32/nds32.c
1554 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
1555 * config/nds32/nds32.opt (minline-asm-r15): New option.
1557 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1559 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
1561 * config/nds32/nds32.md (abssi2): New pattern.
1563 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
1565 * config/i386/i386.md (rex64namesuffix): New mode attribute.
1566 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
1567 Merge insn pattern from sse_cvtsi2ss<round_name> and
1568 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
1569 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
1570 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
1571 using SWI48 mode iterator.
1572 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
1573 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
1574 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
1575 pattern from sse_cvttss2si<round_saeonly_name>
1576 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
1577 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
1578 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
1579 using SWI48 mode iterator.
1580 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
1581 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
1582 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
1583 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
1584 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
1585 using SWI48 mode iterator.
1586 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
1587 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
1588 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
1589 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
1590 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
1591 SWI48 mode iterator.
1592 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
1593 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
1594 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
1595 pattern from sse_cvttsd2si<round_saeonly_name>
1596 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
1598 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1600 * config/nds32/nds32-md-auxiliary.c
1601 (nds32_valid_smw_lwm_base_p): Refine.
1602 (nds32_output_smw_single_word): Refine.
1603 (nds32_output_smw_double_word): New.
1604 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
1606 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1608 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
1609 (nds32_output_stack_pop): Refine.
1610 (nds32_expand_unaligned_load): Refine.
1611 (nds32_expand_unaligned_store): Refine.
1613 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1614 Chung-Ju Wu <jasonwucj@gmail.com>
1616 * config/nds32/constants.md: Add TP_REGNUM constant.
1617 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
1618 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
1620 * config/nds32/nds32-doubleword.md: Consider flag_pic.
1621 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
1622 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
1623 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
1624 and PIC code generation.
1625 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
1627 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
1629 * config/nds32/nds32.md: Support TLS and PIC.
1630 * config/nds32/nds32.c: Support TLS and PIC.
1631 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
1632 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
1635 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
1637 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
1638 mode with E_ prefix.
1640 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1641 Chung-Ju Wu <jasonwucj@gmail.com>
1643 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
1644 * config/nds32/nds32-md-auxiliary.c
1645 (symbolic_reference_mentioned_p): New.
1646 (nds32_legitimize_ict_address): New.
1647 (nds32_expand_ict_move): New.
1648 (nds32_indirect_call_referenced_p): New.
1649 (nds32_symbol_binds_local_p): Delete.
1650 (nds32_long_call_p): Modify.
1651 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
1652 * config/nds32/nds32-protos.h
1653 (symbolic_reference_mentioned_p): Declare.
1654 (nds32_legitimize_ict_address): Declare.
1655 (nds32_expand_ict_move): Declare.
1656 (nds32_indirect_call_referenced_p): Declare.
1657 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
1658 (nds32_relax_group): Use nds32_ict_const_p as condition.
1659 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
1660 (nds32_asm_file_start): Output ict_model directive in asm code.
1661 (nds32_legitimate_address_p): Consider indirect call.
1662 (nds32_print_operand): Consider indirect call.
1663 (nds32_print_operand_address): Consider indirect call.
1664 (nds32_insert_attributes): Handle "indirect_call" attribute.
1665 (TARGET_LEGITIMATE_ADDRESS_P): Define.
1666 (TARGET_LEGITIMATE_CONSTANT_P): Define.
1667 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
1668 (TARGET_DELEGITIMIZE_ADDRESS): Define.
1669 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
1670 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
1671 (TARGET_ICT_MODEL_SMALL): Define.
1672 (TARGET_ICT_MODEL_LARGE): Define.
1673 * config/nds32/nds32.md (movsi): Consider ict model.
1674 (call, call_value): Consider ict model.
1675 (sibcall, sibcall_value): Consider ict model.
1676 * config/nds32/nds32.opt (mict-model): New option.
1677 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
1680 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
1681 Monk Chiang <sh.chiang04@gmail.com>
1682 Jim Wilson <jimw@sifive.com>
1684 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
1685 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
1686 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
1687 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
1688 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
1689 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
1690 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
1691 compute save_libcall_adjustment properly.
1692 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
1693 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
1694 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
1695 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
1696 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
1697 (ABI_SPEC): Handle mabi=ilp32e.
1698 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
1699 (RVE): Add RVE mask.
1700 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
1701 <-march>: Add rv32e as an example.
1703 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
1706 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
1707 (intra_create_variable_infos): Handle C++ constructors.
1709 2018-05-18 Martin Liska <mliska@suse.cz>
1711 * passes.def: Remove a redundant pass.
1713 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
1716 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
1718 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1720 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
1722 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
1724 (armv2, armv2a, armv3, armv3m): Delete architectures.
1725 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
1726 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
1727 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
1729 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
1730 (*mulsidi3adddi): Likewise.
1731 (mulsidi3): Likewise.
1732 (*mulsidi3_nov6): Likewise.
1733 (umulsidi3): Likewise.
1734 (umulsidi3_nov6): Likewise.
1735 (umaddsidi4): Likewise.
1736 (*umulsidi3adddi): Likewise.
1737 (smulsi3_highpart): Likewise.
1738 (*smulsi3_highpart_nov6): Likewise.
1739 (umulsi3_highpart): Likewise.
1740 (*umulsi3_highpart_nov6): Likewise.
1741 * config/arm/arm.h (arm_arch3m): Delete.
1742 * config/arm/arm.c (arm_arch3m): Delete.
1743 (arm_option_override_internal): Update armv3-related comment.
1744 (arm_configure_build_target): Delete use of isa_bit_mode32.
1745 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
1746 (arm_rtx_costs_internal): Delete check of arm_arch3m.
1747 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
1749 (mulusa3): Likewise.
1750 * config/arm/arm-protos.h (arm_arch3m): Delete.
1751 * config/arm/arm-tables.opt: Regenerate.
1752 * config/arm/arm-tune.md: Likewise.
1753 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
1754 deleted architectures.
1756 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1758 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
1759 (armv5t, armv5te): New features.
1760 (ARMv5, ARMv5e): Delete fgroups.
1761 (ARMv5t, ARMv5te): Adjust for above changes.
1763 (armv5, armv5e): Delete arches.
1764 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
1766 (*call_reg_arm): Likewise.
1767 (*call_value_reg_armv5): Likewise.
1768 (*call_value_reg_arm): Likewise.
1769 (*call_symbol): Likewise.
1770 (*call_value_symbol): Likewise.
1771 (*sibcall_insn): Likewise.
1772 (*sibcall_value_insn): Likewise.
1774 (prefetch): Likewise.
1775 (define_split and define_peephole2 dependent on arm_arch5):
1777 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
1779 (TARGET_ARM_QBIT): Likewise.
1780 (TARGET_DSP_MULTIPLY): Likewise.
1781 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
1782 (arm_arch5, arm_arch5e): Delete.
1783 (arm_arch5t, arm_arch5te): Declare.
1784 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
1785 (arm_arch5t): Declare.
1786 (arm_option_reconfigure_globals): Update for the above.
1787 (arm_options_perform_arch_sanity_checks): Update comment, replace
1788 use of arm_arch5 with arm_arch5t.
1789 (use_return_insn): Likewise.
1790 (arm_emit_call_insn): Likewise.
1791 (output_return_instruction): Likewise.
1792 (arm_final_prescan_insn): Likewise.
1793 (arm_coproc_builtin_available): Likewise.
1794 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
1795 arm_arch5e with arm_arch5t and arm_arch5te.
1796 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
1797 (arm_arch5t, arm_arch5te): Declare.
1798 * config/arm/arm-tables.opt: Regenerate.
1799 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
1800 * config/arm/t-multilib: Likewise.
1801 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
1802 instead of arm_arch5.
1803 (*call_reg_thumb1): Likewise.
1804 (*call_value_reg_thumb1_v5): Likewise.
1805 (*call_value_reg_thumb1): Likewise.
1806 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
1808 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
1810 2018-05-18 Martin Liska <mliska@suse.cz>
1812 PR gcov-profile/84846
1813 * doc/gcov.texi: Document -t option of gcov tool.
1815 2018-05-18 Martin Liska <mliska@suse.cz>
1817 PR gcov-profile/84846
1818 * gcov.c (print_usage): Add new -t option.
1819 (process_args): Handle the option.
1820 (generate_results): Use stdout as output when requested by
1823 2018-05-18 Martin Liska <mliska@suse.cz>
1825 PR gcov-profile/84846
1826 * coverage.c (coverage_init): Write PWD to .gcno file.
1827 * doc/gcov.texi: Document how working directory is printed.
1828 * gcov-dump.c (dump_gcov_file): Print PWD.
1829 * gcov.c (output_intermediate_file): Likewise.
1830 (read_graph_file): Read PWD string.
1831 (output_lines): Print PWD.
1833 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1836 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
1837 for retval and return false if all args to phi are zero.
1839 2018-05-18 Richard Biener <rguenther@suse.de>
1841 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
1843 (evrp_dom_walker::before_dom_children): Call it.
1845 2018-05-18 Richard Biener <rguenther@suse.de>
1847 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
1848 results when processing array refs with variable index.
1850 2018-05-18 Toon Moene <toon@moene.org>
1852 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
1853 directly after that of -floop-interchange. Indicate that both
1854 options are enabled by default when specifying -O3.
1856 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1858 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
1859 iterator. Delete separate integer-mode vec_set<mode> expander.
1860 (aarch64_simd_vec_setv2di): Delete.
1861 (vec_setv2di): Delete.
1862 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
1863 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
1864 the "w, r" alternative.
1866 2018-05-18 Martin Liska <mliska@suse.cz>
1868 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
1869 * tree-pass.h (make_pass_lower_switch_O0): New function.
1870 * tree-switch-conversion.c (node_has_low_bound): Remove.
1871 (node_has_high_bound): Likewise.
1872 (node_is_bounded): Likewise.
1873 (class pass_lower_switch): Make it a template type and create
1875 (pass_lower_switch::execute): Add template argument.
1876 (make_pass_lower_switch): New function.
1877 (make_pass_lower_switch_O0): New function.
1878 (do_jump_if_equal): Remove.
1879 (emit_case_nodes): Simplify to just handle all 3 cases and leave
1880 all the hard work to tree optimization passes.
1882 2018-05-18 Martin Liska <mliska@suse.cz>
1884 * dbgcnt.c (limit_low): Renamed from limit.
1885 (limit_high): New variable.
1886 (dbg_cnt_is_enabled): Check for upper limit.
1887 (dbg_cnt): Adjust dumping.
1888 (dbg_cnt_set_limit_by_index): Add new argument for high
1890 (dbg_cnt_set_limit_by_name): Likewise.
1891 (dbg_cnt_process_single_pair): Parse new format.
1892 (dbg_cnt_process_opt): Use strtok.
1893 (dbg_cnt_list_all_counters): Remove 'value' and add
1895 * doc/invoke.texi: Document changes.
1897 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
1899 * doc/sourcebuild.texi (scalar_all_fma): Document.
1900 * tree.def (FMA_EXPR): Delete.
1901 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
1902 * internal-fn.c (ternary_direct): New macro.
1903 (expand_ternary_optab_fn): Likewise.
1904 (direct_ternary_optab_supported_p): Likewise.
1905 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
1906 * builtins.c (fold_builtin_fma): Delete.
1907 (fold_builtin_3): Don't call it.
1908 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
1909 * expr.c (expand_expr_real_2): Likewise.
1910 * fold-const.c (operand_equal_p): Likewise.
1911 (fold_ternary_loc): Likewise.
1912 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
1913 * gimple.c (DEFTREECODE): Likewise.
1914 * gimplify.c (gimplify_expr): Likewise.
1915 * optabs-tree.c (optab_for_tree_code): Likewise.
1916 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
1917 * tree-eh.c (operation_could_trap_p): Likewise.
1918 (stmt_could_throw_1_p): Likewise.
1919 * tree-inline.c (estimate_operator_cost): Likewise.
1920 * tree-pretty-print.c (dump_generic_node): Likewise.
1921 (op_code_prio): Likewise.
1922 * tree-ssa-loop-im.c (stmt_cost): Likewise.
1923 * tree-ssa-operands.c (get_expr_operands): Likewise.
1924 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
1925 * fold-const-call.h (fold_fma): Delete.
1926 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
1927 CFN_FNMA and CFN_FNMS.
1929 * genmatch.c (combined_fn): New enum.
1930 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
1931 (commutative_op): New function.
1932 (commutate): Use it. Handle more than 2 operands.
1933 (dt_operand::gen_gimple_expr): Use commutative_op.
1934 (parser::parse_expr): Allow :c to be used with non-binary
1935 operators if the commutative operand is known.
1936 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
1937 CFN_FMS, CFN_FNMA and CFN_FNMS.
1938 (backprop::process_assign_use): Remove FMA_EXPR handling.
1939 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
1940 (gen_hsa_fma): New function.
1941 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
1942 IFN_FNMA and IFN_FNMS.
1943 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
1944 * gimple-fold.h (follow_all_ssa_edges): Declare.
1945 * gimple-fold.c (follow_all_ssa_edges): New function.
1946 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
1947 gimple_build interface and use follow_all_ssa_edges to fold the result.
1948 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
1949 instead of checking for optabs directly.
1950 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
1951 rather than FMA_EXPRs.
1952 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
1953 call to IFN_FMA instead of an FMA_EXPR.
1955 2018-05-17 Jim Wilson <jimw@sifive.com>
1957 * expr.c (do_tablejump): When converting index to Pmode, if we have a
1958 sign extended promoted subreg, and the range does not have the sign bit
1959 set, then do a sign extend.
1961 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
1962 test, check for sign extended subreg and/or constant operands, and
1963 do a sign extend in that case.
1965 2018-05-17 Steve Ellcey <sellcey@cavium.com>
1967 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
1968 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
1970 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
1971 Change logics_shift_reg to logics_shift_imm.
1972 (thunderx2t99_fp_loadpair_basic): Delete.
1973 (thunderx2t99_fp_storepair_basic): Delete.
1974 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
1975 (thunderx2t99_asimd_polynomial): Delete.
1976 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
1977 and neon_fp_mul_d_scalar_q.
1978 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
1979 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
1980 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
1981 (thunderx2t99_asimd_lut): Add missing tbl types.
1982 (thunderx2t99_asimd_ext): Delete.
1983 (thunderx2t99_asimd_load1_1_mult): Delete.
1984 (thunderx2t99_asimd_load1_2_mult): Delete.
1985 (thunderx2t99_asimd_load1_ldp): New.
1986 (thunderx2t99_asimd_load1): New.
1987 (thunderx2t99_asimd_load2): Add missing *load2* types.
1988 (thunderx2t99_asimd_load3): New.
1989 (thunderx2t99_asimd_load4): New.
1990 (thunderx2t99_asimd_store1_1_mult): Delete.
1991 (thunderx2t99_asimd_store1_2_mult): Delete.
1992 (thunderx2t99_asimd_store2_mult): Delete.
1993 (thunderx2t99_asimd_store2_onelane): Delete.
1994 (thunderx2t99_asimd_store_stp): New.
1995 (thunderx2t99_asimd_store1): New.
1996 (thunderx2t99_asimd_store2): New.
1997 (thunderx2t99_asimd_store3): New.
1998 (thunderx2t99_asimd_store4): New.
2000 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
2002 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
2003 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
2005 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
2006 Segher Boessenkool <segher@kernel.crashing.org>
2009 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
2012 2018-05-17 Richard Biener <rguenther@suse.de>
2014 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
2015 for pruning loop and prune defs feeding only already visited PHIs.
2017 2018-05-17 Richard Biener <rguenther@suse.de>
2019 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
2021 2018-05-17 Bin Cheng <bin.cheng@arm.com>
2022 Richard Biener <rguenther@suse.de>
2024 PR tree-optimization/85793
2025 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
2026 for VMAT_ELEMENTWISE.
2028 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2030 * internal-fn.h (lookup_internal_fn): Declare
2031 * internal-fn.c (lookup_internal_fn): New function.
2032 * gimple.c (gimple_build_call_from_tree): Handle calls to
2034 * gimple-pretty-print.c (dump_gimple_call): Print "." before
2035 internal function names.
2036 * tree-pretty-print.c (dump_generic_node): Likewise.
2037 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
2039 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
2041 * gimple-fold.h (gimple_build): Make the function forms take
2042 combined_fn rather than built_in_function.
2043 (gimple_simplify): Likewise.
2044 * gimple-match-head.c (gimple_simplify): Likewise.
2045 * gimple-fold.c (gimple_build): Likewise.
2046 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
2047 rather than gimple_build_call_internal.
2048 (get_initial_defs_for_reduction): Likewise.
2049 (vect_create_epilog_for_reduction): Likewise.
2050 (vectorizable_live_operation): Likewise.
2052 2018-05-17 Martin Liska <mliska@suse.cz>
2054 * gimple-ssa-sprintf.c (format_directive): Do not use
2055 space in between 'G_' and '('.
2057 2018-05-17 Jakub Jelinek <jakub@redhat.com>
2060 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
2061 even if the mask is not all ones.
2064 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
2066 (ix86_gimple_fold_builtin): Likewise.
2069 * config/i386/i386.c: Include tree-vector-builder.h.
2070 (ix86_vector_shift_count): New function.
2071 (ix86_fold_builtin): Fold shift builtins by scalar count.
2072 (ix86_gimple_fold_builtin): Likewise.
2074 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
2075 _mm512_setzero): New intrinsics.
2077 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
2078 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2080 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
2081 code generation for cases where splatting a value is not useful.
2082 * simplify-rtx.c (simplify_ternary_operation): Simplify
2083 vec_merge across a vec_duplicate and a paradoxical subreg forming a vector
2084 mode to a vec_concat.
2086 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
2088 * config.gcc: Support "goldmont-plus".
2089 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2091 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2092 PROCESSOR_GOLDMONT_PLUS.
2093 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
2094 (processor_target_table): Add "goldmont-plus".
2095 (PTA_GOLDMONT_PLUS): Define.
2096 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
2097 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
2098 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
2099 (fold_builtin_cpu): Add "goldmont-plus".
2100 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
2101 (ix86_option_override_internal): Add "goldmont-plus".
2102 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
2103 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
2104 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
2105 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
2107 2018-05-17 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/85757
2110 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
2111 remove defs that only feed that PHI from further processing.
2113 2018-05-16 Jim Wilson <jimw@sifive.com>
2115 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
2117 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
2119 2018-05-16 Mark Wielaard <mark@klomp.org>
2121 * dwarf2out.c (count_index_strings): New function.
2122 (output_indirect_strings): Call count_index_strings and generate
2123 header for dwarf_version >= 5.
2125 2018-05-16 Mark Wielaard <mark@klomp.org>
2127 * dwarf2out.c (dwarf_FORM): New function.
2128 (set_indirect_string): Use dwarf_FORM.
2129 (reset_indirect_string): Likewise.
2130 (size_of_die): Likewise.
2131 (value_format): Likewise.
2132 (output_die): Likewise.
2133 (add_skeleton_AT_string): Likewise.
2134 (output_macinfo_op): Likewise.
2135 (index_string): Likewise.
2136 (output_index_string_offset): Likewise.
2137 (output_index_string): Likewise.
2138 (count_index_strings): Likewise.
2140 2018-05-16 Carl Love <cel@us.ibm.com>
2142 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2143 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
2145 2018-05-16 Martin Jambor <mjambor@suse.cz>
2147 * ipa-prop.c (ipa_free_all_edge_args): Remove.
2148 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
2150 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
2152 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
2153 (fnma<mode>4): Likewise.
2154 (fms<mode>4): Likewise.
2155 (fnms<mode>4): Likewise.
2156 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
2157 (aarch64_fnma<mode>4): Likewise.
2158 (aarch64_fms<mode>4): Likewise.
2159 (aarch64_fnms<mode>4): Likewise.
2160 (aarch64_fnmadd<mode>4): Likewise.
2162 2018-05-16 Jason Merrill <jason@redhat.com>
2164 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
2166 2018-05-16 Richard Biener <rguenther@suse.de>
2168 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
2169 (dump_stmt_cost): Declare.
2170 (add_stmt_cost): Dump cost we add.
2171 (add_stmt_costs): New function.
2172 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
2174 (vect_analyze_stmt): Adjust prototype.
2175 (vectorizable_condition): Likewise.
2176 (vectorizable_live_operation): Likewise.
2177 (vectorizable_reduction): Likewise.
2178 (vectorizable_induction): Likewise.
2179 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
2180 cost vector to pass to vectorizable_ and record afterwards.
2181 (vect_model_reduction_cost): Take cost vector argument and adjust.
2182 (vect_model_induction_cost): Likewise.
2183 (vectorizable_reduction): Likewise.
2184 (vectorizable_induction): Likewise.
2185 (vectorizable_live_operation): Likewise.
2186 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
2187 SLP_TREE_NUMBER_OF_VEC_STMTS.
2188 (vect_analyze_slp_cost_1): Remove.
2189 (vect_analyze_slp_cost): Likewise.
2190 (vect_slp_analyze_node_operations): Take visited args and
2191 a target cost vector. Avoid processing already visited stmt sets.
2192 (vect_slp_analyze_operations): Use a local cost vector to gather
2193 costs and register those of non-discarded instances.
2194 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
2195 (vect_schedule_slp_instance): Remove copying of
2196 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
2198 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
2199 adding cost. Record cost entry location.
2200 (vect_prologue_cost_for_slp_op): Function to compute cost of
2201 a constant or invariant generated for SLP vect in the prologue,
2202 split out from vect_analyze_slp_cost_1.
2203 (vect_model_simple_cost): Make static. Adjust for SLP costing.
2204 (vect_model_promotion_demotion_cost): Likewise.
2205 (vect_model_store_cost): Likewise, make static.
2206 (vect_model_load_cost): Likewise.
2207 (vectorizable_bswap): Add cost vector arg and adjust.
2208 (vectorizable_call): Likewise.
2209 (vectorizable_simd_clone_call): Likewise.
2210 (vectorizable_conversion): Likewise.
2211 (vectorizable_assignment): Likewise.
2212 (vectorizable_shift): Likewise.
2213 (vectorizable_operation): Likewise.
2214 (vectorizable_store): Likewise.
2215 (vectorizable_load): Likewise.
2216 (vectorizable_condition): Likewise.
2217 (vectorizable_comparison): Likewise.
2218 (can_vectorize_live_stmts): Likewise.
2219 (vect_analyze_stmt): Likewise.
2220 (vect_transform_stmt): Adjust calls to vectorizable_*.
2221 * tree-vectorizer.c: Include gimple-pretty-print.h.
2222 (dump_stmt_cost): New function.
2224 2018-05-16 Richard Biener <rguenther@suse.de>
2226 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
2227 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
2228 * tree-ssa-dse.c: Include tree-ssa-loop.h.
2229 (check_name): New callback.
2230 (dse_classify_store): Track cycles via a visited bitmap of PHI
2231 defs and simplify handling of in-loop and across loop dead stores
2232 and properly fail for loop-variant refs. Handle byte-tracking with
2233 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
2236 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
2238 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
2239 (vect_get_mask_type_for_stmt): Likewise.
2240 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
2242 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
2243 to determine the statement's vector type and the vector type that
2244 should be used for calculating nunits. Deal with cases in which
2245 the type has to be deferred.
2246 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
2247 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
2248 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
2249 (vect_determine_vf_for_stmt): New functions, split out from...
2250 (vect_determine_vectorization_factor): ...here.
2251 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
2252 (vect_get_mask_type_for_stmt): New functions, split out from
2253 vect_determine_vectorization_factor.
2255 2018-05-16 Richard Biener <rguenther@suse.de>
2257 * tree-cfg.c (verify_gimple_assign_ternary): Properly
2258 verify the [VEC_]COND_EXPR embedded comparison.
2260 2018-05-15 Martin Sebor <msebor@redhat.com>
2262 PR tree-optimization/85753
2263 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
2264 RECORD_TYPE in addition to ARRAY_TYPE.
2266 2018-05-15 Martin Sebor <msebor@redhat.com>
2269 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
2271 2018-05-15 Richard Biener <rguenther@suse.de>
2273 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
2274 add by_clobber_p one. Change algorithm to collect all defs
2275 representing uses we need to walk and try reducing them to
2276 a single one before failing.
2277 (dse_dom_walker::dse_optimize_stmt): Adjust.
2279 2018-05-13 Mark Wielaard <mark@klomp.org>
2281 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
2282 (size_of_loc_descr): Likewise.
2283 (output_loc_operands): Likewise.
2284 (output_loc_operands_raw): Likewise.
2285 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
2286 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
2287 (hash_loc_operands): Likewise.
2288 (compare_loc_operands): Likewise.
2290 2018-05-14 Mark Wielaard <mark@klomp.org>
2292 * dwarf2out.c (count_index_addrs): New function.
2293 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
2295 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2297 PR tree-optimization/83648
2298 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
2299 return value as malloc candidate.
2301 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
2304 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite param
2305 as true in call to suggest_attribute.
2307 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
2309 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
2310 -mreadonly-in-sdata.
2312 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2314 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
2316 (aarch64_crypto_aesd_fused): Likewise.
2318 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
2320 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
2321 (movsi_aarch64): Likewise.
2322 (load_pairsi): Likewise.
2323 (load_pairdi): Likewise.
2324 (store_pairsi): Likewise.
2325 (store_pairdi): Likewise.
2326 (load_pairsf): Likewise.
2327 (load_pairdf): Likewise.
2328 (store_pairsf): Likewise.
2329 (store_pairdf): Likewise.
2330 (zero_extend): Likewise.
2331 (trunc): Swap alternatives.
2332 (fcvt_target): Add '?' to prefer w over r.
2334 2018-05-14 Jakub Jelinek <jakub@redhat.com>
2337 * config/i386/i386.md: Disallow non-commutative arithmetics in
2338 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
2339 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
2340 in the peephole2 before it.
2342 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
2344 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
2345 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
2346 (ix86_handle_option): Handle -mcldemote.
2347 * config.gcc: New header.
2348 * config/i386/cldemoteintrin.h: New file.
2349 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
2350 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
2352 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2353 OPTION_MASK_ISA_CLDEMOTE.
2354 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
2355 (ix86_valid_target_attribute_inner_p): Ditto.
2356 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
2357 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
2358 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
2359 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
2360 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
2362 * config/i386/i386.opt: Add -mcldemote.
2363 * config/i386/x86intrin.h: New header.
2364 * doc/invoke.texi: Add -mcldemote.
2366 2018-05-14 Richard Biener <rguenther@suse.de>
2368 * doc/match-and-simplify.texi: Adjust :s documentation.
2370 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
2372 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
2373 intended memcpy size.
2374 (REORDER_45): Likewise.
2376 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
2378 * sort.cc: New file.
2379 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
2380 * vec.c (qsort_chk): Use gcc_qsort.
2381 * Makefile.in (OBJS-libcommon): Add sort.o.
2382 (build/sort.o): New target. Use it...
2383 (BUILD_RTL): ... here, and...
2384 (build/gencfn-macros): ... here, and...
2385 (build/genmatch): ... here.
2387 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2388 Chung-Ju Wu <jasonwucj@gmail.com>
2390 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
2391 * config/nds32/nds32-graywolf.md: New file.
2392 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
2393 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
2395 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
2396 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
2397 * config/nds32/nds32.md (pipeline_model): Add graywolf.
2398 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
2399 * config/nds32/pipelines.md: Include n15 settings.
2401 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2402 Chung-Ju Wu <jasonwucj@gmail.com>
2404 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
2405 * config/nds32/nds32-n13.md: New file.
2406 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
2407 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
2409 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
2410 * config/nds32/nds32.md (pipeline_model): Add n13.
2411 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
2412 * config/nds32/pipelines.md: Include n13 settings.
2414 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
2415 Chung-Ju Wu <jasonwucj@gmail.com>
2417 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
2418 * config/nds32/nds32-n10.md: New file.
2419 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
2420 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
2422 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
2423 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
2424 * config/nds32/nds32.md (pipeline_model): Add n10.
2425 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
2426 * config/nds32/pipelines.md: Include n10 settings.
2428 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
2429 Kito Cheng <kito.cheng@gmail.com>
2430 Chung-Ju Wu <jasonwucj@gmail.com>
2432 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
2433 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
2434 Add enum values for DSP extension instructions.
2435 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
2437 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
2438 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
2440 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
2441 * config/nds32/nds32-dspext.md: New file for DSP implementation.
2442 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
2443 * config/nds32/nds32-intrinsic.md: Likewise.
2444 * config/nds32/nds32_intrinsic.h: Likewise.
2445 * config/nds32/nds32-md-auxiliary.c: Likewise.
2446 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
2447 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
2448 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
2449 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
2450 * config/nds32/nds32-protos.h: New declarations for DSP extension.
2451 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
2452 TYPE_DMAC in switch statement.
2453 * config/nds32/nds32.c: New checking and implementation for DSP
2454 extension instructions.
2455 * config/nds32/nds32.h: Likewise.
2456 * config/nds32/nds32.md: Likewise.
2457 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
2458 * config/nds32/predicates.md: Implement new predicates for DSP
2461 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
2463 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
2464 Reformat alternatives and attributes so it is easier to identify
2465 which constraints/attributes go with which instruction.
2466 (mov<mode>_hardfloat32, FMOVE64): Likewise.
2467 (mov<mode>_softfloat32, FMOVE64): Likewise.
2468 (mov<mode>_hardfloat64, FMOVE64): Likewise.
2469 (mov<mode>_softfloat64, FMOVE64): Likewise.
2471 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
2473 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2475 (Basic PowerPC Built-in Functions): The new name of the
2476 subsection previously known as "PowerPC Built-in Functions".
2477 (Basic PowerPC Built-in Functions Available on all Configurations):
2479 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
2480 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
2481 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
2482 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
2484 2018-05-11 Martin Jambor <mjambor@suse.cz>
2487 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
2490 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2493 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
2495 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
2497 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
2498 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
2499 (ix86_handle_option): Handle -mwaitpkg.
2500 * config.gcc: New header.
2501 * config/i386/cpuid.h (bit_WAITPKG): New bit.
2502 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
2503 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
2505 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2506 OPTION_MASK_ISA_WAITPKG.
2507 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
2508 (ix86_option_override_internal): Add PTA_WAITPKG.
2509 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
2510 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
2511 IX86_BUILTIN_TPAUSE.
2512 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
2513 __builtin_ia32_umwait and __builtin_ia32_tpause.
2514 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
2515 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
2516 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
2517 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
2518 UNSPECV_TPAUSE): New.
2519 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
2520 * config/i386/i386.opt: Add -mwaitpkg.
2521 * config/i386/waitpkgintrin.h: New file.
2522 * config/i386/x86intrin.h: New header.
2523 * doc/invoke.texi: Add -mwaitpkg.
2525 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
2528 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
2530 (cortex-m0): Use armv6s-m isa.
2531 (cortex-m0plus): Likewise.
2532 (cortex-m1): Likewise.
2533 (cortex-m0.small-multiply): Likewise.
2534 (cortex-m0plus.small-multiply): Likewise.
2535 (cortex-m1.small-multiply): Likewise.
2537 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
2538 Jakub Jelinek <jakub@redhat.com>
2540 PR tree-optimization/85692
2541 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
2542 source permute as well.
2544 2018-05-11 Martin Liska <mliska@suse.cz>
2547 * doc/extend.texi: Document LLVM style format for no_sanitize
2550 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
2552 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
2553 mode_supports_vsx_dform_quad to mode_supports_dq_form.
2554 (mode_supports_vsx_dform_quad): Likewise.
2555 (mode_supports_vmx_dform): Move these functions to be next to the
2556 other mode_supports functions.
2557 (mode_supports_dq_form): Likewise.
2558 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
2559 mode_supports_dq_form.
2560 (reg_offset_addressing_ok_p): Likewise.
2561 (offsettable_ok_by_alignment): Likewise.
2562 (rs6000_legitimate_offset_address_p): Likewise.
2563 (legitimate_lo_sum_address_p): Likewise.
2564 (rs6000_legitimize_address): Likewise.
2565 (rs6000_legitimize_reload_address): Likewise.
2566 (rs6000_secondary_reload_inner): Likewise.
2567 (rs6000_preferred_reload_class): Likewise.
2568 (rs6000_output_move_128bit): Likewise.
2570 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2572 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
2573 Generate SImode target register for null target.
2574 <case IX86_BUILTIN_XGETBV>: Ditto.
2575 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
2576 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
2578 2018-05-10 Carl Love <cel@us.ibm.com>
2580 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
2581 dcbtt and dcbtstt if operands[2] is 0.
2583 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
2586 * config/i386/sse.md (usadv64qi): New expander.
2588 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
2590 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
2591 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
2592 -maltivec=be support.
2593 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
2594 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
2595 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
2596 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
2597 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
2598 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
2599 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
2600 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
2601 altivec_vsumsws): Adjust.
2602 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
2603 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
2604 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
2606 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
2607 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
2608 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
2609 (altivec_lve<VI_char>x): Delete expand.
2610 (*altivec_lve<VI_char>x_internal): Rename to...
2611 (altivec_lve<VI_char>x): ... this.
2612 (altivec_lvxl_<mode>): Delete expand.
2613 (*altivec_lvxl_<mode>_internal): Rename to ...
2614 (altivec_lvxl_<mode>): ... this.
2615 (altivec_stvxl_<mode>): Delete expand.
2616 (*altivec_stvxl_<mode>_internal): Rename to ...
2617 (altivec_stvxl_<mode>): ... this.
2618 (altivec_stve<VI_char>x): Delete expand.
2619 (*altivec_stve<VI_char>x_internal): Rename to ...
2620 (altivec_stve<VI_char>x): ... this.
2621 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
2622 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
2623 reduc_plus_scal_<mode>): Adjust.
2624 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
2626 (rs6000_cpu_cpp_builtins): Adjust.
2627 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
2628 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
2629 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
2630 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
2631 -maltivec=be support.
2632 (rs6000_split_vec_extract_var): Adjust.
2633 (rs6000_split_v4si_init): Adjust.
2634 (swap_selector_for_mode): Delete.
2635 (altivec_expand_lvx_be, altivec_expand_stvx_be,
2636 altivec_expand_stvex_be): Delete.
2637 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
2638 -maltivec=be support.
2639 (rs6000_gimple_fold_builtin): Ditto.
2640 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
2642 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
2643 (TARGET_DIRECT_MOVE_64BIT): Adjust.
2644 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
2645 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
2646 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
2647 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
2648 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
2649 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
2650 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
2651 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
2652 anonymous split): Adjust.
2653 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
2654 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
2656 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
2658 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
2659 when --with-gxx-include-dir is also specified.
2660 * configure: Regenerate.
2662 2018-05-09 Jim Wilson <jimw@sifive.com>
2665 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
2666 * config/riscv/t-withmultilib: New.
2667 * config/riscv/withmultilib.h: New.
2668 * doc/install.texi: Document RISC-V --with-multilib-list support.
2670 2018-05-09 Richard Biener <rguenther@suse.de>
2672 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
2674 (vect_bb_vectorization_profitable_p): Adjust. Compute
2675 actual scalar cost using the cost vector and the add_stmt_cost
2678 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2680 PR rtl-optimization/85645
2681 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
2682 in the REG_CFA_REGISTER note for LR, don't leave it empty.
2684 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2686 PR rtl-optimization/85645
2687 * shrink-wrap.c (spread_components): Return a boolean saying if
2688 anything was changed.
2689 (try_shrink_wrapping_separate): Iterate spread_components until
2690 nothing changes anymore.
2692 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2694 PR rtl-optimization/85645
2695 * regrename.c (build_def_use): Also kill the chains that include the
2696 destination of a REG_CFA_REGISTER note.
2698 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
2700 PR rtl-optimization/85645
2701 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
2702 insn that has a REG_CFA_REGISTER note.
2704 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
2706 * cfgexpand.c (expand_clobber): New function.
2707 (expand_gimple_stmt_1): Use it.
2708 * tree-vect-stmts.c (vect_clobber_variable): New function,
2710 (vectorizable_simd_clone_call): ...here.
2711 (vectorizable_store): Emit a clobber either side of an
2712 IFN_STORE_LANES sequence.
2713 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
2715 2018-05-09 Tom de Vries <tom@codesourcery.com>
2718 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
2719 (define_insn "trap_if_false"): Add exit after trap.
2721 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
2723 PR rtl-optimization/85638
2724 * bb-reorder.c: Include common/common-target.h.
2725 (create_forwarder_block): New function extracted from...
2726 (fix_up_crossing_landing_pad): ...here. Rename into...
2727 (dw2_fix_up_crossing_landing_pad): ...this.
2728 (sjlj_fix_up_crossing_landing_pad): New function.
2729 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
2730 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
2731 from both partitions and exit the loop after one iteration.
2733 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2736 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2738 (Basic PowerPC Built-in Functions): The new name of the
2739 subsection previously known as "PowerPC Built-in Functions".
2740 (Basic PowerPC Built-in Functions Available on all Configurations):
2742 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2744 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2746 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2748 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2751 2018-05-08 Jim Wilson <jimw@sifive.com>
2753 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
2754 (LD_EMUL_SUFFIX): New.
2755 (LINK_SPEC): Use it.
2757 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
2759 * doc/extend.texi (PowerPC Built-in Functions): Rename this
2761 (Basic PowerPC Built-in Functions): The new name of the
2762 subsection previously known as "PowerPC Built-in Functions".
2763 (Basic PowerPC Built-in Functions Available on all Configurations):
2765 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
2767 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
2769 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
2771 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
2774 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2777 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
2778 after cmpelim optimization.
2780 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
2782 * config.gcc: Support "goldmont".
2783 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
2784 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
2786 * config/i386/i386.c (m_GOLDMONT): Define.
2787 (processor_target_table): Add "goldmont".
2788 (PTA_GOLDMONT): Define.
2789 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
2790 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
2791 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
2792 (fold_builtin_cpu): Add "goldmont".
2793 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
2794 (ix86_option_override_internal): Add "goldmont".
2795 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
2796 (processor_type): Add PROCESSOR_GOLDMONT.
2797 * config/i386/i386.md: Add CPU "glm".
2798 * config/i386/glm.md: New file.
2799 * config/i386/x86-tune.def: Add m_GOLDMONT.
2800 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
2802 2018-05-08 Jakub Jelinek <jakub@redhat.com>
2805 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
2807 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
2808 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
2809 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
2812 * config/i386/i386.c (ix86_fold_builtin): Handle
2813 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
2816 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
2817 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
2819 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
2822 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
2823 (check_arch): Likewise.
2824 (check_fpu): Return the result rather than printing it.
2825 (end arch): Fix operator precedence.
2826 (end cpu): Likewise.
2827 (END): Print the result from check_fpu.
2829 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2830 Alan Hayward <alan.hayward@arm.com>
2831 David Sherwood <david.sherwood@arm.com>
2833 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
2834 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
2835 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
2836 (*fcmuo<mode>_and): New patterns.
2838 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
2840 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
2841 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
2842 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
2843 (cmp_op, sve_imm_con): New code attributes.
2844 (SVE_COND_INT_CMP, imm_con): Delete.
2845 (cmp_op): Remove above unspecs from int attribute.
2846 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
2848 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
2849 comparison-specific unspecs.
2850 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
2851 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
2852 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
2853 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
2854 (*vec_fcm<cmp_op><mode>): Rename to...
2855 (*fcm<cmp_op><mode>): ...this and adjust likewise.
2856 (*vec_fcmuo<mode>): Rename to...
2857 (*fcmuo<mode>): ...this and adjust likewise.
2858 (*pred_fcm<cmp_op><mode>): New pattern.
2859 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
2860 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
2862 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
2864 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
2865 (aarch64_emit_sve_predicated_cond): New function.
2866 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
2867 (aarch64_emit_unspec_cond_or): Replace with...
2868 (aarch64_emit_sve_or_conds): ...this new function. Use
2869 aarch64_emit_sve_ptrue_op for the individual comparisons and
2870 aarch64_emit_binop to OR them together.
2871 (aarch64_emit_inverted_unspec_cond): Replace with...
2872 (aarch64_emit_sve_inverted_cond): ...this new function. Use
2873 aarch64_emit_sve_ptrue_op for the comparison and
2874 aarch64_emit_unop to invert the result.
2875 (aarch64_expand_sve_vec_cmp_float): Update after the above
2876 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
2878 2018-05-07 Nathan Sidwell <nathan@acm.org>
2880 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
2881 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
2882 (Backwards Compatibility): Likewise.
2884 2018-05-07 Luis Machado <luis.machado@linaro.org>
2888 2018-05-07 Luis Machado <luis.machado@linaro.org>
2890 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2891 <prefetch_dynamic_strides>: New const bool field.
2892 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2893 prefetch_dynamic_strides.
2894 (exynosm1_prefetch_tune): Likewise.
2895 (thunderxt88_prefetch_tune): Likewise.
2896 (thunderx_prefetch_tune): Likewise.
2897 (thunderx2t99_prefetch_tune): Likewise.
2898 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2899 (aarch64_override_options_internal): Update to set
2900 PARAM_PREFETCH_DYNAMIC_STRIDES.
2901 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2902 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2903 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2904 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2905 prefetch-dynamic-strides setting.
2907 2018-05-07 Luis Machado <luis.machado@linaro.org>
2909 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2910 <minimum_stride>: New const int field.
2911 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2912 minimum_stride field.
2913 (exynosm1_prefetch_tune): Likewise.
2914 (thunderxt88_prefetch_tune): Likewise.
2915 (thunderx_prefetch_tune): Likewise.
2916 (thunderx2t99_prefetch_tune): Likewise.
2917 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2918 (aarch64_override_options_internal): Update to set
2919 PARAM_PREFETCH_MINIMUM_STRIDE.
2920 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2921 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2922 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2923 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2924 stride is constant and is below the minimum stride threshold.
2926 2018-05-07 Luis Machado <luis.machado@linaro.org>
2928 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
2931 2018-05-07 Luis Machado <luis.machado@linaro.org>
2933 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2934 <prefetch_dynamic_strides>: New const bool field.
2935 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2936 prefetch_dynamic_strides.
2937 (exynosm1_prefetch_tune): Likewise.
2938 (thunderxt88_prefetch_tune): Likewise.
2939 (thunderx_prefetch_tune): Likewise.
2940 (thunderx2t99_prefetch_tune): Likewise.
2941 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to false.
2942 (aarch64_override_options_internal): Update to set
2943 PARAM_PREFETCH_DYNAMIC_STRIDES.
2944 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2945 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2946 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2947 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2948 prefetch-dynamic-strides setting.
2950 2018-05-07 Luis Machado <luis.machado@linaro.org>
2952 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2953 <minimum_stride>: New const int field.
2954 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2955 minimum_stride field.
2956 (exynosm1_prefetch_tune): Likewise.
2957 (thunderxt88_prefetch_tune): Likewise.
2958 (thunderx_prefetch_tune): Likewise.
2959 (thunderx2t99_prefetch_tune): Likewise.
2960 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
2961 (aarch64_override_options_internal): Update to set
2962 PARAM_PREFETCH_MINIMUM_STRIDE.
2963 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2964 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2965 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2966 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2967 stride is constant and is below the minimum stride threshold.
2969 2018-05-06 Jakub Jelinek <jakub@redhat.com>
2972 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
2973 the type is addressable. Don't force op into register if it has
2976 2018-05-05 Roland McGrath <mcgrathr@google.com>
2979 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
2980 any section for which we don't know a specific type it should have,
2981 regardless of name. Previously this was done only for the exact
2982 names ".init_array", ".fini_array", and ".preinit_array".
2983 (default_elf_asm_named_section): Add comment about
2984 relationship with default_section_type_flags and SECTION_NOTYPE.
2985 (get_section): Don't consider it a type conflict if one side has
2986 SECTION_NOTYPE and the other doesn't, as long as neither has the
2987 SECTION_BSS et al used in the default_section_type_flags logic.
2989 2018-05-05 Tom de Vries <tom@codesourcery.com>
2992 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
2993 (workaround_barsyncs): New function.
2994 (nvptx_reorg): Use workaround_barsyncs.
2995 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
2996 (define_expand "nvptx_membar_cta"): New define_expand.
2997 (define_insn "*nvptx_membar_cta"): New insn.
2999 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
3001 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
3002 To improve optimization opportunities.
3003 * builtin-types.def: The new needed builtin types for the above.
3005 2018-05-04 Richard Biener <rguenther@suse.de>
3007 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
3008 * gimple-ssa-store-merging.c
3009 (imm_store_chain_info::output_merged_store): Remove redundant create,
3010 release split_store vector contents on failure.
3011 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
3012 scalar stmt vector on cache hit.
3014 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
3016 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
3018 * config.gcc (powerpc-xilinx-eabi*): Remove.
3019 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
3021 (fusion_addis_mem_combo_load): Ditto.
3022 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
3024 (rs6000_cpu_cpp_builtins): Ditto.
3025 * config/rs6000/rs6000-linux.c
3026 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
3027 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
3028 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
3030 (rs6000_setup_reg_addr_masks): Ditto.
3031 (rs6000_init_hard_regno_mode_ok): Ditto.
3032 (rs6000_option_override_internal): Ditto.
3033 (legitimate_lo_sum_address_p): Ditto.
3034 (rs6000_legitimize_address): Ditto.
3035 (rs6000_legitimize_reload_address): Ditto.
3036 (rs6000_legitimate_address_p): Ditto.
3037 (abi_v4_pass_in_fpr): Ditto.
3038 (setup_incoming_varargs): Ditto.
3039 (rs6000_gimplify_va_arg): Ditto.
3040 (rs6000_split_multireg_move): Ditto.
3041 (rs6000_savres_strategy): Ditto.
3042 (rs6000_emit_prologue_components): Ditto.
3043 (rs6000_emit_epilogue_components): Ditto.
3044 (rs6000_emit_prologue): Ditto.
3045 (rs6000_emit_epilogue): Ditto.
3046 (rs6000_elf_file_end): Ditto.
3047 (rs6000_function_value): Ditto.
3048 (rs6000_libcall_value): Ditto.
3049 * config/rs6000/rs6000.h: Ditto.
3050 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
3051 (TARGET_MINMAX): ... this. New.
3052 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
3053 * config/rs6000/rs6000.md: Remove Xilinx FP support.
3054 (*movsi_internal1_single): Delete.
3055 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
3056 mfpu=, mxilinx-fpu): Delete.
3057 * config/rs6000/singlefp.h: Delete.
3058 * config/rs6000/sysv4.h: Remove Xilinx FP support.
3059 * config/rs6000/t-rs6000: Ditto.
3060 * config/rs6000/t-xilinx: Delete.
3061 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
3062 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
3063 (VStype_simple): Delete.
3064 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
3065 * config/rs6000/xfpu.h: Delete.
3066 * config/rs6000/xfpu.md: Delete.
3067 * config/rs6000/xilinx.h: Delete.
3068 * config/rs6000/xilinx.opt: Delete.
3069 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
3070 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
3072 2018-05-04 Tom de Vries <tom@codesourcery.com>
3075 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
3078 2018-05-04 Richard Biener <rguenther@suse.de>
3081 * tree-complex.c (update_complex_assignment): We are always in SSA form.
3082 (expand_complex_div_wide): Likewise.
3083 (expand_complex_operations_1): Likewise.
3084 (expand_complex_libcall): Preserve EH info of the original stmt.
3085 (tree_lower_complex): Handle removed blocks.
3086 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
3087 on complex multiplication and division libcall builtins.
3089 2018-05-04 Richard Biener <rguenther@suse.de>
3092 * fold-const.c (negate_expr_p): Restrict negation of operand
3093 zero of a division to when we know that can happen without
3095 (fold_negate_expr_1): Likewise.
3097 2018-05-04 Jakub Jelinek <jakub@redhat.com>
3100 * real.h (real_nextafter): Declare.
3101 * real.c (real_nextafter): New function.
3102 * fold-const-call.c (fold_const_nextafter): New function.
3103 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
3104 CASE_CFN_NEXTTOWARD.
3105 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
3106 even when arg1_mode is different from arg0_mode.
3108 2018-05-03 Nathan Sidwell <nathan@acm.org>
3110 * doc/extend.texi (Deprecated Features): Remove
3112 (Backwards Compatibility): Likewise.
3113 * doc/invoke.texi (C++ Language Options): Likewise.
3114 (C++ Dialect Options): Likewise.
3116 2018-05-03 Jakub Jelinek <jakub@redhat.com>
3119 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
3120 _mm512_mask_mullox_epi64): New intrinsics.
3122 2018-05-03 Tom de Vries <tom@codesourcery.com>
3125 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3126 dump files): Add offload-tree.
3128 2018-05-03 Richard Biener <rguenther@suse.de>
3130 PR tree-optimization/85615
3131 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
3132 to loops not nested in BBs loop father to avoid creating multi-entry
3135 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3137 PR tree-optimization/70291
3138 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
3139 arguments. Change return type to tree. Emit libcall as a new
3140 statement rather than replacing existing one when inplace_p is true.
3141 (expand_complex_multiplication_components): New function.
3142 (expand_complex_multiplication): Expand floating-point complex
3143 multiplication using the above.
3144 (expand_complex_division): Rename inner_type parameter to type.
3145 Update expand_complex_libcall call-site.
3146 (expand_complex_operations_1): Update expand_complex_multiplication
3147 and expand_complex_division call-sites.
3149 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3152 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3153 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3154 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
3155 the highest significant bit of the shift count mask is clear. In
3156 check whether and[sq]i3 is needed verify that all significant bits
3157 of the shift count other than the highest are set.
3159 2018-05-02 Tom de Vries <tom@codesourcery.com>
3162 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
3163 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
3164 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
3165 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
3166 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
3167 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
3168 __builtin_goacc_parlevel_size.
3170 2018-05-02 Richard Biener <rguenther@suse.de>
3172 PR tree-optimization/85597
3173 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
3174 do not use split vect_get_vec_defs call but call vect_get_slp_defs
3177 2018-05-02 Tom de Vries <tom@codesourcery.com>
3180 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3181 dump files): Add ltrans-tree.
3183 2018-05-02 Tom de Vries <tom@codesourcery.com>
3186 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
3187 dump files): Add wpa-ipa.
3189 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
3191 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
3192 powerpc*-*-linux*paired* target.
3193 * config/rs6000/750cl.h: Delete.
3194 * config/rs6000/paired.h: Delete.
3195 * config/rs6000/paired.md: Delete.
3196 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
3198 * config/rs6000/rs6000-builtin.def: Remove paired float support.
3199 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
3200 comment. Remove paired float support.
3201 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
3202 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
3204 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
3205 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
3207 * config/rs6000/rs6000.c: Remove paired float support.
3208 (paired_expand_vector_init, paired_expand_vector_move,
3209 paired_emit_vector_compare, paired_emit_vector_cond_expr,
3210 (paired_expand_lv_builtin, paired_expand_stv_builtin,
3211 paired_expand_builtin, paired_expand_predicate_builtin,
3212 paired_init_builtins): Delete.
3213 * config/rs6000/rs6000.h: Remove paired float support.
3214 * config/rs6000/rs6000.md: Remove paired float support.
3215 (move_from_CR_ov_bit): Delete.
3216 * config/rs6000/rs6000.opt (mpaired): Delete.
3217 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
3218 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
3220 2018-05-02 Richard Biener <rguenther@suse.de>
3223 * gimplify.c (gimplify_save_expr): When in SSA form allow
3224 SAVE_EXPRs to compute to SSA vars.
3226 2018-05-02 Jakub Jelinek <jakub@redhat.com>
3229 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
3230 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
3231 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
3232 clobber operands[2], instead use a new pseudo. Formatting fixes.
3234 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
3236 PR tree-optimization/85586
3237 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
3238 exit early for statements in the same group if the accesses are
3241 2018-05-02 Tom de Vries <tom@codesourcery.com>
3244 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
3247 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
3249 PR tree-optimization/85143
3250 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
3252 2018-05-01 Tom de Vries <tom@codesourcery.com>
3255 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
3256 not found" error message.
3258 2018-05-01 Tom de Vries <tom@codesourcery.com>
3261 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
3262 * vec.c (test_ordered_remove_if): New function.
3263 (vec_c_tests): Call test_ordered_remove_if.
3264 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
3265 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
3266 * tree-vect-patterns.c (vect_pattern_recog_1): Use
3267 VEC_ORDERED_REMOVE_IF.
3269 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3271 PR tree-optimization/82665
3272 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
3273 pointer subtraction where arguments come from a memchr call.
3275 2018-05-01 Jakub Jelinek <jakub@redhat.com>
3277 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
3278 --push-state --as-needed and --pop-state instead of --as-needed and
3279 --no-as-needed if ld supports it.
3280 * configure: Regenerated.
3283 * doc/install.texi2html: Replace _002d with - and _002a with * in
3284 generated html files using sed.
3286 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3289 * gcc-rich-location.c (blank_line_before_p): New function.
3290 (use_new_line): New function.
3291 (gcc_rich_location::add_fixit_insert_formatted): New function.
3292 * gcc-rich-location.h
3293 (gcc_rich_location::add_fixit_insert_formatted): New function.
3295 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3297 * selftest.c (assert_streq): Rename "expected" and "actual" to
3298 "val1" and "val2". Extend NULL-handling to cover both inputs
3299 symmetrically, while still requiring both to be non-NULL for a pass.
3300 * selftest.h (assert_streq): Rename "expected" and "actual" to
3302 (ASSERT_EQ): Likewise.
3303 (ASSERT_EQ_AT): Likewise.
3304 (ASSERT_KNOWN_EQ): Likewise.
3305 (ASSERT_KNOWN_EQ_AT): Likewise.
3306 (ASSERT_NE): Likewise.
3307 (ASSERT_MAYBE_NE): Likewise.
3308 (ASSERT_MAYBE_NE_AT): Likewise.
3309 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
3310 the assertion to pass.
3311 (ASSERT_STREQ_AT): Likewise.
3313 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3315 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
3316 interaction with -pie.
3318 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3320 * selftest.h: Fix alphabetization of per-source-file selftest
3323 2018-04-30 Jason Merrill <jason@redhat.com>
3325 PR c++/61982 - dead stores to destroyed objects.
3326 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
3329 2018-04-30 Jason Merrill <jason@redhat.com>
3331 * tree.c (build_clobber): New.
3332 * tree.h: Declare it.
3333 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
3335 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3337 * diagnostic-show-locus.c (layout::layout): Update for
3338 location_get_source_line returning a char_span.
3339 (struct char_span): Move to input.h.
3340 (struct correction): Update for fields in char_span becoming
3342 (struct source_line): Update for location_get_source_line
3343 returning a char_span.
3344 (layout::print_line): Likewise.
3345 * edit-context.c (edited_file::print_content): Likewise.
3346 (edited_file::print_diff_hunk): Likewise.
3347 (edited_file::print_run_of_changed_lines): Likewise.
3348 (edited_file::get_num_lines): Likewise.
3349 (edited_line::edited_line): Likewise.
3350 * final.c (asm_show_source): Likewise.
3351 * input.c (location_get_source_line): Convert return type
3352 from const char * to char_span, losing the final "line_len"
3354 (dump_location_info): Update for the above.
3355 (get_substring_ranges_for_loc): Likewise. Use a char_span
3356 when handling the literal within the line.
3357 (test_reading_source_line): Update for location_get_source_line
3358 returning a char_span.
3359 * input.h (class char_span): Move here from
3360 diagnostic-show-locus.c, converting from a struct to a class.
3361 Make data members private.
3362 (char_span::operator bool): New.
3363 (char_span::length): New.
3364 (char_span::get_buffer): New.
3365 (char_span::operator[]): New.
3366 (char_span::subspan): Make const.
3367 (char_span::xstrdup): New.
3368 (location_get_source_line): Convert return type from const char *
3369 to char_span, losing the final "line_size" param.
3371 2018-04-30 Jan Hubicka <jh@suse.cz>
3373 * lto-wrapper.c (ltrans_priorities): New static var.
3374 (cmp_priority): New.
3375 (run_gcc): Read priorities and if doing parallel build order
3376 the Makefile by them.
3378 2018-04-30 David Malcolm <dmalcolm@redhat.com>
3380 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
3382 2018-04-30 Richard Biener <rguenther@suse.de>
3384 * tree-cfg.c (verify_address): Remove base argument, add
3385 flag whether to check TREE_ADDRESSABLE and do that.
3386 (verify_expr): Remove.
3387 (verify_types_in_gimple_reference): Add pieces from verify_expr.
3388 (verify_gimple_assign_single): Likewise.
3389 (verify_gimple_switch): Likewise.
3390 (verify_expr_location_1): Dereference tp once. Add (disabled)
3391 piece from verify_expr.
3392 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
3394 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3396 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
3398 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3400 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
3401 (small_data_pattern): Likewise.
3402 (arc_rewrite_small_data): Likewise.
3403 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
3404 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
3405 (get_symbol_alignment): New function.
3406 (legitimate_small_data_address_p): Likewise.
3407 (legitimate_scaled_address): Update, call
3408 legitimate_small_data_address_p.
3409 (output_sdata): New static variable.
3410 (arc_print_operand): Update how we handle small data operands.
3411 (arc_print_operand_address): Likewise.
3412 (arc_legitimate_address_p): Update, use
3413 legitimate_small_data_address_p.
3414 (arc_rewrite_small_data_p): Remove.
3415 (arc_rewrite_small_data_1): Likewise.
3416 (arc_rewrite_small_data): Likewise.
3417 (small_data_pattern): Likewise.
3418 (compact_sda_memory_operand): Update to use
3419 legitimate_small_data_address_p and get_symbol_alignment.
3420 (prepare_move_operands): Don't rewite sdata pattern.
3421 (prepare_extend_operands): Remove.
3422 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
3424 (zero_extendqisi2): Likewise.
3425 (zero_extendhisi2): Likewise.
3426 (extendqihi2): Likewise.
3427 (extendqisi2): Likewise.
3428 (extendhisi2): Likewise.
3432 * config/arc/constraints.md (Usd): Change it to memory constraint.
3434 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
3436 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
3437 as source of std instructions.
3438 * config/arc/arc.md (movsi_insn): Update pattern predicate to
3439 allow 6-bit constants as source for store instructions.
3440 (movdi_insn): Update instruction pattern to allow 6-bit constants
3441 as source for store instructions.
3443 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
3445 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
3447 2018-04-30 Nathan Sidwell <nathan@acm.org>
3448 Sandra Loosemore <sandra@codesourcery.com>
3450 * dumpfile.c (dump_open): Allow '-' for stdout.
3451 * doc/invoke.texi (Developer Options): Document dump filename
3452 determination early. Document stdin/stdout selection.
3454 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
3456 Microblaze Target: PIC data text relative
3458 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
3459 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
3461 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
3462 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
3463 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
3464 New addressing mode for data-text relative position indepenedent code.
3465 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
3466 'ADDRESS_SYMBOLIC_TXT_REL'.
3467 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
3468 (microblaze_legitimate_pic_operand): Exclude function calls from
3469 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
3470 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
3472 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3473 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
3474 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
3475 for 'address + offset'.
3476 (microblaze_expand_prologue): Add new function prologue call for
3478 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
3479 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
3480 table in case of TARGET_PIC_DATA_TEXT_REL.
3481 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
3482 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
3483 Add new macros 'UNSPEC_TEXT',
3484 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
3485 + exclude function calls from 'UNSPEC_PLT' in case of data text
3487 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3488 new target hook for generating address diff vector tables in case of
3490 * doc/tm.texi : Regenerate.
3491 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
3492 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
3493 of addr diff vector generation.
3494 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
3495 target hook definition.
3496 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
3497 Add default function for generate_pic_addr_diff_vec -> flag_pic.
3498 * doc/invoke.texi (Add new pic option): Add new microblaze pic
3499 option for data text relative.
3501 2018-04-30 Richard Biener <rguenther@suse.de>
3503 * tree-chrec.h (evolution_function_is_constant_p): Remove
3505 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
3507 2018-04-30 Richard Biener <rguenther@suse.de>
3510 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
3512 2018-04-30 Richard Biener <rguenther@suse.de>
3514 PR tree-optimization/28364
3515 PR tree-optimization/85275
3516 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
3517 copying first exit test.
3519 2018-04-28 Mark Wielaard <mark@klomp.org>
3521 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
3523 (dwarf_AT): Handle DW_AT_addr_base.
3524 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
3526 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
3529 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
3530 (*ashl<dwi>3_doubleword_mask_1): Ditto.
3531 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
3532 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
3534 2018-04-28 Richard Biener <rguenther@suse.de>
3536 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
3537 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
3538 to reflect use. Only add interesting stmts.
3540 2018-04-27 Martin Jambor <mjambor@suse.cz>
3543 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
3544 the jump function allows for passing through aggregate values.
3546 2018-04-27 David Malcolm <dmalcolm@redhat.com>
3548 * input.h (in_system_header_at): Convert from macro to inline
3550 (from_macro_expansion_at): Likewise.
3551 (from_macro_definition_at): Likewise.
3553 2018-04-27 Jeff Law <law@redhat.com>
3555 * config.gcc: Mark tile* targets as deprecated/obsolete.
3557 2018-04-27 Richard Biener <rguenther@suse.de>
3559 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
3562 2018-04-27 Richard Biener <rguenther@suse.de>
3564 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
3566 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
3568 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
3569 with Yd constraint. Set "preferred_for_speed" attribute from
3570 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
3572 (*movdi_internal): Ditto.
3573 (movti_interunit splitters): Remove
3574 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
3575 (movdi_interunit splitters): Ditto.
3576 * config/i386/constraints.md (Ye): Remove.
3577 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
3579 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3582 * config/aarch64/constraints.md (Usg): Limit to 31.
3585 2018-04-27 Jakub Jelinek <jakub@redhat.com>
3587 PR tree-optimization/85529
3588 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
3589 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
3590 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
3591 zero extension or masking of the MSB bit.
3592 (optimize_range_tests): Add FIRST_BB argument, pass it through
3593 to optimize_range_tests_var_bound.
3594 (maybe_optimize_range_tests, reassociate_bb): Adjust
3595 optimize_range_tests callers.
3597 2018-04-26 Richard Biener <rguenther@suse.de>
3598 Jakub Jelinek <jakub@redhat.com>
3600 * cgraph.h (symbol_table): Just declare debug method here.
3601 * symtab.c (symbol_table::debug): Define.
3603 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
3605 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
3607 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
3609 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
3610 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
3611 (*movdi_internal): Substitute Yi and Yj constraint with x
3612 and Ym and Yn constraint with y constraint. Update "isa"
3613 attribute and set "preferred_for_speed" attribute from
3614 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
3615 (*movsi_internal): Ditto.
3616 (*movdf_internal): Ditto.
3617 (*movsf_internal): Ditto.
3618 (*zero_extendsidi2): Ditto.
3619 * config/i386/sse.md (vec_set<mode>_0): Ditto.
3620 (sse2_loadld): Ditto.
3621 (*vec_extract<ssevecmodelower>_0): Ditto.
3622 (*vec_extractv4si_0_zext_sse4): Ditto.
3623 (vec_concatv2di): Ditto.
3624 (*vec_dup<mode>): Ditto.
3625 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
3626 * config/i386/constraints.md (Yi): Remove.
3631 2018-04-26 Nathan Sidwell <nathan@acm.org>
3633 * dumpfile.c (dump_open): New.
3634 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
3635 (dump_finish): Detect stdio/stderr by value not name.
3637 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
3639 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
3641 2018-04-26 Tom de Vries <tom@codesourcery.com>
3644 * config/nvptx/nvptx.c (verify_neutering_jumps)
3645 (verify_neutering_labels): New function
3646 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
3648 2018-04-26 Tom de Vries <tom@codesourcery.com>
3651 * config/nvptx/nvptx.c (needs_neutering_p): New function.
3652 (nvptx_single): Use needs_neutering_p to skip over insns that do not
3655 2018-04-26 Richard Biener <rguenther@suse.de>
3656 Tom de Vries <tom@codesourcery.com>
3659 * lto-streamer-out.c (output_function): Fixup loops if required to match
3660 discovery done in the reader.
3662 2018-04-26 Richard Biener <rguenther@suse.de>
3664 PR tree-optimization/85116
3665 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
3666 have a loop exit from the single latch predecessor. Remove
3667 case of header with just condition.
3668 (ch_base::copy_headers): Exclude infinite loops from any
3670 (pass_ch::execute): Record exits.
3672 2018-04-26 Richard Biener <rguenther@suse.de>
3674 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
3675 prologue cost vector and pass it to vect_get_load_cost.
3676 (vect_get_peeling_costs_all_drs): Likewise.
3677 (vect_peeling_hash_get_lowest_cost): Likewise.
3678 (vect_enhance_data_refs_alignment): Likewise.
3680 2018-04-26 Richard Biener <rguenther@suse.de>
3683 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
3684 checking of integer<->pointer conversions.
3685 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
3686 sign-/zero-extending pointer types.
3687 (expand_omp_for_static_chunk): Likewise.
3689 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
3690 Jean Lee <xiaoyur347@gmail.com>
3692 * config/mips/mips.c (mips_asan_shadow_offset): New function.
3693 (TARGET_ASAN_SHADOW_OFFSET): Define.
3694 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
3695 true for -fsanitize=address.
3697 2018-04-25 Mark Wielaard <mark@klomp.org>
3699 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
3702 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3704 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
3705 than "alu", remove explicit "memory" and "imm_disp" attributes.
3706 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
3709 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
3710 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
3711 gen_lowpart_no_emit.
3713 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
3716 * config/i386/i386.c (ix86_expand_builtin): Change memory
3717 operand to XI, extend p0 to Pmode.
3718 * config/i386/i386.md: Change unspec volatile and operand
3719 1 mode to XI, change operand 0 mode to P.
3721 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3723 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
3724 GET_MODE_MASK before any checking.
3725 (nds32_can_use_bset_p): Likewise.
3726 (nds32_can_use_btgl_p): Likewise.
3728 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3730 * config/nds32/nds32-doubleword.md: New define_split pattern for
3731 illegal register number.
3733 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3735 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
3737 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
3739 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
3741 2018-04-25 Richard Biener <rguenther@suse.de>
3743 * lto-streamer.h (LTO_major_version): Bump to 8.
3745 2018-04-25 Jakub Jelinek <jakub@redhat.com>
3747 * BASE-VER: Set to 9.0.0.
3749 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
3751 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
3752 in __abskf2 and __powikf2.
3754 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3757 * config/aarch64/constraints.md (Usg, Usj): New constraints.
3758 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
3759 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
3760 Use the above on operand 2. Reindent.
3761 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
3763 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3766 * common/config/i386/i386-common.c (ix86_handle_option): Don't
3768 * config/i386/i386.opt (mcet): Removed.
3769 * doc/install.texi: Remove -mcet documentation.
3770 * doc/invoke.texi: Likewise.
3772 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
3775 * doc/install.texi: Remove -mcet from bootstrap-cet.
3777 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3780 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
3781 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
3785 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
3786 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
3787 containing a CONST_VECTOR.
3789 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
3791 * doc/install.texi: Update newlib dependency for nvptx.
3793 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3796 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
3797 instead of INTVAL when shifting x left.
3799 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
3801 PR tree-optimization/85478
3802 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
3803 vect_grouped_store_supported for single element vectors.
3805 2018-04-24 Richard Biener <rguenther@suse.de>
3808 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
3809 load cost increase to the case of non-constant step.
3811 2018-04-24 Jakub Jelinek <jakub@redhat.com>
3814 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
3815 destination if any_malformed_asm.
3817 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
3820 * expr.c (store_field): In the bitfield case, if the value comes from
3821 a function call and is returned in registers by means of a PARALLEL,
3822 do not change the mode of the temporary unless BLKmode and VOIDmode.
3824 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
3826 PR rtl-optimization/85423
3827 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
3828 dependencies to debug insns when the previous insn is non-debug.
3830 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3832 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
3833 enums into a single definition.
3834 (fls): Fix predicates and printing.
3837 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
3839 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
3840 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
3841 and short u6 immediate.
3842 (check_if_valid_sleep_operand): Remove.
3843 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
3845 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3847 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
3848 flag_always_save_lp condition.
3849 * config/nds32/nds32.opt (malways-save-lp): New option.
3851 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3853 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
3854 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
3855 * config/nds32/nds32.h
3856 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
3857 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
3859 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3861 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
3862 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
3864 2018-04-22 Shiva Chen <shiva0217@gmail.com>
3865 Chung-Ju Wu <jasonwucj@gmail.com>
3867 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
3869 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
3870 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
3872 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3874 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
3876 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3878 * config/nds32/nds32-protos.h (nds32_data_alignment,
3879 nds32_local_alignment): Declare.
3880 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
3881 nds32_local_alignment): New functions.
3882 (TARGET_CONSTANT_ALIGNMENT): Define.
3883 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
3885 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3887 * config/nds32/nds32.c
3888 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
3889 (TARGET_MODES_TIEABLE_P): Likewise.
3891 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
3893 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
3896 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
3897 Chung-Ju Wu <jasonwucj@gmail.com>
3899 * config/nds32/constants.md (unspec_volatile_element): Add enum values
3900 for unaligned access.
3901 * config/nds32/nds32-intrinsic.c: Implementation of expanding
3903 * config/nds32/nds32-intrinsic.md: Likewise.
3904 * config/nds32/nds32_intrinsic.h: Likewise.
3905 * config/nds32/nds32.h (nds32_builtins): Likewise.
3906 * config/nds32/nds32.opt (munaligned-access): New option.
3907 * config/nds32/nds32.c (nds32_asm_file_start): Display
3908 flag_unaligned_access status.
3910 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3912 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
3913 -mno-relax is present.
3914 * config/riscv/linux.h (LINK_SPEC): Ditto.
3916 2018-04-20 Martin Sebor <msebor@redhat.com>
3919 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
3921 (gimple_fold_builtin_stxcpy_chk): Same.
3922 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
3924 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
3927 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
3928 __powikf2 when long double is IEEE 128-bit.
3930 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
3932 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
3933 step to make sure stack always aligned.
3935 2018-04-20 Carl Love <cel@us.ibm.com>
3938 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
3939 size check for arg0.
3941 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
3942 Tom de Vries <tom@codesourcery.com>
3945 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
3946 Emit insns for calls too.
3947 (nvptx_find_par): Always look for worker-level predecessor insn.
3948 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
3950 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
3951 (nvptx_process_pars): Propagate frames for calls.
3953 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
3956 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
3958 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
3959 (ix86_handle_option): Don't handle OPT_mibt.
3960 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
3962 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
3964 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
3965 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
3966 (ix86_target_macros): Define __CET__ with flag_cf_protection
3967 for -fcf-protection.
3968 * config/i386/i386.c (isa2_opts): Remove -mibt.
3969 * config/i386/i386.h (TARGET_IBT): Removed.
3970 (TARGET_IBT_P): Likewise.
3971 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
3972 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
3973 * config/i386/i386.opt (mcet): Update help message.
3976 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
3977 -mcet as an alias for -mshstk.
3979 2018-04-20 Richard Biener <rguenther@suse.de>
3982 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
3983 complexity by forcing a single use of the multiply operand.
3985 2018-04-20 Martin Jambor <mjambor@suse.cz>
3988 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
3989 recursion dependency to only apply to non-clones.
3991 2018-04-20 Martin Jambor <mjambor@suse.cz>
3994 * ipa-cp.c (create_specialized_node): Check that clones of
3995 self-recursive edges exist during IPA-CP.
3997 2018-04-19 Toon Moene <toon@moene.org>
3999 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
4002 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4004 PR tree-optimization/85467
4005 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
4006 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
4007 VECTOR_CST element to type.
4009 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4012 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
4013 * config/i386/i386.md (builtin_setjmp_setup): Removed.
4014 (builtin_longjmp): Likewise.
4015 (save_stack_nonlocal): New pattern.
4016 (restore_stack_nonlocal): Likewise.
4018 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4021 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4022 Replace ASM_OUTPUT_LABEL with fprintf.
4024 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
4027 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
4028 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
4029 * config/i386/i386-c.c (ix86_target_macros_internal): Also
4030 define __IBT__ and __SHSTK__ for -fcf-protection.
4031 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
4033 (ix86_trampoline_init): Likewise.
4034 (x86_output_mi_thunk): Likewise.
4035 (ix86_notrack_prefixed_insn_p): Likewise.
4036 (ix86_option_override_internal): Don't disallow -fcf-protection.
4037 * config/i386/i386.md (rdssp<mode>): Also enable for
4039 (incssp<mode>): Likewise.
4040 (nop_endbr): Likewise.
4041 * config/i386/i386.opt (mcet): Change help message to built-in
4045 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
4046 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
4047 enable CET built-in functions.
4049 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
4051 * common/config/i386/i386-common.c
4052 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
4053 OPTION_MASK_ISA_MOVDIRI_UNSET,
4054 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
4055 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
4056 * config.gcc (movdirintrin.h): New header.
4057 * config/i386/cpuid.h (bit_MOVDIRI,
4058 bit_MOVDIR64B): New bits.
4059 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
4061 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
4062 (VOID, PVOID, PCVOID)): New function types.
4063 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
4064 __builtin_ia32_directstoreu_u64,
4065 __builtin_ia32_movdir64b): New builtins.
4066 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
4067 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
4069 (ix86_valid_target_attribute_inner_p): Ditto.
4070 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
4071 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
4072 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
4073 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
4074 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
4075 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
4076 (movdiri<mode>, movdir64b_<mode>): New.
4077 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
4078 * config/i386/immintrin.h: Include movdirintrin.h.
4079 * config/i386/movdirintrin.h: New file.
4080 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
4082 2018-04-19 Richard Biener <rguenther@suse.de>
4085 * cfg.c (clear_bb_flags): When loop state says we have
4086 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
4088 2018-04-19 Richard Biener <rguenther@suse.de>
4090 PR tree-optimization/84737
4091 * tree-vect-data-refs.c (vect_copy_ref_info): New function
4092 copying restrict info.
4093 (vect_setup_realignment): Use it.
4094 * tree-vectorizer.h (vect_copy_ref_info): Declare.
4095 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
4096 the first DR to all generated stores.
4097 (vectorizable_load): Likewise for loads.
4099 2018-04-19 Jakub Jelinek <jakub@redhat.com>
4101 PR tree-optimization/85446
4102 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
4103 the integral and pointer types to have the same precision.
4105 * doc/install.texi: Document --disable-cet being the default and
4108 2018-04-18 Martin Liska <mliska@suse.cz>
4110 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
4113 2018-04-18 Martin Liska <mliska@suse.cz>
4116 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4119 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4120 arguments if they are comparable.
4122 2018-04-18 Martin Liska <mliska@suse.cz>
4125 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
4128 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
4131 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
4134 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
4135 ENDBR after calling __morestack.
4137 2018-04-18 David Malcolm <dmalcolm@redhat.com>
4140 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
4141 by using gcc_base_ver to generate a gcc_driver_version, and use
4142 it when generating GCC_DRIVER_NAME.
4143 * configure: Regenerate.
4145 2018-04-18 Jakub Jelinek <jakub@redhat.com>
4148 * config.gcc: Obsolete powerpc*-*-*spe*.
4150 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4153 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
4154 (stabstr_D): Change type of unum from unsigned int to
4155 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
4158 2018-04-17 Jim Wilson <jimw@sifive.com>
4161 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
4162 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
4163 Set arg_pointer_offset after using pretend_args_size.
4165 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4167 PR rtl-optimization/85431
4168 * dse.c (record_store): Ignore zero width stores.
4171 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
4172 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
4173 __builtin_stack_restore rather than after it.
4174 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
4175 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
4176 argument instead of virtual_dynamic_stack_rtx.
4178 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
4180 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
4182 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
4183 Add note to error message to explain internal mapping of overloaded
4184 built-in function name to non-overloaded built-in function name.
4185 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
4188 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
4191 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
4192 where the inputs overlap with the output.
4194 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4196 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
4197 (=v, v) alternative and explicit "memory" attribute.
4198 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
4199 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4201 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
4202 "sselog1" type instead of "sselog".
4203 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
4204 "sselog". Remove explicit "memory" attribute.
4205 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
4206 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4208 (vec_extract_hi_v32hi): Merge all alternatives into one, use
4209 "sselog1" type instead of "sselog". Remove explicit "memory"
4211 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
4212 use "sselog1" type instead of "sselog". Remove explicit "memory"
4214 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
4215 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
4217 (vec_extract_hi_v64qi): Merge all alternatives into one, use
4218 "sselog1" type instead of "sselog". Remove explicit "memory"
4220 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
4221 use "sselog1" type instead of "sselog". Remove explicit "memory"
4225 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
4228 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
4231 2018-04-17 Martin Jambor <mjambor@suse.cz>
4234 * ipa-cp.c (create_specialized_node): Call
4235 expand_all_artificial_thunks if necessary.
4237 2018-04-17 Martin Liska <mliska@suse.cz>
4240 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
4241 in message, remote space in between '_G' and '('.
4243 2018-04-17 Jakub Jelinek <jakub@redhat.com>
4246 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
4247 avx512f_vmcmp<mode>3<round_saeonly_name>,
4248 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
4249 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
4250 avx512f_rndscale<mode><round_saeonly_name>,
4251 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
4252 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
4253 Use %<iptr>2 instead of %2 for -masm=intel.
4254 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
4255 avx512f_vcvttss2usi<round_saeonly_name>,
4256 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
4258 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
4259 avx512f_vcvttsd2usi<round_saeonly_name>,
4260 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
4261 Use %q1 instead of %1 for -masm=intel.
4262 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
4263 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
4264 of %3 for -masm=intel.
4265 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
4267 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
4269 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
4271 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
4272 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
4274 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
4276 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
4277 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
4278 %g1 and one with %0 and %1.
4279 (avx512er_vmrcp28<mode><round_saeonly_name>,
4280 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
4282 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
4283 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
4284 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
4285 of %0 and %{%4%} for -masm=intel.
4286 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
4287 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
4288 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
4289 order of %0 and %{%5%}%{z%} for -masm=intel.
4291 2018-04-17 Jan Hubicka <jh@suse.cz>
4294 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
4296 2018-04-17 Martin Liska <mliska@suse.cz>
4299 * multiple_target.c (create_dispatcher_calls): Set apostrophes
4300 for target_clone error message. Make default implementation
4301 clone to be a local declaration.
4302 (separate_attrs): Add new argument and check for an empty
4304 (expand_target_clones): Handle it.
4305 (ipa_target_clone): Make redirection just for target_clones
4308 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
4309 Tom de Vries <tom@codesourcery.com>
4312 * omp-expand.c (expand_oacc_for): Add dummy false branch for
4313 tiled basic blocks without omp continue statements.
4315 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
4318 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
4319 vec_extract expression as having side effects to make sure it gets
4322 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
4325 * config/i386/i386.c (get_builtin_code_for_version): Check
4328 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
4331 * gcc/config.gcc: Support "skylake".
4332 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
4334 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
4335 (processor_target_table): Add "skylake".
4336 (ix86_option_override_internal): Add "skylake".
4337 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
4338 PROCESSOR_CANNONLAKE.
4339 (get_builtin_code_for_version): Fix priority for
4340 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
4341 PROCESSOR_SKYLAKE-AVX512.
4342 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
4343 (processor_type): Add PROCESSOR_SKYLAKE.
4345 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
4346 Jason Merrill <jason@redhat.com>
4349 * convert.c (convert_to_integer_1): Use direct recursion for
4350 enumeral types and types with a precision less than the number
4351 of bits in their mode.
4353 2018-04-16 Julia Koval <julia.koval@intel.com>
4356 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
4357 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
4359 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4362 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
4363 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
4364 and -mno-direct-move.
4366 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
4369 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
4370 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
4371 Ensure negative shifts result in {0}.
4373 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
4375 PR rtl-optimization/79916
4376 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
4377 regs (if any) to define how to gnerate SD moves when LRA is in
4380 2018-04-13 Jakub Jelinek <jakub@redhat.com>
4382 PR rtl-optimization/85393
4383 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
4384 * except.c (expand_dw2_landing_pad_for_region): Make static.
4385 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
4386 a label and unconditional jump to old_bb, rather than
4387 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
4390 PR rtl-optimization/85376
4391 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
4392 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
4393 instead of a specific value.
4395 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4396 Bin Cheng <bin.cheng@arm.com>
4398 PR tree-optimization/82965
4399 PR tree-optimization/83991
4400 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
4401 by_profile_only parameter.
4402 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
4403 information if the loop was predicted to iterate too many times.
4404 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
4406 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
4409 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
4412 2018-04-13 Martin Liska <mliska@suse.cz>
4413 Jakub Jelinek <jakub@redhat.com>
4416 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
4417 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
4418 * builtins.c (expand_builtin_memory_copy_args): Use
4419 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
4420 handle dest_addr == pc_rtx.
4422 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
4425 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
4426 asked to not generate direct moves.
4427 (fix_trunc<mode>si2_stfiwx): Similar.
4428 (fix_trunc<mode>si2_internal): Similar.
4430 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4433 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
4434 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
4435 lookup if dest in some wider mode is known to be const0_rtx and
4436 if so, record permanent equivalence for it to be ZERO_EXTEND of
4437 the narrower mode destination.
4439 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4441 * lto-streamer-out.c (output_function): Revert 259346.
4442 * omp-expand.c (expand_oacc_for): Likewise.
4444 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
4446 PR rtl-optimization/85354
4447 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
4448 * sel-sched.c (sel_global_init): ... here.
4450 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
4453 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
4454 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
4455 mode for PE-COFF targets.
4456 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
4457 (i386_pe_asm_lto_end): Likewise.
4458 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
4459 (TARGET_ASM_LTO_END): Likewise.
4460 * config/i386/winnt.c (saved_debug_info_level): New static variable.
4461 (i386_pe_asm_lto_start): New function.
4462 (i386_pe_asm_lto_end): Likewise.
4464 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
4465 Richard Biener <rguenther@suse.de>
4468 * lto-streamer-out.c (output_function): Fix CFG loop state before
4470 * omp-expand.c (expand_oacc_for): Handle calls to internal
4471 functions like regular functions.
4473 2018-04-12 Richard Biener <rguenther@suse.de>
4476 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
4477 for the early LTO debug to properly generate references to it
4478 during DIE emission. Do not re-use that for the skeleton for
4480 (dwarf2out_early_finish): Likewise.
4482 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4485 * config/i386/sse.md
4486 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
4487 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
4488 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
4489 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
4490 and output is a reg, avoid creating invalid lowpart subreg, but
4491 instead split into a 512-bit move. Don't split if not AVX512VL,
4492 input is xmm16+ reg and output is a mem.
4493 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
4494 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
4495 xmm16+ reg and output is a mem.
4497 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4499 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
4500 also for flag_dwarf2_cfi_asm.
4502 2018-04-12 Jakub Jelinek <jakub@redhat.com>
4504 PR rtl-optimization/85342
4505 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
4506 a bool scalar var inside of the loop instead. Don't try to update
4507 recog_data.operand after failed apply_change_group.
4509 2018-04-12 Tom de Vries <tom@codesourcery.com>
4512 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
4513 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
4514 array with flexible array member as array without given dimension.
4515 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
4516 argument for undefined param to true.
4518 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
4521 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4522 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
4523 from PowerPC section.
4524 * config/rs6000/sysv4.opt (mcall-): Improve help text.
4525 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
4526 help text that is too long.
4527 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
4528 help text that is too long.
4529 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
4530 help text that is too long.
4532 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
4534 * config/alpha/alpha.md (stack_probe_internal): Rename
4535 from "probe_stack". Update all callers.
4537 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4539 PR rtl-optimization/84566
4540 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
4541 sched_macro_fuse_insns.
4543 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4546 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
4547 (compute_block_dependences): ... from here.
4549 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4551 PR tree-optimization/85331
4552 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
4553 from int to HOST_WIDE_INT.
4555 2018-04-11 Martin Jambor <mjambor@suse.cz>
4558 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
4559 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
4560 not the same as the source val.
4561 (cgraph_edge_brings_value_p): New parameter.
4562 (gather_edges_for_value): Pass destination value to
4563 cgraph_edge_brings_value_p.
4564 (perhaps_add_new_callers): Likewise.
4565 (get_info_about_necessary_edges): Likewise and exclude values brought
4566 only by self-recursive edges.
4567 (create_specialized_node): Redirect only clones of self-calling edges.
4568 (+self_recursive_pass_through_p): New function.
4569 (find_more_scalar_values_for_callers_subset): Use it.
4570 (find_aggregate_values_for_callers_subset): Likewise.
4571 (known_aggs_to_agg_replacement_list): Removed.
4572 (decide_whether_version_node): Re-calculate known constants for all
4573 remaining context clones.
4575 2018-04-11 Richard Biener <rguenther@suse.de>
4578 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
4579 from early DWARF output.
4580 (dwarf2out_early_finish): Output line info unconditionally into
4581 early DWARF and add reference to it.
4583 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4586 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
4587 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
4588 other than V2DFmode using iptr mode attribute.
4589 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
4591 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
4593 PR rtl-optimization/84659
4594 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
4596 2018-04-11 Jakub Jelinek <jakub@redhat.com>
4599 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
4601 (output_loc_list): Pass address of a dummy size variable even in the
4602 locview handling loop.
4603 (index_location_lists): Add comment on why skip_loc_list_entry can't
4606 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
4609 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
4612 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
4615 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
4616 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
4617 and -mstring-compare-inline-limit.
4619 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4622 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
4623 for stack clash protection in a register whenever we need it to be in
4626 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4628 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
4629 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
4631 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
4634 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
4636 (mlong-double-): Ditto.
4637 * config/rs6000/sysv4.opt (msdata=): Ditto.
4638 (mtls-size=): Ditto.
4640 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4642 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4643 erroneous entries for
4644 "vector int vec_ldl (int, long int *)", and
4645 "vector unsigned int vec_ldl (int, unsigned long int *)".
4646 Add comments and entries for
4647 "vector bool char vec_ldl (int, bool char *)",
4648 "vector bool short vec_ldl (int, bool short *)",
4649 "vector bool int vec_ldl (int, bool int *)",
4650 "vector bool long long vec_ldl (int, bool long long *)",
4651 "vector pixel vec_ldl (int, pixel *)",
4652 "vector long long vec_ldl (int, long long *)",
4653 "vector unsigned long long vec_ldl (int, unsigned long long *)".
4654 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
4655 type tree bool_long_long_type_node and correct definition of
4656 bool_V2DI_type_node to make reference to this new type tree.
4657 (rs6000_mangle_type): Replace erroneous reference to
4658 bool_long_type_node with bool_long_long_type_node.
4659 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
4660 comments to emphasize sign distinctions for char and int types and
4661 replace RS6000_BTI_bool_long constant with
4662 RS6000_BTI_bool_long_long constant. Also add comment to restrict
4663 use of RS6000_BTI_pixel.
4664 (bool_long_type_node): Remove this macro definition.
4665 (bool_long_long_type_node): New macro definition
4667 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4669 PR rtl-optimization/85300
4670 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
4671 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
4672 simplify_unary_operation fails.
4674 2018-04-10 Martin Liska <mliska@suse.cz>
4676 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
4677 cgraph_edge and ipa_ref.
4679 2018-04-10 Jakub Jelinek <jakub@redhat.com>
4683 * config/i386/sse.md
4684 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
4685 computation of the VEC_MERGE selector from mask.
4686 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
4687 Fix decoding of the VEC_MERGE selector into mask.
4689 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4691 PR tree-optimization/85286
4692 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
4694 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
4696 * final.c (final_1): Set insn_last_address as well as
4697 insn_current_address.
4699 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4702 * explow.c (emit_stack_probe): Call validize_mem on memory location
4703 before passing it to gen_probe_stack. Create address operand and
4704 legitimize it for the probe_stack_address case.
4706 2018-04-09 Jan Hubicka <jh@suse.cz>
4709 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
4710 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
4711 * tree.c (free_lang_data_in_type): Fix handling of binfos;
4713 (free_lang_data): Rebuild type inheritance graph.
4715 2018-04-09 Martin Sebor <msebor@redhat.com>
4717 * invoke.texi (-finline-small-functions): Mention other optimization
4719 (-findirect-inlining, -fpartial-inlining): Same.
4720 (-finline-functions-called-once): Same.
4721 (-freorder-blocks-and-partition): Same.
4723 2018-04-09 Jan Hubicka <jh@suse.cz>
4726 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
4727 jumps; choose last target that matches the criteria (i.e.
4728 no partition changes for non-crossing jumps).
4729 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
4730 support for redirecting crossing jumps to non-crossing.
4732 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
4734 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
4735 also for naked functions.
4737 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
4739 * config/arc/arc.md (add_shift): New pattern.
4740 (add_shift2): Likewise.
4741 (sub_shift): Likewise.
4742 (sub_shift_cmp0_noout): Likewise.
4743 (compare_si_ashiftsi): Likewise.
4744 (xbfu_cmp0_noout): New combine pattern.
4745 (xbfu_cmp0"): Likewise.
4746 (movsi_set_cc_insn): Place the predicable variant first.
4747 (commutative_binary_cmp0_noout): Remove clobber.
4748 (commutative_binary_cmp0): New pattern.
4749 (noncommutative_binary_cmp0): Likewise.
4750 (noncommutative_binary_cmp0_noout): Likewise.
4751 (noncommutative_binary_comparison_result_used): Removed.
4752 (rsub_cmp0): New pattern.
4753 (rsub_cmp0_noout): Likewise.
4754 (extzvsi): Changed, keep only meaningful variants.
4755 (SQH, SEZ): New iterators.
4756 (SQH_postfix): New mode attribute.
4757 (SEZ_prefix): New code attribute.
4758 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
4759 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
4760 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
4762 (noncommutative_operator): Check the availability of barrel
4765 2018-04-09 Richard Biener <rguenther@suse.de>
4767 PR tree-optimization/85284
4768 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
4769 Only use the niter constraining form of simple_iv when the exit
4772 2018-04-09 Tom de Vries <tom@codesourcery.com>
4775 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4776 (define_expand "*memory_barrier"): New define_expand.
4777 (define_insn "memory_barrier"): New insn.
4779 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4781 PR rtl-optimization/80463
4782 PR rtl-optimization/83972
4783 PR rtl-optimization/83480
4785 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
4786 correct producer for the insn.
4787 (tidy_control_flow): Fixup seqnos in case of debug insns.
4789 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4791 PR rtl-optimization/83913
4793 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
4794 different sched-times when merging exprs.
4796 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4798 PR rtl-optimization/83962
4800 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
4801 tidy_fallthru_edge and tidy_control_flow.
4803 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
4805 PR rtl-optimization/83530
4807 * sel-sched.c (force_next_insn): New global variable.
4808 (remove_insn_for_debug): When force_next_insn is true, also leave only
4809 next insn in the ready list.
4810 (sel_sched_region): When the region wasn't scheduled, make another pass
4811 over it with force_next_insn set to 1.
4813 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4815 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
4817 * config/nds32/constants.md (unspec_volatile_element): Add enum values
4818 for interrupt control.
4819 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4820 functions for interrupt control.
4821 * config/nds32/nds32-intrinsic.md: Likewise.
4822 * config/nds32/nds32_intrinsic.h: Likewise.
4823 * config/nds32/nds32.h (nds32_builtins): Likewise.
4825 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4827 * config/nds32/nds32.c (nds32_init_machine_status,
4828 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
4829 strict_aligned_p field.
4830 (nds32_expand_to_rtl_hook): New function.
4831 (TARGET_EXPAND_TO_RTL_HOOK): Define.
4832 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
4834 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4835 Chung-Ju Wu <jasonwucj@gmail.com>
4837 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
4838 * config/nds32/nds32-n7.md: New file.
4839 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
4840 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
4842 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
4843 * config/nds32/nds32.md (pipeline_model): Add n7.
4844 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
4845 * config/nds32/pipelines.md: Include n7 settings.
4847 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4848 Chung-Ju Wu <jasonwucj@gmail.com>
4850 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
4851 * config/nds32/nds32-e8.md: New file.
4852 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
4853 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
4855 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
4856 * config/nds32/nds32.md (pipeline_model): Add e8.
4857 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
4858 * config/nds32/pipelines.md: Include e8 settings.
4860 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4861 Chung-Ju Wu <jasonwucj@gmail.com>
4863 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
4864 * config/nds32/nds32-n8.md: New file.
4865 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
4866 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
4868 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
4869 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
4870 * config/nds32/nds32.md (pipeline_model): Add n8.
4871 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
4872 * config/nds32/pipelines.md: Include n8 settings.
4874 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
4875 Chung-Ju Wu <jasonwucj@gmail.com>
4877 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
4878 * config/nds32/nds32-n9-2r1w.md: New file.
4879 * config/nds32/nds32-n9-3r2w.md: New file.
4880 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
4881 nds32_register_ports): New or modify for cpu n9.
4882 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
4884 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
4885 * config/nds32/nds32-utils.c: New file.
4886 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
4887 TARGET_MUL_SLOW): Define.
4888 * config/nds32/nds32.md (pipeline_model): New attribute.
4889 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
4890 New options that support cpu n9.
4891 * config/nds32/pipelines.md: Include n9 settings.
4892 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
4894 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
4896 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
4897 information if necessary.
4898 (output_cond_branch_compare_zero): Likewise.
4899 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
4900 (nds32_target_alignment): Refine for alignment.
4901 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
4902 (FUNCTION_BOUNDARY): Modify.
4903 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
4905 * config/nds32/nds32.opt (malways-align, malign-functions): New.
4907 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
4909 * config/nds32/constants.md (unspec_volatile_element): Add values for
4910 TLB operation and data prefetch.
4911 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
4912 functions for TLB operation and data prefetch.
4913 * config/nds32/nds32-intrinsic.md: Likewise.
4914 * config/nds32/nds32_intrinsic.h: Likewise.
4915 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
4916 (nds32_print_operand): Likewise.
4917 * config/nds32/nds32.h (nds32_builtins): Likewise.
4919 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
4920 Andrew Pinski <pinsika@gcc.gnu.org>
4923 * match.pd: Use constant_boolean_node of correct type instead of
4924 boolean_true_node or boolean_false_node for simplifying
4925 pointer comparisons to zero.
4927 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4929 PR tree-optimization/80021
4930 * tree.c (verify_type_variant): Make error call in verify_variant_match
4931 translatable and remove final full stop.
4933 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4935 * config/nds32/constants.md (unspec_volatile_element): Add
4936 UNSPEC_VOLATILE_EH_RETURN.
4937 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
4938 nds32_output_stack_pop): Support dwarf exception handling process.
4939 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
4940 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
4941 exception handling process.
4942 (nds32_compute_stack_frame): Likewise.
4943 (nds32_return_addr_rtx): Likewise.
4944 (nds32_initial_elimination_offset): Likewise.
4945 (nds32_expand_prologue): Likewise.
4946 (nds32_expand_epilogue): Likewise.
4947 (nds32_dynamic_chain_address): New function.
4948 * config/nds32/nds32.h (machine_function): Add fields for dwarf
4950 (DYNAMIC_CHAIN_ADDRESS): Define.
4951 (EH_RETURN_DATA_REGNO): Define.
4952 (EH_RETURN_STACKADJ_RTX): Define.
4953 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
4954 patterns for dwarf exception handling.
4956 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
4958 * config/nds32/nds32.h: Clean up obsolete macros.
4960 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4962 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4963 Add enum values for particular instructions.
4964 * config/nds32/nds32-intrinsic.c: Implementation of expanding
4965 particular intrinsic functions.
4966 * config/nds32/nds32-intrinsic.md: Likewise.
4967 * config/nds32/nds32_intrinsic.h: Likewise.
4968 * config/nds32/nds32.h (nds32_builtins): Likewise.
4969 * config/nds32/nds32.md (type): Add pbsad and pbsada.
4970 (btst, ave): New patterns for particular instructions.
4972 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4974 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4975 Add enum values for atomic load/store and memory sync.
4976 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
4978 * config/nds32/nds32-intrinsic.md: Likewise.
4979 * config/nds32/nds32_intrinsic.h: Likewise.
4980 * config/nds32/nds32.h (nds32_builtins): Likewise.
4982 2018-04-07 Jakub Jelinek <jakub@redhat.com>
4984 PR tree-optimization/85257
4985 * fold-const.c (native_encode_vector): If not all elts could fit
4986 and off is -1, return 0 rather than offset.
4987 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
4988 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
4989 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
4990 adjust buffer in native_interpret_expr call.
4992 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
4994 * config/nds32/constants.md (unspec_volatile_element): Add cache
4995 control enum values.
4996 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
4997 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
4998 * config/nds32/nds32.c (nds32_cctl_names): New.
4999 (nds32_print_operand): Handle cache control register names.
5000 * config/nds32/nds32.h (nds32_builtins): New enum values.
5001 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
5003 * config/nds32/nds32.md (type): Add mmu.
5004 * config/nds32/pipelines.md (simple_insn): Add mmu.
5006 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5008 * config/nds32/nds32.md (type): Remove call.
5009 * config/nds32/pipelines.md (simple_insn): Likewise.
5011 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5013 * config/nds32/constants.md (unspec_volatile_element): Add
5014 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
5015 UNSPEC_VOLATILE_FMFCFG.
5016 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
5017 description for fmfcfg and fmfcsr.
5018 (bdesc_1arg): Add fmtcsr.
5019 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
5020 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
5021 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
5022 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
5023 unspec_fmfcfg): New patterns.
5024 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
5025 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
5026 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
5027 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
5028 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
5029 __nds32__fmfcfg): Define.
5031 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
5033 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
5034 intrinsic register names.
5035 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
5036 intrinsic register enum values and macros.
5038 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
5040 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
5041 for load/store addressing form.
5042 (nds32_print_operand_address): Likewise.
5044 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
5047 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
5048 based on LABEL_REF. Remove useless assertion.
5049 (pic_address_needs_scratch): Fix formatting.
5050 (sparc_legitimize_pic_address): Minor tweaks.
5051 (sparc_delegitimize_address): Adjust assertion accordingly.
5052 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
5053 into symbolic_operand.
5054 (movsi_high_pic_label_ref): Likewise.
5055 (movsi_lo_sum_pic_label_ref): Likewise.
5056 (movdi_pic_label_ref): Likewise.
5057 (movdi_high_pic_label_ref): Likewise.
5058 (movdi_lo_sum_pic_label_ref): Likewise.
5060 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
5062 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
5063 custom LIB_SPEC setup.
5065 2018-04-06 Ruslan Bukin <br@bsdpad.com>
5066 Kito Cheng <kito.cheng@gmail.com>
5068 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
5069 * config/riscv/freebsd.h: New.
5071 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5073 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
5074 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
5077 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5078 Kito Cheng <kito.cheng@gmail.com>
5080 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
5081 nds32_output_call, nds32_symbol_binds_local_p): New functions.
5082 * config/nds32/nds32-protos.h (nds32_output_call,
5083 nds32_output_return): Declare.
5084 * config/nds32/nds32.md: Refine all the call and return patterns.
5086 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5089 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
5090 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
5092 PR rtl-optimization/84872
5093 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
5094 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
5097 2018-04-06 Tamar Christina <tamar.christina@arm.com>
5099 * expr.c (copy_blkmode_to_reg): Revert 254862.
5100 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
5102 2018-04-06 Richard Biener <rguenther@suse.de>
5105 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
5106 after seeing a component reference with an adjacent field. Treat
5107 refs to arrays at struct end of external decls similar to
5108 refs to unconstrained commons.
5110 2018-04-06 Jakub Jelinek <jakub@redhat.com>
5113 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
5114 look through SAVE_EXPRs with non-side-effects argument. Adjust
5116 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
5119 2018-04-06 Richard Biener <rguenther@suse.de>
5122 * alias.c (find_base_term): New wrapper around find_base_term
5123 unwinding CSELIB_VAL_PTR changes.
5124 (find_base_term): Do not restore CSELIB_VAL_PTR during the
5127 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5129 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
5131 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
5132 constant definitions.
5133 ("nop"): lr 0,0 -> nopr r0
5134 ("nop_lr0", "nop_lr1"): New insn definitions.
5136 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
5138 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
5139 NDS32_V3PUSH_AVAILABLE_P macro.
5141 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
5142 Chung-Ju Wu <jasonwucj@gmail.com>
5144 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
5145 (nds32*-*-*): Add float and fpu_config into supported_defaults.
5146 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
5147 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
5148 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
5149 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
5150 * config/nds32/constraints.md: New constraints and checking for hard
5151 float configuration.
5152 * config/nds32/iterators.md: New mode iterator and attribute for hard
5153 float configuration.
5154 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
5156 * config/nds32/nds32-fpu.md: New file.
5157 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
5158 deal with hard float code generation.
5159 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
5161 (abi_type, float_reg_number): New enum type.
5162 * config/nds32/nds32-predicates.c: New predicates for hard float.
5163 * config/nds32/nds32-protos.h: Declare functions for hard float.
5164 * config/nds32/nds32.c: Implementation for hard float configuration.
5165 * config/nds32/nds32.h: Definitions for hard float configuration.
5166 * config/nds32/nds32.md: Include hard float machine description and
5167 modify patterns for hard float configuration.
5168 * config/nds32/nds32.opt: New options for hard float configuration.
5169 * config/nds32/predicates.md: New predicates for hard float
5172 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5174 * common/config/nds32/nds32-common.c
5175 (nds32_option_optimization_table): Enable -mreleax-hint by default.
5177 2018-04-05 Jakub Jelinek <jakub@redhat.com>
5180 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
5181 CONSTRUCTOR_ELT (ctor, ...)->value.
5183 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
5186 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
5188 2018-04-05 Tom de Vries <tom@codesourcery.com>
5191 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
5194 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5195 Kito Cheng <kito.cheng@gmail.com>
5197 * config/nds32/constraints.md (U33): Fine-tune checking condition.
5198 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
5199 * config/nds32/nds32.h (nds32_16bit_address_type): Add
5200 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
5202 2018-04-05 Shiva Chen <shiva0217@gmail.com>
5203 Kito Cheng <kito.cheng@gmail.com>
5205 * config/nds32/constraints.md (Ufe): New memory constraint.
5206 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
5207 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
5208 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
5210 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
5211 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
5213 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5215 * config/nds32/nds32.md: Use optimize_size in the condition for
5216 alu-shift instructions.
5218 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5220 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
5222 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5224 * config/nds32/nds32.md (negsi2): Refine pattern.
5226 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5227 Chung-Ju Wu <jasonwucj@gmail.com>
5229 * config/nds32/iterators.md (shift_rotate): New code iterator.
5230 (shift): New code attribute.
5231 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
5232 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
5233 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
5234 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
5235 bit-wise operations.
5236 (andsi3, *andsi3): Ditto.
5237 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
5238 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
5239 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
5240 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
5241 nds32_ior_operand, nds32_xor_operand): New predicates.
5243 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5245 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
5246 (addsi3, subsi3): ... this.
5248 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5250 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
5252 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5254 * config/nds32/nds32.md: Adjust indention.
5256 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
5258 * config/nds32/nds32.md (feature): New attribute.
5260 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
5262 * config/nds32/nds32.md (subtype): New attribute.
5264 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
5267 * config/arm/arm-builtins.c (arm_expand_builtin): Change
5268 expansion to perform a bitwise AND of the argument followed by a
5269 boolean negation of the result.
5271 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
5273 PR rtl-optimization/84878
5274 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
5275 the basic block. Assert the use reference is not artificial and that
5276 it has an associated insn.
5278 2018-04-04 Michael Matz <matz@suse.de>
5280 * builtins.c (compute_objsize): Pass correct operand
5281 to array_at_struct_end_p.
5283 2018-04-04 Richard Biener <rguenther@suse.de>
5286 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
5287 from contexts for DINFO_LEVEL_TERSE and below.
5289 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5291 * config/nds32/nds32-doubleword.md (move_<mode>): Require
5292 resiter_operand condition.
5293 * config/nds32/nds32.md (*move<mode>): Ditto.
5295 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5296 Monk Chiang <sh.chiang04@gmail.com>
5298 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
5300 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5302 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
5304 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5305 Kito Cheng <kito.cheng@gmail.com>
5307 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
5308 nds32_cond_code_str, output_cond_branch,
5309 output_cond_branch_compare_zero, nds32_expand_cbranch,
5310 nds32_expand_cstore, nds32_expand_movcc,
5311 nds32_output_cbranchsi4_equality_zero,
5312 nds32_output_cbranchsi4_equality_reg,
5313 nds32_output_cbranchsi4_equality_reg_or_const_int,
5314 nds32_output_cbranchsi4_greater_less_zero: New functions.
5315 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
5316 nds32_expand_cstore, nds32_expand_movcc,
5317 nds32_output_cbranchsi4_equality_zero,
5318 nds32_output_cbranchsi4_equality_reg,
5319 nds32_output_cbranchsi4_equality_reg_or_const_int,
5320 nds32_output_cbranchsi4_greater_less_zero): Declare.
5321 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
5322 nds32_rimm11s_operand): New predicates.
5323 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
5324 * config/nds32/nds32.md: Rewrite all the branch and conditional move
5327 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
5329 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
5330 * config/nds32/nds32.md: Ditto.
5331 * config/nds32/pipelines.md: Ditto.
5333 2018-04-04 Richard Biener <rguenther@suse.de>
5335 PR tree-optimization/85168
5336 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
5337 propagating abnormals.
5339 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5341 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
5343 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
5344 Kito Cheng <kito.cheng@gmail.com>
5346 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
5347 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
5348 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
5349 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
5350 * config/nds32/nds32.md (sibcall_internal): New.
5351 (sibcall_register): Remove.
5352 (sibcall_immediate): Remove.
5353 (sibcall_value_internal): New.
5354 (sibcall_value_register): Remove.
5355 (sibcall_value_immediate): Remove.
5356 * config/nds32/predicates.md (nds32_general_register_operand): New.
5357 (nds32_call_address_operand): New.
5359 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5361 PR rtl-optimization/85167
5362 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
5363 bb_defs if *split_p, instead preinitialize it to NULL.
5365 PR tree-optimization/85156
5366 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
5367 evaluating the argument multiple times.
5369 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
5371 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
5373 (_mm_cvtpd_ps): Likewise.
5374 (_mm_cvttpd_epi32): Likewise.
5375 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
5376 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
5377 vector, pixel, and bool following altivec.h include.
5379 2018-04-03 Martin Sebor <msebor@redhat.com>
5381 * doc/extend.texi (Common Function Attributes): Clarify.
5382 (const attribute): Likewise.
5383 (pure attribute): Likewise.
5385 2018-04-03 Jakub Jelinek <jakub@redhat.com>
5388 * config/i386/i386.c (ix86_expand_vector_set): Use
5389 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
5391 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
5393 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
5394 instructions when changing rounding bits to preserve precision bits
5395 in the x87 control word.
5397 2018-04-03 Martin Liska <mliska@suse.cz>
5399 PR tree-optimization/82491
5400 * rtl.h (strip_offset_and_add): Replace += suboffset with
5401 poly_uint64 () + suboffset.
5403 2018-03-29 Martin Liska <mliska@suse.cz>
5404 Martin Jambor <mjambor@suse.cz>
5407 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
5408 param_type is not an integral or pointer type.
5410 2018-04-03 Richard Biener <rguenther@suse.de>
5412 * sese.h (recompute_all_dominators): Remove.
5414 2018-04-02 Martin Sebor <msebor@redhat.com>
5416 * doc/invoke.texi (-Wrestrict): Fix typos.
5418 2018-04-02 Jim Wilson <jimw@sifive.com>
5420 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
5421 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
5422 (<optab>di3, <optab>si3_extend): Likewise.
5423 (<optab>si3_mask, <optab>si3_mask_1): New.
5424 (<optab>di3_mask, <optab>di3_mask_1): New.
5425 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
5426 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
5427 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
5429 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
5431 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
5434 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
5437 (nds32_canonicalize_comparison): New function.
5439 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5440 Kito Cheng <kito.cheng@gmail.com>
5441 Kuan-Lin Chen <kuanlinchentw@gmail.com>
5443 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
5444 * config/nds32/constants.md (unspec_volatile_element): Add
5445 UNSPEC_VOLATILE_RELAX_GROUP.
5446 * config/nds32/nds32-relax-opt.c: New file.
5447 * config/nds32/nds32-predicates.c
5448 (nds32_symbol_load_store_p): New function.
5449 * config/nds32/nds32-protos.h
5450 (nds32_symbol_load_store_p): Declare function.
5451 (make_pass_nds32_relax_opt): Declare new rtl pass function.
5452 * config/nds32/nds32.c
5453 (nds32_register_pass): New function to register pass.
5454 (nds32_register_passes): New function to register passes.
5455 * config/nds32/nds32.md (relax_group): New pattern.
5456 * config/nds32/nds32.opt (mrelax-hint): New option.
5457 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
5459 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
5461 * config/nds32/t-nds32: Modify files dependency.
5463 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5465 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
5466 (PROFILE_HOOK): Define its implementation.
5468 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
5470 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
5471 type and 32-bit size.
5473 2018-04-01 Jakub Jelinek <jakub@redhat.com>
5476 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
5477 (V_128_256): New mode iterator.
5478 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
5479 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
5480 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
5482 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
5483 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
5485 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
5488 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
5489 NaN inputs correctly.
5491 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
5494 * config/rs6000/vsx.md (??r): New mode attribute.
5495 (*vsx_mov<mode>_64bit): Use it.
5496 (*vsx_mov<mode>_32bit): Likewise.
5498 2018-03-30 Martin Sebor <msebor@redhat.com>
5500 PR tree-optimization/84818
5501 * builtins.c (check_access): Use warning_n.
5503 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5506 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
5508 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
5511 2018-03-30 Julia Koval <julia.koval@intel.com>
5514 * x86-tune.def (movx, partial_reg_dependency): Enable for
5517 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
5520 * lra-constraints.c (process_alt_operands): Move setting
5521 this_alternative_matches below.
5523 2018-03-29 Martin Liska <mliska@suse.cz>
5526 * doc/invoke.texi: Document how LTO works with debug info.
5527 Describe auto-load support of binutils. Mention 'x86-64'
5528 as valid option value of -march option.
5530 2018-03-29 Jakub Jelinek <jakub@redhat.com>
5532 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
5535 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
5536 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
5537 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
5540 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5543 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
5544 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
5545 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
5546 for RS6000_BTM_POWERPC64.
5547 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
5548 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
5549 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
5554 2018-03-28 Carl Love <cel@us.ibm.com>
5557 2017-09-27 Carl Love <cel@us.ibm.com>
5559 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
5560 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
5561 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
5564 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5566 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
5567 instead of __vector bool.
5568 (_mm_max_pu8): Likewise.
5569 (_mm_min_pi16): Likewise.
5571 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
5574 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
5575 (DIVWEUO): Likewise.
5577 (DIVDEUO): Likewise.
5578 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
5579 DIVWEUO and DIVDEUO.
5580 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
5581 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
5582 (div_extend): Likewise.
5583 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
5585 (__builtin_divweuo): Likewise.
5586 (__builtin_divdeo): Likewise.
5587 (__builtin_divdeuo): Likewise.
5589 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5592 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
5593 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
5595 PR tree-optimization/82004
5596 * gimple-match-head.c (optimize_pow_to_exp): New function.
5597 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
5598 Don't fold to exp if optimize_pow_to_exp is false.
5600 2018-03-28 Martin Liska <mliska@suse.cz>
5603 * calls.c (initialize_argument_information): Fix trailing space.
5604 * common.opt: Fix typo and provide better explanation for
5605 -fsanitize-coverage option.
5606 * config/i386/i386.opt: Fix typo.
5608 2018-03-28 Jakub Jelinek <jakub@redhat.com>
5609 Martin Liska <mliska@suse.cz>
5612 * gimplify.c (asan_poison_variable): Don't do the check for
5613 gimplify_omp_ctxp here.
5614 (gimplify_decl_expr): Do it here.
5615 (gimplify_target_expr): Likewise.
5617 2018-03-28 Martin Liska <mliska@suse.cz>
5620 * config/i386/i386.c (ix86_function_arg_advance): Do not call
5621 chkp_type_bounds_count if MPX is not enabled.
5623 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
5625 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
5627 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
5630 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
5631 function to create the function decl for complex long double
5632 multiply and divide for -mabi=ieeelongdouble.
5633 (init_float128_ieee): Call it.
5635 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
5638 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
5639 -fcf-protection=branch -mibt.
5640 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
5642 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5645 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
5647 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
5650 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
5651 extern array declarations.
5653 2018-03-27 Richard Biener <rguenther@suse.de>
5656 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
5657 explicit single_use checks.
5659 2018-03-27 Richard Biener <rguenther@suse.de>
5661 PR tree-optimization/85082
5662 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
5665 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
5667 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
5668 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
5669 Turn on fasynchronous-unwind-tables and funwind-tables.
5671 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
5674 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
5675 (*bmi_blsr_<mode>_ccz): Ditto.
5677 2018-03-26 Tom de Vries <tom@codesourcery.com>
5679 PR tree-optimization/85063
5680 * omp-general.c (offloading_function_p): New function. Factor out
5682 * omp-offload.c (pass_omp_target_link::gate): ... here.
5683 * omp-general.h (offloading_function_p): Declare.
5684 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
5685 with attribute omp declare target for offloading functions.
5687 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
5689 PR tree-optimization/84005
5690 * tree-data-ref.h (get_base_for_alignment): Declare.
5691 * tree-data-ref.c (get_base_for_alignment_1): New function.
5692 (get_base_for_alignment): Likewise.
5693 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
5694 get_base_for_alignment to find a suitable base object, instead
5695 of always using drb->base_address.
5697 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5700 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
5701 known size by default.
5703 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
5706 * lra-constraints.c (process_alt_operands): Don't match BLKmode
5707 and non BLKmode operands.
5709 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5712 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
5713 Clean up attributes.
5715 2018-03-23 Richard Biener <rguenther@suse.de>
5718 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
5719 we are going to emit early debug for LTO.
5721 2018-03-23 Jakub Jelinek <jakub@redhat.com>
5724 * function.c (match_asm_constraints_1): Don't optimize if input
5725 doesn't satisfy general_operand predicate for output's mode.
5728 * alias.c (write_dependence_p): Don't require for x_canonicalized
5729 non-VOIDmode if x has VOIDmode.
5732 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
5733 just don't try to optimize it rather than assert it never happens.
5735 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5737 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
5738 macro expansions for definition of ST_INTERNAL_<mode> and
5739 LD_INTERNAL_<mode> builtins.
5740 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
5742 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
5744 (altivec_expand_st_builtin): Likewise.
5745 (altivec_expand_builtin): Remove calls to deleted functions.
5746 (rs6000_address_for_altivec): Delete this function.
5747 * config/rs6000/vector.md: Remove expands for
5748 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
5750 2018-03-22 Sudakshina Das <sudi.das@arm.com>
5753 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
5754 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
5755 re-computing once computed.
5756 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
5757 (arm_init_machine_status): Initialize
5758 machine->static_chain_stack_bytes.
5760 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
5763 * doc/extend.texi: Add four new prototypes for vec_ld.
5764 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
5765 definitions for more logical presentation.
5766 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
5767 entries for V1TI variants of __builtin_altivec_ld builtin.
5768 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
5769 handling of V1TI variant of LVX icode pattern.
5770 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
5771 (rs6000_gimple_fold_builtin): Likewise.
5772 (altivec_init_builtins): Add code to define
5773 __builtin_altivec_lvx_v1ti function.
5775 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5778 * function.c (match_asm_constraints_1): Don't do the optimization
5779 if input isn't a REG, SUBREG, MEM or constant.
5781 2018-03-22 Tom de Vries <tom@codesourcery.com>
5783 PR tree-optimization/84956
5784 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
5785 bb_has_abnormal_pred.
5787 2018-03-22 Jakub Jelinek <jakub@redhat.com>
5790 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
5791 DECL_INITIAL (decl) to decl at the end.
5792 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
5795 2018-03-21 Joseph Myers <joseph@codesourcery.com>
5797 * doc/extend.texi (__builtin_tgmath): Document when complex
5798 integer types are treated as _Complex _Float64.
5800 2018-03-21 Tom de Vries <tom@codesourcery.com>
5802 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
5804 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5806 PR tree-optimization/84960
5807 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
5808 if it is ENTRY block, move them into single succ of ENTRY in that case.
5810 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
5812 PR tree-optimization/84811
5813 * poly-int.h (poly_span_traits): Remove the T3 parameter and
5814 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
5815 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
5816 (known_subrange_p): Update accordingly. Cast each value involved
5817 in the size comparison, rather than casting the result of the
5820 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5822 PR tree-optimization/84982
5823 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
5824 by flipping the least significant bit rather than all bits from
5825 bitpos to bitpos + bitsize - 1.
5827 2018-03-21 Nathan Sidwell <nathan@acm.org>
5829 * doc/extend.texi (Deprecated Features): Remove mention of
5830 long-deleted deprecations.
5832 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5835 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
5836 * configure: Regenerate.
5838 2018-03-21 Tom de Vries <tom@codesourcery.com>
5840 PR tree-optimization/83126
5841 * tree-parloops.c (num_phis): New function.
5842 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
5844 2018-03-21 Nathan Sidwell <nathan@acm.org>
5846 * doc/extend.texi (Deprecated Features): Update deprecated flags,
5847 mention anon-struct/union members and trailing attributes.
5849 2018-03-21 Bin Cheng <bin.cheng@arm.com>
5851 PR tree-optimization/84969
5852 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
5853 builtin memset partitions if they set different rhs values.
5855 2018-03-21 Jakub Jelinek <jakub@redhat.com>
5857 PR rtl-optimization/84989
5858 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
5859 VEC_DUPLICATE with scalar result mode.
5861 2018-03-21 Martin Liska <mliska@suse.cz>
5864 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
5865 not intended return statement.
5867 2018-03-21 Martin Liska <mliska@suse.cz>
5870 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
5871 (chkp_find_bound_slots_1): Limit number of iterations.
5873 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
5876 * Minor grammar fixes for x86 options.
5878 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5881 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
5882 holding REG_CFA_RESTORE notes, instead turn them into a USE.
5884 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
5887 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
5888 (altivec_lvx_<mode>_1op): Likewise.
5889 (altivec_stvx_<mode>_2op): Likewise.
5890 (altivec_stvx_<mode>_1op): Likewise.
5891 (altivec_lvx_<VM2:mode>): New define_expand.
5892 (altivec_stvx_<VM2:mode>): Likewise.
5893 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
5894 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5895 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
5896 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
5897 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
5898 (rs6000_gen_lvx): Likewise.
5899 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
5900 (altivec_expand_stv_builtin): Likewise.
5901 (altivec_expand_builtin): Likewise.
5902 * config/rs6000/vector.md: Likewise.
5904 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5907 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
5910 2018-03-20 Richard Biener <rguenther@suse.de>
5913 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
5914 sign-conversions as zero, fall back to standard scalar_stmt
5917 2018-03-20 Martin Liska <mliska@suse.cz>
5920 * predict.c (rebuild_frequencies): Handle case when we have
5921 PROFILE_ABSENT, but flag_guess_branch_prob is false.
5923 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5926 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
5927 flag_section_anchors.
5928 * varasm.c (use_blocks_for_decl_p): Remove hack for
5929 dw2_force_const_mem.
5932 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
5934 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
5935 be created, use lowpart_subreg of operands[0] rather than operands[0]
5937 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
5938 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
5939 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
5940 and n constraint instead of aarch64_shift_imm_di and Usd.
5941 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
5942 (*aarch64_<optab>_reg_minus<mode>3): ... this.
5944 2018-03-20 Sudakshina Das <sudi.das@arm.com>
5947 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
5948 to favor GPR over NEON registers.
5949 (<shift>di3_neon): Likewise.
5951 2018-03-20 Tom de Vries <tom@codesourcery.com>
5954 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
5955 (nvptx_process_pars): Emit bar.sync asap and alap.
5957 2018-03-20 Tom de Vries <tom@codesourcery.com>
5960 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
5961 seen_label if seen_label is already set.
5963 2018-03-20 Jakub Jelinek <jakub@redhat.com>
5966 * config/i386/i386.c (fold_builtin_cpu): For features above 31
5967 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
5968 Use 1U instead of 1. Formatting fixes.
5971 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
5972 instead of TREE_TYPE (s1) for the return value.
5974 2018-03-19 Jakub Jelinek <jakub@redhat.com>
5976 PR tree-optimization/84946
5977 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
5978 bitsize + bitsize in poly_uint64 rather than poly_int64.
5981 * dwarf2asm.c: Include fold-const.c.
5982 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
5983 of decl rather than decl itself.
5985 PR rtl-optimization/84643
5986 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
5988 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
5991 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
5992 calling assemble_variable.
5994 2018-03-19 Sudakshina Das <sudi.das@arm.com>
5997 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
5998 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
6000 2018-03-19 Jim Wilson <jimw@sifive.com>
6003 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
6004 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
6005 (riscv_first_stack_step): Likewise.
6006 (riscv_option_override): Use STACK_BOUNDARY instead of
6008 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
6010 (BIGGEST_ALIGNMENT): Set to 128.
6011 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
6012 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
6015 2018-03-19 Richard Biener <rguenther@suse.de>
6017 PR tree-optimization/84933
6018 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
6019 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
6021 2018-03-19 Richard Biener <rguenther@suse.de>
6023 PR tree-optimization/84859
6024 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
6025 (cond_if_else_store_replacement): Perform sinking operation on
6026 single-store BBs regardless of MAX_STORES_TO_SINK setting.
6027 Generalize what a BB with a single eligible store is.
6029 2018-03-19 Richard Biener <rguenther@suse.de>
6031 PR tree-optimization/84929
6032 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
6033 chrec_is_positive against non-chrec arg.
6035 2018-03-19 Tamar Christina <tamar.christina@arm.com>
6038 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
6040 2018-03-18 Martin Liska <mliska@suse.cz>
6042 PR rtl-optimization/84635
6043 * regrename.c (build_def_use): Use matches_mode only when
6046 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
6048 PR tree-optimization/84913
6049 * tree-vect-loop.c (vectorizable_reduction): Don't try to
6050 vectorize chains of COND_EXPRs.
6052 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6054 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
6056 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6058 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
6060 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
6062 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
6064 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6065 Kito Cheng <kito.cheng@gmail.com>
6067 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
6068 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
6069 (nds32_adjust_reg_alloc_order): New function.
6070 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
6072 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
6074 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
6075 nds32_print_operand, nds32_print_operand_address): Use
6076 HOST_WIDE_INT_PRINT_DEC instead.
6078 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
6080 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
6082 2018-03-17 Jakub Jelinek <jakub@redhat.com>
6085 * config/i386/i386.c (initial_ix86_tune_features,
6086 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
6088 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
6089 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
6090 rather than 1u << ix86_tune. Formatting fix.
6091 (ix86_option_override_internal): Change ix86_arch_mask from
6092 unsigned int to unsigned HOST_WIDE_INT, initialize to
6093 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
6094 (ix86_function_specific_restore): Likewise.
6096 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6099 * postreload.c (reload_combine_recognize_pattern): Perform
6100 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
6101 truncate_int_for_mode the result for the destination's mode.
6104 * hsa-gen.c (mem_type_for_type): Fix comment typo.
6105 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
6107 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6110 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
6113 * lra-assigns.c (lra_split_hard_reg_for): Don't use
6114 regno_allocno_class_array and sorted_pseudos.
6115 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
6116 insns where regno is used.
6118 2018-03-16 Martin Liska <mliska@suse.cz>
6121 * multiple_target.c (create_dispatcher_calls): Redirect
6122 reference in the symbol table.
6124 2018-03-16 Martin Liska <mliska@suse.cz>
6127 * multiple_target.c (create_dispatcher_calls): Redirect also
6130 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6134 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
6135 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
6136 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
6138 2018-03-16 Julia Koval <julia.koval@intel.com>
6140 * doc/invoke.texi (Skylake Server): Add CLWB.
6141 Cannonlake): Remove CLWB.
6143 2018-03-16 Jakub Jelinek <jakub@redhat.com>
6145 PR tree-optimization/84841
6146 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
6148 (FLOAT_ONE_CONST_TYPE): Define.
6149 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
6150 (sort_by_operand_rank): Put entries with higher constant_type last
6151 rather than first to match comments.
6153 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
6155 * config/nios2/nios2.md (movsi_internal): Fix thinko in
6158 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6161 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
6162 check and use address of target always.
6164 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
6167 * config/i386/i386.c (indirect_thunk_needed): Update comments.
6168 (indirect_thunk_bnd_needed): Likewise.
6169 (indirect_thunks_used): Likewise.
6170 (indirect_thunks_bnd_used): Likewise.
6171 (indirect_return_needed): New.
6172 (indirect_return_bnd_needed): Likewise.
6173 (output_indirect_thunk_function): Add a bool argument for
6175 (output_indirect_thunk_function): Don't generate alias for
6176 function return thunk.
6177 (ix86_code_end): Call output_indirect_thunk_function to generate
6178 function return thunks.
6179 (ix86_output_function_return): Set indirect_return_bnd_needed
6180 and indirect_return_needed instead of indirect_thunk_bnd_needed
6181 and indirect_thunk_needed.
6183 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
6185 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
6186 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
6187 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
6189 2018-03-15 David Malcolm <dmalcolm@redhat.com>
6190 Paul Hua <paul.hua.gm@gmail.com>
6193 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
6195 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
6197 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
6198 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
6201 2018-03-15 Tamar Christina <tamar.christina@arm.com>
6204 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6205 instead of GET_MODE_SIZE when comparing Units.
6207 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6210 * varasm.c (hash_section): Return an unchangeble hash value
6211 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
6212 Return !aarch64_can_use_per_function_literal_pools_p ().
6214 2018-03-15 Jakub Jelinek <jakub@redhat.com>
6217 * optabs.c (emit_conditional_move): Pass address of cmode's copy
6218 rather than address of cmode as last argument to prepare_cmp_insn.
6220 2018-03-15 Julia Koval <julia.koval@intel.com>
6222 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
6223 F_AVX512VNNI, F_AVX512BITALG): New.
6225 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
6228 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
6229 insn for floating-point loads and stores.
6231 2018-03-14 Carl Love <cel@us.ibm.com>
6233 * config/rs6000/rs6000-c.c: Add macro definitions for
6234 ALTIVEC_BUILTIN_VEC_PERMXOR.
6235 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
6236 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
6237 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
6238 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
6240 * config/doc/extend.texi: Add prototypes for vec_permxor.
6242 2018-03-14 David Malcolm <dmalcolm@redhat.com>
6245 * diagnostic-show-locus.c (class layout_point): Convert m_line
6246 from int to linenum_type.
6247 (line_span::comparator): Use linenum "compare" function when
6248 comparing line numbers.
6249 (test_line_span): New function.
6250 (layout_range::contains_point): Convert param "row" from int to
6252 (layout_range::intersects_line_p): Likewise.
6253 (layout::will_show_line_p): Likewise.
6254 (layout::print_source_line): Likewise.
6255 (layout::should_print_annotation_line_p): Likewise.
6256 (layout::print_annotation_line): Likewise.
6257 (layout::print_leading_fixits): Likewise.
6258 (layout::annotation_line_showed_range_p): Likewise.
6259 (struct line_corrections): Likewise for field m_row.
6260 (line_corrections::line_corrections): Likewise for param "row".
6261 (layout::print_trailing_fixits): Likewise.
6262 (layout::get_state_at_point): Likewise.
6263 (layout::get_x_bound_for_row): Likewise.
6264 (layout::print_line): Likewise.
6265 (diagnostic_show_locus): Likewise for locals "last_line" and
6267 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
6268 * input.c (selftest::test_linenum_comparisons): New function.
6269 (selftest::input_c_tests): Call it.
6270 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
6271 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
6272 * selftest.h (ASSERT_GT): New macro.
6273 (ASSERT_GT_AT): New macro.
6274 (ASSERT_LT): New macro.
6275 (ASSERT_LT_AT): New macro.
6277 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
6279 PR rtl-optimization/84780
6280 * combine.c (distribute_links): Don't make a link based on pc_rtx.
6282 2018-03-14 Martin Liska <mliska@suse.cz>
6284 * tree.c (record_node_allocation_statistics): Use
6285 get_stats_node_kind.
6286 (get_stats_node_kind): New function extracted from
6287 record_node_allocation_statistics.
6288 (free_node): Use get_stats_node_kind.
6290 2018-03-14 Richard Biener <rguenther@suse.de>
6292 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
6293 that the value-set of ANTIC_IN doesn't grow.
6296 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
6298 (BB_VISITED_WITH_VISITED_SUCCS): New define.
6299 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
6301 2018-03-14 Julia Koval <julia.koval@intel.com>
6303 * config.gcc (icelake-client, icelake-server): New.
6305 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
6306 (initial_ix86_arch_features): Ditto.
6307 (PTA_SKYLAKE): Add SGX.
6308 (PTA_ICELAKE): Remove.
6309 (PTA_ICELAKE_CLIENT): New.
6310 (PTA_ICELAKE_SERVER): New.
6311 (ix86_option_override_internal): Split up icelake on icelake client and
6313 (get_builtin_code_for_version): Ditto.
6314 (fold_builtin_cpu): Ditto.
6315 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
6316 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
6317 * config/i386/i386.h (processor_type): Ditto.
6318 * doc/invoke.texi: Ditto.
6320 2018-03-14 Jakub Jelinek <jakub@redhat.com>
6323 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
6324 INTEGER_CST offset, add it together with bitpos / 8 and
6325 sign extend based on POINTER_SIZE.
6329 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
6332 * config/i386/constraints.md (Yc): New register constraint.
6333 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
6334 Use Yc constraint for alternative 2 of operand 0. Remove
6335 preferred_for_speed attribute.
6337 2018-03-14 Richard Biener <rguenther@suse.de>
6339 PR tree-optimization/84830
6340 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
6341 with the old one to avoid oscillations.
6343 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
6346 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
6348 (assign_by_spills): Return a flag of reload assignment failure.
6349 Do not process the reload assignment failures. Do not spill other
6350 reload pseudos if they has the same reg class. Update n if
6352 (lra_assign): Add a return arg. Set up from the result of
6353 assign_by_spills call.
6354 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6355 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6356 usage_insns if it is not NULL.
6357 (spill_hard_reg_in_range): New function.
6358 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6359 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6360 function prototypes.
6361 (lra_assign): Change prototype.
6362 * lra.c (lra): Add code to deal with fails by splitting hard reg
6365 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
6367 * config/riscv/riscv.opt (mrelax): New option.
6368 * config/riscv/riscv.c (riscv_file_start): Emit ".option
6369 "norelax" when riscv_mrelax is disabled.
6370 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
6372 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6375 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
6376 reassociation for int modes.
6378 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6380 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
6381 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
6383 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
6384 * config/aarch64/aarch64-sve.md
6385 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
6386 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
6387 (*extend<mode><Vwide>2): Rename to...
6388 (aarch64_sve_extend<mode><Vwide>2): ...this.
6389 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
6390 renaming the old pattern to...
6391 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
6393 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
6394 define_expand, renaming the old pattern to...
6395 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
6396 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
6397 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
6398 account when deciding which SVE instruction the optab should use.
6399 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
6401 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6403 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
6404 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
6405 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
6406 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
6407 (tlsdesc_small_<mode>): Turn a define_expand and use
6408 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
6409 (tlsdesc_small_advsimd_<mode>): ...this.
6410 (tlsdesc_small_sve_<mode>): New pattern.
6412 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
6414 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
6415 (UNSPEC_UMUL_HIGHPART): New constants.
6416 (MUL_HIGHPART): New int iteraor.
6417 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
6418 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
6420 (*<su>mul<mode>3_highpart): New define_insn.
6422 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
6425 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
6428 2018-03-13 Martin Liska <mliska@suse.cz>
6431 * (sem_item_optimizer::sem_item_optimizer): Initialize new
6433 (sem_item_optimizer::~sem_item_optimizer): Release it.
6434 (sem_item_optimizer::merge_classes): Register variable aliases.
6435 (sem_item_optimizer::fixup_pt_set): New function.
6436 (sem_item_optimizer::fixup_points_to_sets): Likewise.
6437 * ipa-icf.h: Declare new variables and functions.
6439 2018-03-13 Jakub Jelinek <jakub@redhat.com>
6442 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
6443 integer_pow2p@2 and test integer_pow2p in condition.
6444 (A < 0 ? C : 0): Similarly for @1.
6447 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
6448 characters starting at p contain '\0' character, don't look beyond
6452 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
6453 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
6456 * reg-stack.c (change_stack): Change update_end var from int to
6457 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
6458 also call set_block_for_insn on the newly added insns and rescan.
6461 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
6462 on the last operand.
6465 * tree.c (stabilize_reference_1): Return save_expr (e) for
6466 STATEMENT_LIST even if it doesn't have side-effects.
6468 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
6470 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
6472 2018-03-12 Renlin Li <renlin.li@arm.com>
6474 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
6475 aarch64_output_scalar_simd_mov_immediate.
6477 2018-03-12 Martin Sebor <msebor@redhat.com>
6479 PR tree-optimization/83456
6480 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
6481 for perfectly overlapping calls to memcpy.
6482 (gimple_fold_builtin_memory_chk): Same.
6483 (gimple_fold_builtin_strcpy): Handle no-warning.
6484 (gimple_fold_builtin_stxcpy_chk): Same.
6485 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
6487 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6489 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
6490 parameter. Use it for SFmode.
6491 (rs6000_function_arg_advance_1): Adjust.
6492 (rs6000_function_arg): Adjust.
6493 (rs6000_gimplify_va_arg): Pass false for that new parameter.
6495 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
6497 PR rtl-optimization/84169
6498 PR rtl-optimization/84780
6499 * combine.c (can_combine_p): Check for a 2-insn combination whether
6500 the destination register is used between the two insns, too.
6502 2018-03-12 Richard Biener <rguenther@suse.de>
6504 PR tree-optimization/84803
6505 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
6506 for refs DR analysis didn't process.
6508 2018-03-12 Richard Biener <rguenther@suse.de>
6510 PR tree-optimization/84777
6511 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
6512 force-vectorize loops ignore whether we are optimizing for size.
6514 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
6516 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
6517 (TARGET_MD_ASM_ADJUST): Define.
6519 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
6520 Kito Cheng <kito.cheng@gmail.com>
6521 Chung-Ju Wu <jasonwucj@gmail.com>
6523 * config/nds32/nds32.c (nds32_compute_stack_frame,
6524 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
6525 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
6526 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
6527 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
6528 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
6529 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
6530 * config/nds32/nds32.md (prologue, epilogue): Use macro
6531 NDS32_V3PUSH_AVAILABLE_P to do checking.
6533 2018-03-11 Jakub Jelinek <jakub@redhat.com>
6536 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
6537 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
6538 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
6539 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
6540 addition of most attributes on !orig_type_die or the attribute not
6541 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
6543 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6544 Chung-Ju Wu <jasonwucj@gmail.com>
6546 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
6548 * config/nds32/nds32.opt (mvh): New option.
6550 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6551 Chung-Ju Wu <jasonwucj@gmail.com>
6553 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
6555 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
6556 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
6559 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
6560 Chung-Ju Wu <jasonwucj@gmail.com>
6562 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
6564 * config/nds32/nds32-multiple.md (strlensi): New pattern.
6565 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
6567 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
6568 Kito Cheng <kito.cheng@gmail.com>
6569 Chung-Ju Wu <jasonwucj@gmail.com>
6571 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
6572 UNSPEC_FFMISM and UNSPEC_FLMISM.
6573 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
6574 for ffb, ffmism and flmism.
6575 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
6576 (unspec_ffmism): Ditto.
6577 (unspec_flmism): Ditto.
6578 (nds32_expand_builtin_impl): Check if string extension is available.
6579 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
6580 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
6582 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
6585 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6588 * lra-assigns.c (assign_by_spills): Return a flag of reload
6589 assignment failure. Do not process the reload assignment
6590 failures. Do not spill other reload pseudos if they has the same
6592 (lra_assign): Add a return arg. Set up from the result of
6593 assign_by_spills call.
6594 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6595 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6596 usage_insns if it is not NULL.
6597 (spill_hard_reg_in_range): New function.
6598 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6599 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6600 function prototypes.
6601 (lra_assign): Change prototype.
6602 * lra.c (lra): Add code to deal with fails by splitting hard reg
6605 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
6608 * config/i386/i386.opt: Replace Enforcment with Enforcement.
6610 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
6613 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
6614 (dw_val_node): Add val_symbolic_view.
6615 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
6616 (symview_upper_bound): New.
6617 (new_line_info_table): Initialize symviews_since_reset.
6618 (dwarf2out_source_line): Count symviews_since_reset and set
6619 symview_upper_bound.
6620 (dw_val_equal_p): Handle symview.
6621 (add_AT_symview): New.
6622 (print_dw_val): Handle symview.
6623 (attr_checksum, attr_checksum_ordered): Likewise.
6624 (same_dw_val_p, size_of_die): Likewise.
6625 (value_format, output_die): Likewise.
6626 (add_high_low_attributes): Use add_AT_symview for entry_view.
6627 (dwarf2out_finish): Reset symview_upper_bound, clear
6630 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
6633 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
6634 Add strict argument and use it.
6635 (rs6000_split_multireg_move): Update for new strict argument.
6636 (mem_operand_gpr): Disallow all non-offsettable addresses.
6637 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
6639 2018-03-09 Jakub Jelinek <jakub@redhat.com>
6642 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
6643 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
6644 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
6647 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
6648 decl, use remap_type if we want to use the type.
6650 2018-03-09 Martin Sebor <msebor@redhat.com>
6652 PR tree-optimization/84526
6653 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
6655 (builtin_access::generic_overlap): Be prepared to handle non-array
6658 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
6660 PR rtl-optimization/84682
6661 * lra-constraints.c (process_address_1): Check is_address flag
6662 for address constraints.
6663 (process_alt_operands): Likewise.
6664 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
6665 preprocess_constraints.
6666 * recog.h (preprocess_constraints): Add oploc parameter.
6668 * recog.c (preprocess_constraints): Test address_operand for
6669 CT_ADDRESS constraints.
6671 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
6674 * lra-assigns.c (assign_by_spills): Return a flag of reload
6675 assignment failure. Do not process the reload assignment
6676 failures. Do not spill other reload pseudos if they has the same
6678 (lra_assign): Add a return arg. Set up from the result of
6679 assign_by_spills call.
6680 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
6681 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
6682 usage_insns if it is not NULL.
6683 (spill_hard_reg_in_range): New function.
6684 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
6685 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
6686 function prototypes.
6687 (lra_assign): Change prototype.
6688 * lra.c (lra): Add code to deal with fails by splitting hard reg
6691 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6694 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
6695 Accept complain bool parameter. Only emit errors if it is true.
6696 (arm_parse_cpu_option_name): Likewise.
6697 (arm_target_thumb_only): Adjust callers of the above.
6698 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
6699 prototype to take a default true bool parameter.
6700 (arm_parse_arch_option_name): Likewise.
6702 2018-03-09 David Malcolm <dmalcolm@redhat.com>
6703 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
6707 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
6708 * configure: Regenerate.
6709 * configure.ac ("linker --version-script option"): New.
6710 ("linker soname option"): New.
6712 2018-03-09 Richard Biener <rguenther@suse.de>
6714 PR tree-optimization/84775
6715 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
6716 immediate uses of predicate stmts and mark them modified.
6719 PR tree-optimization/84178
6720 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6722 (version_loop_for_if_conversion): Delay update_ssa call.
6723 (tree_if_conversion): Delay update_ssa until after predicate
6726 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
6729 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
6730 when the function accesses prior frames.
6732 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6735 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
6736 gen_llsym, otherwise call maybe_gen_llsym.
6739 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
6740 has ',' character inside of it.
6742 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6745 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
6746 as clobbering CC_REGNUM.
6748 2018-03-08 Richard Biener <rguenther@suse.de>
6751 * tree-scalar-evolution.c: Include tree-into-ssa.h.
6752 (follow_copies_to_constant): Do not follow SSA names registered
6755 2018-03-08 Richard Biener <rguenther@suse.de>
6757 PR tree-optimization/84178
6758 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6760 (version_loop_for_if_conversion): Delay update_ssa call.
6761 (tree_if_conversion): Delay update_ssa until after predicate
6764 2018-03-08 David Malcolm <dmalcolm@redhat.com>
6766 PR tree-optimization/84178
6767 * tree-if-conv.c (release_bb_predicate): Remove the
6768 the assertion that the stmts have NULL use_ops.
6769 Discard the statements, asserting that they haven't
6770 yet been added to a BB.
6772 2018-03-08 Richard Biener <rguenther@suse.de>
6774 PR tree-optimization/84746
6775 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
6776 (phi_translate): Pass in destination ANTIC_OUT set.
6777 (phi_translate_1): Likewise. For a simplified result lookup
6778 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
6779 (phi_translate_set): Adjust.
6780 (do_pre_regular_insertion): Likewise.
6781 (do_pre_partial_partial_insertion): Likewise.
6783 2018-03-08 Martin Liska <mliska@suse.cz>
6785 PR gcov-profile/84735
6786 * doc/gcov.texi: Document usage of profile files.
6787 * gcov-io.h: Document changes in the format.
6789 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
6793 * dwarf2out.c (struct dw_line_info_table): Update comments for
6795 (FORCE_RESET_NEXT_VIEW): New.
6796 (FORCE_RESETTING_VIEW_P): New.
6797 (RESETTING_VIEW_P): Check for -1 too.
6798 (ZERO_VIEW_P): Likewise.
6799 (new_line_info_table): Force-reset next view.
6800 (dwarf2out_begin_function): Likewise.
6801 (dwarf2out_source_line): Simplify zero_view_p initialization.
6802 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
6803 view directly. Omit view when omitting .loc at line 0.
6805 2018-03-08 Jakub Jelinek <jakub@redhat.com>
6807 PR tree-optimization/84740
6808 * tree-switch-conversion.c (process_switch): Call build_constructors
6809 only if info.phi_count is non-zero.
6811 PR tree-optimization/84739
6812 * tree-tailcall.c (find_tail_calls): Check call arguments against
6813 DECL_ARGUMENTS (current_function_decl) rather than
6814 DECL_ARGUMENTS (func) when checking for tail recursion.
6816 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6818 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
6819 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
6820 Volker Reichelt's entry and add entries for people that perform
6821 GCC fuzzy testing and report numerous bugs.
6823 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
6826 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
6827 readonly data in sdata, if that is disabled.
6828 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
6829 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
6830 -mreadonly-in-sdata option.
6832 2018-03-07 Martin Sebor <msebor@redhat.com>
6834 PR tree-optimization/84468
6835 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
6836 basic block when looking for nul assignment.
6838 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
6841 * except.h (output_function_exception_table): Adjust prototype.
6842 * except.c (output_function_exception_table): Remove FNNAME parameter
6843 and add SECTION parameter. Ouput one part of the table at a time.
6844 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
6845 the first part of the exception table and emit unwind directives.
6846 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
6847 (i386_pe_seh_cold_init): Likewise.
6848 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
6849 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
6850 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
6851 (ix86_output_call_insn): Emit a nop in one more case for SEH.
6852 * config/i386/winnt.c: Include except.h.
6853 (struct seh_frame_state): Add reg_offset, after_prologue and
6854 in_cold_section fields.
6855 (i386_pe_seh_end_prologue): Set seh->after_prologue.
6856 (i386_pe_seh_cold_init): New function.
6857 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
6858 to seh->in_cold_section.
6859 (seh_emit_push): Record the offset of the push.
6860 (seh_emit_save): Record the offet of the save.
6861 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
6862 Test seh->after_prologue to disregard the epilogue.
6863 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
6864 (i386_pe_end_cold_function): New function.
6866 2018-03-07 Jakub Jelinek <jakub@redhat.com>
6869 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
6870 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
6873 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6875 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
6876 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
6879 * multiple_target.c: Include tree-inline.h and intl.h.
6880 (expand_target_clones): Diagnose and fail if node->definition and
6881 !tree_versionable_function_p (node->decl).
6883 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
6885 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
6887 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
6888 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
6889 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
6891 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6894 * combine.c (try_combine): Use reg_or_subregno instead of handling
6895 just paradoxical SUBREGs and REGs.
6897 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
6899 * config/arc/arc.c (arc_finalize_pic): Remove function.
6900 (arc_must_save_register): We use single base PIC register, remove
6901 checks to save/restore the PIC register.
6902 (arc_expand_prologue): Likewise.
6903 * config/arc/arc-protos.h (arc_set_default_type_attributes):
6905 (arc_verify_short): Likewise.
6906 (arc_attr_type): Likewise.
6907 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
6908 (walk_stores): Likewise.
6909 (arc_address_cost): Make it static.
6910 (arc_verify_short): Likewise.
6911 (branch_dest): Likewise.
6912 (arc_attr_type): Likewise.
6913 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
6914 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
6915 (arc_final_prescan_insn): Remove inserting the nops due to
6916 hardware hazards. It is done in reorg step.
6917 (insn_length_variant_t): Remove.
6918 (insn_length_parameters_t): Likewise.
6919 (arc_insn_length_parameters): Likewise.
6920 (arc_get_insn_variants): Likewise.
6921 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
6923 2018-03-06 Jakub Jelinek <jakub@redhat.com>
6926 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
6929 PR tree-optimization/84687
6930 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6932 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
6934 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6936 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
6937 Rename to ppc_speculation_barrier.
6938 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
6939 __builtin_ppc_speculation_barrier.
6941 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6944 * combine.c (combine_simplify_rtx): Don't try to simplify if
6945 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
6948 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
6950 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
6951 to 32 bytes when compiling for POWER9.
6953 2018-03-05 Jakub Jelinek <jakub@redhat.com>
6956 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
6957 regparm >= 3 with no arg reg available also for calls with
6958 flag_force_indirect_call. Pass decl to ix86_function_regparm.
6961 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
6963 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
6965 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
6968 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
6970 2018-03-05 Richard Biener <rguenther@suse.de>
6972 PR tree-optimization/84486
6973 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
6974 When inserting a __builtin_assume_aligned call set the LHS
6975 SSA name alignment info accordingly.
6977 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
6979 PR tree-optimization/84114
6980 * config/aarch64/aarch64.c (aarch64_reassociation_width)
6981 Avoid reassociation of FLOAT_MODE addition.
6983 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
6985 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
6986 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
6987 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
6988 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
6989 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
6990 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
6991 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
6993 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
6994 __builtin_ia32_wbinvd): New builtins.
6995 (SPECIAL_ARGS2): New.
6996 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
6997 (SPECIAL_ARGS2): New.
6998 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
6999 (ix86_valid_target_attribute_inner_p): Ditto.
7000 (ix86_init_mmx_sse_builtins): Add special_args2.
7001 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
7002 TARGET_WBNOINVD_P): New.
7003 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
7004 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
7005 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
7006 * config/i386/immintrin.h (_wbinvd): New intrinsic.
7007 * config/i386/pconfigintrin.h: New file.
7008 * config/i386/wbnoinvdintrin.h: Ditto.
7009 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
7010 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
7012 2018-03-05 Richard Biener <rguenther@suse.de>
7014 PR tree-optimization/84670
7015 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7017 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7018 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7019 (compute_antic_aux): Only assert the number of values in ANTIC_IN
7020 doesn't grow if all successors (recursively) were visited at least
7023 2018-03-05 Richard Biener <rguenther@suse.de>
7025 PR tree-optimization/84650
7026 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
7027 if executed in the loop pipeline.
7029 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
7031 * doc/configfiles.texi (Configuration Files): Move info about
7032 conditionalizing $target-protos.h to...
7033 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
7034 differs from $target-protos.h.
7036 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
7037 Chung-Ju Wu <jasonwucj@gmail.com>
7039 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
7040 * config/nds32/nds32-multiple.md (setmemsi): Define.
7041 * config/nds32/nds32-memory-manipulation.c
7042 (nds32_gen_dup_4_byte_to_word_value): New.
7043 (emit_setmem_word_loop): New.
7044 (emit_setmem_byte_loop): New.
7045 (nds32_expand_setmem_loop): New.
7046 (nds32_expand_setmem_loop_v3m): New.
7047 (nds32_expand_setmem_unroll): New.
7048 (nds32_expand_setmem): New.
7050 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7051 Chung-Ju Wu <jasonwucj@gmail.com>
7053 * config/nds32/nds32-memory-manipulation.c
7054 (nds32_emit_load_store): New.
7055 (nds32_emit_post_inc_load_store): New.
7056 (nds32_emit_mem_move): New.
7057 (nds32_emit_mem_move_block): New.
7058 (nds32_expand_movmemsi_loop_unknown_size): New.
7059 (nds32_expand_movmemsi_loop_known_size): New.
7060 (nds32_expand_movmemsi_loop): New.
7061 (nds32_expand_movmemsi_unroll): New.
7062 (nds32_expand_movmemqi): Rename ...
7063 (nds32_expand_movmemsi): ... to this.
7064 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
7065 (movmemsi): ... to this.
7066 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
7067 (nds32_expand_movmemsi): ... to this.
7069 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7070 Monk Chiang <sh.chiang04@gmail.com>
7071 Chung-Ju Wu <jasonwucj@gmail.com>
7073 * config/nds32/nds32-protos.h
7074 (nds32_expand_load_multiple): New arguments.
7075 (nds32_expand_store_multiple): Ditto.
7076 (nds32_valid_multiple_load_store): Rename ...
7077 (nds32_valid_multiple_load_store_p): ... to this.
7078 * config/nds32/nds32-memory-manipulation.c
7079 (nds32_expand_load_multiple): Refine implementation.
7080 (nds32_expand_store_multiple): Ditto.
7081 * config/nds32/nds32-multiple.md
7082 (load_multiple): Update nds32_expand_load_multiple interface.
7083 (store_multiple): Update nds32_expand_store_multiple interface.
7084 * config/nds32/nds32-predicates.c
7085 (nds32_valid_multiple_load_store): Rename ...
7086 (nds32_valid_multiple_load_store_p): ... to this and refine
7088 * config/nds32/predicates.md
7089 (nds32_load_multiple_and_update_address_operation): New predicate.
7090 (nds32_store_multiple_and_update_address_operation): New predicate.
7092 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
7093 Chung-Ju Wu <jasonwucj@gmail.com>
7095 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
7096 (combo): New attribute.
7097 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
7099 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
7101 * config/nds32/nds32.opt: Change -mcmodel= default value.
7103 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
7104 Monk Chiang <sh.chiang04@gmail.com>
7105 Chung-Ju Wu <jasonwucj@gmail.com>
7107 * config/nds32/constants.md (unspec_element): New enum.
7108 * config/nds32/constraints.md (Umw): New constraint.
7109 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
7110 * config/nds32/nds32-intrinsic.md: Likewise.
7111 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
7112 (nds32_valid_smw_lwm_base_p): New.
7113 (nds32_output_smw_single_word): New.
7114 (nds32_output_lmw_single_word): New.
7115 (nds32_expand_unaligned_load): New.
7116 (nds32_expand_unaligned_store): New.
7117 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
7118 (nds32_output_smw_single_word): Declare.
7119 (nds32_output_lmw_single_word): Declare.
7120 (nds32_expand_unaligned_load): Declare.
7121 (nds32_expand_unaligned_store): Declare.
7122 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
7123 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
7124 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
7125 NDS32_BUILTIN_UASTORE_DW.
7126 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
7129 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
7130 Kito Cheng <kito.cheng@gmail.com>
7131 Chung-Ju Wu <jasonwucj@gmail.com>
7133 * config/nds32/nds32-intrinsic.c
7134 (nds32_expand_builtin_null_ftype_reg): Delete.
7135 (nds32_expand_builtin_reg_ftype_imm): Ditto.
7136 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
7137 (nds32_read_argument): New.
7138 (nds32_legitimize_target): Ditto.
7139 (nds32_legitimize_argument): Ditto.
7140 (nds32_check_constant_argument): Ditto.
7141 (nds32_expand_unop_builtin): Ditto.
7142 (nds32_expand_unopimm_builtin): Ditto.
7143 (nds32_expand_binop_builtin): Ditto.
7144 (nds32_builtin_decl_impl): Ditto.
7145 (builtin_description): Ditto.
7146 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
7147 (nds32_init_builtins_impl): Ditto.
7148 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
7149 (nds32_builtin_decl): New.
7150 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
7151 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
7153 2018-03-02 Jeff Law <law@redhat.com>
7155 * reorg.c (stop_search_p): Handle DEBUG_INSN.
7156 (redundant_insn, fill_simple_delay_slots): Likewise.
7157 (fill_slots_from_thread): Likewise.
7158 * resource.c (mark_referenced_resources): Likewise.
7159 (mark_set_resources, find_dead_or_set_registers): Likewise.
7161 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7163 * substring-locations.h (format_warning_va): Formatting fix for
7165 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
7167 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
7168 * substring-locations.c: Include intl.h.
7169 (format_warning_va): Turned into small wrapper around
7170 format_warning_n_va, renamed to ...
7171 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
7172 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
7174 (format_warning_at_substring_n): New function.
7175 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
7176 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
7177 format_warning_at_substring with just a shorter name instead of
7178 const function pointer.
7179 (fmtwarn_n): New function.
7180 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
7181 appropriate, get rid of all the fmtstr temporaries, move conditionals
7182 with G_() wrapped string literals directly into fmtwarn arguments,
7183 cast dir.len to (int), formatting fixes.
7185 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
7187 * doc/invoke.texi: Remove "Cilk Plus" references.
7189 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7190 Richard Biener <rguenther@suse.de>
7193 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
7194 for error or warning attributes if CALL_FROM_THUNK_P is set.
7197 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7200 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
7201 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
7204 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
7205 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
7207 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
7208 instead of -1U in last predictors element's probability member.
7210 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
7213 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
7214 arguments if they are comparable.
7216 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7218 PR tree-optimization/84634
7219 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
7220 masks and masked_loop_p with a single loop_masks, making sure it's
7221 null for bb vectorization.
7223 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
7225 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
7226 (vect_analyze_data_ref_access): Use loop->safe_len rather than
7227 loop->force_vectorize to check whether there is no alias.
7229 2018-03-02 Jakub Jelinek <jakub@redhat.com>
7232 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
7234 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
7236 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
7237 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
7238 instead of a loop around prev_real_insn.
7239 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
7243 * config/i386/i386.c (ix86_print_operand): Use conditional
7244 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
7247 2018-03-02 Richard Biener <rguenther@suse.de>
7249 PR tree-optimization/84427
7250 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
7251 (bitmap_set_subtract_values): Rewrite to handle multiple
7254 (prune_clobbered_mems): Likewise.
7255 (phi_translate): Take edge instead of pred/phiblock.
7256 (phi_translate_1): Likewise.
7257 (phi_translate_set): Likewise. Insert all translated
7258 exprs for a value into the set, keeping possibly multiple
7259 expressions per value.
7260 (compute_antic_aux): Adjust for phi_translate changes.
7261 When intersecting union the expressions and prune those
7262 not in the final value set, keeping possibly multiple
7263 expressions per value. Do not use value-insertion
7264 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
7265 all expressions. Add verification that the value-sets
7266 only shrink during iteration.
7267 (compute_partial_antic_aux): Adjust for the phi_translate changes.
7268 (do_pre_regular_insertion): Likewise.
7269 (do_pre_partial_partial_insertion): Likewise.
7271 2018-03-02 Richard Biener <rguenther@suse.de>
7274 * config/darwin.c (saved_debug_info_level): New static global.
7275 (darwin_asm_lto_start): Disable debug info generation for LTO out.
7276 (darwin_asm_lto_end): Restore debug info generation settings.
7278 2018-03-01 Martin Liska <mliska@suse.cz>
7281 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
7284 2018-03-01 Richard Biener <rguenther@suse.de>
7287 * dwarf2out.c (gen_variable_die): Properly handle late VLA
7288 type annotation with LTO when debug was disabled at compile-time.
7290 2018-03-01 Matthew Fortune <mfortune@gmail.com>
7292 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
7294 (mips_final_postscan_insn): Likewise.
7296 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
7298 PR rtl-optimization/84528
7299 * alias.c (init_alias_target): Add commentary.
7300 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
7301 a unique base value if the frame pointer is not eliminated
7302 to the stack pointer.
7304 2018-03-01 Tom de Vries <tom@codesourcery.com>
7306 PR rtl-optimization/83327
7307 * lra-int.h (hard_regs_spilled_into): Declare.
7308 * lra.c (hard_regs_spilled_into): Define.
7309 (init_reg_info): Init hard_regs_spilled_into.
7310 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
7311 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
7312 (process_bb_lives): Handle hard_regs_spilled_into.
7313 (lra_create_live_ranges_1): Before doing liveness propagation, clear
7314 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
7316 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
7318 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
7319 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
7320 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
7321 * config/rs6000/aix72.h: New file.
7323 2018-02-28 Jakub Jelinek <jakub@redhat.com>
7325 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
7326 instead of warning_at with conditional singular and plural messages
7330 * stor-layout.c (update_alignment_for_field): For
7331 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
7332 && !DECL_PACKED (field), do the alignment update, just use
7333 only desired_align instead of MAX (type_align, desired_align)
7335 (place_field): Don't do known_align < desired_align handling
7336 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
7337 is non-NULL, instead do it after rli->prev_field handling and
7338 only if not within a bitfield word. For DECL_PACKED (field)
7339 use type_align of BITS_PER_UNIT.
7341 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
7343 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
7344 superfluous parentheses and trailing spaces.
7346 2018-02-28 Richard Biener <rguenther@suse.de>
7348 PR tree-optimization/84584
7349 * graphite-scop-detection.c (scop_detection::add_scop): Discard
7350 SCoPs with fake exit edge.
7352 2018-02-28 Martin Liska <mliska@suse.cz>
7355 * timevar.c (timer::print): Fix format to properly print 100%
7358 2018-02-28 Richard Biener <rguenther@suse.de>
7361 * genmatch.c (capture_info::walk_match): Do not mark
7362 captured expressions without operands as expr_p given
7363 they act more like predicates and should be subject to
7364 "lost tail" side-effect preserving.
7366 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
7368 PR rtl-optimization/81611
7369 * auto-inc-dec.c (attempt_change): Move dead note from
7370 mem_insn if it's the next use of regno
7371 (find_address): Take address use of reg holding
7372 non-incremented value. Add parm to limit search to the named
7374 (merge_in_block): Attempt to use a mem insn that is the next
7375 use of the original regno.
7377 2018-02-27 Martin Sebor <msebor@redhat.com>
7380 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
7381 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
7383 2018-02-27 Martin Sebor <msebor@redhat.com>
7385 PR translation/84207
7386 * diagnostic-core.h (warning_n, error_n, inform_n): Change
7387 n argument to unsigned HOST_WIDE_INT.
7388 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
7389 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
7390 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
7391 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
7393 2018-02-27 Richard Biener <rguenther@suse.de>
7395 PR tree-optimization/84512
7396 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
7397 Do not use the estimate returned from record_stmt_cost for
7398 the scalar iteration cost but sum properly using add_stmt_cost.
7400 2018-02-27 Richard Biener <rguenther@suse.de>
7402 PR tree-optimization/84466
7403 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
7404 Adjust last change to less strictly validate use operands.
7406 2018-02-27 Martin Liska <mliska@suse.cz>
7408 PR gcov-profile/84548
7409 * gcov.c (process_file): Allow partial overlap and consider it
7410 also as group functions.
7411 (output_lines): Properly calculate range of lines for a group.
7413 2018-02-27 Martin Liska <mliska@suse.cz>
7415 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
7416 'ggc' suffixes. Change first column width.
7417 (timer::print): Fix formatting of the column.
7419 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
7421 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
7422 preserve inline entry blocks for the sake of debug inline
7423 entry point markers alone.
7424 (remove_unused_locals): Suggest in comments a better place to
7425 force the preservation of inline entry blocks that are
7426 otherwise unused, but do not preserve them.
7428 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7430 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
7432 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7435 * config/i386/constraints.md (Bs): Replace
7436 ix86_indirect_branch_register with
7437 TARGET_INDIRECT_BRANCH_REGISTER.
7439 * config/i386/i386.md (indirect_jump): Likewise.
7440 (tablejump): Likewise.
7441 (*sibcall_memory): Likewise.
7442 (*sibcall_value_memory): Likewise.
7443 Peepholes of indirect call and jump via memory: Likewise.
7444 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
7445 (*sibcall_value_GOT_32): Likewise.
7446 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7447 (GOT_memory_operand): Likewise.
7448 (call_insn_operand): Likewise.
7449 (sibcall_insn_operand): Likewise.
7450 (GOT32_symbol_operand): Likewise.
7451 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
7453 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7455 PR rtl-optimization/83496
7456 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
7457 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
7458 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
7459 redundant insn, if any.
7460 (relax_delay_slots): Likewise.
7461 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
7463 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
7465 PR tree-optimization/83965
7466 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
7467 that grouped statements are part of a reduction chain. Return
7468 true if the statement is not marked as a reduction itself but
7470 (vect_recog_dot_prod_pattern): Don't check whether the statement
7471 is part of a group here.
7472 (vect_recog_sad_pattern): Likewise.
7473 (vect_recog_widen_sum_pattern): Likewise.
7475 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
7478 * final.c (rest_of_clean_state): Also look for calls inside sequences.
7480 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
7483 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
7485 (ix86_output_indirect_function_return): New prototype.
7486 (ix86_split_simple_return_pop_internal): Likewise.
7487 * config/i386/i386.c (indirect_return_via_cx): New.
7488 (indirect_return_via_cx_bnd): Likewise.
7489 (indirect_thunk_name): Handle return va CX_REG.
7490 (output_indirect_thunk_function): Create alias for
7491 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
7492 (ix86_output_indirect_jmp): Remove the bool argument.
7493 (ix86_output_indirect_function_return): New function.
7494 (ix86_split_simple_return_pop_internal): Likewise.
7495 * config/i386/i386.md (*indirect_jump): Don't pass false
7496 to ix86_output_indirect_jmp.
7497 (*tablejump_1): Likewise.
7498 (simple_return_pop_internal): Change it to define_insn_and_split.
7499 Call ix86_split_simple_return_pop_internal to split it for
7501 (simple_return_indirect_internal): Call
7502 ix86_output_indirect_function_return instead of
7503 ix86_output_indirect_jmp.
7505 2018-02-26 Jakub Jelinek <jakub@redhat.com>
7508 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
7509 memset and value initialization afterwards.
7511 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
7513 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
7515 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7518 * common/config/aarch64/aarch64-common.c
7519 (aarch_option_optimization_table[]): Switch
7520 off fomit-frame-pointer
7522 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7523 Chung-Ju Wu <jasonwucj@gmail.com>
7525 * config/nds32/nds32-multiple.md (load_multiple): Disallow
7527 (store_multiple): Ditto.
7529 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
7531 * config.gcc: Add --with-cpu support for nds32 target.
7532 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
7533 * config/nds32/nds32.opt: Add -mcpu= option.
7535 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
7537 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
7538 isel=yes): Warn for these deprecated options.
7540 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
7542 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
7543 ISA_2_5_MASKS_EMBEDDED.
7545 2018-02-23 Jakub Jelinek <jakub@redhat.com>
7547 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
7548 p->max as pointers rather than using iterative_hash_expr.
7550 2018-02-23 Carl Love <cel@us.ibm.com>
7552 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
7553 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
7555 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
7556 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
7557 P8V_BUILTIN_VEC_VUNSIGNED2.
7559 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
7562 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
7563 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
7565 * lra-constraints.c (curr_insn_transform): Set up
7566 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
7568 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
7569 * lra-eliminations.c (spill_pseudos): Ditto.
7570 (process_insn_for_elimination): Ditto.
7571 * lra-lives.c (reg_early_clobber_p): Use the new macros.
7572 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
7573 LRA_NON_CLOBBERED_ALT.
7575 2018-02-22 Martin Sebor <msebor@redhat.com>
7577 PR tree-optimization/84480
7578 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
7579 to maybe_diag_stxncpy_trunc. Call it.
7580 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
7581 from gimple_fold_builtin_strcpy. Print inlining stack.
7582 (handle_builtin_stxncpy): Print inlining stack.
7583 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
7585 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
7588 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
7589 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
7590 and -fcheck-pointer-bounds are used together.
7591 (indirect_thunk_prefix): New enum.
7592 (indirect_thunk_need_prefix): New function.
7593 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
7594 "_nt" instead of "_bnd" for NOTRACK prefix.
7595 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
7596 (output_indirect_thunk_function): Likewise.
7598 (ix86_code_end): Update output_indirect_thunk_function calls.
7599 (ix86_output_indirect_branch_via_reg): Replace
7600 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
7601 (ix86_output_indirect_branch_via_push): Likewise.
7602 (ix86_output_function_return): Likewise.
7603 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
7604 incompatible with -fcf-protection=branch and
7605 -fcheck-pointer-bounds.
7607 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7610 * config/aarch64/aarch64.c (aarch64_print_address_internal):
7611 Change gcc_assert call to output_operand_lossage.
7613 2018-02-22 Steve Ellcey <sellcey@cavium.com>
7615 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
7617 2018-02-22 DJ Delorie <dj@redhat.com>
7618 Sebastian Perta <sebastian.perta@renesas.com>
7619 Oleg Endo <olegendo@gcc.gnu.org>
7621 * config/rx/rx.c (rx_rtx_costs): New function.
7622 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
7624 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
7626 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
7628 2018-02-22 Martin Liska <mliska@suse.cz>
7631 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
7632 Add "native" as a possible value.
7634 2018-02-22 Martin Liska <mliska@suse.cz>
7637 * config/i386/i386.c (ix86_option_override_internal):
7638 Add "native" as a possible value for -march and -mtune.
7640 2018-02-22 Jakub Jelinek <jakub@redhat.com>
7643 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
7644 to all type variants.
7646 PR tree-optimization/84503
7647 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
7648 width as info->bitpos + info->bitsize - start.
7649 (merged_store_group::merge_overlapping): Simplify width computation.
7650 (check_no_overlap): New function.
7651 (imm_store_chain_info::try_coalesce_bswap): Compute expected
7652 start + width and last_order of the group, fail if check_no_overlap
7654 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
7655 to group if check_no_overlap fails.
7657 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7659 * config/rs6000/altivec.md: Delete contraint arguments to
7660 define_expand, define_split, and define_peephole2, and in
7661 define_insn_and_split if always unused.
7662 * config/rs6000/darwin.md: Ditto.
7663 * config/rs6000/dfp.md: Ditto.
7664 * config/rs6000/rs6000.md: Ditto.
7665 * config/rs6000/sync.md: Ditto.
7666 * config/rs6000/vector.md: Ditto.
7667 * config/rs6000/vsx.md: Ditto.
7669 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
7671 * config/rs6000/altivec.md: Write output control strings as braced
7672 blocks instead of double-quoted strings.
7673 * config/rs6000/darwin.md: Ditto.
7674 * config/rs6000/rs6000.md: Ditto.
7675 * config/rs6000/vector.md: Ditto.
7676 * config/rs6000/vsx.md: Ditto.
7678 2018-02-21 Jason Merrill <jason@redhat.com>
7680 PR c++/84314 - ICE with templates and fastcall attribute.
7681 * attribs.c (build_type_attribute_qual_variant): Remove assert.
7683 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7685 * ipa-cp.c (determine_versionability): Fix comment typos.
7687 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
7690 * ipa-cp.c (determine_versionability): Do not version functions caling
7693 2018-02-21 Martin Liska <mliska@suse.cz>
7696 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
7697 Add "native" as a possible value.
7698 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
7699 the macro when native cpu detection is available.
7701 2018-02-21 Martin Liska <mliska@suse.cz>
7704 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
7705 Add "native" as a possible value.
7706 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
7707 when native cpu detection is available.
7709 2018-02-21 Jakub Jelinek <jakub@redhat.com>
7710 Martin Sebor <msebor@redhat.com>
7712 PR tree-optimization/84478
7713 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
7715 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
7716 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
7717 support which is conservatively correct, for 2 only stay conservative
7718 for maxlen. Formatting and comment capitalization fixes. Add STRICT
7719 argument to the 2 argument get_range_strlen, adjust 6 arg
7720 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
7722 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
7723 (gimple_fold_builtin_strlen): Pass true as last argument to
7726 2018-02-20 Martin Sebor <msebor@redhat.com>
7729 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
7730 (builtin_memref::set_base_and_offset): Same. Handle inner references.
7731 (builtin_memref::builtin_memref): Factor out parts into
7732 set_base_and_offset and call it.
7734 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7737 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
7738 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
7739 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
7740 search at the associated MODE_INT.
7742 2018-02-20 Jeff Law <law@redhat.com>
7745 PR tree-optimization/81592
7747 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
7748 for range data rather than using global data.
7749 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
7750 range data rather than using global data.
7751 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
7752 pass it to children as needed.
7753 (struct directive::fmtresult): Similarly.
7754 (struct directive::set_width): Similarly.
7755 (struct directive::set_precision): Similarly.
7756 (format_integer, format_directive, parse_directive): Similarly.
7757 (format_none): Accept unnamed vr_values parameter.
7758 (format_percent, format_floating, format_character): Similarly.
7759 (format_string, format_plain): Similarly.
7760 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
7761 the EVRP range analyzer for range data rather than using global data.
7762 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
7763 gimple-ssa-evrp-analyze.h
7764 (class sprintf_dom_walker): Add after_dom_children member function.
7765 Add evrp_range_analyzer member.
7766 (sprintf_dom_walker::before_dom_children): Call into the EVRP
7767 range analyzer as needed.
7768 (sprintf_dom_walker::after_dom_children): New member function.
7769 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
7771 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
7772 (evrp_range_analyzer::pop_to_marker): Likewise.
7774 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
7776 PR tree-optimization/84419
7777 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
7778 with the required type if its current type is compatible but
7781 2018-02-20 Jakub Jelinek <jakub@redhat.com>
7784 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
7785 after vectorization.
7787 2018-02-20 Martin Liska <mliska@suse.cz>
7790 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
7791 possible values if we don't have a hint.
7793 2018-02-20 Martin Liska <mliska@suse.cz>
7797 * final.c (shorten_branches): Build align_tab array with one
7799 * opts.c (finish_options): Add alignment option limit check.
7800 (MAX_CODE_ALIGN): Likewise.
7801 (MAX_CODE_ALIGN_VALUE): Likewise.
7802 * doc/invoke.texi: Document maximum allowed option value for
7803 all -falign-* options.
7805 2018-02-19 Jakub Jelinek <jakub@redhat.com>
7808 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
7809 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
7810 * var-tracking.c (emit_note_insn_var_location): Remove all references
7811 to NOTE_INSN_CALL_ARG_LOCATION.
7812 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
7813 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
7814 Use copy_rtx_if_shared.
7815 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
7816 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
7817 (dwarf2out_var_location): Remove handling of
7818 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
7820 * final.c (final_scan_insn): Remove all references to
7821 NOTE_INSN_CALL_ARG_LOCATION.
7822 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
7823 before dumping final insns.
7824 * except.c (emit_note_eh_region_end): Remove all references to
7825 NOTE_INSN_CALL_ARG_LOCATION.
7826 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
7827 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
7828 * config/arc/arc.c (hwloop_optimize): Likewise.
7829 * config/arm/arm.c (create_fix_barrier): Likewise.
7830 * config/s390/s390.c (s390_chunkify_start): Likewise.
7831 * config/sh/sh.c (find_barrier): Likewise.
7832 * config/i386/i386.c (rest_of_insert_endbranch,
7833 ix86_seh_fixup_eh_fallthru): Likewise.
7834 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
7835 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
7836 * config/frv/frv.c (frv_function_prologue): Likewise.
7837 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
7839 (note_outside_basic_block_p): Remove all references to
7840 NOTE_INSN_CALL_ARG_LOCATION.
7841 * gengtype.c (adjust_field_rtx_def): Likewise.
7842 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
7844 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
7845 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
7848 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
7851 PR tree-optimization/84452
7852 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
7853 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
7856 2018-02-19 Martin Liska <mliska@suse.cz>
7859 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
7861 2018-02-19 Martin Liska <mliska@suse.cz>
7862 Richard Sandiford <richard.sandiford@linaro.org>
7864 PR tree-optimization/82491
7865 * gimple-fold.c (get_base_constructor): Make earlier bail out
7868 2018-02-19 Carl Love <cel@us.ibm.com>
7870 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
7871 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
7873 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
7874 P8V_BUILTIN_VEC_NEG.
7876 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
7878 * config/rl78/rl78.md (movdf): New define expand.
7880 2018-02-19 Martin Liska <mliska@suse.cz>
7883 * doc/invoke.texi: Fix typo.
7884 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
7886 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
7888 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
7889 handle rs6000_single_float and rs6000_double_float specially for
7892 2018-02-16 Jeff Law <law@redhat.com>
7894 * config/rx/rx.c (add_pop_cfi_notes): New function.;
7897 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7900 * ipa-inline.c (inline_small_functions): Fix a typo.
7902 2018-02-16 Nathan Sidwell <nathan@acm.org>
7904 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
7906 2018-02-16 Carl Love <cel@us.ibm.com>
7908 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
7909 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
7910 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
7911 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
7912 expansion to P8V_BUILTIN_VEC_FLOAT2.
7914 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
7916 PR rtl-optimization/70023
7917 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
7918 src_regno into account.
7920 2018-02-16 Carl Love <cel@us.ibm.com>
7922 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
7923 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
7924 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
7925 * config/rs6000/rs6000.c: Remove case statements for
7926 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
7927 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
7928 and P9V_BUILTIN_VEC_VINSERT4B.
7929 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
7930 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
7931 * config/rs6000/vsx.md:
7932 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
7935 2018-02-16 Carl Love <cel@us.ibm.com>
7937 * config/rs6000/altivec.h: Add builtin names vec_extract4b
7939 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
7941 * config/rs6000/rs6000-c.c: Add the definitions for
7942 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
7943 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
7944 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
7945 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
7946 definition for insert4b and define insn *insert3b_internal.
7947 * doc/extend.texi: Add documentation for vec_extract4b.
7949 2018-02-16 Nathan Sidwell <nathan@acm.org>
7951 * doc/extend.texi (Backwards Compatibility): Mention friend
7952 injection. Note for-scope is deprecated.
7953 * doc/invoke.texi (-ffriend-injection): Deprecate.
7955 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
7957 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
7958 that moved to I2, also allow destinations that are a paradoxical
7959 subreg (instead of a normal reg).
7961 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
7964 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
7967 2018-02-16 Richard Biener <rguenther@suse.de>
7969 PR tree-optimization/84037
7970 PR tree-optimization/84016
7972 * config/i386/i386.c (ix86_builtin_vectorization_cost):
7973 Adjust vec_construct for the fact we need additional higher latency
7974 128bit inserts for AVX256 and AVX512 vector builds.
7975 (ix86_add_stmt_cost): Scale vector construction cost for
7978 2018-02-16 Richard Biener <rguenther@suse.de>
7980 PR tree-optimization/84417
7981 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
7982 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
7983 (non_rewritable_lvalue_p): Likewise, use poly-ints.
7985 2018-02-16 Martin Liska <mliska@suse.cz>
7988 * internal-fn.def (ASAN_CHECK): Set proper flags.
7989 (ASAN_MARK): Likewise.
7991 2018-02-16 Julia Koval <julia.koval@intel.com>
7993 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
7994 from PTA_CANNONLAKE.
7996 2018-02-16 Jakub Jelinek <jakub@redhat.com>
7999 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
8000 Use ++iter rather than iter++ for std::list iterators.
8001 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
8002 defer deleting them until all nodes in the forest are processed. Do
8003 free even leaf nodes. Change to_process into auto_vec.
8006 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
8007 * vec.h (vec_default_construct): Use memset instead of placement new
8008 if BROKEN_VALUE_INITIALIZATION is defined.
8009 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
8010 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
8013 PR rtl-optimization/83723
8014 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
8015 * lra.c (lra_substitute_pseudo): Likewise. If true, use
8016 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
8018 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
8020 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
8022 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
8024 PR rtl-optimization/81443
8025 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
8026 from inner REGs to paradoxical SUBREGs.
8028 2018-02-16 Richard Biener <rguenther@suse.de>
8030 PR tree-optimization/84399
8031 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
8032 For operands we can analyze at their definition make sure we can
8033 analyze them at each use as well.
8035 2018-02-16 Richard Biener <rguenther@suse.de>
8037 PR tree-optimization/84190
8038 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
8039 volatile accesses if the decl isn't volatile.
8041 2018-02-15 Jason Merrill <jason@redhat.com>
8043 PR c++/84314 - ICE with templates and fastcall attribute.
8044 * attribs.c (build_type_attribute_qual_variant): Don't clobber
8045 TYPE_CANONICAL on an existing type.
8047 2018-02-15 Jakub Jelinek <jakub@redhat.com>
8049 PR tree-optimization/84383
8050 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
8051 dstoff nor call operand_equal_p if dstbase is NULL.
8053 PR tree-optimization/84334
8054 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
8055 also a CONSTANT_CLASS_P, punt.
8057 2018-02-14 Jim Wilson <jimw@sifive.com>
8059 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
8060 first SMALL_OPERAND check. New local min_second_step. Move assert
8061 to where locals are set. Add TARGET_RVC support.
8062 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
8064 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
8066 * doc/invoke.texi: Correct -Wformat-overflow code sample.
8068 2018-02-14 Martin Sebor <msebor@redhat.com>
8070 PR tree-optimization/83698
8071 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
8072 arrays constrain the offset range to their bounds.
8073 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
8074 (builtin_access::overlap): Avoid setting the size of overlap if it's
8076 (maybe_diag_overlap): Also consider arrays when deciding what values
8077 of offsets to include in diagnostics.
8079 2018-02-14 Martin Sebor <msebor@redhat.com>
8082 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
8083 that correspond to the kind of a declaration.
8085 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
8088 * config/pa/pa.md: Load address of PIC label using the linkage table
8089 if the label is nonlocal.
8091 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
8093 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
8094 warning message if user requests -maltivec=be.
8095 * doc/invoke.texi: Document deprecation of -maltivec=be.
8097 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
8100 * config/rs6000/rs6000-c.c: Update definitions for
8101 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
8102 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
8104 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
8107 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
8108 add _get_ssp intrinsics. Remove argument from
8109 __builtin_ia32_rdssp[d|q].
8110 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
8111 * config/i386/i386-builtin.def: Remove argument from
8112 __builtin_ia32_rdssp[d|q].
8113 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
8114 ix86_expand_special_args_builtin for _rdssp[d|q].
8115 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
8116 Clear register before usage.
8117 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
8118 Add documentation for new _get_ssp and _inc_ssp intrinsics.
8120 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
8122 PR tree-optimization/84357
8123 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
8124 operand 1 of an ARRAY_REF too.
8126 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
8129 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
8130 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8132 (set_of_reg): New struct.
8133 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
8134 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
8135 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
8137 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
8138 Split into bitclr, bitset, bitinvert patterns if appropriate.
8139 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
8140 use rx_fuse_in_memory_bitop.
8141 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
8142 to named insn, correct maximum insn length.
8144 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
8147 * machmode.def: Define a complex mode for PARTIAL_INT.
8148 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
8150 * doc/rtl.texi: Document CSPImode.
8151 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
8153 (msp430_hard_regno_nregs_with_padding): Likewise.
8155 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
8158 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
8160 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
8162 PR rtl-optimization/84169
8163 * combine.c (try_combine): New variable split_i2i3. Set it to true if
8164 we generated a parallel as new i3 and we split that to new i2 and i3
8165 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
8166 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
8167 those to i2, not i1. Partially rewrite this scan code.
8169 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8172 * stor-layout.c (place_field): For variable length fields, adjust
8173 offset_align afterwards not just based on the field's alignment,
8174 but also on the size.
8177 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
8178 of exps and logs in the use_exp2 case.
8180 2018-02-13 Jeff Law <law@redhat.com>
8182 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
8185 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
8188 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
8192 * common.opt (gas-loc-support, gas-locview-support): New.
8193 (ginline-points, ginternal-reset-location-views): New.
8194 * doc/invoke.texi: Document them. Use @itemx where intended.
8195 (gvariable-location-views): Adjust.
8196 * target.def (reset_location_view): New.
8197 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
8198 (TARGET_RESET_LOCATION_VIEW): New.
8199 * doc/tm.texi: Rebuilt.
8200 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
8201 (dwarf2out_default_as_locview_support): New.
8202 (output_asm_line_debug_info): Use option variables.
8203 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
8204 (output_loc_list): Likewise.
8205 (add_high_low_attributes): Check option variables.
8206 Don't output entry view attribute in strict mode.
8207 (gen_inlined_subroutine_die): Check option variables.
8208 (dwarf2out_inline_entry): Likewise.
8209 (init_sections_and_labels): Likewise.
8210 (dwarf2out_early_finish): Likewise.
8211 (maybe_reset_location_view): New, from...
8212 (dwarf2out_var_location): ... here. Call it.
8213 * debug.h (dwarf2out_default_as_loc_support): Declare.
8214 (dwarf2out_default_as_locview_support): Declare.
8215 * hooks.c (hook_int_rtx_insn_0): New.
8216 * hooks.h (hook_int_rtx_insn_0): Declare.
8217 * toplev.c (process_options): Take -gas-loc-support and
8218 -gas-locview-support from dwarf2out. Enable
8219 -gvariable-location-views by default only with locview
8220 assembler support. Enable -ginternal-reset-location-views by
8221 default only if the target defines the corresponding hook.
8222 Enable -ginline-points by default if location views are
8223 enabled; force it disabled if statement frontiers are
8225 * tree-inline.c (expand_call_inline): Check option variables.
8226 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
8228 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
8230 PR tree-optimization/84321
8231 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
8232 handling. Also check whether the anti-range contains any values
8233 that satisfy the mask; switch to a VR_RANGE if not.
8235 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
8238 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
8240 2018-02-13 Martin Jambor <mjambor@suse.cz>
8243 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
8244 of call statements, also set location of a load to a temporary.
8246 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
8248 * config/rl78/rl78.c (add_vector_labels): New function.
8249 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
8250 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
8251 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
8252 which checks that no arguments are passed.
8253 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
8254 * doc/extend.texi: Documentation for the new attribute.
8256 2018-02-13 Andreas Schwab <schwab@suse.de>
8258 * config/riscv/linux.h (CPP_SPEC): Define.
8260 2018-02-13 Jakub Jelinek <jakub@redhat.com>
8263 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
8264 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
8265 OPTION_MASK_ISA_AES as first argument to def_builtin_const
8266 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
8267 instead of OPTION_MASK_ISA_PCLMUL as first argument to
8268 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
8269 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
8270 temporarily for AES and PCLMUL builtins.
8272 PR tree-optimization/84339
8273 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
8274 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
8278 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
8279 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
8280 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
8282 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
8284 * omp-simd-clone.h: New file.
8285 * omp-simd-clone.c: Include omp-simd-clone.h.
8286 (expand_simd_clones): No longer static.
8287 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
8288 cgraph.h and omp-simd-clone.h.
8289 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
8290 (vect_recog_widen_shift_pattern): Formatting fix.
8291 (vect_pattern_recog_1): Don't check optab for calls.
8294 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
8295 operands[2] into a REG before using gen_lowpart on it.
8297 2018-02-12 Jeff Law <law@redhat.com>
8300 * config/sh/sh.c (find_barrier): Consider a sibling call
8303 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
8304 successfully back substituting a reg.
8306 2018-02-12 Richard Biener <rguenther@suse.de>
8308 PR tree-optimization/84037
8309 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
8310 parameter, move visited init to caller.
8311 (vect_slp_analyze_operations): Separate cost from validity
8312 check, initialize visited once for all instances.
8313 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
8315 * tree-vect-stmts.c (vect_model_simple_cost): Make early
8317 (vect_model_promotion_demotion_cost): Likewise.
8318 (vectorizable_bswap): Guard cost modeling with !slp_node
8319 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
8321 (vectorizable_call): Likewise.
8322 (vectorizable_conversion): Likewise.
8323 (vectorizable_assignment): Likewise.
8324 (vectorizable_shift): Likewise.
8325 (vectorizable_operation): Likewise.
8326 (vectorizable_store): Likewise.
8327 (vectorizable_load): Likewise.
8328 (vectorizable_condition): Likewise.
8329 (vectorizable_comparison): Likewise.
8331 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
8334 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
8335 (ASAN_MARK): Fix fnspec to account for return value, change pointer
8336 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
8338 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8341 * params.def (inline-min-speedup): Increase from 8 to 15.
8342 (max-inline-insns-auto): Decrease from 40 to 30.
8343 * ipa-split.c (consider_split): Add some buffer for function to
8344 be considered inlining candidate.
8345 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
8348 2018-02-12 Richard Biener <rguenther@suse.de>
8350 PR tree-optimization/84037
8351 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
8352 matched stmts if we cannot swap the non-matched ones.
8354 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8356 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
8357 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
8358 _mm_maskz_scalef_round_ss): New intrinsics.
8359 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
8360 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
8361 __builtin_ia32_scalefss_round): Remove.
8362 (__builtin_ia32_scalefsd_mask_round,
8363 __builtin_ia32_scalefss_mask_round): New intrinsics.
8364 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
8365 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
8366 ((match_operand:VF_128 2 "<round_nimm_predicate>"
8367 "<round_constraint>")): Changed to ...
8368 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
8369 "<round_scalar_constraint>")): ... this.
8370 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
8371 %0, %1, %2<round_op3>}"): Changed to ...
8372 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
8373 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
8374 %2<round_scalar_mask_op3>}"): ... this.
8375 * config/i386/subst.md (round_scalar_nimm_predicate): New.
8377 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
8379 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
8380 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
8381 (_mm_maskz_sqrt_round_ss): New intrinsics.
8382 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
8383 (__builtin_ia32_sqrtsd_mask_round)
8384 (__builtin_ia32_sqrtss_mask_round): New builtins.
8385 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
8386 (__builtin_ia32_sqrtss_round): Remove.
8387 (__builtin_ia32_sqrtsd_mask_round)
8388 (__builtin_ia32_sqrtss_mask_round): New builtins.
8389 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
8390 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
8391 ((match_operand:VF_128 1 "vector_operand"
8392 "xBm,<round_constraint>")): Changed to ...
8393 ((match_operand:VF_128 1 "vector_operand"
8394 "xBm,<round_scalar_constraint>")): ... this.
8395 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
8396 %0, %2, %<iptr>1<round_op3>}): Changed to ...
8397 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
8398 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
8399 %<iptr>1<round_scalar_mask_op3>}): ... this.
8400 ((set_attr "prefix" "<round_prefix>")): Changed to ...
8401 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
8403 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
8406 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
8407 Cast vec_cmpeq result to correct type.
8408 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
8409 Cast vec_cmpgt result to correct type.
8411 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
8413 * final.c (final_scan_insn_1): Renamed from...
8414 (final_scan_insn): ... this. New wrapper, to recover
8415 seen from the outermost call in recursive ones.
8416 * config/sparc/sparc.c (output_return): Drop seen from call.
8417 (output_sibcall): Likewise.
8418 * config/visium/visium.c (output_branch): Likewise.
8420 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
8422 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
8425 2018-02-10 Alan Modra <amodra@gmail.com>
8428 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
8429 Specify LR as an input.
8431 2018-02-10 Jakub Jelinek <jakub@redhat.com>
8434 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
8435 remove_member_access_dummy_vars): New functions.
8436 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
8437 lower_omp_1, execute_lower_omp): Use them.
8439 PR rtl-optimization/84308
8440 * shrink-wrap.c (spread_components): Release todo vector.
8442 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
8444 PR rtl-optimization/57193
8445 * ira-color.c (struct allocno_color_data): Add member
8446 conflict_allocno_hard_prefs.
8447 (update_conflict_allocno_hard_prefs): New.
8448 (bucket_allocno_compare_func): Add a preference based on
8449 conflict_allocno_hard_prefs.
8450 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
8451 (color_allocnos): Remove a dead code. Initiate
8452 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
8454 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8457 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
8458 constraint from =wa to wa. Avoid a subreg on the output operand,
8459 instead use a pseudo and subreg it in a move.
8460 (p9_xxbrd_<mode>): Changed to ...
8461 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
8462 (p9_xxbrd_v2df): New expander.
8463 (p9_xxbrw_<mode>): Changed to ...
8464 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
8465 (p9_xxbrw_v4sf): New expander.
8467 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8469 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
8471 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
8474 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
8475 multiply in 32-bit mode.
8476 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
8477 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
8480 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
8482 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
8483 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
8484 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
8485 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
8487 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
8490 * dwarf2out.c (is_trivial_indirect_ref): New function.
8491 (dwarf2out_late_global_decl): Do not generate a location
8492 attribute for variables that have a non-trivial DECL_VALUE_EXPR
8493 and that are not defined in the current unit.
8495 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8497 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
8498 instead of a libcall for UNORDERED.
8500 2018-02-09 Tamar Christina <tamar.christina@arm.com>
8503 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
8504 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
8506 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8509 * config/s390/s390.c (s390_set_current_function): Invoke
8510 s390_indirect_branch_settings also if fndecl didn't change.
8512 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8514 * config/rs6000/rs6000.md (blockage): Set length to zero.
8516 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
8518 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
8520 2018-02-09 Jakub Jelinek <jakub@redhat.com>
8523 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
8524 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
8528 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
8529 PARALLEL incoming that failed vt_get_decl_and_offset check.
8532 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
8533 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
8535 (get_variable_section): For decls in named .bss* sections pass true as
8536 second argument to bss_initializer_p.
8538 2018-02-09 Marek Polacek <polacek@redhat.com>
8539 Jakub Jelinek <jakub@redhat.com>
8542 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
8543 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
8544 Sync some changes from cxx_fold_indirect_ref.
8546 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
8548 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
8550 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
8551 (BLOCK_INLINE_ENTRY_LABEL): New.
8552 (dwarf2out_var_location): Disregard inline entry markers.
8553 (inline_entry_data): New struct.
8554 (inline_entry_data_hasher): New hashtable type.
8555 (inline_entry_data_hasher::hash): New.
8556 (inline_entry_data_hasher::equal): New.
8557 (inline_entry_data_table): New variable.
8558 (add_high_low_attributes): Add DW_AT_entry_pc and
8559 DW_AT_GNU_entry_view attributes if a pending entry is found
8560 in inline_entry_data_table. Add old entry_pc attribute only
8561 if debug nonbinding markers are disabled.
8562 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
8563 markers are enabled.
8564 (block_within_block_p, dwarf2out_inline_entry): New.
8565 (dwarf2out_finish): Check that no entries remained in
8566 inline_entry_data_table.
8567 * final.c (reemit_insn_block_notes): Handle inline entry notes.
8568 (final_scan_insn, notice_source_line): Likewise.
8569 (rest_of_clean_state): Skip inline entry markers.
8570 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
8572 * gimple.c (gimple_build_debug_inline_entry): New.
8573 * gimple.h (enum gimple_debug_subcode): Add
8574 GIMPLE_DEBUG_INLINE_ENTRY.
8575 (gimple_build_debug_inline_entry): Declare.
8576 (gimple_debug_inline_entry_p): New.
8577 (gimple_debug_nonbind_marker_p): Adjust.
8578 * insn-notes.def (INLINE_ENTRY): New.
8579 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
8580 inline entry marker notes.
8581 (print_insn): Likewise.
8582 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
8583 (INSN_DEBUG_MARKER_KIND): Likewise.
8584 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
8585 * tree-inline.c (expand_call_inline): Build and insert
8586 debug_inline_entry stmt.
8587 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
8588 inline entry blocks early, if nonbind markers are enabled.
8589 (dump_scope_block): Dump fragment info.
8590 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
8591 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
8592 (gimple_build_debug_inline_entry): New.
8593 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
8594 Enable/disable inline entry points too.
8595 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
8596 (DEBUG_INSN): Describe inline entry markers.
8598 * common.opt (gvariable-location-views): New.
8599 (gvariable-location-views=incompat5): New.
8600 * config.in: Rebuilt.
8601 * configure: Rebuilt.
8602 * configure.ac: Test assembler for view support.
8603 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
8604 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
8605 * dwarf2out.c (var_loc_view): New typedef.
8606 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
8607 (dwarf2out_locviews_in_attribute): New.
8608 (dwarf2out_locviews_in_loclist): New.
8609 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
8610 (enum dw_line_info_opcode): Add LI_adv_address.
8611 (struct dw_line_info_table): Add view.
8612 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
8613 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
8614 (zero_view_p): New variable.
8615 (ZERO_VIEW_P): New macro.
8616 (output_asm_line_debug_info): New.
8617 (struct var_loc_node): Add view.
8618 (add_AT_view_list, AT_loc_list): New.
8619 (add_var_loc_to_decl): Add view param. Test it against last.
8620 (new_loc_list): Add view params. Record them.
8621 (AT_loc_list_ptr): Handle loc and view lists.
8622 (view_list_to_loc_list_val_node): New.
8623 (print_dw_val): Handle dw_val_class_view_list.
8624 (size_of_die): Likewise.
8625 (value_format): Likewise.
8626 (loc_list_has_views): New.
8627 (gen_llsym): Set vl_symbol too.
8628 (maybe_gen_llsym, skip_loc_list_entry): New.
8629 (dwarf2out_maybe_output_loclist_view_pair): New.
8630 (output_loc_list): Output view list or entries too.
8631 (output_view_list_offset): New.
8632 (output_die): Handle dw_val_class_view_list.
8633 (output_dwarf_version): New.
8634 (output_compilation_unit_header): Use it.
8635 (output_skeleton_debug_sections): Likewise.
8636 (output_rnglists, output_line_info): Likewise.
8637 (output_pubnames, output_aranges): Update version comments.
8638 (output_one_line_info_table): Output view numbers in asm comments.
8639 (dw_loc_list): Determine current endview, pass it to new_loc_list.
8640 Call maybe_gen_llsym.
8641 (loc_list_from_tree_1): Adjust.
8642 (add_AT_location_description): Create view list attribute if
8643 needed, check it's absent otherwise.
8644 (convert_cfa_to_fb_loc_list): Adjust.
8645 (maybe_emit_file): Call output_asm_line_debug_info for test.
8646 (dwarf2out_var_location): Reset views as needed. Precompute
8647 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
8648 attribute. Set view.
8649 (new_line_info_table): Reset next view.
8650 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
8651 (dwarf2out_source_line): Likewise. Output view resets and labels to
8652 the assembler, or select appropriate line info opcodes.
8653 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
8654 (optimize_string_length): Catch it. Adjust.
8655 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
8656 dw_val_class_view_list, and remove it if no longer needed.
8657 (hash_loc_list): Hash view numbers.
8658 (loc_list_hasher::equal): Compare them.
8659 (optimize_location_lists): Check whether a view list symbol is
8660 needed, and whether the locview attribute is present, and
8661 whether they match. Remove the locview attribute if no longer
8663 (index_location_lists): Call skip_loc_list_entry for test.
8664 (dwarf2out_finish): Call output_asm_line_debug_info for test.
8665 Use output_dwarf_version.
8666 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
8667 (struct dw_val_node): Add val_view_list.
8668 * final.c (SEEN_NEXT_VIEW): New.
8669 (set_next_view_needed): New.
8670 (clear_next_view_needed): New.
8671 (maybe_output_next_view): New.
8672 (final_start_function): Rename to...
8673 (final_start_function_1): ... this. Take pointer to FIRST,
8674 add SEEN parameter. Emit param bindings in the initial view.
8675 (final_start_function): Reintroduce SEEN-less interface.
8676 (final): Rename to...
8677 (final_1): ... this. Take SEEN parameter. Output final pending
8678 next view at the end.
8679 (final): Reintroduce seen-less interface.
8680 (final_scan_insn): Output pending next view before switching
8681 sections or ending a block. Mark the next view as needed when
8682 outputting variable locations. Notify debug backend of section
8683 changes, and of location view changes.
8684 (rest_of_handle_final): Adjust.
8685 * toplev.c (process_options): Autodetect value for debug variable
8686 location views option. Warn on incompat5 without -gdwarf-5.
8687 * doc/invoke.texi (gvariable-location-views): New.
8688 (gvariable-location-views=incompat5): New.
8689 (gno-variable-location-views): New.
8691 2018-02-08 David Malcolm <dmalcolm@redhat.com>
8693 PR tree-optimization/84136
8694 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
8695 that the result of find_edge is non-NULL.
8697 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8700 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
8701 storing integer register in SImode. Fix cost of 256 and 512
8702 byte aligned SSE register store.
8704 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
8706 * config/i386/i386.c (ix86_multiplication_cost): Fix
8707 multiplication cost for TARGET_AVX512DQ.
8709 2018-02-08 Marek Polacek <polacek@redhat.com>
8711 PR tree-optimization/84238
8712 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
8715 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8717 PR tree-optimization/84265
8718 * tree-vect-stmts.c (vectorizable_store): Don't treat
8719 VMAT_CONTIGUOUS accesses as grouped.
8720 (vectorizable_load): Likewise.
8722 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8724 PR tree-optimization/81635
8725 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
8726 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
8727 (test_round_for_mask): New functions.
8728 (wide_int_cc_tests): Call test_round_for_mask.
8729 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
8730 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
8731 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
8732 range returned by get_range_info.
8734 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
8737 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
8738 * symtab.c: Include builtins.h
8739 (symtab_node::output_to_lto_symbol_table_p): Move here
8740 from lto-streamer-out.c:output_symbol_p.
8741 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
8742 (output_symbol_p): Move all logic to symtab.c
8743 (produce_symtab): Update.
8745 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
8747 * config/s390/s390-opts.h (enum indirect_branch): Define.
8748 * config/s390/s390-protos.h (s390_return_addr_from_memory)
8749 (s390_indirect_branch_via_thunk)
8750 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
8751 (enum s390_indirect_branch_type): Define.
8752 * config/s390/s390.c (struct s390_frame_layout, struct
8753 machine_function): Remove.
8754 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
8755 (indirect_branch_table_label_no, indirect_branch_table_name):
8757 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
8758 (enum s390_indirect_branch_option): Define.
8759 (s390_return_addr_from_memory): New function.
8760 (s390_handle_string_attribute): New function.
8761 (s390_attribute_table): Add new attribute handler.
8762 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
8763 (s390_indirect_branch_via_thunk): New function.
8764 (s390_indirect_branch_via_inline_thunk): New function.
8765 (s390_function_ok_for_sibcall): When jumping via thunk disallow
8766 sibling call optimization for non z10 compiles.
8767 (s390_emit_call): Force indirect branch target to be a single
8768 register. Add r1 clobber for non-z10 compiles.
8769 (s390_emit_epilogue): Emit return jump via return_use expander.
8770 (s390_reorg): Handle JUMP_INSNs as execute targets.
8771 (s390_option_override_internal): Perform validity checks for the
8772 new command line options.
8773 (s390_indirect_branch_attrvalue): New function.
8774 (s390_indirect_branch_settings): New function.
8775 (s390_set_current_function): Invoke s390_indirect_branch_settings.
8776 (s390_output_indirect_thunk_function): New function.
8777 (s390_code_end): Implement target hook.
8778 (s390_case_values_threshold): Implement target hook.
8779 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
8781 * config/s390/s390.h (struct s390_frame_layout)
8782 (struct machine_function): Move here from s390.c.
8783 (TARGET_INDIRECT_BRANCH_NOBP_RET)
8784 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
8785 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
8786 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
8787 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
8788 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
8789 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
8790 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
8791 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
8792 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
8793 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
8794 (mnemonic attribute): Add values which aren't recognized
8796 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
8797 pattern for branch conversion. Fix mnemonic attribute.
8798 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
8799 indirect branch via thunk if requested.
8800 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
8801 ("*indirect_jump"): Disable for branch conversion using out of
8803 ("indirect_jump_via_thunk<mode>_z10")
8804 ("indirect_jump_via_thunk<mode>")
8805 ("indirect_jump_via_inlinethunk<mode>_z10")
8806 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
8807 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
8808 ("casesi_jump_via_inlinethunk<mode>_z10")
8809 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
8810 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
8811 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
8812 ("*indirect2_jump"): Disable for branch conversion.
8813 ("casesi_jump"): Turn into expander and expand patterns for branch
8815 ("return_use"): New expander.
8816 ("*return"): Emit return via thunk and rename it to ...
8817 ("*return<mode>"): ... this one.
8818 * config/s390/s390.opt: Add new options and and enum for the
8821 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
8823 * lra-constraints.c (match_reload): Unconditionally use
8824 gen_lowpart_SUBREG, rather than selecting between that
8825 and equivalent gen_rtx_SUBREG code.
8827 2018-02-08 Richard Biener <rguenther@suse.de>
8829 PR tree-optimization/84233
8830 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
8831 changed flag instead of boguously re-using phi_inserted.
8833 2018-02-08 Martin Jambor <mjambor@suse.cz>
8835 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
8836 static local variables.
8838 2018-02-08 Richard Biener <rguenther@suse.de>
8840 PR tree-optimization/84278
8841 * tree-vect-stmts.c (vectorizable_store): When looking for
8842 smaller vector types to perform grouped strided loads/stores
8843 make sure the mode is supported by the target.
8844 (vectorizable_load): Likewise.
8846 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8848 * config/aarch64/aarch64.c (aarch64_components_for_bb):
8849 Increase LDP/STP opportunities by adding adjacent callee-saves.
8851 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
8853 PR rtl-optimization/84068
8854 PR rtl-optimization/83459
8855 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
8857 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
8859 PR tree-optimization/84224
8860 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
8861 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
8864 2018-02-07 Iain Sandoe <iain@codesourcery.com>
8867 * config/rs6000/altivec.md (*restore_world): Remove LR use.
8868 * config/rs6000/predicates.md (restore_world_operation): Adjust op
8869 count, remove one USE.
8871 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
8873 * doc/install.texi (Configuration): Document the
8874 --with-long-double-format={ibm,ieee} PowerPC configuration
8878 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
8879 Convert from define_expand to be define_insn_and_split. Rework
8880 float/double/_Float128 conversions to QI/HI/SImode to work with
8881 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
8882 conversions to QI/HImode types did a store and then a load to
8883 truncate the value. For conversions to VSX registers, don't split
8884 the insn, instead emit the code directly. Use the code iterator
8885 any_fix to combine signed and unsigned conversions.
8886 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
8887 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
8888 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8889 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
8890 (fix_<mode>di2_hw): Likewise.
8891 (fixuns_<mode>di2_hw): Likewise.
8892 (fix_<mode>si2_hw): Likewise.
8893 (fixuns_<mode>si2_hw): Likewise.
8894 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
8895 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
8896 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
8897 fix<uns>_trunc<SFDF:mode>si2_p8.
8898 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
8900 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
8901 (fix<uns>_<mode>_mem): Likewise.
8902 (fctiw<u>z_<mode>_mem): Likewise.
8903 (fix<uns>_<mode>_mem): Likewise.
8904 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
8905 the register allocator from doing a direct move to the GPRs to do
8906 a store, and instead use the ISA 3.0 store byte/half-word from
8907 vector register instruction. For IEEE 128-bit floating point,
8908 also optimize stores of 32-bit ints.
8909 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
8911 2018-02-07 Alan Hayward <alan.hayward@arm.com>
8913 * genextract.c (push_pathstr_operand): New function to support
8915 (walk_rtx): Call push_pathstr_operand.
8916 (print_path): Support [a-zA-Z].
8918 2018-02-07 Richard Biener <rguenther@suse.de>
8920 PR tree-optimization/84037
8921 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
8922 (cse_and_gimplify_to_preheader): Declare.
8923 (vect_get_place_in_interleaving_chain): Likewise.
8924 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8926 (_loop_vec_info::~_loop_vec_info): Delete it.
8927 (cse_and_gimplify_to_preheader): New function.
8928 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
8929 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
8930 (vectorizable_load): Likewise. For grouped stores always base
8931 the IV on the first element.
8932 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
8933 condition before gimplifying.
8935 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8937 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
8938 *DIV_EXPR and *MOD_EXPR.
8940 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
8943 * config/i386/i386.c (ix86_option_override_internal): Mask out
8944 the CF_SET bit when checking -fcf-protection.
8946 2018-02-07 Tom de Vries <tom@codesourcery.com>
8949 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
8952 2018-02-07 Richard Biener <rguenther@suse.de>
8954 PR tree-optimization/84204
8955 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
8958 PR tree-optimization/84205
8959 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
8960 special-case isl_ast_op_zdiv_r.
8962 PR tree-optimization/84223
8963 * graphite-scop-detection.c (gather_bbs::before_dom_children):
8964 Only add conditions from within the region.
8965 (gather_bbs::after_dom_children): Adjust.
8967 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
8970 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
8971 * config/avr/avr.md: Only post-reload split REG-REG moves if
8972 either register is GENERAL_REG_P.
8974 2018-02-07 Jakub Jelinek <jakub@redhat.com>
8976 PR tree-optimization/84235
8977 * tree-ssa-scopedtables.c
8978 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
8979 if the subtraction is performed in floating point type where NaNs are
8980 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
8981 build 1. Formatting fix.
8983 2018-02-06 Jakub Jelinek <jakub@redhat.com>
8986 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
8987 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
8988 and skip it regardless of bb boundaries. Use CALL_P macro,
8989 don't test INSN_P (insn) together with CALL_P or JUMP_P check
8990 unnecessarily, formatting fix.
8992 2018-02-06 Michael Collison <michael.collison@arm.com>
8994 * config/arm/thumb2.md:
8995 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
8996 (*thumb_mov_notscc): Ditto.
8998 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
9001 * config/rs6000/rs6000.md (su code attribute): Use "u" for
9002 unsigned_fix, not "s".
9004 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9006 * configure.ac (gcc_fn_eh_frame_ro): New function.
9007 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
9008 correct .eh_frame permissions.
9009 * configure: Regenerate.
9011 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
9013 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
9016 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9018 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9019 Display warning message for -mno-speculate-indirect-jumps.
9021 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
9023 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
9025 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
9027 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
9029 PR tree-optimization/84225
9030 * tree-eh.c (find_trapping_overflow): Only call
9031 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
9033 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9036 * config/i386/i386.c: Reimplement the check of possible options
9037 -mibt/-mshstk conbination. Change error messages.
9038 * doc/invoke.texi: Fix a typo: remove extra '='.
9040 2018-02-06 Marek Polacek <polacek@redhat.com>
9042 PR tree-optimization/84228
9043 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
9045 2018-02-06 Tamar Christina <tamar.christina@arm.com>
9048 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
9049 emitted arch directives.
9050 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
9051 __ARM_FEATURE_COPROC before changing architectures.
9053 2018-02-06 Richard Biener <rguenther@suse.de>
9055 * config/i386/i386.c (print_reg): Fix typo.
9056 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
9058 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
9060 * configure: Regenerate.
9062 2018-02-05 Martin Sebor <msebor@redhat.com>
9064 PR tree-optimization/83369
9065 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
9068 2018-02-05 Martin Liska <mliska@suse.cz>
9070 * doc/invoke.texi: Cherry-pick upstream r323995.
9072 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
9074 * ira.c (ira_init_register_move_cost): Adjust comment.
9076 2018-02-05 Martin Liska <mliska@suse.cz>
9078 PR gcov-profile/84137
9079 * doc/gcov.texi: Fix typo in documentation.
9081 2018-02-05 Martin Liska <mliska@suse.cz>
9083 PR gcov-profile/83879
9084 * doc/gcov.texi: Document necessity of --dynamic-list-data when
9085 using dlopen functionality.
9087 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
9089 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
9090 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
9091 _mm_maskz_range_ss, _mm_mask_range_round_ss,
9092 _mm_maskz_range_round_ss): New intrinsics.
9093 (__builtin_ia32_rangesd128_round)
9094 (__builtin_ia32_rangess128_round): Remove.
9095 (__builtin_ia32_rangesd128_mask_round,
9096 __builtin_ia32_rangess128_mask_round): New builtins.
9097 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
9098 __builtin_ia32_rangess128_round): Remove.
9099 (__builtin_ia32_rangesd128_mask_round,
9100 __builtin_ia32_rangess128_mask_round): New builtins.
9101 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
9102 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
9103 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
9104 "<round_saeonly_constraint>")): Changed to ...
9105 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
9106 "<round_saeonly_scalar_constraint>")): ... this.
9107 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
9108 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
9109 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
9110 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
9111 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
9113 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
9115 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
9117 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
9118 Remove all values except native, 8540 and 8548.
9120 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
9122 * config/i386/i386.c (ix86_output_function_return): Pass
9123 INVALID_REGNUM, instead of -1, as invalid register number to
9124 indirect_thunk_name and output_indirect_thunk.
9126 2018-02-02 Julia Koval <julia.koval@intel.com>
9128 * config.gcc: Add -march=icelake.
9129 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
9130 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
9131 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
9132 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
9133 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
9134 (processor_target_table): Add icelake.
9135 (ix86_option_override_internal): Handle new PTAs.
9136 (get_builtin_code_for_version): Handle icelake.
9137 (M_INTEL_COREI7_ICELAKE): New.
9138 (fold_builtin_cpu): Handle icelake.
9139 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
9140 * doc/invoke.texi: Add -march=icelake.
9142 2018-02-02 Julia Koval <julia.koval@intel.com>
9144 * config/i386/i386.c (ix86_option_override_internal): Change flags type
9145 to wide_int_bitmask.
9146 * wide-int-bitmask.h: New.
9148 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9151 * config/i386/i386.md: Replace Pmode with word_mode in
9152 builtin_setjmp_setup and builtin_longjmp to support x32.
9154 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
9158 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
9160 (rs6000_supported_cpu_names): New static variable.
9161 (linux_cpu_translation_table): Likewise.
9162 (elf_platform) <cpu>: Define new static variable and use it.
9163 Translate kernel AT_PLATFORM name to canonical name if needed.
9164 Error if platform name is unknown.
9166 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
9169 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
9171 2018-02-01 Jeff Law <law@redhat.com>
9174 * config/i386/i386.c (release_scratch_register_on_entry): Add new
9175 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
9176 the scratch if RELEASE_VIA_POP is false.
9177 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
9178 If we have to save a temporary register, decrement SIZE appropriately.
9179 Pass new arguments to release_scratch_register_on_entry.
9180 (ix86_adjust_stack_and_probe): Likewise.
9181 (ix86_emit_probe_stack_range): Pass new arguments to
9182 release_scratch_register_on_entry.
9184 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
9186 PR rtl-optimization/84157
9187 * combine.c (change_zero_ext): Use REG_P predicate in
9188 front of HARD_REGISTER_P predicate.
9190 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
9192 * config/avr/avr.c (avr_option_override): Move disabling of
9193 -fdelete-null-pointer-checks to...
9194 * common/config/avr/avr-common.c (avr_option_optimization_table):
9197 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9199 PR tree-optimization/81635
9200 * tree-data-ref.c (split_constant_offset_1): For types that
9201 wrap on overflow, try to use range info to prove that wrapping
9204 2018-02-01 Renlin Li <renlin.li@arm.com>
9207 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
9209 (aarch64_register_move_cost): Likewise.
9210 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
9212 (REG_CLASS_NAMES): Likewise.
9213 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
9214 TAILCALL_ADDR_REGS. Remove IP registers.
9215 * config/aarch64/aarch64.md (Ucs): Update register constraint.
9217 2018-02-01 Richard Biener <rguenther@suse.de>
9219 * domwalk.h (dom_walker::dom_walker): Add additional constructor
9220 for specifying RPO order and allow NULL for that.
9221 * domwalk.c (dom_walker::dom_walker): Likewise.
9222 (dom_walker::walk): Handle NULL RPO order.
9223 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
9225 (rewrite_update_dom_walker): Likewise.
9226 (mark_def_dom_walker): Likewise.
9228 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9230 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
9231 (aarch64_maybe_expand_sve_subreg_move): Declare.
9232 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
9233 * config/aarch64/predicates.md (aarch64_any_register_operand): New
9235 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
9236 that are semantically a reverse operation.
9237 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
9238 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
9239 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
9241 (aarch64_can_change_mode_class): For big-endian, forbid changes
9242 between two SVE modes if they have different element sizes.
9244 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9246 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
9247 the TImode handling for big-endian targets.
9249 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9251 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
9252 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
9254 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
9255 Remove BSWAP handing for big-endian targets and use the form of
9256 LD1RQ appropariate for the mode.
9258 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9260 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
9261 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
9264 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
9267 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
9268 check for operands that need to go through aarch64_sve_reload_be.
9270 2018-02-01 Jakub Jelinek <jakub@redhat.com>
9272 PR tree-optimization/81661
9273 PR tree-optimization/84117
9274 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
9275 * tree-eh.c: Include gimplify.h.
9276 (find_trapping_overflow, replace_trapping_overflow,
9277 rewrite_to_non_trapping_overflow): New functions.
9278 * tree-vect-loop.c: Include tree-eh.h.
9279 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
9280 * tree-data-ref.c: Include tree-eh.h.
9281 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
9283 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
9285 PR rtl-optimization/84123
9286 * combine.c (change_zero_ext): Check if hard register satisfies
9287 can_change_dest_mode before calling gen_lowpart_SUBREG.
9289 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
9292 * ira.c (ira_init_register_move_cost): Remove assert.
9294 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9296 PR rtl-optimization/84071
9297 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
9298 * doc/tm.texi: Regenerate.
9300 2018-01-31 Richard Biener <rguenther@suse.de>
9302 PR tree-optimization/84132
9303 * tree-data-ref.c (analyze_miv_subscript): Properly
9304 check whether evolution_function_is_affine_multivariate_p
9305 before calling gcd_of_steps_may_divide_p.
9307 2018-01-31 Julia Koval <julia.koval@intel.com>
9310 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
9311 * config/i386/i386.md (rdpid_rex64) New.
9312 (rdpid): Make 32bit only.
9314 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
9317 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
9318 an IDENTIFIER_NODE for FUNCTION_TYPE's.
9320 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9323 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9325 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9327 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
9329 PR rtl-optimization/84071
9330 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
9331 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
9333 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9335 * config/arc/arc.c (arc_handle_aux_attribute): New function.
9336 (arc_attribute_table): Add 'aux' attribute.
9337 (arc_in_small_data_p): Consider aux like variables.
9338 (arc_is_aux_reg_p): New function.
9339 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
9340 (arc_get_aux_arg): New function.
9341 (prepare_move_operands): Handle aux-register access.
9342 (arc_handle_aux_attribute): New function.
9343 * doc/extend.texi (ARC Variable attributes): Add subsection.
9345 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
9347 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
9348 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
9349 (arc_attribute_table): Add 'uncached' attribute.
9350 (arc_print_operand): Print '.di' flag for uncached memory
9352 (arc_in_small_data_p): Do not consider for small data the uncached
9354 (arc_is_uncached_mem_p): New function.
9355 * config/arc/predicates.md (compact_store_memory_operand): Check
9356 for uncached memory accesses.
9357 (nonvol_nonimm_operand): Likewise.
9358 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
9360 2018-01-31 Jakub Jelinek <jakub@redhat.com>
9363 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
9364 falign-loops=): Add Optimization flag.
9366 2018-01-30 Jeff Law <law@redhat.com>
9369 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
9370 INT_REGISTERS_SAVED. Check it prior to calling
9371 get_scratch_register_on_entry.
9372 (ix86_adjust_stack_and_probe): Similarly.
9373 (ix86_emit_probe_stack_range): Similarly.
9374 (ix86_expand_prologue): Corresponding changes.
9376 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9379 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
9380 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
9382 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
9385 * lra-constraints.c (curr_insn_transform): Process AND in the
9388 2018-01-30 Jakub Jelinek <jakub@redhat.com>
9390 PR rtl-optimization/83986
9391 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
9392 dependence against last_pending_memory_flush in addition to
9395 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
9397 PR tree-optimization/81611
9398 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
9401 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9404 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
9407 2018-01-30 Richard Biener <rguenther@suse.de>
9408 Jakub Jelinek <jakub@redhat.com>
9410 PR tree-optimization/84111
9411 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
9412 inner loops added during recursion, as they don't have up-to-date
9415 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9418 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
9419 (can_inline_edge_by_limits_p): ... here.
9420 (can_early_inline_edge_p, check_callers,
9421 update_caller_keys, update_callee_keys, recursive_inlining,
9422 add_new_edges_to_heap, speculation_useful_p,
9423 inline_small_functions,
9424 inline_small_functions, flatten_function,
9425 inline_to_all_callers_1): Update.
9427 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
9429 * profile-count.c (profile_count::combine_with_ipa_count): Handle
9432 2018-01-30 Richard Biener <rguenther@suse.de>
9434 PR tree-optimization/83008
9435 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
9436 invariant and constant vector uses in stmts when they need
9439 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9442 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
9443 * configure: Regenerate.
9445 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9447 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
9449 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
9450 Use gen_rtx_REG rather than gen_lowpart.
9452 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9454 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
9455 rather than 0 when creating partial subregs.
9457 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
9459 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
9462 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
9465 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
9466 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
9467 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
9468 flags. This restores the settings used before the 2017-07-24.
9469 Turning off pre increment/decrement/modify allows IVOPTS to
9470 optimize DF/SF loops where the index is an int.
9472 2018-01-29 Richard Biener <rguenther@suse.de>
9473 Kelvin Nilsen <kelvin@gcc.gnu.org>
9476 * tree-vect-stmts.c (vectorizable_call): Don't call
9477 targetm.vectorize_builtin_md_vectorized_function if callee is
9480 2018-01-22 Carl Love <cel@us.ibm.com>
9482 * doc/extend.tex: Fix typo in second arg in
9483 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
9485 2018-01-29 Richard Biener <rguenther@suse.de>
9487 PR tree-optimization/84086
9488 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
9489 (flush_ssaname_freelist): When SSA names were released reset
9490 the SCEV hash table.
9492 2018-01-29 Richard Biener <rguenther@suse.de>
9494 PR tree-optimization/84057
9495 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
9496 removed paths when removing edges.
9498 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
9500 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
9501 -mfunction-return=@var{choice}.
9503 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
9506 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
9508 (layout::print_source_line): Likewise.
9509 (test_get_line_width_without_trailing_whitespace): Add test cases.
9511 2018-01-27 Jakub Jelinek <jakub@redhat.com>
9514 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
9517 2018-01-26 Jim Wilson <jimw@sifive.com>
9519 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
9521 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
9524 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9526 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
9527 and CMP + SUB-immediate -> SUBS.
9529 2018-01-26 Martin Sebor <msebor@redhat.com>
9531 PR tree-optimization/83896
9532 * tree-ssa-strlen.c (get_string_len): Rename...
9533 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
9534 Avoid assuming length is constant.
9535 (handle_char_store): Use HOST_WIDE_INT for string length.
9537 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
9540 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
9541 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
9543 2018-01-26 Richard Biener <rguenther@suse.de>
9545 PR rtl-optimization/84003
9546 * dse.c (record_store): Only record redundant stores when
9547 the earlier store aliases at least all accesses the later one does.
9549 2018-01-26 Jakub Jelinek <jakub@redhat.com>
9551 PR rtl-optimization/83985
9552 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
9553 REG_CFA_RESTORE insns.
9554 (delete_unmarked_insns): Don't ignore separate shrink wrapping
9555 REG_CFA_RESTORE insns here.
9558 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
9559 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
9561 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9563 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
9564 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
9565 (arc_init): Likewise.
9566 (arc_override_options): Likewise.
9567 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
9569 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
9571 * config/arc/arc.h (TARGET_DBNZ): Define.
9572 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
9573 properly set the tune attribute.
9574 (dbnz): Use TARGET_DBNZ guard.
9575 * config/arc/arc.opt (mtune): Add core3 option.
9577 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9579 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
9580 recognize new pic like addresses.
9581 (arc_delegitimize_address): Clean up.
9583 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9585 * config/arc/arc-arches.def: Option mrf16 valid for all
9587 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
9588 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
9589 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
9590 * config/arc/arc-tables.opt: Regenerate.
9591 * config/arc/arc.c (arc_conditional_register_usage): Handle
9592 reduced register file case.
9593 (arc_file_start): Set must have build attributes.
9594 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
9596 * config/arc/arc.opt (mrf16): Add new option.
9597 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
9598 * config/arc/genmultilib.awk: Handle new mrf16 option.
9599 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
9600 * config/arc/t-multilib: Regenerate.
9601 * doc/invoke.texi (ARC Options): Document mrf16 option.
9603 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9605 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
9606 * config/arc/arc.c (arc_handle_secure_attribute): New function.
9607 (arc_attribute_table): Add 'secure_call' attribute.
9608 (arc_print_operand): Print secure call operand.
9609 (arc_function_ok_for_sibcall): Don't optimize tail calls when
9611 (arc_is_secure_call_p): New function. * config/arc/arc.md
9612 (call_i): Add support for sjli instruction.
9613 (call_value_i): Likewise.
9614 * config/arc/constraints.md (Csc): New constraint.
9616 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
9617 John Eric Martin <John.Martin@emmicro-us.com>
9619 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
9620 * config/arc/arc.c (_arc_jli_section): New struct.
9621 (arc_jli_section): New type.
9622 (rc_jli_sections): New static variable.
9623 (arc_handle_jli_attribute): New function.
9624 (arc_attribute_table): Add jli_always and jli_fixed attribute.
9625 (arc_file_end): New function.
9626 (TARGET_ASM_FILE_END): Define.
9627 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
9628 (arc_add_jli_section): New function.
9629 (jli_call_scan): Likewise.
9630 (arc_reorg): Call jli_call_scan.
9631 (arc_output_addsi): Remove 'S' from printing asm operand.
9632 (arc_is_jli_call_p): New function.
9633 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
9635 (movhi_insn): Likewise.
9636 (movsi_insn): Likewise.
9637 (movsi_set_cc_insn): Likewise.
9638 (loadqi_update): Likewise.
9639 (load_zeroextendqisi_update): Likewise.
9640 (load_signextendqisi_update): Likewise.
9641 (loadhi_update): Likewise.
9642 (load_zeroextendhisi_update): Likewise.
9643 (load_signextendhisi_update): Likewise.
9644 (loadsi_update): Likewise.
9645 (loadsf_update): Likewise.
9646 (movsicc_insn): Likewise.
9647 (bset_insn): Likewise.
9648 (bxor_insn): Likewise.
9649 (bclr_insn): Likewise.
9650 (bmsk_insn): Likewise.
9651 (bicsi3_insn): Likewise.
9652 (cmpsi_cc_c_insn): Likewise.
9653 (movsi_ne): Likewise.
9654 (movsi_cond_exec): Likewise.
9655 (clrsbsi2): Likewise.
9666 (call_i): Remove 'S' asm letter, add jli instruction.
9667 (call_value_i): Likewise.
9668 * config/arc/arc.op (mjli-always): New option.
9669 * config/arc/constraints.md (Cji): New constraint.
9670 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
9672 (subsf3_fpx): Likewise.
9673 (mulsf3_fpx): Likewise.
9674 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
9676 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
9678 * doc/invoke.texi (ARC): Document mjli-always option.
9680 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
9682 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
9683 avoid addition with 0 and use incw and decw where possible.
9685 2018-01-26 Richard Biener <rguenther@suse.de>
9687 PR tree-optimization/81082
9688 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
9689 association if it requires casting to unsigned.
9690 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
9691 from fold_plusminus_mult_expr to catch important cases late when
9692 range info is available.
9694 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9696 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
9697 * configure.ac (hidden_linkonce): New test.
9698 * configure: Regenerate.
9699 * config.in: Regenerate.
9701 2018-01-26 Julia Koval <julia.koval@intel.com>
9703 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
9704 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
9705 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
9706 _mm_mask_bitshuffle_epi64_mask): Fix type.
9707 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
9708 USI_FTYPE_V4DI_V4DI_USI): Remove.
9709 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
9710 __builtin_ia32_vpshufbitqmb256_mask,
9711 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
9712 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
9713 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
9715 2018-01-26 Alan Modra <amodra@gmail.com>
9718 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
9719 UNSPEC_VBPERMQ. Sort other unspecs.
9721 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
9723 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
9725 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
9728 * predict.c (drop_profile): Do not push/pop cfun; update also
9730 (handle_missing_profiles): Fix logic looking for zero profiles.
9732 2018-01-25 Jakub Jelinek <jakub@redhat.com>
9735 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
9736 on functions with #pragma omp declare simd or functions with simd
9738 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
9739 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
9740 Remove trailing \n from warning_at calls.
9742 2018-01-25 Tom de Vries <tom@codesourcery.com>
9745 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9746 for neutered workers.
9748 2018-01-24 Joseph Myers <joseph@codesourcery.com>
9751 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
9752 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
9754 2018-01-24 Jeff Law <law@redhat.com>
9757 * i386.c (get_probe_interval): Move to earlier point.
9758 (ix86_compute_frame_layout): If -fstack-clash-protection and
9759 the frame is larger than the probe interval, then use pushes
9760 to save registers rather than reg->mem moves.
9761 (ix86_expand_prologue): Remove conditional for int_registers_saved
9764 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
9767 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
9768 min/max for never referenced object.
9770 2018-01-24 Jakub Jelinek <jakub@redhat.com>
9773 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
9775 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
9776 attributes from DECL_ATTRIBUTES (decl) without affecting
9777 DECL_ATTRIBUTES (current_function_decl).
9778 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
9779 functions with non-NULL DECL_ABSTRACT_ORIGIN.
9781 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
9783 PR tree-optimization/83979
9784 * fold-const.c (fold_comparison): Use constant_boolean_node
9785 instead of boolean_{true,false}_node.
9787 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
9789 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
9792 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9794 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
9795 Simplify the clause that sets the length attribute.
9796 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
9797 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
9798 clause that sets the length attribute.
9799 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
9801 2018-01-24 Tom de Vries <tom@codesourcery.com>
9804 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
9805 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
9806 Add strict parameter.
9807 (prevent_branch_around_nothing): Insert dummy insn between branch to
9808 label and label with no ptx insn inbetween.
9809 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
9811 2018-01-24 Tom de Vries <tom@codesourcery.com>
9814 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
9815 for neutered threads in warp.
9816 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
9818 2018-01-24 Richard Biener <rguenther@suse.de>
9820 PR tree-optimization/83176
9821 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
9824 2018-01-24 Richard Biener <rguenther@suse.de>
9826 PR tree-optimization/82819
9827 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
9828 code generating pluses that are no-ops in the target precision.
9830 2018-01-24 Richard Biener <rguenther@suse.de>
9833 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
9835 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9837 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
9838 to merge probabilities.
9839 * predict.c (probably_never_executed): Also mark as cold functions
9840 with global 0 profile and guessed local profile.
9841 * profile-count.c (profile_probability::combine_with_count): New
9843 * profile-count.h (profile_probability::operator*,
9844 profile_probability::operator*=, profile_probability::operator/,
9845 profile_probability::operator/=): Reduce precision to adjusted
9846 and set value to guessed on contradictory divisions.
9847 (profile_probability::combine_with_freq): Remove.
9848 (profile_probability::combine_wiht_count): Declare.
9849 (profile_count::force_nonzero):: Set to adjusted.
9850 (profile_count::probability_in):: Set quality to adjusted.
9851 * tree-ssa-tail-merge.c (replace_block_by): Use
9854 2018-01-23 Andrew Waterman <andrew@sifive.com>
9855 Jim Wilson <jimw@sifive.com>
9857 * config/riscv/riscv.c (riscv_stack_boundary): New.
9858 (riscv_option_override): Set riscv_stack_boundary. Handle
9859 riscv_preferred_stack_boundary_arg.
9860 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
9861 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
9862 (STACK_BOUNDARY): Set to riscv_stack_boundary.
9863 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
9864 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
9865 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
9867 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
9870 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
9871 of struct ix86_frame.
9872 (ix86_expand_epilogue): Likewise. Add a local variable for
9873 the reg_save_offset field in struct ix86_frame.
9875 2018-01-23 Bin Cheng <bin.cheng@arm.com>
9877 PR tree-optimization/82604
9878 * tree-loop-distribution.c (enum partition_kind): New enum item
9879 PKIND_PARTIAL_MEMSET.
9880 (partition_builtin_p): Support above new enum item.
9881 (generate_code_for_partition): Ditto.
9882 (compute_access_range): Differentiate cases that equality can be
9883 proven at all loops, the innermost loops or no loops.
9884 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
9885 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
9886 (finalize_partitions, distribute_loop): Don't fuse partition of
9887 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
9888 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
9891 2018-01-23 Martin Liska <mliska@suse.cz>
9893 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
9894 order to ignore the predictor.
9895 (PRED_POLYMORPHIC_CALL): Likewise.
9896 (PRED_RECURSIVE_CALL): Likewise.
9898 2018-01-23 Martin Liska <mliska@suse.cz>
9900 * tree-profile.c (tree_profiling): Print function header to
9901 aware reader which function we are working on.
9902 * value-prof.c (gimple_find_values_to_profile): Do not print
9903 not interesting value histograms.
9905 2018-01-23 Martin Liska <mliska@suse.cz>
9907 * profile-count.h (enum profile_quality): Add
9908 profile_uninitialized as the first value. Do not number values
9909 as they are zero based.
9910 (profile_count::verify): Update sanity check.
9911 (profile_probability::verify): Likewise.
9913 2018-01-23 Nathan Sidwell <nathan@acm.org>
9915 * doc/invoke.texi (ffor-scope): Deprecate.
9917 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9919 PR tree-optimization/83510
9920 * domwalk.c (set_all_edges_as_executable): New function.
9921 (dom_walker::dom_walker): Convert bool param
9922 "skip_unreachable_blocks" to enum reachability. Move setup of
9923 edge flags to set_all_edges_as_executable and only do it when
9924 reachability is REACHABLE_BLOCKS.
9925 * domwalk.h (enum dom_walker::reachability): New enum.
9926 (dom_walker::dom_walker): Convert bool param
9927 "skip_unreachable_blocks" to enum reachability.
9928 (set_all_edges_as_executable): New decl.
9929 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
9930 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
9932 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
9933 but converting true to REACHABLE_BLOCKS.
9934 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
9936 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
9937 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
9938 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
9940 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
9941 if check_all_array_refs will be called.
9943 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9945 * tree.c (selftest::test_location_wrappers): Add more test
9948 2018-01-23 David Malcolm <dmalcolm@redhat.com>
9950 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
9951 (selftest::test_bit_in_range): Likewise.
9953 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9956 * doc/sourcebuild.texi (vect_float): Say that the selector
9957 only describes the situation when -funsafe-math-optimizations is on.
9958 (vect_float_strict): Document.
9960 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
9962 PR tree-optimization/83965
9963 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9964 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
9965 instead of checking only for a reduction.
9966 (vect_recog_widen_sum_pattern): Likewise.
9968 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9970 * predict.c (probably_never_executed): Only use precise profile info.
9971 (compute_function_frequency): Skip after inlining hack since we now
9972 have quality checking.
9974 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
9976 * profile-count.h (profile_probability::very_unlikely,
9977 profile_probability::unlikely, profile_probability::even): Set
9978 precision to guessed.
9980 2018-01-23 Richard Biener <rguenther@suse.de>
9982 PR tree-optimization/83963
9983 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
9984 Properly terminate dominator walk when crossing the exit edge not
9985 when visiting its source block.
9987 2018-01-23 Jakub Jelinek <jakub@redhat.com>
9990 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
9991 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
9993 2018-01-22 Jakub Jelinek <jakub@redhat.com>
9995 PR tree-optimization/83957
9996 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
9997 semicolon after for body surrounded by braces.
9999 PR tree-optimization/83081
10000 * profile-count.h (profile_probability::split): New method.
10001 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
10002 Use profile_probability::split.
10003 (do_compare_rtx_and_jump): Fix adjustment of probabilities
10004 when splitting a single conditional jump into 2.
10006 2018-01-22 David Malcolm <dmalcolm@redhat.com>
10008 PR tree-optimization/69452
10009 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
10012 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10014 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
10015 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
10016 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
10018 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10020 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
10021 * config/rl78/rl78.md (movdi): New define_expand.
10022 * config/rl78/rl78.c (rl78_split_movdi): New function.
10024 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
10027 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
10029 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
10030 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
10031 128-bit to produce an UNSPEC move to get the double word with the
10032 signbit and then a shift directly to do signbit.
10033 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
10034 implementation with a new version that just does either a direct
10035 move or a regular move. Move memory interface to separate insns.
10036 Move insns so they are next to the expander.
10037 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
10038 with signbit move. Split big and little endian case.
10039 (signbit<mode>2_dm_mem_le): Likewise.
10040 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
10041 (signbit<mode>2_dm2): Likewise.
10043 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10045 * config/rl78/rl78.md (anddi3): New define_expand.
10047 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10049 * config/rl78/rl78.md (umindi3): New define_expand.
10051 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10053 * config/rl78/rl78.md (smindi3): New define_expand.
10055 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10057 * config/rl78/rl78.md (smaxdi3): New define_expand.
10059 2018-01-22 Carl Love <cel@us.ibm.com>
10061 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
10062 LVX_V1TI): Add macro expansion.
10063 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
10064 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
10065 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
10066 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
10067 Change check to determine if the instruction is a byte reversing
10068 entry. Fix typo in comment.
10069 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
10070 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
10071 Add def_builtin calls for new builtins.
10072 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
10073 Add define_insn expansion.
10075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10077 * config/rl78/rl78.md (umaxdi3): New define_expand.
10079 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
10081 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
10082 for non-QImode registers.
10084 2018-01-22 Richard Biener <rguenther@suse.de>
10086 PR tree-optimization/83963
10087 * graphite-scop-detection.c (scop_detection::get_sese): Delay
10088 including the loop exit block.
10089 (scop_detection::merge_sese): Likewise.
10090 (scop_detection::add_scop): Do it here instead.
10092 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10094 * doc/sourcebuild.texi (arm_softfloat): Document.
10096 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
10099 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
10100 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
10101 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
10103 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10104 David Edelsohn <dje.gcc@gmail.com>
10107 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10108 Change "crset eq" to "crset 2".
10109 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10110 (*call_indirect_aix<mode>_nospec): Likewise.
10111 (*call_value_indirect_aix<mode>_nospec): Likewise.
10112 (*call_indirect_elfv2<mode>_nospec): Likewise.
10113 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
10114 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
10115 change assembly output from . to $.
10116 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10117 (indirect_jump<mode>_nospec): Change assembly output from . to $.
10118 (*tablejump<mode>_internal1_nospec): Likewise.
10120 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
10123 * config/sh/sh_optimize_sett_clrt.cc:
10124 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
10126 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
10128 PR tree-optimization/83940
10129 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
10130 offset_dt to vect_constant_def rather than vect_unknown_def_type.
10131 (vect_check_load_store_mask): Add a mask_dt_out parameter and
10132 use it to pass back the definition type.
10133 (vect_check_store_rhs): Likewise rhs_dt_out.
10134 (vect_build_gather_load_calls): Add a mask_dt argument and use
10135 it instead of a call to vect_is_simple_use.
10136 (vectorizable_store): Update calls to vect_check_load_store_mask
10137 and vect_check_store_rhs. Use the dt returned by the latter instead
10138 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
10139 instead of calls to vect_is_simple_use. Pass the scalar rather
10140 than the vector operand to vect_is_simple_use when handling
10141 second and subsequent copies of an rhs value.
10142 (vectorizable_load): Update calls to vect_check_load_store_mask
10143 and vect_build_gather_load_calls. Use the cached mask_dt and
10144 gs_info.offset_dt instead of calls to vect_is_simple_use.
10146 2018-01-20 Jakub Jelinek <jakub@redhat.com>
10148 PR middle-end/83945
10149 * tree-emutls.c: Include gimplify.h.
10150 (lower_emutls_2): New function.
10151 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
10152 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
10153 it before further processing.
10156 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
10157 UINTVAL (trueop1) instead of INTVAL (op1).
10159 2018-01-19 Jakub Jelinek <jakub@redhat.com>
10163 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
10164 INCOMING_FRAME_SP_OFFSET if not defined.
10165 (scan_trace): Add ENTRY argument. If true and
10166 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
10167 emit a note to adjust the CFA offset.
10168 (create_cfi_notes): Adjust scan_trace callers.
10169 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
10170 INCOMING_FRAME_SP_OFFSET in the CIE.
10171 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
10172 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
10174 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
10175 * doc/tm.texi: Regenerated.
10177 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10179 PR rtl-optimization/83147
10180 * lra-constraints.c (remove_inheritance_pseudos): Use
10181 lra_substitute_pseudo_within_insn.
10183 2018-01-19 Tom de Vries <tom@codesourcery.com>
10184 Cesar Philippidis <cesar@codesourcery.com>
10187 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
10189 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
10192 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
10193 spaces for function labels.
10195 2018-01-19 Martin Liska <mliska@suse.cz>
10197 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
10198 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
10199 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
10200 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
10201 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
10202 (PRED_CONST_RETURN): Change from 69 to 65.
10203 (PRED_NULL_RETURN): Change from 91 to 71.
10204 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
10205 (PRED_LOOP_GUARD): Change from 66 to 73.
10207 2018-01-19 Martin Liska <mliska@suse.cz>
10209 * predict.c (predict_insn_def): Add new assert.
10210 (struct branch_predictor): Change type to signed integer.
10211 (test_prediction_value_range): Amend test to cover
10212 PROB_UNINITIALIZED.
10213 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
10214 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
10215 (PRED_LOOP_ITERATIONS_MAX): Likewise.
10216 (PRED_LOOP_IV_COMPARE): Likewise.
10217 * predict.h (PROB_UNINITIALIZED): Define new constant.
10219 2018-01-19 Martin Liska <mliska@suse.cz>
10221 * predict.c (dump_prediction): Add new format for
10222 analyze_brprob.py script which is enabled with -details
10224 * profile-count.h (precise_p): New function.
10226 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10228 PR tree-optimization/83922
10229 * tree-vect-loop.c (vect_verify_full_masking): Return false if
10230 there are no statements that need masking.
10231 (vect_active_double_reduction_p): New function.
10232 (vect_analyze_loop_operations): Use it when handling phis that
10233 are not in the loop header.
10235 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
10237 PR tree-optimization/83914
10238 * tree-vect-loop.c (vectorizable_induction): Don't convert
10239 init_expr or apply the peeling adjustment for inductions
10240 that are nested within the vectorized loop.
10242 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10244 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
10247 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10251 * function.h (gimplify_parameters): Add gimple_seq * argument.
10252 * function.c: Include gimple.h and options.h.
10253 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
10254 for the added local temporaries if needed.
10255 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
10256 if there are any parameter cleanups, wrap whole body into a
10257 try/finally with the cleanups.
10259 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
10262 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
10263 Use GET_MODE_CLASS for scalar floating point.
10265 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10269 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
10270 Fix call of call_cgraph_insertion_hooks.
10272 2018-01-18 Martin Sebor <msebor@redhat.com>
10274 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
10276 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
10279 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
10282 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
10285 * common.opt: (-ffile-prefix-map): New option.
10286 * opts.c (common_handle_option): Defer it.
10287 * opts-global.c (handle_common_deferred_options): Handle it.
10288 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
10289 * file-prefix-map.h: New file.
10290 (remap_debug_filename, add_debug_prefix_map): ...here.
10291 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
10292 * final.c (debug_prefix_map, add_debug_prefix_map
10293 remap_debug_filename): Move to...
10294 * file-prefix-map.c: New file.
10295 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
10296 generalize, get rid of alloca(), use strrchr() instead of strchr().
10297 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
10298 Implement in terms of add_prefix_map().
10299 (remap_macro_filename, remap_debug_filename): Implement in term of
10301 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
10302 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
10303 * dbxout.c: Include file-prefix-map.h.
10304 * varasm.c: Likewise.
10305 * vmsdbgout.c: Likewise.
10306 * xcoffout.c: Likewise.
10307 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
10308 * doc/cppopts.texi (-fmacro-prefix-map): Document.
10309 * doc/invoke.texi (-ffile-prefix-map): Document.
10310 (-fdebug-prefix-map): Update description.
10312 2018-01-18 Martin Liska <mliska@suse.cz>
10314 * config/i386/i386.c (indirect_thunk_name): Document that also
10316 (output_indirect_thunk): Document why both instructions
10317 (pause and lfence) are generated.
10319 2018-01-18 Richard Biener <rguenther@suse.de>
10321 PR tree-optimization/83887
10322 * graphite-scop-detection.c
10323 (scop_detection::get_nearest_dom_with_single_entry): Remove.
10324 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
10325 (scop_detection::merge_sese): Re-implement with a flood-fill
10326 algorithm that properly finds a SESE region if it exists.
10328 2018-01-18 Jakub Jelinek <jakub@redhat.com>
10331 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
10332 pointer_diff optimizations use view_convert instead of convert.
10334 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10336 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
10337 Generate different code for -mno-speculate-indirect-jumps.
10338 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
10339 (*call_indirect_aix<mode>): Disable for
10340 -mno-speculate-indirect-jumps.
10341 (*call_indirect_aix<mode>_nospec): New define_insn.
10342 (*call_value_indirect_aix<mode>): Disable for
10343 -mno-speculate-indirect-jumps.
10344 (*call_value_indirect_aix<mode>_nospec): New define_insn.
10345 (*sibcall_nonlocal_sysv<mode>): Generate different code for
10346 -mno-speculate-indirect-jumps.
10347 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
10349 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
10351 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
10352 long double type, set the flags for noting the default long double
10353 type, even if we don't pass or return a long double type.
10355 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
10358 * ipa-inline.c (flatten_function): Do not overwrite final inlining
10361 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
10363 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
10364 support for merge[hl].
10365 (fold_mergehl_helper): New helper function.
10366 (tree-vector-builder.h): New #include for tree_vector_builder usage.
10367 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
10368 (altivec_vmrglw_direct): Add xxmrglw insn.
10370 2018-01-17 Andrew Waterman <andrew@sifive.com>
10372 * config/riscv/riscv.c (riscv_conditional_register_usage): If
10373 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
10375 2018-01-17 David Malcolm <dmalcolm@redhat.com>
10378 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
10379 call the lto_location_cache before reading the
10380 DECL_SOURCE_LOCATION of the types.
10382 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
10383 Richard Sandiford <richard.sandiford@linaro.org>
10385 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
10386 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
10387 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
10388 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
10389 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
10391 * config/aarch64/constraints.md (aarch64_movti_operand):
10393 * config/aarch64/predicates.md (Uti): Add new constraint.
10395 2018-01-17 Carl Love <cel@us.ibm.com>
10397 * config/rs6000/vsx.md (define_expand xl_len_r,
10398 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
10399 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
10401 (define_expand, define_insn): Move the shift left from the
10402 define_insn to the define_expand for lxvl and stxvl instructions.
10403 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
10404 and XL_LEN_R definitions to PURE.
10406 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
10408 * config/i386/i386.c (indirect_thunk_name): Declare regno
10409 as unsigned int. Compare regno with INVALID_REGNUM.
10410 (output_indirect_thunk): Ditto.
10411 (output_indirect_thunk_function): Ditto.
10412 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
10413 in the call to output_indirect_thunk_function.
10415 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
10417 PR middle-end/83884
10418 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
10419 rather than the size of inner_type to determine the stack slot size
10420 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
10422 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
10425 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
10428 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
10430 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
10431 endian Linux systems to optionally enable multilibs for selecting
10432 the long double type if the user configured an explicit type.
10433 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
10434 have no long double multilibs if not defined.
10435 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
10436 warn if the user used -mabi={ieee,ibm}longdouble and we built
10437 multilibs for long double.
10438 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
10439 appropriate multilib option.
10440 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
10442 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
10443 for building long double multilibs.
10444 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
10446 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
10448 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
10451 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
10453 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
10456 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
10459 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
10462 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10464 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
10465 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
10467 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
10469 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
10470 different rtl trees depending on TARGET_64BIT.
10471 (rs6000_gen_lvx): Likewise.
10473 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
10475 * config/visium/visium.md (nop): Tweak comment.
10476 (hazard_nop): Likewise.
10478 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10480 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
10481 -mspeculate-indirect-jumps.
10482 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
10483 for -mno-speculate-indirect-jumps.
10484 (*call_indirect_elfv2<mode>_nospec): New define_insn.
10485 (*call_value_indirect_elfv2<mode>): Disable for
10486 -mno-speculate-indirect-jumps.
10487 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
10488 (indirect_jump): Emit different RTL for
10489 -mno-speculate-indirect-jumps.
10490 (*indirect_jump<mode>): Disable for
10491 -mno-speculate-indirect-jumps.
10492 (*indirect_jump<mode>_nospec): New define_insn.
10493 (tablejump): Emit different RTL for
10494 -mno-speculate-indirect-jumps.
10495 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
10496 (tablejumpsi_nospec): New define_expand.
10497 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
10498 (tablejumpdi_nospec): New define_expand.
10499 (*tablejump<mode>_internal1): Disable for
10500 -mno-speculate-indirect-jumps.
10501 (*tablejump<mode>_internal1_nospec): New define_insn.
10502 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
10505 2018-01-16 Artyom Skrobov tyomitch@gmail.com
10507 * caller-save.c (insert_save): Drop unnecessary parameter. All
10510 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10511 Richard Biener <rguenth@suse.de>
10514 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
10515 return early, inline manually is_gimple_sizepos. Make sure if we
10516 call gimplify_expr we don't end up with a gimple constant.
10517 * tree.c (variably_modified_type_p): Don't return true for
10518 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
10519 * gimplify.h (is_gimple_sizepos): Remove.
10521 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10523 PR tree-optimization/83857
10524 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
10525 vectorizable_live_operation for pure SLP statements.
10526 (vectorizable_live_operation): Handle PHIs.
10528 2018-01-16 Richard Biener <rguenther@suse.de>
10530 PR tree-optimization/83867
10531 * tree-vect-stmts.c (vect_transform_stmt): Precompute
10532 nested_in_vect_loop_p since the scalar stmt may get invalidated.
10534 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10537 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
10538 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
10539 If off is not INTEGER_CST, issue a may not be aligned warning
10540 rather than isn't aligned. Use isn%'t rather than isn't.
10541 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
10543 <case MULT_EXPR>: Improve the case when bottom and one of the
10544 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
10545 operand, in that case check if the other operand is multiple of
10546 bottom divided by the INTEGER_CST operand.
10548 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10551 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
10552 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
10553 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
10554 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
10555 * config/pa/pa.c (pa_function_arg_advance): Likewise.
10556 (pa_function_arg, pa_arg_partial_bytes): Likewise.
10557 (pa_function_arg_size): New function.
10559 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10561 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
10562 in a separate statement.
10564 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
10566 PR tree-optimization/83847
10567 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
10568 group gathers and scatters.
10570 2018-01-16 Jakub Jelinek <jakub@redhat.com>
10572 PR rtl-optimization/86620
10573 * params.def (max-sched-ready-insns): Bump minimum value to 1.
10575 PR rtl-optimization/83213
10576 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
10577 to last if both are JUMP_INSNs.
10579 PR tree-optimization/83843
10580 * gimple-ssa-store-merging.c
10581 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
10582 store_immediate_info for bswap/nop orig_stores.
10584 2018-01-15 Andrew Waterman <andrew@sifive.com>
10586 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
10588 <UDIV>: Increase cost if !TARGET_DIV.
10590 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
10592 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
10593 (define_attr "cr_logical_3op"): New.
10594 (cceq_ior_compare): Adjust.
10595 (cceq_ior_compare_complement): Adjust.
10596 (*cceq_rev_compare): Adjust.
10597 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
10598 (is_cracked_insn): Adjust.
10599 (insn_must_be_first_in_group): Adjust.
10600 * config/rs6000/40x.md: Adjust.
10601 * config/rs6000/440.md: Adjust.
10602 * config/rs6000/476.md: Adjust.
10603 * config/rs6000/601.md: Adjust.
10604 * config/rs6000/603.md: Adjust.
10605 * config/rs6000/6xx.md: Adjust.
10606 * config/rs6000/7450.md: Adjust.
10607 * config/rs6000/7xx.md: Adjust.
10608 * config/rs6000/8540.md: Adjust.
10609 * config/rs6000/cell.md: Adjust.
10610 * config/rs6000/e300c2c3.md: Adjust.
10611 * config/rs6000/e500mc.md: Adjust.
10612 * config/rs6000/e500mc64.md: Adjust.
10613 * config/rs6000/e5500.md: Adjust.
10614 * config/rs6000/e6500.md: Adjust.
10615 * config/rs6000/mpc.md: Adjust.
10616 * config/rs6000/power4.md: Adjust.
10617 * config/rs6000/power5.md: Adjust.
10618 * config/rs6000/power6.md: Adjust.
10619 * config/rs6000/power7.md: Adjust.
10620 * config/rs6000/power8.md: Adjust.
10621 * config/rs6000/power9.md: Adjust.
10622 * config/rs6000/rs64.md: Adjust.
10623 * config/rs6000/titan.md: Adjust.
10625 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10627 * config/i386/predicates.md (indirect_branch_operand): Rewrite
10628 ix86_indirect_branch_register logic.
10630 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10632 * config/i386/constraints.md (Bs): Update
10633 ix86_indirect_branch_register check. Don't check
10634 ix86_indirect_branch_register with GOT_memory_operand.
10636 * config/i386/predicates.md (GOT_memory_operand): Don't check
10637 ix86_indirect_branch_register here.
10638 (GOT32_symbol_operand): Likewise.
10640 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10642 * config/i386/predicates.md (constant_call_address_operand):
10643 Rewrite ix86_indirect_branch_register logic.
10644 (sibcall_insn_operand): Likewise.
10646 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10648 * config/i386/constraints.md (Bs): Replace
10649 ix86_indirect_branch_thunk_register with
10650 ix86_indirect_branch_register.
10652 * config/i386/i386.md (indirect_jump): Likewise.
10653 (tablejump): Likewise.
10654 (*sibcall_memory): Likewise.
10655 (*sibcall_value_memory): Likewise.
10656 Peepholes of indirect call and jump via memory: Likewise.
10657 * config/i386/i386.opt: Likewise.
10658 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10659 (GOT_memory_operand): Likewise.
10660 (call_insn_operand): Likewise.
10661 (sibcall_insn_operand): Likewise.
10662 (GOT32_symbol_operand): Likewise.
10664 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10666 PR middle-end/83837
10667 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
10668 type rather than type addr's type points to.
10669 (expand_omp_atomic_mutex): Likewise.
10670 (expand_omp_atomic): Likewise.
10672 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
10675 * config/i386/i386.c (output_indirect_thunk_function): Use
10676 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
10677 for __x86_return_thunk.
10679 2018-01-15 Richard Biener <rguenther@suse.de>
10681 PR middle-end/83850
10682 * expmed.c (extract_bit_field_1): Fix typo.
10684 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10687 * config/arm/iterators.md (VF): New mode iterator.
10688 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
10689 Remove integer-related logic from pattern.
10690 (neon_vabd<mode>_3): Likewise.
10692 2018-01-15 Jakub Jelinek <jakub@redhat.com>
10694 PR middle-end/82694
10695 * common.opt (fstrict-overflow): No longer an alias.
10696 (fwrapv-pointer): New option.
10697 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
10698 also for pointer types based on flag_wrapv_pointer.
10699 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
10700 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
10701 opts->x_flag_wrapv got set.
10702 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
10703 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
10704 POINTER_TYPE_OVERFLOW_UNDEFINED.
10705 * match.pd: Likewise in address comparison pattern.
10706 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
10708 2018-01-15 Richard Biener <rguenther@suse.de>
10711 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
10712 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
10713 Reset type names to their identifier if their TYPE_DECL doesn't
10714 have linkage (and thus is used for ODR and devirt).
10715 (save_debug_info_for_decl): Remove.
10716 (save_debug_info_for_type): Likewise.
10717 (add_tree_to_fld_list): Adjust.
10718 * tree-pretty-print.c (dump_generic_node): Make dumping of
10719 type names more robust.
10721 2018-01-15 Richard Biener <rguenther@suse.de>
10723 * BASE-VER: Bump to 8.0.1.
10725 2018-01-14 Martin Sebor <msebor@redhat.com>
10728 * builtins.c (check_access): Avoid warning when the no-warning bit
10731 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
10733 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
10734 * ira-color (allocno_hard_regs_compare): Likewise.
10736 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
10739 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
10740 Use .pushsection/.popsection.
10742 2018-01-14 Martin Sebor <msebor@redhat.com>
10745 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
10747 2018-01-14 Jakub Jelinek <jakub@redhat.com>
10749 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
10750 entry from extra_headers.
10751 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
10752 extra_headers, make the list bitwise identical to the i?86-*-* one.
10754 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10756 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
10757 -mcmodel=large with -mindirect-branch=thunk,
10758 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
10759 -mfunction-return=thunk-extern.
10760 * doc/invoke.texi: Document -mcmodel=large is incompatible with
10761 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
10762 -mfunction-return=thunk and -mfunction-return=thunk-extern.
10764 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10766 * config/i386/i386.c (print_reg): Print the name of the full
10767 integer register without '%'.
10768 (ix86_print_operand): Handle 'V'.
10769 * doc/extend.texi: Document 'V' modifier.
10771 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10773 * config/i386/constraints.md (Bs): Disallow memory operand for
10774 -mindirect-branch-register.
10776 * config/i386/predicates.md (indirect_branch_operand): Likewise.
10777 (GOT_memory_operand): Likewise.
10778 (call_insn_operand): Likewise.
10779 (sibcall_insn_operand): Likewise.
10780 (GOT32_symbol_operand): Likewise.
10781 * config/i386/i386.md (indirect_jump): Call convert_memory_address
10782 for -mindirect-branch-register.
10783 (tablejump): Likewise.
10784 (*sibcall_memory): Likewise.
10785 (*sibcall_value_memory): Likewise.
10786 Disallow peepholes of indirect call and jump via memory for
10787 -mindirect-branch-register.
10788 (*call_pop): Replace m with Bw.
10789 (*call_value_pop): Likewise.
10790 (*sibcall_pop_memory): Replace m with Bs.
10791 * config/i386/i386.opt (mindirect-branch-register): New option.
10792 * doc/invoke.texi: Document -mindirect-branch-register option.
10794 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10796 * config/i386/i386-protos.h (ix86_output_function_return): New.
10797 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
10798 set function_return_type.
10799 (indirect_thunk_name): Add ret_p to indicate thunk for function
10801 (output_indirect_thunk_function): Pass false to
10802 indirect_thunk_name.
10803 (ix86_output_indirect_branch_via_reg): Likewise.
10804 (ix86_output_indirect_branch_via_push): Likewise.
10805 (output_indirect_thunk_function): Create alias for function
10806 return thunk if regno < 0.
10807 (ix86_output_function_return): New function.
10808 (ix86_handle_fndecl_attribute): Handle function_return.
10809 (ix86_attribute_table): Add function_return.
10810 * config/i386/i386.h (machine_function): Add
10811 function_return_type.
10812 * config/i386/i386.md (simple_return_internal): Use
10813 ix86_output_function_return.
10814 (simple_return_internal_long): Likewise.
10815 * config/i386/i386.opt (mfunction-return=): New option.
10816 (indirect_branch): Mention -mfunction-return=.
10817 * doc/extend.texi: Document function_return function attribute.
10818 * doc/invoke.texi: Document -mfunction-return= option.
10820 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
10822 * config/i386/i386-opts.h (indirect_branch): New.
10823 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
10824 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
10825 with local indirect jump when converting indirect call and jump.
10826 (ix86_set_indirect_branch_type): New.
10827 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
10828 (indirectlabelno): New.
10829 (indirect_thunk_needed): Likewise.
10830 (indirect_thunk_bnd_needed): Likewise.
10831 (indirect_thunks_used): Likewise.
10832 (indirect_thunks_bnd_used): Likewise.
10833 (INDIRECT_LABEL): Likewise.
10834 (indirect_thunk_name): Likewise.
10835 (output_indirect_thunk): Likewise.
10836 (output_indirect_thunk_function): Likewise.
10837 (ix86_output_indirect_branch_via_reg): Likewise.
10838 (ix86_output_indirect_branch_via_push): Likewise.
10839 (ix86_output_indirect_branch): Likewise.
10840 (ix86_output_indirect_jmp): Likewise.
10841 (ix86_code_end): Call output_indirect_thunk_function if needed.
10842 (ix86_output_call_insn): Call ix86_output_indirect_branch if
10844 (ix86_handle_fndecl_attribute): Handle indirect_branch.
10845 (ix86_attribute_table): Add indirect_branch.
10846 * config/i386/i386.h (machine_function): Add indirect_branch_type
10847 and has_local_indirect_jump.
10848 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
10850 (tablejump): Likewise.
10851 (*indirect_jump): Use ix86_output_indirect_jmp.
10852 (*tablejump_1): Likewise.
10853 (simple_return_indirect_internal): Likewise.
10854 * config/i386/i386.opt (mindirect-branch=): New option.
10855 (indirect_branch): New.
10858 (thunk-inline): Likewise.
10859 (thunk-extern): Likewise.
10860 * doc/extend.texi: Document indirect_branch function attribute.
10861 * doc/invoke.texi: Document -mindirect-branch= option.
10863 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
10866 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
10868 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
10870 * ipa-inline.c (want_inline_small_function_p): Return false if
10871 inlining has already failed with CIF_FINAL_ERROR.
10872 (update_caller_keys): Call want_inline_small_function_p before
10874 (update_callee_keys): Likewise.
10876 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
10878 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
10880 (rs6000_quadword_masked_address_p): Likewise.
10881 (quad_aligned_load_p): Likewise.
10882 (quad_aligned_store_p): Likewise.
10883 (const_load_sequence_p): Add comment to describe the outer-most loop.
10884 (mimic_memory_attributes_and_flags): New function.
10885 (rs6000_gen_stvx): Likewise.
10886 (replace_swapped_aligned_store): Likewise.
10887 (rs6000_gen_lvx): Likewise.
10888 (replace_swapped_aligned_load): Likewise.
10889 (replace_swapped_load_constant): Capitalize argument name in
10890 comment describing this function.
10891 (rs6000_analyze_swaps): Add a third pass to search for vector loads
10892 and stores that access quad-word aligned addresses and replace
10893 with stvx or lvx instructions when appropriate.
10894 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
10895 New function prototype.
10896 (rs6000_quadword_masked_address_p): Likewise.
10897 (rs6000_gen_lvx): Likewise.
10898 (rs6000_gen_stvx): Likewise.
10899 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
10900 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
10901 when memory address is aligned.
10902 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
10903 this split to select lvx instruction when memory address is aligned.
10904 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
10905 instruction when memory address is aligned.
10906 (*vsx_le_perm_load_v16qi): Likewise.
10907 (four unnamed splitters): Modify to select the stvx instruction
10908 when memory is aligned.
10910 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
10912 * predict.c (determine_unlikely_bbs): Handle correctly BBs
10913 which appears in the queue multiple times.
10915 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10916 Alan Hayward <alan.hayward@arm.com>
10917 David Sherwood <david.sherwood@arm.com>
10919 * tree-vectorizer.h (vec_lower_bound): New structure.
10920 (_loop_vec_info): Add check_nonzero and lower_bounds.
10921 (LOOP_VINFO_CHECK_NONZERO): New macro.
10922 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
10923 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
10924 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
10925 fields. Make seg_len the distance travelled, not including the
10927 (dr_direction_indicator): Declare.
10928 (dr_zero_step_indicator): Likewise.
10929 (dr_known_forward_stride_p): Likewise.
10930 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
10932 (runtime_alias_check_p): Allow runtime alias checks with
10934 (operator ==): Compare access_size and align.
10935 (prune_runtime_alias_test_list): Rework for new distinction between
10936 the access_size and seg_len.
10937 (create_intersect_range_checks_index): Likewise. Cope with polynomial
10939 (get_segment_min_max): New function.
10940 (create_intersect_range_checks): Use it.
10941 (dr_step_indicator): New function.
10942 (dr_direction_indicator): Likewise.
10943 (dr_zero_step_indicator): Likewise.
10944 (dr_known_forward_stride_p): Likewise.
10945 * tree-loop-distribution.c (data_ref_segment_size): Return
10946 DR_STEP * (niters - 1).
10947 (compute_alias_check_pairs): Update call to the dr_with_seg_len
10949 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
10950 (vect_preserves_scalar_order_p): New function, split out from...
10951 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
10952 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
10953 (vect_vfa_access_size): New function.
10954 (vect_vfa_align): Likewise.
10955 (vect_compile_time_alias): Take access_size_a and access_b arguments.
10956 (dump_lower_bound): New function.
10957 (vect_check_lower_bound): Likewise.
10958 (vect_small_gap_p): Likewise.
10959 (vectorizable_with_step_bound_p): Likewise.
10960 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
10961 depencies if the vectorization factor is 1. Convert the checks
10962 for nonzero steps into checks on the bounds of DR_STEP. Try using
10963 a bunds check for variable steps if the minimum required step is
10964 relatively small. Update calls to the dr_with_seg_len
10965 constructor and to vect_compile_time_alias.
10966 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
10968 (vect_loop_versioning): Call it.
10969 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
10971 (vect_estimate_min_profitable_iters): Account for any bounds checks.
10973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
10974 Alan Hayward <alan.hayward@arm.com>
10975 David Sherwood <david.sherwood@arm.com>
10977 * doc/sourcebuild.texi (vect_scatter_store): Document.
10978 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
10980 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
10982 * genopinit.c (main): Add supports_vec_scatter_store and
10983 supports_vec_scatter_store_cached to target_optabs.
10984 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
10985 IFN_MASK_SCATTER_STORE.
10986 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
10988 * internal-fn.h (internal_store_fn_p): Declare.
10989 (internal_fn_stored_value_index): Likewise.
10990 * internal-fn.c (scatter_store_direct): New macro.
10991 (expand_scatter_store_optab_fn): New function.
10992 (direct_scatter_store_optab_supported_p): New macro.
10993 (internal_store_fn_p): New function.
10994 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
10995 IFN_MASK_SCATTER_STORE.
10996 (internal_fn_mask_index): Likewise.
10997 (internal_fn_stored_value_index): New function.
10998 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
10999 for scatter stores.
11000 * optabs-query.h (supports_vec_scatter_store_p): Declare.
11001 * optabs-query.c (supports_vec_scatter_store_p): New function.
11002 * tree-vectorizer.h (vect_get_store_rhs): Declare.
11003 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
11004 true for scatter stores.
11005 (vect_gather_scatter_fn_p): Handle scatter stores too.
11006 (vect_check_gather_scatter): Consider using scatter stores if
11007 supports_vec_scatter_store_p.
11008 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
11009 scatter stores too.
11010 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11011 internal_fn_stored_value_index.
11012 (check_load_store_masking): Handle scatter stores too.
11013 (vect_get_store_rhs): Make public.
11014 (vectorizable_call): Use internal_store_fn_p.
11015 (vectorizable_store): Handle scatter store internal functions.
11016 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
11017 when deciding whether the end of the group has been reached.
11018 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
11019 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
11020 (mask_scatter_store<mode>): New insns.
11022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11023 Alan Hayward <alan.hayward@arm.com>
11024 David Sherwood <david.sherwood@arm.com>
11026 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
11027 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
11028 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
11030 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
11031 Use vect_truncate_gather_scatter_offset if we can't treat the
11032 operation as a normal gather load or scatter store.
11033 (get_group_load_store_type): Take the gather_scatter_info
11034 as argument. Try using a gather load or scatter store for
11035 single-element groups.
11036 (get_load_store_type): Update calls to get_group_load_store_type
11037 and vect_use_strided_gather_scatters_p.
11039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11040 Alan Hayward <alan.hayward@arm.com>
11041 David Sherwood <david.sherwood@arm.com>
11043 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
11044 optional tree argument.
11045 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
11047 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
11048 but continue to use the current value as a fallback.
11049 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
11050 to compare the updates.
11051 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
11052 (get_load_store_type): Use it when handling a strided access.
11053 (vect_get_strided_load_store_ops): New function.
11054 (vect_get_data_ptr_increment): Likewise.
11055 (vectorizable_load): Handle strided gather loads. Always pass
11056 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
11063 (mask_gather_load@var{m}): Likewise.
11064 * genopinit.c (main): Add supports_vec_gather_load and
11065 supports_vec_gather_load_cached to target_optabs.
11066 * optabs-tree.c (init_tree_optimization_optabs): Use
11067 ggc_cleared_alloc to allocate target_optabs.
11068 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
11069 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
11071 * internal-fn.h (internal_load_fn_p): Declare.
11072 (internal_gather_scatter_fn_p): Likewise.
11073 (internal_fn_mask_index): Likewise.
11074 (internal_gather_scatter_fn_supported_p): Likewise.
11075 * internal-fn.c (gather_load_direct): New macro.
11076 (expand_gather_load_optab_fn): New function.
11077 (direct_gather_load_optab_supported_p): New macro.
11078 (direct_internal_fn_optab): New function.
11079 (internal_load_fn_p): Likewise.
11080 (internal_gather_scatter_fn_p): Likewise.
11081 (internal_fn_mask_index): Likewise.
11082 (internal_gather_scatter_fn_supported_p): Likewise.
11083 * optabs-query.c (supports_at_least_one_mode_p): New function.
11084 (supports_vec_gather_load_p): Likewise.
11085 * optabs-query.h (supports_vec_gather_load_p): Declare.
11086 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
11087 and memory_type field.
11088 (NUM_PATTERNS): Bump to 15.
11089 * tree-vect-data-refs.c: Include internal-fn.h.
11090 (vect_gather_scatter_fn_p): New function.
11091 (vect_describe_gather_scatter_call): Likewise.
11092 (vect_check_gather_scatter): Try using internal functions for
11093 gather loads. Recognize existing calls to a gather load function.
11094 (vect_analyze_data_refs): Consider using gather loads if
11095 supports_vec_gather_load_p.
11096 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
11097 (vect_get_gather_scatter_offset_type): Likewise.
11098 (vect_convert_mask_for_vectype): Likewise.
11099 (vect_add_conversion_to_patterm): Likewise.
11100 (vect_try_gather_scatter_pattern): Likewise.
11101 (vect_recog_gather_scatter_pattern): New pattern recognizer.
11102 (vect_vect_recog_func_ptrs): Add it.
11103 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
11104 internal_fn_mask_index and internal_gather_scatter_fn_p.
11105 (check_load_store_masking): Take the gather_scatter_info as an
11106 argument and handle gather loads.
11107 (vect_get_gather_scatter_ops): New function.
11108 (vectorizable_call): Check internal_load_fn_p.
11109 (vectorizable_load): Likewise. Handle gather load internal
11111 (vectorizable_store): Update call to check_load_store_masking.
11112 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
11113 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
11114 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
11115 (aarch64_gather_scale_operand_d): New predicates.
11116 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
11117 (mask_gather_load<mode>): New insns.
11119 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11120 Alan Hayward <alan.hayward@arm.com>
11121 David Sherwood <david.sherwood@arm.com>
11123 * optabs.def (fold_left_plus_optab): New optab.
11124 * doc/md.texi (fold_left_plus_@var{m}): Document.
11125 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
11126 * internal-fn.c (fold_left_direct): Define.
11127 (expand_fold_left_optab_fn): Likewise.
11128 (direct_fold_left_optab_supported_p): Likewise.
11129 * fold-const-call.c (fold_const_fold_left): New function.
11130 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
11131 * tree-parloops.c (valid_reduction_p): New function.
11132 (gather_scalar_reductions): Use it.
11133 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
11134 (vect_finish_replace_stmt): Declare.
11135 * tree-vect-loop.c (fold_left_reduction_fn): New function.
11136 (needs_fold_left_reduction_p): New function, split out from...
11137 (vect_is_simple_reduction): ...here. Accept reductions that
11138 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
11139 (vect_force_simple_reduction): Also store the reduction type in
11140 the assignment's STMT_VINFO_REDUC_TYPE.
11141 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
11142 (merge_with_identity): New function.
11143 (vect_expand_fold_left): Likewise.
11144 (vectorize_fold_left_reduction): Likewise.
11145 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
11146 scalar phi in place for it. Check for target support and reject
11147 cases that would reassociate the operation. Defer the transform
11148 phase to vectorize_fold_left_reduction.
11149 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
11150 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
11151 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
11153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11155 * tree-if-conv.c (predicate_mem_writes): Remove redundant
11156 call to ifc_temp_var.
11158 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11159 Alan Hayward <alan.hayward@arm.com>
11160 David Sherwood <david.sherwood@arm.com>
11162 * target.def (legitimize_address_displacement): Take the original
11163 offset as a poly_int.
11164 * targhooks.h (default_legitimize_address_displacement): Update
11166 * targhooks.c (default_legitimize_address_displacement): Likewise.
11167 * doc/tm.texi: Regenerate.
11168 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
11169 as an argument, moving assert of ad->disp == ad->disp_term to...
11170 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
11171 Try calling targetm.legitimize_address_displacement before expanding
11172 the address rather than afterwards, and adjust for the new interface.
11173 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
11174 Match the new hook interface. Handle SVE addresses.
11175 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
11176 new hook interface.
11178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11180 * Makefile.in (OBJS): Add early-remat.o.
11181 * target.def (select_early_remat_modes): New hook.
11182 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
11183 * doc/tm.texi: Regenerate.
11184 * targhooks.h (default_select_early_remat_modes): Declare.
11185 * targhooks.c (default_select_early_remat_modes): New function.
11186 * timevar.def (TV_EARLY_REMAT): New timevar.
11187 * passes.def (pass_early_remat): New pass.
11188 * tree-pass.h (make_pass_early_remat): Declare.
11189 * early-remat.c: New file.
11190 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
11192 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
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 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
11199 vfm1 with a bound_epilog parameter.
11200 (vect_do_peeling): Update calls accordingly, and move the prologue
11201 call earlier in the function. Treat the base bound_epilog as 0 for
11202 fully-masked loops and retain vf - 1 for other loops. Add 1 to
11203 this base when peeling for gaps.
11204 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
11205 with fully-masked loops.
11206 (vect_estimate_min_profitable_iters): Handle the single peeled
11207 iteration in that case.
11209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11210 Alan Hayward <alan.hayward@arm.com>
11211 David Sherwood <david.sherwood@arm.com>
11213 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
11214 single-element interleaving even if the size is not a power of 2.
11215 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
11216 accesses for single-element interleaving if the group size is
11219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11220 Alan Hayward <alan.hayward@arm.com>
11221 David Sherwood <david.sherwood@arm.com>
11223 * doc/md.texi (fold_extract_last_@var{m}): Document.
11224 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
11225 * optabs.def (fold_extract_last_optab): New optab.
11226 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
11227 * internal-fn.c (fold_extract_direct): New macro.
11228 (expand_fold_extract_optab_fn): Likewise.
11229 (direct_fold_extract_optab_supported_p): Likewise.
11230 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
11231 * tree-vect-loop.c (vect_model_reduction_cost): Handle
11232 EXTRACT_LAST_REDUCTION.
11233 (get_initial_def_for_reduction): Do not create an initial vector
11234 for EXTRACT_LAST_REDUCTION reductions.
11235 (vectorizable_reduction): Leave the scalar phi in place for
11236 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
11237 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
11238 epilogue code for EXTRACT_LAST_REDUCTION and defer the
11239 transform phase to vectorizable_condition.
11240 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
11242 (vect_finish_stmt_generation): ...here.
11243 (vect_finish_replace_stmt): New function.
11244 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
11245 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
11247 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
11249 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11250 Alan Hayward <alan.hayward@arm.com>
11251 David Sherwood <david.sherwood@arm.com>
11253 * doc/md.texi (extract_last_@var{m}): Document.
11254 * optabs.def (extract_last_optab): New optab.
11255 * internal-fn.def (EXTRACT_LAST): New internal function.
11256 * internal-fn.c (cond_unary_direct): New macro.
11257 (expand_cond_unary_optab_fn): Likewise.
11258 (direct_cond_unary_optab_supported_p): Likewise.
11259 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
11260 loops using EXTRACT_LAST.
11261 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
11262 (extract_last_<mode>): ...this optab.
11263 (vec_extract<mode><Vel>): Update accordingly.
11265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11266 Alan Hayward <alan.hayward@arm.com>
11267 David Sherwood <david.sherwood@arm.com>
11269 * target.def (empty_mask_is_expensive): New hook.
11270 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
11271 * doc/tm.texi: Regenerate.
11272 * targhooks.h (default_empty_mask_is_expensive): Declare.
11273 * targhooks.c (default_empty_mask_is_expensive): New function.
11274 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
11275 if the target says that empty masks are expensive.
11276 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
11278 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
11280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11281 Alan Hayward <alan.hayward@arm.com>
11282 David Sherwood <david.sherwood@arm.com>
11284 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
11285 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
11286 (vect_use_loop_mask_for_alignment_p): New function.
11287 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
11288 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
11289 niters_skip argument. Make sure that the first niters_skip elements
11290 of the first iteration are inactive.
11291 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
11292 Update call to vect_set_loop_masks_directly.
11293 (get_misalign_in_elems): New function, split out from...
11294 (vect_gen_prolog_loop_niters): ...here.
11295 (vect_update_init_of_dr): Take a code argument that specifies whether
11296 the adjustment should be added or subtracted.
11297 (vect_update_init_of_drs): Likewise.
11298 (vect_prepare_for_masked_peels): New function.
11299 (vect_do_peeling): Skip prologue peeling if we're using a mask
11300 instead. Update call to vect_update_inits_of_drs.
11301 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11303 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
11304 alignment. Do not include the number of peeled iterations in
11305 the minimum threshold in that case.
11306 (vectorizable_induction): Adjust the start value down by
11307 LOOP_VINFO_MASK_SKIP_NITERS iterations.
11308 (vect_transform_loop): Call vect_prepare_for_masked_peels.
11309 Take the number of skipped iterations into account when calculating
11311 * tree-vect-stmts.c (vect_gen_while_not): New function.
11313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11314 Alan Hayward <alan.hayward@arm.com>
11315 David Sherwood <david.sherwood@arm.com>
11317 * doc/sourcebuild.texi (vect_fully_masked): Document.
11318 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
11319 default value to 0.
11320 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
11322 (vect_analyze_loop_2): ...here. Don't check the vectorization
11323 factor against the number of loop iterations if the loop is
11326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11327 Alan Hayward <alan.hayward@arm.com>
11328 David Sherwood <david.sherwood@arm.com>
11330 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
11331 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
11332 (dump_groups): Update accordingly.
11333 (iv_use::mem_type): New member variable.
11334 (address_p): New function.
11335 (record_use): Add a mem_type argument and initialize the new
11337 (record_group_use): Add a mem_type argument. Use address_p.
11338 Remove obsolete null checks of base_object. Update call to record_use.
11339 (find_interesting_uses_op): Update call to record_group_use.
11340 (find_interesting_uses_cond): Likewise.
11341 (find_interesting_uses_address): Likewise.
11342 (get_mem_type_for_internal_fn): New function.
11343 (find_address_like_use): Likewise.
11344 (find_interesting_uses_stmt): Try find_address_like_use before
11345 calling find_interesting_uses_op.
11346 (addr_offset_valid_p): Use the iv mem_type field as the type
11347 of the addressed memory.
11348 (add_autoinc_candidates): Likewise.
11349 (get_address_cost): Likewise.
11350 (split_small_address_groups_p): Use address_p.
11351 (split_address_groups): Likewise.
11352 (add_iv_candidate_for_use): Likewise.
11353 (autoinc_possible_for_pair): Likewise.
11354 (rewrite_groups): Likewise.
11355 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
11356 (determine_group_iv_cost): Update after split of USE_ADDRESS.
11357 (get_alias_ptr_type_for_ptr_address): New function.
11358 (rewrite_use_address): Rewrite address uses in calls that were
11359 identified by find_address_like_use.
11361 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11362 Alan Hayward <alan.hayward@arm.com>
11363 David Sherwood <david.sherwood@arm.com>
11365 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
11367 * gimple-expr.h (is_gimple_addressable: Likewise.
11368 * gimple-expr.c (is_gimple_address): Likewise.
11369 * internal-fn.c (expand_call_mem_ref): New function.
11370 (expand_mask_load_optab_fn): Use it.
11371 (expand_mask_store_optab_fn): Likewise.
11373 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11374 Alan Hayward <alan.hayward@arm.com>
11375 David Sherwood <david.sherwood@arm.com>
11377 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
11378 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
11379 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
11380 (cond_umax@var{mode}): Document.
11381 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
11382 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
11383 (cond_umin_optab, cond_umax_optab): New optabs.
11384 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
11385 (COND_IOR, COND_XOR): New internal functions.
11386 * internal-fn.h (get_conditional_internal_fn): Declare.
11387 * internal-fn.c (cond_binary_direct): New macro.
11388 (expand_cond_binary_optab_fn): Likewise.
11389 (direct_cond_binary_optab_supported_p): Likewise.
11390 (get_conditional_internal_fn): New function.
11391 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
11392 Cope with reduction statements that are vectorized as calls rather
11394 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
11395 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
11396 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
11397 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11398 (UNSPEC_COND_EOR): New unspecs.
11399 (optab): Add mappings for them.
11400 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
11401 (sve_int_op, sve_fp_op): New int attributes.
11403 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11404 Alan Hayward <alan.hayward@arm.com>
11405 David Sherwood <david.sherwood@arm.com>
11407 * optabs.def (while_ult_optab): New optab.
11408 * doc/md.texi (while_ult@var{m}@var{n}): Document.
11409 * internal-fn.def (WHILE_ULT): New internal function.
11410 * internal-fn.h (direct_internal_fn_supported_p): New override
11411 that takes two types as argument.
11412 * internal-fn.c (while_direct): New macro.
11413 (expand_while_optab_fn): New function.
11414 (convert_optab_supported_p): Likewise.
11415 (direct_while_optab_supported_p): New macro.
11416 * wide-int.h (wi::udiv_ceil): New function.
11417 * tree-vectorizer.h (rgroup_masks): New structure.
11418 (vec_loop_masks): New typedef.
11419 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
11420 and fully_masked_p.
11421 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
11422 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
11423 (vect_max_vf): New function.
11424 (slpeel_make_loop_iterate_ntimes): Delete.
11425 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
11426 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
11427 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
11428 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
11429 internal-fn.h, stor-layout.h and optabs-query.h.
11430 (vect_set_loop_mask): New function.
11431 (add_preheader_seq): Likewise.
11432 (add_header_seq): Likewise.
11433 (interleave_supported_p): Likewise.
11434 (vect_maybe_permute_loop_masks): Likewise.
11435 (vect_set_loop_masks_directly): Likewise.
11436 (vect_set_loop_condition_masked): Likewise.
11437 (vect_set_loop_condition_unmasked): New function, split out from
11438 slpeel_make_loop_iterate_ntimes.
11439 (slpeel_make_loop_iterate_ntimes): Rename to..
11440 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
11441 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
11442 (vect_do_peeling): Update call accordingly.
11443 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
11445 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
11446 mask_compare_type, can_fully_mask_p and fully_masked_p.
11447 (release_vec_loop_masks): New function.
11448 (_loop_vec_info): Use it to free the loop masks.
11449 (can_produce_all_loop_masks_p): New function.
11450 (vect_get_max_nscalars_per_iter): Likewise.
11451 (vect_verify_full_masking): Likewise.
11452 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
11453 retries, and free the mask rgroups before retrying. Check loop-wide
11454 reasons for disallowing fully-masked loops. Make the final decision
11455 about whether use a fully-masked loop or not.
11456 (vect_estimate_min_profitable_iters): Do not assume that peeling
11457 for the number of iterations will be needed for fully-masked loops.
11458 (vectorizable_reduction): Disable fully-masked loops.
11459 (vectorizable_live_operation): Likewise.
11460 (vect_halve_mask_nunits): New function.
11461 (vect_double_mask_nunits): Likewise.
11462 (vect_record_loop_mask): Likewise.
11463 (vect_get_loop_mask): Likewise.
11464 (vect_transform_loop): Handle the case in which the final loop
11465 iteration might handle a partial vector. Call vect_set_loop_condition
11466 instead of slpeel_make_loop_iterate_ntimes.
11467 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
11468 (check_load_store_masking): New function.
11469 (prepare_load_store_mask): Likewise.
11470 (vectorizable_store): Handle fully-masked loops.
11471 (vectorizable_load): Likewise.
11472 (supportable_widening_operation): Use vect_halve_mask_nunits for
11474 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
11475 (vect_gen_while): New function.
11476 * config/aarch64/aarch64.md (umax<mode>3): New expander.
11477 (aarch64_uqdec<mode>): New insn.
11479 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11480 Alan Hayward <alan.hayward@arm.com>
11481 David Sherwood <david.sherwood@arm.com>
11483 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
11484 (reduc_xor_scal_optab): New optabs.
11485 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
11486 (reduc_xor_scal_@var{m}): Document.
11487 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
11488 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
11489 internal functions.
11490 * fold-const-call.c (fold_const_call): Handle them.
11491 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
11492 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
11493 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
11494 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
11495 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
11496 (UNSPEC_XORV): New unspecs.
11497 (optab): Add entries for them.
11498 (BITWISEV): New int iterator.
11499 (bit_reduc_op): New int attributes.
11501 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11502 Alan Hayward <alan.hayward@arm.com>
11503 David Sherwood <david.sherwood@arm.com>
11505 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
11506 * internal-fn.def (VEC_SHL_INSERT): New internal function.
11507 * optabs.def (vec_shl_insert_optab): New optab.
11508 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
11509 (duplicate_and_interleave): Likewise.
11510 * tree-vect-loop.c: Include internal-fn.h.
11511 (neutral_op_for_slp_reduction): New function, split out from
11512 get_initial_defs_for_reduction.
11513 (get_initial_def_for_reduction): Handle option 2 for variable-length
11514 vectors by loading the neutral value into a vector and then shifting
11515 the initial value into element 0.
11516 (get_initial_defs_for_reduction): Replace the code argument with
11517 the neutral value calculated by neutral_op_for_slp_reduction.
11518 Use gimple_build_vector for constant-length vectors.
11519 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
11520 but the first group_size elements have a neutral value.
11521 Use duplicate_and_interleave otherwise.
11522 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
11523 Update call to get_initial_defs_for_reduction. Handle SLP
11524 reductions for variable-length vectors by creating one vector
11525 result for each scalar result, with the elements associated
11526 with other scalar results stubbed out with the neutral value.
11527 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
11528 Require IFN_VEC_SHL_INSERT for double reductions on
11529 variable-length vectors, or SLP reductions that have
11530 a neutral value. Require can_duplicate_and_interleave_p
11531 support for variable-length unchained SLP reductions if there
11532 is no neutral value, such as for MIN/MAX reductions. Also require
11533 the number of vector elements to be a multiple of the number of
11534 SLP statements when doing variable-length unchained SLP reductions.
11535 Update call to vect_create_epilog_for_reduction.
11536 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
11537 and remove initial values.
11538 (duplicate_and_interleave): Make public.
11539 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
11540 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
11542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11543 Alan Hayward <alan.hayward@arm.com>
11544 David Sherwood <david.sherwood@arm.com>
11546 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
11547 (can_duplicate_and_interleave_p): New function.
11548 (vect_get_and_check_slp_defs): Take the vector of statements
11549 rather than just the current one. Remove excess parentheses.
11550 Restriction rejectinon of vect_constant_def and vect_external_def
11551 for variable-length vectors to boolean types, or types for which
11552 can_duplicate_and_interleave_p is false.
11553 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
11554 (duplicate_and_interleave): New function.
11555 (vect_get_constant_vectors): Use gimple_build_vector for
11556 constant-length vectors and suitable variable-length constant
11557 vectors. Use duplicate_and_interleave for other variable-length
11558 vectors. Don't defer the update when inserting new statements.
11560 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11561 Alan Hayward <alan.hayward@arm.com>
11562 David Sherwood <david.sherwood@arm.com>
11564 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
11565 min_profitable_iters doesn't go negative.
11567 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11568 Alan Hayward <alan.hayward@arm.com>
11569 David Sherwood <david.sherwood@arm.com>
11571 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
11572 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
11573 * optabs.def (vec_mask_load_lanes_optab): New optab.
11574 (vec_mask_store_lanes_optab): Likewise.
11575 * internal-fn.def (MASK_LOAD_LANES): New internal function.
11576 (MASK_STORE_LANES): Likewise.
11577 * internal-fn.c (mask_load_lanes_direct): New macro.
11578 (mask_store_lanes_direct): Likewise.
11579 (expand_mask_load_optab_fn): Handle masked operations.
11580 (expand_mask_load_lanes_optab_fn): New macro.
11581 (expand_mask_store_optab_fn): Handle masked operations.
11582 (expand_mask_store_lanes_optab_fn): New macro.
11583 (direct_mask_load_lanes_optab_supported_p): Likewise.
11584 (direct_mask_store_lanes_optab_supported_p): Likewise.
11585 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
11587 (vect_load_lanes_supported): Likewise.
11588 * tree-vect-data-refs.c (strip_conversion): New function.
11589 (can_group_stmts_p): Likewise.
11590 (vect_analyze_data_ref_accesses): Use it instead of checking
11591 for a pair of assignments.
11592 (vect_store_lanes_supported): Take a masked_p parameter.
11593 (vect_load_lanes_supported): Likewise.
11594 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
11595 vect_store_lanes_supported and vect_load_lanes_supported.
11596 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11597 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
11598 parameter. Don't allow gaps for masked accesses.
11599 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
11600 and vect_load_lanes_supported.
11601 (get_load_store_type): Take a masked_p parameter and update
11602 call to get_group_load_store_type.
11603 (vectorizable_store): Update call to get_load_store_type.
11604 Handle IFN_MASK_STORE_LANES.
11605 (vectorizable_load): Update call to get_load_store_type.
11606 Handle IFN_MASK_LOAD_LANES.
11608 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11609 Alan Hayward <alan.hayward@arm.com>
11610 David Sherwood <david.sherwood@arm.com>
11612 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
11614 * config/aarch64/aarch64-protos.h
11615 (aarch64_sve_struct_memory_operand_p): Declare.
11616 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
11617 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
11618 (VPRED, vpred): Handle SVE structure modes.
11619 * config/aarch64/constraints.md (Utx): New constraint.
11620 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
11621 (aarch64_sve_struct_nonimmediate_operand): New predicates.
11622 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
11623 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
11624 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
11625 structure modes. Split into pieces after RA.
11626 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
11627 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
11629 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
11630 SVE structure modes.
11631 (aarch64_classify_address): Likewise.
11632 (sizetochar): Move earlier in file.
11633 (aarch64_print_operand): Handle SVE register lists.
11634 (aarch64_array_mode): New function.
11635 (aarch64_sve_struct_memory_operand_p): Likewise.
11636 (TARGET_ARRAY_MODE): Redefine.
11638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11639 Alan Hayward <alan.hayward@arm.com>
11640 David Sherwood <david.sherwood@arm.com>
11642 * target.def (array_mode): New target hook.
11643 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
11644 * doc/tm.texi: Regenerate.
11645 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
11646 * hooks.c (hook_optmode_mode_uhwi_none): New function.
11647 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
11648 targetm.array_mode.
11649 * stor-layout.c (mode_for_array): Likewise. Support polynomial
11652 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11653 Alan Hayward <alan.hayward@arm.com>
11654 David Sherwood <david.sherwood@arm.com>
11656 * fold-const.c (fold_binary_loc): Check the argument types
11657 rather than the result type when testing for a vector operation.
11659 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11661 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
11662 * doc/tm.texi: Regenerate.
11664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11665 Alan Hayward <alan.hayward@arm.com>
11666 David Sherwood <david.sherwood@arm.com>
11668 * doc/invoke.texi (-msve-vector-bits=): Document new option.
11669 (sve): Document new AArch64 extension.
11670 * doc/md.texi (w): Extend the description of the AArch64
11671 constraint to include SVE vectors.
11672 (Upl, Upa): Document new AArch64 predicate constraints.
11673 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
11675 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
11676 (msve-vector-bits=): New option.
11677 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
11678 SVE when these are disabled.
11679 (sve): New extension.
11680 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
11681 modes. Adjust their number of units based on aarch64_sve_vg.
11682 (MAX_BITSIZE_MODE_ANY_MODE): Define.
11683 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
11684 aarch64_addr_query_type.
11685 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
11686 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
11687 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
11688 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
11689 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
11690 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
11691 (aarch64_simd_imm_zero_p): Delete.
11692 (aarch64_check_zero_based_sve_index_immediate): Declare.
11693 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11694 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11695 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11696 (aarch64_sve_float_mul_immediate_p): Likewise.
11697 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11698 rather than an rtx.
11699 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
11700 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
11701 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
11702 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
11703 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
11704 (aarch64_regmode_natural_size): Likewise.
11705 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
11706 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
11708 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
11709 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
11710 for VG and the SVE predicate registers.
11711 (V_ALIASES): Add a "z"-prefixed alias.
11712 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
11713 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
11714 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
11715 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
11716 (REG_CLASS_NAMES): Add entries for them.
11717 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
11718 and the predicate registers.
11719 (aarch64_sve_vg): Declare.
11720 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
11721 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
11722 (REGMODE_NATURAL_SIZE): Define.
11723 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
11725 * config/aarch64/aarch64.c: Include cfgrtl.h.
11726 (simd_immediate_info): Add a constructor for series vectors,
11727 and an associated step field.
11728 (aarch64_sve_vg): New variable.
11729 (aarch64_dbx_register_number): Handle VG and the predicate registers.
11730 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
11731 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
11732 (VEC_ANY_DATA, VEC_STRUCT): New constants.
11733 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
11734 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
11735 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
11736 (aarch64_get_mask_mode): New functions.
11737 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
11738 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
11739 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
11740 predicate modes and predicate registers. Explicitly restrict
11741 GPRs to modes of 16 bytes or smaller. Only allow FP registers
11742 to store a vector mode if it is recognized by
11743 aarch64_classify_vector_mode.
11744 (aarch64_regmode_natural_size): New function.
11745 (aarch64_hard_regno_caller_save_mode): Return the original mode
11747 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
11748 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
11749 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
11750 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
11752 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
11753 does not overlap dest if the function is frame-related. Handle
11755 (aarch64_split_add_offset): New function.
11756 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
11757 them aarch64_add_offset.
11758 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
11759 and update call to aarch64_sub_sp.
11760 (aarch64_add_cfa_expression): New function.
11761 (aarch64_expand_prologue): Pass extra temporary registers to the
11762 functions above. Handle the case in which we need to emit new
11763 DW_CFA_expressions for registers that were originally saved
11764 relative to the stack pointer, but now have to be expressed
11765 relative to the frame pointer.
11766 (aarch64_output_mi_thunk): Pass extra temporary registers to the
11768 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
11769 IP0 and IP1 values for SVE frames.
11770 (aarch64_expand_vec_series): New function.
11771 (aarch64_expand_sve_widened_duplicate): Likewise.
11772 (aarch64_expand_sve_const_vector): Likewise.
11773 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
11774 Handle SVE constants. Use emit_move_insn to move a force_const_mem
11775 into the register, rather than emitting a SET directly.
11776 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
11777 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
11778 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
11779 (offset_9bit_signed_scaled_p): New functions.
11780 (aarch64_replicate_bitmask_imm): New function.
11781 (aarch64_bitmask_imm): Use it.
11782 (aarch64_cannot_force_const_mem): Reject expressions involving
11783 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
11784 (aarch64_classify_index): Handle SVE indices, by requiring
11785 a plain register index with a scale that matches the element size.
11786 (aarch64_classify_address): Handle SVE addresses. Assert that
11787 the mode of the address is VOIDmode or an integer mode.
11788 Update call to aarch64_classify_symbol.
11789 (aarch64_classify_symbolic_expression): Update call to
11790 aarch64_classify_symbol.
11791 (aarch64_const_vec_all_in_range_p): New function.
11792 (aarch64_print_vector_float_operand): Likewise.
11793 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
11794 "vN" for FP registers with SVE modes. Handle (const ...) vectors
11795 and the FP immediates 1.0 and 0.5.
11796 (aarch64_print_address_internal): Handle SVE addresses.
11797 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
11798 (aarch64_regno_regclass): Handle predicate registers.
11799 (aarch64_secondary_reload): Handle big-endian reloads of SVE
11801 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
11802 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
11803 (aarch64_convert_sve_vector_bits): New function.
11804 (aarch64_override_options): Use it to handle -msve-vector-bits=.
11805 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
11806 rather than an rtx.
11807 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
11808 Handle SVE vector and predicate modes. Accept VL-based constants
11809 that need only one temporary register, and VL offsets that require
11810 no temporary registers.
11811 (aarch64_conditional_register_usage): Mark the predicate registers
11812 as fixed if SVE isn't available.
11813 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
11814 Return true for SVE vector and predicate modes.
11815 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
11816 rather than an unsigned int. Handle SVE modes.
11817 (aarch64_preferred_simd_mode): Update call accordingly. Handle
11819 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
11821 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
11822 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
11823 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
11824 (aarch64_sve_float_mul_immediate_p): New functions.
11825 (aarch64_sve_valid_immediate): New function.
11826 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
11827 Explicitly reject structure modes. Check for INDEX constants.
11828 Handle PTRUE and PFALSE constants.
11829 (aarch64_check_zero_based_sve_index_immediate): New function.
11830 (aarch64_simd_imm_zero_p): Delete.
11831 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
11832 vector modes. Accept constants in the range of CNT[BHWD].
11833 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
11834 ask for an Advanced SIMD mode.
11835 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
11836 (aarch64_simd_vector_alignment): Handle SVE predicates.
11837 (aarch64_vectorize_preferred_vector_alignment): New function.
11838 (aarch64_simd_vector_alignment_reachable): Use it instead of
11840 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
11841 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
11843 (MAX_VECT_LEN): Delete.
11844 (expand_vec_perm_d): Add a vec_flags field.
11845 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
11846 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
11847 (aarch64_evpc_ext): Don't apply a big-endian lane correction
11849 (aarch64_evpc_rev): Rename to...
11850 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
11851 (aarch64_evpc_rev_global): New function.
11852 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
11853 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
11855 (aarch64_evpc_sve_tbl): New function.
11856 (aarch64_expand_vec_perm_const_1): Update after rename of
11857 aarch64_evpc_rev. Handle SVE permutes too, trying
11858 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
11859 than aarch64_evpc_tbl.
11860 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
11861 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
11862 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
11863 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
11864 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
11865 (aarch64_expand_sve_vcond): New functions.
11866 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
11867 of aarch64_vector_mode_p.
11868 (aarch64_dwarf_poly_indeterminate_value): New function.
11869 (aarch64_compute_pressure_classes): Likewise.
11870 (aarch64_can_change_mode_class): Likewise.
11871 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
11872 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
11873 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
11874 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
11875 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
11876 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
11877 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
11878 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
11880 (Dn, Dl, Dr): Accept const as well as const_vector.
11881 (Dz): Likewise. Compare against CONST0_RTX.
11882 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
11883 of "vector" where appropriate.
11884 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
11885 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
11886 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
11887 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
11888 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
11889 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
11890 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
11891 (v_int_equiv): Extend to SVE modes.
11892 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
11894 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
11895 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
11896 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
11897 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
11898 (SVE_COND_FP_CMP): New int iterators.
11899 (perm_hilo): Handle the new unpack unspecs.
11900 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
11902 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
11903 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
11904 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
11905 (aarch64_equality_operator, aarch64_constant_vector_operand)
11906 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
11907 (aarch64_sve_nonimmediate_operand): Likewise.
11908 (aarch64_sve_general_operand): Likewise.
11909 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
11910 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
11911 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
11912 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
11913 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
11914 (aarch64_sve_float_arith_immediate): Likewise.
11915 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
11916 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
11917 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
11918 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
11919 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
11920 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
11921 (aarch64_sve_float_arith_operand): Likewise.
11922 (aarch64_sve_float_arith_with_sub_operand): Likewise.
11923 (aarch64_sve_float_mul_operand): Likewise.
11924 (aarch64_sve_vec_perm_operand): Likewise.
11925 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
11926 (aarch64_mov_operand): Accept const_poly_int and const_vector.
11927 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
11928 as well as const_vector.
11929 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
11930 in file. Use CONST0_RTX and CONSTM1_RTX.
11931 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
11932 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
11933 Use aarch64_simd_imm_zero.
11934 * config/aarch64/aarch64-sve.md: New file.
11935 * config/aarch64/aarch64.md: Include it.
11936 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
11937 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
11938 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
11939 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
11940 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
11941 (sve): New attribute.
11942 (enabled): Disable instructions with the sve attribute unless
11944 (movqi, movhi): Pass CONST_POLY_INT operaneds through
11945 aarch64_expand_mov_immediate.
11946 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
11947 CNT[BHSD] immediates.
11948 (movti): Split CONST_POLY_INT moves into two halves.
11949 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
11950 Split additions that need a temporary here if the destination
11951 is the stack pointer.
11952 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
11953 (*add<mode>3_poly_1): New instruction.
11954 (set_clobber_cc): New expander.
11956 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11958 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
11959 parameter and use it instead of GET_MODE_SIZE (innermode). Use
11960 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
11961 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
11962 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
11963 Change innermode from fixed_mode_size to machine_mode.
11964 (simplify_subreg): Update call accordingly. Handle a constant-sized
11965 subreg of a variable-length CONST_VECTOR.
11967 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
11968 Alan Hayward <alan.hayward@arm.com>
11969 David Sherwood <david.sherwood@arm.com>
11971 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
11972 (add_offset_to_base): New function, split out from...
11973 (create_mem_ref): ...here. When handling a scale other than 1,
11974 check first whether the address is valid without the offset.
11975 Add it into the base if so, leaving the index and scale as-is.
11977 2018-01-12 Jakub Jelinek <jakub@redhat.com>
11980 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
11981 fold_for_warn before checking if arg2 is INTEGER_CST.
11983 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
11985 * config/rs6000/predicates.md (load_multiple_operation): Delete.
11986 (store_multiple_operation): Delete.
11987 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
11988 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
11989 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
11990 guarded by TARGET_STRING.
11991 (rs6000_output_load_multiple): Delete.
11992 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
11993 OPTION_MASK_STRING / TARGET_STRING handling.
11994 (print_operand) <'N', 'O'>: Add comment that these are unused now.
11995 (const rs6000_opt_masks) <"string">: Change mask to 0.
11996 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
11997 (MASK_STRING): Delete.
11998 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
12000 (load_multiple): Delete.
12007 (store_multiple): Delete.
12014 (movmemsi_8reg): Delete.
12015 (corresponding unnamed define_insn): Delete.
12016 (movmemsi_6reg): Delete.
12017 (corresponding unnamed define_insn): Delete.
12018 (movmemsi_4reg): Delete.
12019 (corresponding unnamed define_insn): Delete.
12020 (movmemsi_2reg): Delete.
12021 (corresponding unnamed define_insn): Delete.
12022 (movmemsi_1reg): Delete.
12023 (corresponding unnamed define_insn): Delete.
12024 * config/rs6000/rs6000.opt (mno-string): New.
12025 (mstring): Replace by deprecation warning stub.
12026 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
12028 2018-01-12 Jakub Jelinek <jakub@redhat.com>
12030 * regrename.c (regrename_do_replace): If replacing the same
12031 reg multiple times, try to reuse last created gen_raw_REG.
12034 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
12035 main to workaround a bug in GDB.
12037 2018-01-12 Tom de Vries <tom@codesourcery.com>
12040 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
12042 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
12044 PR rtl-optimization/80481
12045 * ira-color.c (get_cap_member): New function.
12046 (allocnos_conflict_by_live_ranges_p): Use it.
12047 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
12048 (setup_slot_coalesced_allocno_live_ranges): Ditto.
12050 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
12053 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
12054 (*saddl_se_1): Ditto.
12055 (*ssubsi_1): Ditto.
12056 (*ssubl_se_1): Ditto.
12058 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12060 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
12061 rather than wi::to_widest for DR_INITs.
12062 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
12063 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
12064 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
12066 (vect_analyze_group_access_1): Note that here.
12068 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12070 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
12071 polynomial type sizes.
12073 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
12075 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
12076 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
12077 (gimple_add_tmp_var): Likewise.
12079 2018-01-12 Martin Liska <mliska@suse.cz>
12081 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
12082 (gimple_alloc_sizes): Likewise.
12083 (dump_gimple_statistics): Use PRIu64 in printf format.
12084 * gimple.h: Change uint64_t to int.
12086 2018-01-12 Martin Liska <mliska@suse.cz>
12088 * tree-core.h: Use uint64_t instead of int.
12089 * tree.c (tree_node_counts): Likewise.
12090 (tree_node_sizes): Likewise.
12091 (dump_tree_statistics): Use PRIu64 in printf format.
12093 2018-01-12 Martin Liska <mliska@suse.cz>
12095 * Makefile.in: As qsort_chk is implemented in vec.c, add
12096 vec.o to linkage of gencfn-macros.
12097 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
12098 passing the info to record_node_allocation_statistics.
12099 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
12101 * ggc-common.c (struct ggc_usage): Add operator== and use
12102 it in operator< and compare function.
12103 * mem-stats.h (struct mem_usage): Likewise.
12104 * vec.c (struct vec_usage): Remove operator< and compare
12105 function. Can be simply inherited.
12107 2018-01-12 Martin Jambor <mjambor@suse.cz>
12110 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
12111 * tree-ssa-math-opts.c: Include domwalk.h.
12112 (convert_mult_to_fma_1): New function.
12113 (fma_transformation_info): New type.
12114 (fma_deferring_state): Likewise.
12115 (cancel_fma_deferring): New function.
12116 (result_of_phi): Likewise.
12117 (last_fma_candidate_feeds_initial_phi): Likewise.
12118 (convert_mult_to_fma): Added deferring logic, split actual
12119 transformation to convert_mult_to_fma_1.
12120 (math_opts_dom_walker): New type.
12121 (math_opts_dom_walker::after_dom_children): New method, body moved
12122 here from pass_optimize_widening_mul::execute, added deferring logic
12124 (pass_optimize_widening_mul::execute): Moved most of code to
12125 math_opts_dom_walker::after_dom_children.
12126 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
12127 * config/i386/i386.c (ix86_option_override_internal): Added
12128 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
12130 2018-01-12 Richard Biener <rguenther@suse.de>
12133 * dwarf2out.c (gen_variable_die): Do not reset old_die for
12134 inline instance vars.
12136 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
12139 * config/rx/rx.c (rx_is_restricted_memory_address):
12140 Handle SUBREG case.
12142 2018-01-12 Richard Biener <rguenther@suse.de>
12144 PR tree-optimization/80846
12145 * target.def (split_reduction): New target hook.
12146 * targhooks.c (default_split_reduction): New function.
12147 * targhooks.h (default_split_reduction): Declare.
12148 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
12149 target requests first reduce vectors by combining low and high
12151 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
12152 (get_vectype_for_scalar_type_and_size): Export.
12153 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
12154 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
12155 * doc/tm.texi: Regenerate.
12156 * config/i386/i386.c (ix86_split_reduction): Implement
12157 TARGET_VECTORIZE_SPLIT_REDUCTION.
12159 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12162 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
12163 in PIC mode except for TARGET_VXWORKS_RTP.
12164 * config/sparc/sparc.c: Include cfgrtl.h.
12165 (TARGET_INIT_PIC_REG): Define.
12166 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
12167 (sparc_pic_register_p): New predicate.
12168 (sparc_legitimate_address_p): Use it.
12169 (sparc_legitimize_pic_address): Likewise.
12170 (sparc_delegitimize_address): Likewise.
12171 (sparc_mode_dependent_address_p): Likewise.
12172 (gen_load_pcrel_sym): Remove 4th parameter.
12173 (load_got_register): Adjust call to above. Remove obsolete stuff.
12174 (sparc_expand_prologue): Do not call load_got_register here.
12175 (sparc_flat_expand_prologue): Likewise.
12176 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
12177 (sparc_use_pseudo_pic_reg): New function.
12178 (sparc_init_pic_reg): Likewise.
12179 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
12180 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
12182 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
12184 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
12185 Add item for branch_cost.
12187 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
12189 PR rtl-optimization/83565
12190 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
12191 not extend the result to a larger mode for rotate operations.
12192 (num_sign_bit_copies1): Likewise.
12194 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
12197 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
12199 Use values-Xc.o for -pedantic.
12200 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
12202 2018-01-12 Martin Liska <mliska@suse.cz>
12205 * ipa-devirt.c (final_warning_record::grow_type_warnings):
12207 (possible_polymorphic_call_targets): Use it.
12208 (ipa_devirt): Likewise.
12210 2018-01-12 Martin Liska <mliska@suse.cz>
12212 * profile-count.h (enum profile_quality): Use 0 as invalid
12213 enum value of profile_quality.
12215 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
12217 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
12218 -mext-string options.
12220 2018-01-12 Richard Biener <rguenther@suse.de>
12222 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
12223 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
12224 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
12226 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
12228 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
12230 * configure.ac (--with-long-double-format): Add support for the
12231 configuration option to change the default long double format on
12233 * config.gcc (powerpc*-linux*-*): Likewise.
12234 * configure: Regenerate.
12235 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
12236 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
12237 used without modification.
12239 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12241 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
12242 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
12243 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
12244 MISC_BUILTIN_SPEC_BARRIER.
12245 (rs6000_init_builtins): Likewise.
12246 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
12248 (speculation_barrier): New define_insn.
12249 * doc/extend.texi: Document __builtin_speculation_barrier.
12251 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12254 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
12255 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
12256 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
12258 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
12259 integral modes instead of "ss" and "sd".
12260 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
12261 vectors with 32-bit and 64-bit elements.
12262 (vecdupssescalarmodesuffix): New mode attribute.
12263 (vec_dup<mode>): Use it.
12265 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
12268 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
12269 frame if argument is passed on stack.
12271 2018-01-11 Jakub Jelinek <jakub@redhat.com>
12274 * ree.c (combine_reaching_defs): Optimize also
12275 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
12276 reg2=any_extend(exp); reg1=reg2;, formatting fix.
12278 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12280 PR middle-end/83189
12281 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
12283 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
12285 PR middle-end/83718
12286 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
12287 after they are computed.
12289 2018-01-11 Bin Cheng <bin.cheng@arm.com>
12291 PR tree-optimization/83695
12292 * gimple-loop-linterchange.cc
12293 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
12294 reset cached scev information after interchange.
12295 (pass_linterchange::execute): Remove call to scev_reset_htab.
12297 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12299 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
12300 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
12301 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
12302 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
12303 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
12304 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
12305 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
12306 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
12307 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
12308 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
12309 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
12310 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
12311 (V_lane_reg): Likewise.
12312 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
12314 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
12315 (vfmal_lane_low<mode>_intrinsic,
12316 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
12317 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
12318 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
12319 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
12320 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
12321 vfmsl_lane_high<mode>_intrinsic): New define_insns.
12323 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12325 * config/arm/arm-cpus.in (fp16fml): New feature.
12326 (ALL_SIMD): Add fp16fml.
12327 (armv8.2-a): Add fp16fml as an option.
12328 (armv8.3-a): Likewise.
12329 (armv8.4-a): Add fp16fml as part of fp16.
12330 * config/arm/arm.h (TARGET_FP16FML): Define.
12331 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
12333 * config/arm/arm-modes.def (V2HF): Define.
12334 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
12335 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
12336 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
12337 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
12338 vfmsl_low, vfmsl_high): New set of builtins.
12339 * config/arm/iterators.md (PLUSMINUS): New code iterator.
12340 (vfml_op): New code attribute.
12341 (VFMLHALVES): New int iterator.
12342 (VFML, VFMLSEL): New mode attributes.
12343 (V_reg): Define mapping for V2HF.
12344 (V_hi, V_lo): New mode attributes.
12345 (VF_constraint): Likewise.
12346 (vfml_half, vfml_half_selector): New int attributes.
12347 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
12349 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
12350 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
12352 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
12353 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
12354 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
12355 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
12357 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
12358 Document new effective target and option set.
12360 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12362 * config/arm/arm-cpus.in (armv8_4): New feature.
12363 (ARMv8_4a): New fgroup.
12364 (armv8.4-a): New arch.
12365 * config/arm/arm-tables.opt: Regenerate.
12366 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
12367 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
12368 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
12369 Add matching rules for -march=armv8.4-a and extensions.
12370 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
12372 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
12375 * config/rx/rx.md (BW): New mode attribute.
12376 (sync_lock_test_and_setsi): Add mode suffix to insn output.
12378 2018-01-11 Richard Biener <rguenther@suse.de>
12380 PR tree-optimization/83435
12381 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
12382 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
12383 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
12385 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12386 Alan Hayward <alan.hayward@arm.com>
12387 David Sherwood <david.sherwood@arm.com>
12389 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
12391 (aarch64_classify_address): Initialize it. Track polynomial offsets.
12392 (aarch64_print_address_internal): Use it to check for a zero offset.
12394 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12395 Alan Hayward <alan.hayward@arm.com>
12396 David Sherwood <david.sherwood@arm.com>
12398 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
12399 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
12400 Return a poly_int64 rather than a HOST_WIDE_INT.
12401 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
12402 rather than a HOST_WIDE_INT.
12403 * config/aarch64/aarch64.h (aarch64_frame): Protect with
12404 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
12405 hard_fp_offset, frame_size, initial_adjust, callee_offset and
12406 final_offset from HOST_WIDE_INT to poly_int64.
12407 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
12408 to_constant when getting the number of units in an Advanced SIMD
12410 (aarch64_builtin_vectorized_function): Check for a constant number
12412 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
12414 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
12415 attribute instead of GET_MODE_NUNITS.
12416 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
12417 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
12418 GET_MODE_SIZE for fixed-size registers.
12419 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
12420 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
12421 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
12422 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
12423 (aarch64_print_operand, aarch64_print_address_internal)
12424 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
12425 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
12426 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
12427 Handle polynomial GET_MODE_SIZE.
12428 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
12429 wider than SImode without modification.
12430 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
12431 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
12432 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
12433 passing and returning SVE modes.
12434 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
12435 rather than GEN_INT.
12436 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
12437 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
12438 (aarch64_allocate_and_probe_stack_space): Likewise.
12439 (aarch64_layout_frame): Cope with polynomial offsets.
12440 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
12441 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
12442 polynomial offsets.
12443 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
12444 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
12445 poly_int64 rather than a HOST_WIDE_INT.
12446 (aarch64_get_separate_components, aarch64_process_components)
12447 (aarch64_expand_prologue, aarch64_expand_epilogue)
12448 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
12449 (aarch64_anchor_offset): New function, split out from...
12450 (aarch64_legitimize_address): ...here.
12451 (aarch64_builtin_vectorization_cost): Handle polynomial
12452 TYPE_VECTOR_SUBPARTS.
12453 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
12455 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
12456 number of elements from the PARALLEL rather than the mode.
12457 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
12458 rather than GET_MODE_BITSIZE.
12459 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
12460 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
12461 (aarch64_expand_vec_perm_const_1): Handle polynomial
12462 d->perm.length () and d->perm elements.
12463 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
12464 Apply to_constant to d->perm elements.
12465 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
12466 polynomial CONST_VECTOR_NUNITS.
12467 (aarch64_move_pointer): Take amount as a poly_int64 rather
12469 (aarch64_progress_pointer): Avoid temporary variable.
12470 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
12471 the mode attribute instead of GET_MODE.
12473 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12474 Alan Hayward <alan.hayward@arm.com>
12475 David Sherwood <david.sherwood@arm.com>
12477 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
12478 x exists before using it.
12479 (aarch64_add_constant_internal): Rename to...
12480 (aarch64_add_offset_1): ...this. Replace regnum with separate
12481 src and dest rtxes. Handle the case in which they're different,
12482 including when the offset is zero. Replace scratchreg with an rtx.
12483 Use 2 additions if there is no spare register into which we can
12484 move a 16-bit constant.
12485 (aarch64_add_constant): Delete.
12486 (aarch64_add_offset): Replace reg with separate src and dest
12487 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
12488 Use aarch64_add_offset_1.
12489 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
12490 an rtx rather than an int. Take the delta as a poly_int64
12491 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
12492 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
12493 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
12494 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
12495 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
12496 and aarch64_add_sp.
12497 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
12498 aarch64_add_constant.
12500 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12502 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
12503 Use scalar_float_mode.
12505 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12507 * config/aarch64/aarch64-simd.md
12508 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
12509 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
12510 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
12511 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
12512 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
12513 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
12514 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
12515 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
12516 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
12517 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
12519 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12522 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
12523 targ_options->x_arm_arch_string is non NULL.
12525 2018-01-11 Tamar Christina <tamar.christina@arm.com>
12527 * config/aarch64/aarch64.h
12528 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
12530 2018-01-11 Sudakshina Das <sudi.das@arm.com>
12533 * expmed.c (emit_store_flag_force): Swap if const op0
12534 and change VOIDmode to mode of op0.
12536 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
12538 PR rtl-optimization/83761
12539 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
12540 than bytes to mode_for_size.
12542 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12544 PR middle-end/83189
12545 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
12546 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
12549 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12551 PR middle-end/83575
12552 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
12553 when in layout mode.
12554 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
12555 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
12558 2018-01-10 Michael Collison <michael.collison@arm.com>
12560 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
12561 * config/aarch64/aarch64-option-extension.def: Add
12562 AARCH64_OPT_EXTENSION of 'fp16fml'.
12563 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12564 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
12565 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
12566 * config/aarch64/constraints.md (Ui7): New constraint.
12567 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
12568 (VFMLA_SEL_W): Ditto.
12571 (VFMLA16_LOW): New int iterator.
12572 (VFMLA16_HIGH): Ditto.
12573 (UNSPEC_FMLAL): New unspec.
12574 (UNSPEC_FMLSL): Ditto.
12575 (UNSPEC_FMLAL2): Ditto.
12576 (UNSPEC_FMLSL2): Ditto.
12577 (f16mac): New code attribute.
12578 * config/aarch64/aarch64-simd-builtins.def
12579 (aarch64_fmlal_lowv2sf): Ditto.
12580 (aarch64_fmlsl_lowv2sf): Ditto.
12581 (aarch64_fmlalq_lowv4sf): Ditto.
12582 (aarch64_fmlslq_lowv4sf): Ditto.
12583 (aarch64_fmlal_highv2sf): Ditto.
12584 (aarch64_fmlsl_highv2sf): Ditto.
12585 (aarch64_fmlalq_highv4sf): Ditto.
12586 (aarch64_fmlslq_highv4sf): Ditto.
12587 (aarch64_fmlal_lane_lowv2sf): Ditto.
12588 (aarch64_fmlsl_lane_lowv2sf): Ditto.
12589 (aarch64_fmlal_laneq_lowv2sf): Ditto.
12590 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
12591 (aarch64_fmlalq_lane_lowv4sf): Ditto.
12592 (aarch64_fmlsl_lane_lowv4sf): Ditto.
12593 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
12594 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
12595 (aarch64_fmlal_lane_highv2sf): Ditto.
12596 (aarch64_fmlsl_lane_highv2sf): Ditto.
12597 (aarch64_fmlal_laneq_highv2sf): Ditto.
12598 (aarch64_fmlsl_laneq_highv2sf): Ditto.
12599 (aarch64_fmlalq_lane_highv4sf): Ditto.
12600 (aarch64_fmlsl_lane_highv4sf): Ditto.
12601 (aarch64_fmlalq_laneq_highv4sf): Ditto.
12602 (aarch64_fmlsl_laneq_highv4sf): Ditto.
12603 * config/aarch64/aarch64-simd.md:
12604 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
12605 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12606 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
12607 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
12608 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
12609 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
12610 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
12611 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
12612 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
12613 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
12614 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
12615 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
12616 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
12617 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
12618 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
12619 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
12620 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
12621 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
12622 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
12623 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
12624 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
12625 (vfmlsl_low_u32): Ditto.
12626 (vfmlalq_low_u32): Ditto.
12627 (vfmlslq_low_u32): Ditto.
12628 (vfmlal_high_u32): Ditto.
12629 (vfmlsl_high_u32): Ditto.
12630 (vfmlalq_high_u32): Ditto.
12631 (vfmlslq_high_u32): Ditto.
12632 (vfmlal_lane_low_u32): Ditto.
12633 (vfmlsl_lane_low_u32): Ditto.
12634 (vfmlal_laneq_low_u32): Ditto.
12635 (vfmlsl_laneq_low_u32): Ditto.
12636 (vfmlalq_lane_low_u32): Ditto.
12637 (vfmlslq_lane_low_u32): Ditto.
12638 (vfmlalq_laneq_low_u32): Ditto.
12639 (vfmlslq_laneq_low_u32): Ditto.
12640 (vfmlal_lane_high_u32): Ditto.
12641 (vfmlsl_lane_high_u32): Ditto.
12642 (vfmlal_laneq_high_u32): Ditto.
12643 (vfmlsl_laneq_high_u32): Ditto.
12644 (vfmlalq_lane_high_u32): Ditto.
12645 (vfmlslq_lane_high_u32): Ditto.
12646 (vfmlalq_laneq_high_u32): Ditto.
12647 (vfmlslq_laneq_high_u32): Ditto.
12648 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
12649 (AARCH64_FL_FOR_ARCH8_4): New.
12650 (AARCH64_ISA_F16FML): New ISA flag.
12651 (TARGET_F16FML): New feature flag for fp16fml.
12652 (doc/invoke.texi): Document new fp16fml option.
12654 2018-01-10 Michael Collison <michael.collison@arm.com>
12656 * config/aarch64/aarch64-builtins.c:
12657 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
12658 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12659 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
12660 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
12661 (AARCH64_ISA_SHA3): New ISA flag.
12662 (TARGET_SHA3): New feature flag for sha3.
12663 * config/aarch64/iterators.md (sha512_op): New int attribute.
12664 (CRYPTO_SHA512): New int iterator.
12665 (UNSPEC_SHA512H): New unspec.
12666 (UNSPEC_SHA512H2): Ditto.
12667 (UNSPEC_SHA512SU0): Ditto.
12668 (UNSPEC_SHA512SU1): Ditto.
12669 * config/aarch64/aarch64-simd-builtins.def
12670 (aarch64_crypto_sha512hqv2di): New builtin.
12671 (aarch64_crypto_sha512h2qv2di): Ditto.
12672 (aarch64_crypto_sha512su0qv2di): Ditto.
12673 (aarch64_crypto_sha512su1qv2di): Ditto.
12674 (aarch64_eor3qv8hi): Ditto.
12675 (aarch64_rax1qv2di): Ditto.
12676 (aarch64_xarqv2di): Ditto.
12677 (aarch64_bcaxqv8hi): Ditto.
12678 * config/aarch64/aarch64-simd.md:
12679 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
12680 (aarch64_crypto_sha512su0qv2di): Ditto.
12681 (aarch64_crypto_sha512su1qv2di): Ditto.
12682 (aarch64_eor3qv8hi): Ditto.
12683 (aarch64_rax1qv2di): Ditto.
12684 (aarch64_xarqv2di): Ditto.
12685 (aarch64_bcaxqv8hi): Ditto.
12686 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
12687 (vsha512h2q_u64): Ditto.
12688 (vsha512su0q_u64): Ditto.
12689 (vsha512su1q_u64): Ditto.
12690 (veor3q_u16): Ditto.
12691 (vrax1q_u64): Ditto.
12692 (vxarq_u64): Ditto.
12693 (vbcaxq_u16): Ditto.
12694 * config/arm/types.md (crypto_sha512): New type attribute.
12695 (crypto_sha3): Ditto.
12696 (doc/invoke.texi): Document new sha3 option.
12698 2018-01-10 Michael Collison <michael.collison@arm.com>
12700 * config/aarch64/aarch64-builtins.c:
12701 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
12702 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12703 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
12704 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
12705 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
12706 (AARCH64_ISA_SM4): New ISA flag.
12707 (TARGET_SM4): New feature flag for sm4.
12708 * config/aarch64/aarch64-simd-builtins.def
12709 (aarch64_sm3ss1qv4si): Ditto.
12710 (aarch64_sm3tt1aq4si): Ditto.
12711 (aarch64_sm3tt1bq4si): Ditto.
12712 (aarch64_sm3tt2aq4si): Ditto.
12713 (aarch64_sm3tt2bq4si): Ditto.
12714 (aarch64_sm3partw1qv4si): Ditto.
12715 (aarch64_sm3partw2qv4si): Ditto.
12716 (aarch64_sm4eqv4si): Ditto.
12717 (aarch64_sm4ekeyqv4si): Ditto.
12718 * config/aarch64/aarch64-simd.md:
12719 (aarch64_sm3ss1qv4si): Ditto.
12720 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
12721 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
12722 (aarch64_sm4eqv4si): Ditto.
12723 (aarch64_sm4ekeyqv4si): Ditto.
12724 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
12725 (sm3part_op): Ditto.
12726 (CRYPTO_SM3TT): Ditto.
12727 (CRYPTO_SM3PART): Ditto.
12728 (UNSPEC_SM3SS1): New unspec.
12729 (UNSPEC_SM3TT1A): Ditto.
12730 (UNSPEC_SM3TT1B): Ditto.
12731 (UNSPEC_SM3TT2A): Ditto.
12732 (UNSPEC_SM3TT2B): Ditto.
12733 (UNSPEC_SM3PARTW1): Ditto.
12734 (UNSPEC_SM3PARTW2): Ditto.
12735 (UNSPEC_SM4E): Ditto.
12736 (UNSPEC_SM4EKEY): Ditto.
12737 * config/aarch64/constraints.md (Ui2): New constraint.
12738 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
12739 * config/arm/types.md (crypto_sm3): New type attribute.
12740 (crypto_sm4): Ditto.
12741 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
12742 (vsm3tt1aq_u32): Ditto.
12743 (vsm3tt1bq_u32): Ditto.
12744 (vsm3tt2aq_u32): Ditto.
12745 (vsm3tt2bq_u32): Ditto.
12746 (vsm3partw1q_u32): Ditto.
12747 (vsm3partw2q_u32): Ditto.
12748 (vsm4eq_u32): Ditto.
12749 (vsm4ekeyq_u32): Ditto.
12750 (doc/invoke.texi): Document new sm4 option.
12752 2018-01-10 Michael Collison <michael.collison@arm.com>
12754 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
12755 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
12756 (AARCH64_FL_FOR_ARCH8_4): New.
12757 (AARCH64_FL_V8_4): New flag.
12758 (doc/invoke.texi): Document new armv8.4-a option.
12760 2018-01-10 Michael Collison <michael.collison@arm.com>
12762 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12763 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
12764 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
12765 * config/aarch64/aarch64-option-extension.def: Add
12766 AARCH64_OPT_EXTENSION of 'sha2'.
12767 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
12768 (crypto): Disable sha2 and aes if crypto disabled.
12769 (crypto): Enable aes and sha2 if enabled.
12770 (simd): Disable sha2 and aes if simd disabled.
12771 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
12773 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
12774 (TARGET_SHA2): New feature flag for sha2.
12775 (TARGET_AES): New feature flag for aes.
12776 * config/aarch64/aarch64-simd.md:
12777 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
12778 conditional on TARGET_AES.
12779 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
12780 (aarch64_crypto_sha1hsi): Make pattern conditional
12782 (aarch64_crypto_sha1hv4si): Ditto.
12783 (aarch64_be_crypto_sha1hv4si): Ditto.
12784 (aarch64_crypto_sha1su1v4si): Ditto.
12785 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
12786 (aarch64_crypto_sha1su0v4si): Ditto.
12787 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
12788 (aarch64_crypto_sha256su0v4si): Ditto.
12789 (aarch64_crypto_sha256su1v4si): Ditto.
12790 (doc/invoke.texi): Document new aes and sha2 options.
12792 2018-01-10 Martin Sebor <msebor@redhat.com>
12794 PR tree-optimization/83781
12795 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
12798 2018-01-11 Martin Sebor <msebor@gmail.com>
12799 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
12801 PR tree-optimization/83501
12802 PR tree-optimization/81703
12804 * tree-ssa-strlen.c (get_string_cst): Rename...
12805 (get_string_len): ...to this. Handle global constants.
12806 (handle_char_store): Adjust.
12808 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
12809 Jim Wilson <jimw@sifive.com>
12811 * config/riscv/riscv-protos.h (riscv_output_return): New.
12812 * config/riscv/riscv.c (struct machine_function): New naked_p field.
12813 (riscv_attribute_table, riscv_output_return),
12814 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
12815 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
12816 (riscv_compute_frame_info): Only compute frame->mask if not a naked
12818 (riscv_expand_prologue): Add early return for naked function.
12819 (riscv_expand_epilogue): Likewise.
12820 (riscv_function_ok_for_sibcall): Return false for naked function.
12821 (riscv_set_current_function): New.
12822 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
12823 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
12824 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
12825 * doc/extend.texi (RISC-V Function Attributes): New.
12827 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
12829 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
12830 check for 128-bit long double before checking TCmode.
12831 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
12832 128-bit long doubles before checking TFmode or TCmode.
12833 (FLOAT128_IBM_P): Likewise.
12835 2018-01-10 Martin Sebor <msebor@redhat.com>
12837 PR tree-optimization/83671
12838 * builtins.c (c_strlen): Unconditionally return zero for the empty
12840 Use -Warray-bounds for warnings.
12841 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
12842 for non-constant array indices with COMPONENT_REF, arrays of
12843 arrays, and pointers to arrays.
12844 (gimple_fold_builtin_strlen): Determine and set length range for
12845 non-constant character arrays.
12847 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
12849 PR middle-end/81897
12850 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
12853 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
12855 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
12857 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12860 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
12861 VECTOR_MEM_ALTIVEC_OR_VSX_P.
12862 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
12863 indexed_or_indirect_operand predicate.
12864 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
12865 (*vsx_le_perm_load_v8hi): Likewise.
12866 (*vsx_le_perm_load_v16qi): Likewise.
12867 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
12868 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
12869 (*vsx_le_perm_store_v8hi): Likewise.
12870 (*vsx_le_perm_store_v16qi): Likewise.
12871 (eight unnamed splitters): Likewise.
12873 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
12875 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
12876 * config/rs6000/emmintrin.h: Likewise.
12877 * config/rs6000/mmintrin.h: Likewise.
12878 * config/rs6000/xmmintrin.h: Likewise.
12880 2018-01-10 David Malcolm <dmalcolm@redhat.com>
12883 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
12885 * tree.c (tree_nop_conversion): Return true for location wrapper
12887 (maybe_wrap_with_location): New function.
12888 (selftest::check_strip_nops): New function.
12889 (selftest::test_location_wrappers): New function.
12890 (selftest::tree_c_tests): Call it.
12891 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
12892 (maybe_wrap_with_location): New decl.
12893 (EXPR_LOCATION_WRAPPER_P): New macro.
12894 (location_wrapper_p): New inline function.
12895 (tree_strip_any_location_wrapper): New inline function.
12897 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
12900 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
12901 stack_realign_offset for the largest alignment of stack slot
12903 (ix86_find_max_used_stack_alignment): New function.
12904 (ix86_finalize_stack_frame_flags): Use it. Set
12905 max_used_stack_alignment if we don't realign stack.
12906 * config/i386/i386.h (machine_function): Add
12907 max_used_stack_alignment.
12909 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
12911 * config/arm/arm.opt (-mbranch-cost): New option.
12912 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
12915 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
12918 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
12919 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
12921 2018-01-10 Richard Biener <rguenther@suse.de>
12924 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
12925 early out so it also covers the case where we have a non-NULL
12928 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12930 PR tree-optimization/83753
12931 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
12932 for non-strided grouped accesses if the number of elements is 1.
12934 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
12937 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
12938 * i386.h (TARGET_USE_GATHER): Define.
12939 * x86-tune.def (X86_TUNE_USE_GATHER): New.
12941 2018-01-10 Martin Liska <mliska@suse.cz>
12944 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
12945 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
12947 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
12948 CLEANUP_NO_PARTITIONING is not set.
12950 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
12952 * doc/rtl.texi: Remove documentation of (const ...) wrappers
12953 for vectors, as a partial revert of r254296.
12954 * rtl.h (const_vec_p): Delete.
12955 (const_vec_duplicate_p): Don't test for vector CONSTs.
12956 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
12957 * expmed.c (make_tree): Likewise.
12960 * common.md (E, F): Use CONSTANT_P instead of checking for
12962 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
12963 checking for CONST_VECTOR.
12965 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
12967 PR middle-end/83575
12968 * predict.c (force_edge_cold): Handle in more sane way edges
12969 with no prediction.
12971 2018-01-09 Carl Love <cel@us.ibm.com>
12973 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
12975 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
12976 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
12977 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
12978 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
12979 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
12980 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
12981 * config/rs6000/rs6000-protos.h: Add extern defition for
12982 rs6000_generate_float2_double_code.
12983 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
12985 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
12986 (float2_v2df): Add define_expand.
12988 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
12991 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
12992 op_mode in the force_to_mode call.
12994 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
12996 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
12997 instead of checking each element individually.
12998 (aarch64_evpc_uzp): Likewise.
12999 (aarch64_evpc_zip): Likewise.
13000 (aarch64_evpc_ext): Likewise.
13001 (aarch64_evpc_rev): Likewise.
13002 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
13003 instead of checking each element individually. Return true without
13005 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
13006 whether all selected elements come from the same input, instead of
13007 checking each element individually. Remove calls to gen_rtx_REG,
13008 start_sequence and end_sequence and instead assert that no rtl is
13011 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13013 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
13014 order of HIGH and CONST checks.
13016 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
13018 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
13019 if the destination isn't an SSA_NAME.
13021 2018-01-09 Richard Biener <rguenther@suse.de>
13023 PR tree-optimization/83668
13024 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
13026 (canonicalize_loop_form): ... here, renamed from ...
13027 (canonicalize_loop_closed_ssa_form): ... this and amended to
13028 swap successor edges for loop exit blocks to make us use
13029 the RPO order we need for initial schedule generation.
13031 2018-01-09 Joseph Myers <joseph@codesourcery.com>
13033 PR tree-optimization/64811
13034 * match.pd: When optimizing comparisons with Inf, avoid
13035 introducing or losing exceptions from comparisons with NaN.
13037 2018-01-09 Martin Liska <mliska@suse.cz>
13040 * asan.c (shadow_mem_size): Add gcc_assert.
13042 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
13044 Don't save registers in main().
13047 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
13048 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
13049 * config/avr/avr.c (avr_set_current_function): Don't error if
13050 naked, OS_task or OS_main are specified at the same time.
13051 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
13053 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
13055 * common/config/avr/avr-common.c (avr_option_optimization_table):
13056 Switch on -mmain-is-OS_task for optimizing compilations.
13058 2018-01-09 Richard Biener <rguenther@suse.de>
13060 PR tree-optimization/83572
13061 * graphite.c: Include cfganal.h.
13062 (graphite_transform_loops): Connect infinite loops to exit
13063 and remove fake edges at the end.
13065 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13067 * ipa-inline.c (edge_badness): Revert accidental checkin.
13069 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
13072 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
13073 symbols; not inline clones.
13075 2018-01-09 Jakub Jelinek <jakub@redhat.com>
13078 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
13079 hard registers. Formatting fixes.
13081 PR preprocessor/83722
13082 * gcc.c (try_generate_repro): Pass
13083 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
13084 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
13087 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
13088 Kito Cheng <kito.cheng@gmail.com>
13090 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
13091 (riscv_leaf_function_p): Delete.
13092 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
13094 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13096 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
13098 (do_ifelse): New function.
13099 (do_isel): New function.
13100 (do_sub3): New function.
13101 (do_add3): New function.
13102 (do_load_mask_compare): New function.
13103 (do_overlap_load_compare): New function.
13104 (expand_compare_loop): New function.
13105 (expand_block_compare): Call expand_compare_loop() when appropriate.
13106 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
13107 option description.
13108 (-mblock-compare-inline-loop-limit): New option.
13110 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13113 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
13114 Reverse order of second and third operands in first alternative.
13115 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
13116 of first and second elements in UNSPEC_VPERMR vector.
13117 (altivec_expand_vec_perm_le): Likewise.
13119 2018-01-08 Jeff Law <law@redhat.com>
13121 PR rtl-optimizatin/81308
13122 * tree-switch-conversion.c (cfg_altered): New file scoped static.
13123 (process_switch): If group_case_labels makes a change, then set
13125 (pass_convert_switch::execute): If a switch is converted, then
13126 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
13128 PR rtl-optimization/81308
13129 * recog.c (split_all_insns): Conditionally cleanup the CFG after
13132 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
13134 PR target/83663 - Revert r255946
13135 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
13136 generation for cases where splatting a value is not useful.
13137 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
13138 across a vec_duplicate and a paradoxical subreg forming a vector
13139 mode to a vec_concat.
13141 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13143 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
13144 -march=armv8.3-a variants.
13145 * config/arm/t-multilib: Likewise.
13146 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
13148 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
13150 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
13152 (cceq_ior_compare_complement): Give it a name so I can use it, and
13153 change boolean_or_operator predicate to boolean_operator so it can
13154 be used to generate a crand.
13155 (eqne): New code iterator.
13156 (bd/bd_neg): New code_attrs.
13157 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
13158 a single define_insn.
13159 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
13160 decrement (bdnzt/bdnzf/bdzt/bdzf).
13161 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
13162 with the new names of the branch decrement patterns, and added the
13163 names of the branch decrement conditional patterns.
13165 2018-01-08 Richard Biener <rguenther@suse.de>
13167 PR tree-optimization/83563
13168 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
13171 2018-01-08 Richard Biener <rguenther@suse.de>
13173 PR middle-end/83713
13174 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
13176 2018-01-08 Richard Biener <rguenther@suse.de>
13178 PR tree-optimization/83685
13179 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
13180 references to abnormals.
13182 2018-01-08 Richard Biener <rguenther@suse.de>
13185 * dwarf2out.c (output_indirect_strings): Handle empty
13186 skeleton_debug_str_hash.
13187 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
13189 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13191 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
13192 (emit_store_direct): Likewise.
13193 (arc_trampoline_adjust_address): Likewise.
13194 (arc_asm_trampoline_template): New function.
13195 (arc_initialize_trampoline): Use asm_trampoline_template.
13196 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
13197 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
13198 * config/arc/arc.md (flush_icache): Delete pattern.
13200 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
13202 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
13203 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
13206 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13209 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
13210 by not USED_FOR_TARGET.
13211 (make_pass_resolve_sw_modes): Likewise.
13213 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
13215 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
13218 2018-01-08 Richard Biener <rguenther@suse.de>
13220 PR middle-end/83580
13221 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
13223 2018-01-08 Richard Biener <rguenther@suse.de>
13225 PR middle-end/83517
13226 * match.pd ((t * 2) / 2) -> t): Add missing :c.
13228 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
13230 PR middle-end/81897
13231 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
13232 basic blocks with a small number of successors.
13233 (convert_control_dep_chain_into_preds): Improve handling of
13235 (dump_predicates): Split apart into...
13236 (dump_pred_chain): ...here...
13237 (dump_pred_info): ...and here.
13238 (can_one_predicate_be_invalidated_p): Add debugging printfs.
13239 (can_chain_union_be_invalidated_p): Improve check for invalidation
13241 (uninit_uses_cannot_happen): Avoid unnecessary if
13242 convert_control_dep_chain_into_preds yielded nothing.
13244 2018-01-06 Martin Sebor <msebor@redhat.com>
13246 PR tree-optimization/83640
13247 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
13248 subtracting negative offset from size.
13249 (builtin_access::overlap): Adjust offset bounds of the access to fall
13250 within the size of the object if possible.
13252 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
13254 PR rtl-optimization/83699
13255 * expmed.c (extract_bit_field_1): Restrict the vector usage of
13256 extract_bit_field_as_subreg to cases in which the extracted
13257 value is also a vector.
13259 * lra-constraints.c (process_alt_operands): Test for the equivalence
13260 substitutions when detecting a possible reload cycle.
13262 2018-01-06 Jakub Jelinek <jakub@redhat.com>
13265 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
13266 by default if flag_selective_schedling{,2}. Formatting fixes.
13268 PR rtl-optimization/83682
13269 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
13270 if it has non-VECTOR_MODE element mode.
13271 (vec_duplicate_p): Likewise.
13273 PR middle-end/83694
13274 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
13275 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
13277 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13280 * config/i386/i386-builtin.def
13281 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
13282 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
13283 Require also OPTION_MASK_ISA_AVX512F in addition to
13284 OPTION_MASK_ISA_GFNI.
13285 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
13286 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
13287 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
13288 to OPTION_MASK_ISA_GFNI.
13289 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
13290 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
13291 OPTION_MASK_ISA_AVX512BW.
13292 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
13293 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
13294 addition to OPTION_MASK_ISA_GFNI.
13295 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
13296 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
13297 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
13298 to OPTION_MASK_ISA_GFNI.
13299 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
13300 a requirement for all ISAs rather than any of them with a few
13302 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
13304 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
13305 bitmasks to be enabled with 3 exceptions, instead of requiring any
13306 enabled ISA with lots of exceptions.
13307 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
13308 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
13309 Change avx512bw in isa attribute to avx512f.
13310 * config/i386/sgxintrin.h: Add license boilerplate.
13311 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
13312 to __AVX512F__ and __AVX512VL to __AVX512VL__.
13313 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
13314 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
13316 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
13317 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
13318 temporarily sse2 rather than sse if not enabled already.
13321 * config/i386/sse.md (VI248_VLBW): Rename to ...
13322 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
13323 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
13324 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
13325 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
13326 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
13327 mode iterator instead of VI248_VLBW.
13329 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
13331 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
13332 (record_modified): Skip clobbers; add debug output.
13333 (param_change_prob): Use sreal frequencies.
13335 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13337 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
13338 punt for user-aligned variables.
13340 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
13342 * tree-chrec.c (chrec_contains_symbols): Return true for
13345 2018-01-05 Sudakshina Das <sudi.das@arm.com>
13348 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
13349 of (x|y) == x for BICS pattern.
13351 2018-01-05 Jakub Jelinek <jakub@redhat.com>
13353 PR tree-optimization/83605
13354 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
13355 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
13358 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
13360 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
13361 * config/epiphany/rtems.h: New file.
13363 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13364 Uros Bizjak <ubizjak@gmail.com>
13367 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
13368 QIreg_operand instead of register_operand predicate.
13369 * config/i386/i386.c (ix86_rop_should_change_byte_p,
13370 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
13371 comments instead of -fmitigate[-_]rop.
13373 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13376 * cgraphunit.c (symbol_table::compile): Switch to text_section
13377 before calling assembly_start debug hook.
13378 * run-rtl-passes.c (run_rtl_passes): Likewise.
13381 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13383 * tree-vrp.c (extract_range_from_binary_expr_1): Check
13384 range_int_cst_p rather than !symbolic_range_p before calling
13385 extract_range_from_multiplicative_op_1.
13387 2018-01-04 Jeff Law <law@redhat.com>
13389 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
13390 redundant test in assertion.
13392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13394 * doc/rtl.texi: Document machine_mode wrapper classes.
13396 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13398 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
13399 using tree_to_uhwi.
13401 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13403 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
13404 the VEC_PERM_EXPR fold to fail.
13406 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13409 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
13410 to switched_sections.
13412 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13415 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
13416 test for d.testing.
13418 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
13421 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
13422 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
13424 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13427 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
13428 is BLKmode and bitpos not zero or mode change is needed.
13430 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
13433 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
13436 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
13439 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
13440 instead of MULT rtx. Update all corresponding splitters.
13441 (*saddl_se): Ditto.
13442 (*ssub<modesuffix>): Ditto.
13443 (*ssubl_se): Ditto.
13444 (*cmp_sadd_di): Update split patterns.
13445 (*cmp_sadd_si): Ditto.
13446 (*cmp_sadd_sidi): Ditto.
13447 (*cmp_ssub_di): Ditto.
13448 (*cmp_ssub_si): Ditto.
13449 (*cmp_ssub_sidi): Ditto.
13450 * config/alpha/predicates.md (const23_operand): New predicate.
13451 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
13452 Look for ASHIFT, not MULT inner operand.
13453 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
13455 2018-01-04 Martin Liska <mliska@suse.cz>
13457 PR gcov-profile/83669
13458 * gcov.c (output_intermediate_file): Add version to intermediate
13460 * doc/gcov.texi: Document new field 'version' in intermediate
13461 file format. Fix location of '-k' option of gcov command.
13463 2018-01-04 Martin Liska <mliska@suse.cz>
13466 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
13468 2018-01-04 Jakub Jelinek <jakub@redhat.com>
13470 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
13472 2018-01-03 Martin Sebor <msebor@redhat.com>
13474 PR tree-optimization/83655
13475 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
13476 checking calls with invalid arguments.
13478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13480 * tree-vect-stmts.c (vect_get_store_rhs): New function.
13481 (vectorizable_mask_load_store): Delete.
13482 (vectorizable_call): Return false for masked loads and stores.
13483 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
13484 instead of gimple_assign_rhs1.
13485 (vectorizable_load): Handle IFN_MASK_LOAD.
13486 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
13488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13490 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
13492 (vectorizable_mask_load_store): ...here.
13493 (vectorizable_load): ...and here.
13495 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13497 * tree-vect-stmts.c (vect_build_all_ones_mask)
13498 (vect_build_zero_merge_argument): New functions, split out from...
13499 (vectorizable_load): ...here.
13501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13503 * tree-vect-stmts.c (vect_check_store_rhs): New function,
13505 (vectorizable_mask_load_store): ...here.
13506 (vectorizable_store): ...and here.
13508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13510 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
13512 (vectorizable_mask_load_store): ...here.
13514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13516 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
13517 (vect_model_store_cost): Take a vec_load_store_type instead of a
13519 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
13520 (vect_model_store_cost): Take a vec_load_store_type instead of a
13522 (vectorizable_mask_load_store): Update accordingly.
13523 (vectorizable_store): Likewise.
13524 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
13526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13528 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
13529 IFN_MASK_LOAD calls here rather than...
13530 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
13532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13533 Alan Hayward <alan.hayward@arm.com>
13534 David Sherwood <david.sherwood@arm.com>
13536 * expmed.c (extract_bit_field_1): For vector extracts,
13537 fall back to extract_bit_field_as_subreg if vec_extract
13540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13541 Alan Hayward <alan.hayward@arm.com>
13542 David Sherwood <david.sherwood@arm.com>
13544 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
13545 they are variable or constant sized.
13546 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
13547 slots for constant-sized data.
13549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13550 Alan Hayward <alan.hayward@arm.com>
13551 David Sherwood <david.sherwood@arm.com>
13553 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
13554 handling COND_EXPRs with boolean comparisons, try to find a better
13555 basis for the mask type than the boolean itself.
13557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13559 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
13560 is calculated and how it can be overridden.
13561 * genmodes.c (max_bitsize_mode_any_mode): New variable.
13562 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
13564 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
13567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13568 Alan Hayward <alan.hayward@arm.com>
13569 David Sherwood <david.sherwood@arm.com>
13571 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
13572 Remove the mode argument.
13573 (aarch64_simd_valid_immediate): Remove the mode and inverse
13575 * config/aarch64/iterators.md (bitsize): New iterator.
13576 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
13577 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
13578 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
13579 aarch64_simd_valid_immediate.
13580 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
13581 (aarch64_reg_or_bic_imm): Likewise.
13582 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
13583 with an insn_type enum and msl with a modifier_type enum.
13584 Replace element_width with a scalar_mode. Change the shift
13585 to unsigned int. Add constructors for scalar_float_mode and
13586 scalar_int_mode elements.
13587 (aarch64_vect_float_const_representable_p): Delete.
13588 (aarch64_can_const_movi_rtx_p)
13589 (aarch64_simd_scalar_immediate_valid_for_move)
13590 (aarch64_simd_make_constant): Update call to
13591 aarch64_simd_valid_immediate.
13592 (aarch64_advsimd_valid_immediate_hs): New function.
13593 (aarch64_advsimd_valid_immediate): Likewise.
13594 (aarch64_simd_valid_immediate): Remove mode and inverse
13595 arguments. Rewrite to use the above. Use const_vec_duplicate_p
13596 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
13597 and aarch64_float_const_representable_p on the result.
13598 (aarch64_output_simd_mov_immediate): Remove mode argument.
13599 Update call to aarch64_simd_valid_immediate and use of
13600 simd_immediate_info.
13601 (aarch64_output_scalar_simd_mov_immediate): Update call
13604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13605 Alan Hayward <alan.hayward@arm.com>
13606 David Sherwood <david.sherwood@arm.com>
13608 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
13609 (mode_nunits): Likewise CONST_MODE_NUNITS.
13610 * machmode.def (ADJUST_NUNITS): Document.
13611 * genmodes.c (mode_data::need_nunits_adj): New field.
13612 (blank_mode): Update accordingly.
13613 (adj_nunits): New variable.
13614 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
13616 (emit_mode_size_inline): Set need_bytesize_adj for all modes
13617 listed in adj_nunits.
13618 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
13619 listed in adj_nunits. Don't emit case statements for such modes.
13620 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
13621 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
13622 nothing if adj_nunits is nonnull.
13623 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
13624 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
13625 (emit_mode_fbit): Update use of print_maybe_const_decl.
13626 (emit_move_size): Likewise. Treat the array as non-const
13628 (emit_mode_adjustments): Handle adj_nunits.
13630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13632 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
13633 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
13634 (VECTOR_MODES): Use it.
13635 (make_vector_modes): Take the prefix as an argument.
13637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13638 Alan Hayward <alan.hayward@arm.com>
13639 David Sherwood <david.sherwood@arm.com>
13641 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
13642 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
13643 for MODE_VECTOR_BOOL.
13644 * machmode.def (VECTOR_BOOL_MODE): Document.
13645 * genmodes.c (VECTOR_BOOL_MODE): New macro.
13646 (make_vector_bool_mode): New function.
13647 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
13649 * lto-streamer-in.c (lto_input_mode_table): Likewise.
13650 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
13652 * stor-layout.c (int_mode_for_mode): Likewise.
13653 * tree.c (build_vector_type_for_mode): Likewise.
13654 * varasm.c (output_constant_pool_2): Likewise.
13655 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
13656 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
13657 for MODE_VECTOR_BOOL.
13658 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
13659 of mode class checks.
13660 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
13661 instead of a list of mode class checks.
13662 (expand_vector_scalar_condition): Likewise.
13663 (type_for_widest_vector_mode): Handle BImode as an inner mode.
13665 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13666 Alan Hayward <alan.hayward@arm.com>
13667 David Sherwood <david.sherwood@arm.com>
13669 * machmode.h (mode_size): Change from unsigned short to
13671 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
13672 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13673 or if measurement_type is not polynomial.
13674 (fixed_size_mode::includes_p): Check for constant-sized modes.
13675 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
13676 return a poly_uint16 rather than an unsigned short.
13677 (emit_mode_size): Change the type of mode_size from unsigned short
13678 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
13679 (emit_mode_adjustments): Cope with polynomial vector sizes.
13680 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13682 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13684 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
13685 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
13686 * caller-save.c (setup_save_areas): Likewise.
13687 (replace_reg_with_saved_mem): Likewise.
13688 * calls.c (emit_library_call_value_1): Likewise.
13689 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
13690 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
13691 (gen_lowpart_for_combine): Likewise.
13692 * convert.c (convert_to_integer_1): Likewise.
13693 * cse.c (equiv_constant, cse_insn): Likewise.
13694 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
13695 (cselib_subst_to_values): Likewise.
13696 * dce.c (word_dce_process_block): Likewise.
13697 * df-problems.c (df_word_lr_mark_ref): Likewise.
13698 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
13699 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
13700 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
13701 (rtl_for_decl_location): Likewise.
13702 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
13703 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
13704 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
13705 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
13706 (expand_expr_real_1): Likewise.
13707 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
13708 (pad_below): Likewise.
13709 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13710 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
13711 * ira.c (get_subreg_tracking_sizes): Likewise.
13712 * ira-build.c (ira_create_allocno_objects): Likewise.
13713 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
13714 (ira_sort_regnos_for_alter_reg): Likewise.
13715 * ira-costs.c (record_operand_costs): Likewise.
13716 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
13717 (resolve_simple_move): Likewise.
13718 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
13719 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
13720 (lra_constraints): Likewise.
13721 (CONST_POOL_OK_P): Reject variable-sized modes.
13722 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
13723 (add_pseudo_to_slot, lra_spill): Likewise.
13724 * omp-low.c (omp_clause_aligned_alignment): Likewise.
13725 * optabs-query.c (get_best_extraction_insn): Likewise.
13726 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13727 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
13728 (expand_mult_highpart, valid_multiword_target_p): Likewise.
13729 * recog.c (offsettable_address_addr_space_p): Likewise.
13730 * regcprop.c (maybe_mode_change): Likewise.
13731 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
13732 * regrename.c (build_def_use): Likewise.
13733 * regstat.c (dump_reg_info): Likewise.
13734 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
13735 (find_reloads, find_reloads_subreg_address): Likewise.
13736 * reload1.c (eliminate_regs_1): Likewise.
13737 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
13738 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
13739 (simplify_binary_operation_1, simplify_subreg): Likewise.
13740 * targhooks.c (default_function_arg_padding): Likewise.
13741 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
13742 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
13743 (verify_gimple_assign_ternary): Likewise.
13744 * tree-inline.c (estimate_move_cost): Likewise.
13745 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13746 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
13747 (get_address_cost_ainc): Likewise.
13748 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
13749 (vect_supportable_dr_alignment): Likewise.
13750 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13751 (vectorizable_reduction): Likewise.
13752 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
13753 (vectorizable_operation, vectorizable_load): Likewise.
13754 * tree.c (build_same_sized_truth_vector_type): Likewise.
13755 * valtrack.c (cleanup_auto_inc_dec): Likewise.
13756 * var-tracking.c (emit_note_insn_var_location): Likewise.
13757 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
13758 (ADDR_VEC_ALIGN): Likewise.
13760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13761 Alan Hayward <alan.hayward@arm.com>
13762 David Sherwood <david.sherwood@arm.com>
13764 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
13766 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
13767 or if measurement_type is polynomial.
13768 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
13769 * combine.c (make_extraction): Likewise.
13770 * dse.c (find_shift_sequence): Likewise.
13771 * dwarf2out.c (mem_loc_descriptor): Likewise.
13772 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
13773 (extract_bit_field, extract_low_bits): Likewise.
13774 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
13775 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
13776 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
13777 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
13778 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
13779 * reload.c (find_reloads): Likewise.
13780 * reload1.c (alter_reg): Likewise.
13781 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
13782 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
13783 * tree-if-conv.c (predicate_mem_writes): Likewise.
13784 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
13785 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
13786 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
13787 * valtrack.c (dead_debug_insert_temp): Likewise.
13788 * varasm.c (mergeable_constant_section): Likewise.
13789 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
13791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13792 Alan Hayward <alan.hayward@arm.com>
13793 David Sherwood <david.sherwood@arm.com>
13795 * expr.c (expand_assignment): Cope with polynomial mode sizes
13796 when assigning to a CONCAT.
13798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13799 Alan Hayward <alan.hayward@arm.com>
13800 David Sherwood <david.sherwood@arm.com>
13802 * machmode.h (mode_precision): Change from unsigned short to
13804 (mode_to_precision): Return a poly_uint16 rather than an unsigned
13806 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
13807 or if measurement_type is not polynomial.
13808 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
13809 in which the mode is already known to be a scalar_int_mode.
13810 * genmodes.c (emit_mode_precision): Change the type of mode_precision
13811 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
13813 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13814 for GET_MODE_PRECISION.
13815 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13816 for GET_MODE_PRECISION.
13817 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
13819 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
13820 (expand_field_assignment, make_extraction): Likewise.
13821 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
13822 (get_last_value): Likewise.
13823 * convert.c (convert_to_integer_1): Likewise.
13824 * cse.c (cse_insn): Likewise.
13825 * expr.c (expand_expr_real_1): Likewise.
13826 * lra-constraints.c (simplify_operand_subreg): Likewise.
13827 * optabs-query.c (can_atomic_load_p): Likewise.
13828 * optabs.c (expand_atomic_load): Likewise.
13829 (expand_atomic_store): Likewise.
13830 * ree.c (combine_reaching_defs): Likewise.
13831 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
13832 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
13833 * tree.h (type_has_mode_precision_p): Likewise.
13834 * ubsan.c (instrument_si_overflow): Likewise.
13836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13837 Alan Hayward <alan.hayward@arm.com>
13838 David Sherwood <david.sherwood@arm.com>
13840 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
13841 polynomial numbers of units.
13842 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
13843 (valid_vector_subparts_p): New function.
13844 (build_vector_type): Remove temporary shim and take the number
13845 of units as a poly_uint64 rather than an int.
13846 (build_opaque_vector_type): Take the number of units as a
13847 poly_uint64 rather than an int.
13848 * tree.c (build_vector_from_ctor): Handle polynomial
13849 TYPE_VECTOR_SUBPARTS.
13850 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
13851 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
13852 (build_vector_from_val): If the number of units is variable,
13853 use build_vec_duplicate_cst for constant operands and
13854 VEC_DUPLICATE_EXPR otherwise.
13855 (make_vector_type): Remove temporary is_constant ().
13856 (build_vector_type, build_opaque_vector_type): Take the number of
13857 units as a poly_uint64 rather than an int.
13858 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
13860 * cfgexpand.c (expand_debug_expr): Likewise.
13861 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
13862 (store_constructor, expand_expr_real_1): Likewise.
13863 (const_scalar_mask_from_tree): Likewise.
13864 * fold-const-call.c (fold_const_reduction): Likewise.
13865 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
13866 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
13867 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
13868 (fold_relational_const): Likewise.
13869 (native_interpret_vector): Likewise. Change the size from an
13870 int to an unsigned int.
13871 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
13872 TYPE_VECTOR_SUBPARTS.
13873 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
13874 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
13875 duplicating a non-constant operand into a variable-length vector.
13876 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
13877 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
13878 * ipa-icf.c (sem_variable::equals): Likewise.
13879 * match.pd: Likewise.
13880 * omp-simd-clone.c (simd_clone_subparts): Likewise.
13881 * print-tree.c (print_node): Likewise.
13882 * stor-layout.c (layout_type): Likewise.
13883 * targhooks.c (default_builtin_vectorization_cost): Likewise.
13884 * tree-cfg.c (verify_gimple_comparison): Likewise.
13885 (verify_gimple_assign_binary): Likewise.
13886 (verify_gimple_assign_ternary): Likewise.
13887 (verify_gimple_assign_single): Likewise.
13888 * tree-pretty-print.c (dump_generic_node): Likewise.
13889 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
13890 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
13891 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
13892 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
13893 (vect_shift_permute_load_chain): Likewise.
13894 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
13895 (expand_vector_condition, optimize_vector_constructor): Likewise.
13896 (lower_vec_perm, get_compute_type): Likewise.
13897 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
13898 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
13899 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13900 (vect_recog_mask_conversion_pattern): Likewise.
13901 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
13902 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
13903 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
13904 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
13905 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
13906 (vectorizable_shift, vectorizable_operation, vectorizable_store)
13907 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
13908 (supportable_widening_operation): Likewise.
13909 (supportable_narrowing_operation): Likewise.
13910 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
13912 * varasm.c (output_constant): Likewise.
13914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13915 Alan Hayward <alan.hayward@arm.com>
13916 David Sherwood <david.sherwood@arm.com>
13918 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
13919 so that both the length == 3 and length != 3 cases set up their
13920 own permute vectors. Add comments explaining why we know the
13921 number of elements is constant.
13922 (vect_permute_load_chain): Likewise.
13924 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13925 Alan Hayward <alan.hayward@arm.com>
13926 David Sherwood <david.sherwood@arm.com>
13928 * machmode.h (mode_nunits): Change from unsigned char to
13930 (ONLY_FIXED_SIZE_MODES): New macro.
13931 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
13932 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
13933 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
13935 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
13936 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
13937 or if measurement_type is not polynomial.
13938 * genmodes.c (ZERO_COEFFS): New macro.
13939 (emit_mode_nunits_inline): Make mode_nunits_inline return a
13941 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
13942 Use ZERO_COEFFS when emitting initializers.
13943 * data-streamer.h (bp_pack_poly_value): New function.
13944 (bp_unpack_poly_value): Likewise.
13945 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
13946 for GET_MODE_NUNITS.
13947 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
13948 for GET_MODE_NUNITS.
13949 * tree.c (make_vector_type): Remove temporary shim and make
13950 the real function take the number of units as a poly_uint64
13951 rather than an int.
13952 (build_vector_type_for_mode): Handle polynomial nunits.
13953 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
13954 * emit-rtl.c (const_vec_series_p_1): Likewise.
13955 (gen_rtx_CONST_VECTOR): Likewise.
13956 * fold-const.c (test_vec_duplicate_folding): Likewise.
13957 * genrecog.c (validate_pattern): Likewise.
13958 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
13959 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
13960 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
13961 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
13962 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
13963 * rtlanal.c (subreg_get_info): Likewise.
13964 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
13965 (vect_grouped_load_supported): Likewise.
13966 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
13967 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
13968 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
13969 (simplify_const_unary_operation, simplify_binary_operation_1)
13970 (simplify_const_binary_operation, simplify_ternary_operation)
13971 (test_vector_ops_duplicate, test_vector_ops): Likewise.
13972 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
13973 instead of CONST_VECTOR_NUNITS.
13974 * varasm.c (output_constant_pool_2): Likewise.
13975 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
13976 explicit-encoded elements in the XVEC for variable-length vectors.
13978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13980 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
13982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13983 Alan Hayward <alan.hayward@arm.com>
13984 David Sherwood <david.sherwood@arm.com>
13986 * coretypes.h (fixed_size_mode): Declare.
13987 (fixed_size_mode_pod): New typedef.
13988 * builtins.h (target_builtins::x_apply_args_mode)
13989 (target_builtins::x_apply_result_mode): Change type to
13990 fixed_size_mode_pod.
13991 * builtins.c (apply_args_size, apply_result_size, result_vector)
13992 (expand_builtin_apply_args_1, expand_builtin_apply)
13993 (expand_builtin_return): Update accordingly.
13995 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
13997 * cse.c (hash_rtx_cb): Hash only the encoded elements.
13998 * cselib.c (cselib_hash_rtx): Likewise.
13999 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
14000 CONST_VECTOR encoding.
14002 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14003 Jeff Law <law@redhat.com>
14006 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
14007 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
14008 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
14009 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
14012 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
14013 explicitly probe *sp in a noreturn function if there were any callee
14014 register saves or frame pointer is needed.
14016 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14019 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
14020 BLKmode for ternary, binary or unary expressions.
14023 * var-tracking.c (delete_vta_debug_insn): New inline function.
14024 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
14025 insns from get_insns () to NULL instead of each bb separately.
14026 Use delete_vta_debug_insn. No longer static.
14027 (vt_debug_insns_local, variable_tracking_main_1): Adjust
14028 delete_vta_debug_insns callers.
14029 * rtl.h (delete_vta_debug_insns): Declare.
14030 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
14031 instead of variable_tracking_main.
14033 2018-01-03 Martin Sebor <msebor@redhat.com>
14035 PR tree-optimization/83603
14036 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
14037 arguments past the endof the argument list in functions declared
14038 without a prototype.
14039 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
14040 Avoid checking when arguments are null.
14042 2018-01-03 Martin Sebor <msebor@redhat.com>
14045 * doc/extend.texi (attribute const): Fix a typo.
14046 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
14047 issuing -Wsuggest-attribute for void functions.
14049 2018-01-03 Martin Sebor <msebor@redhat.com>
14051 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
14052 offset_int::from instead of wide_int::to_shwi.
14053 (maybe_diag_overlap): Remove assertion.
14054 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
14055 * gimple-ssa-sprintf.c (format_directive): Same.
14056 (parse_directive): Same.
14057 (sprintf_dom_walker::compute_format_length): Same.
14058 (try_substitute_return_value): Same.
14060 2018-01-03 Jeff Law <law@redhat.com>
14062 PR middle-end/83654
14063 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
14064 non-constant residual for zero at runtime and avoid probing in
14065 that case. Reorganize code for trailing problem to mirror handling
14068 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14070 PR tree-optimization/83501
14071 * tree-ssa-strlen.c (get_string_cst): New.
14072 (handle_char_store): Call get_string_cst.
14074 2018-01-03 Martin Liska <mliska@suse.cz>
14076 PR tree-optimization/83593
14077 * tree-ssa-strlen.c: Include tree-cfg.h.
14078 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
14079 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
14080 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
14082 (strlen_dom_walker::before_dom_children): Call
14083 gimple_purge_dead_eh_edges. Dump tranformation with details
14085 (strlen_dom_walker::before_dom_children): Update call by adding
14086 new argument cleanup_eh.
14087 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
14089 2018-01-03 Martin Liska <mliska@suse.cz>
14092 * cif-code.def (VARIADIC_THUNK): New enum value.
14093 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
14096 2018-01-03 Jan Beulich <jbeulich@suse.com>
14098 * sse.md (mov<mode>_internal): Tighten condition for when to use
14099 vmovdqu<ssescalarsize> for TI and OI modes.
14101 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14103 Update copyright years.
14105 2018-01-03 Martin Liska <mliska@suse.cz>
14108 * ipa-visibility.c (function_and_variable_visibility): Skip
14109 functions with noipa attribure.
14111 2018-01-03 Jakub Jelinek <jakub@redhat.com>
14113 * gcc.c (process_command): Update copyright notice dates.
14114 * gcov-dump.c (print_version): Ditto.
14115 * gcov.c (print_version): Ditto.
14116 * gcov-tool.c (print_version): Ditto.
14117 * gengtype.c (create_file): Ditto.
14118 * doc/cpp.texi: Bump @copying's copyright year.
14119 * doc/cppinternals.texi: Ditto.
14120 * doc/gcc.texi: Ditto.
14121 * doc/gccint.texi: Ditto.
14122 * doc/gcov.texi: Ditto.
14123 * doc/install.texi: Ditto.
14124 * doc/invoke.texi: Ditto.
14126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14128 * vector-builder.h (vector_builder::m_full_nelts): Change from
14129 unsigned int to poly_uint64.
14130 (vector_builder::full_nelts): Update prototype accordingly.
14131 (vector_builder::new_vector): Likewise.
14132 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
14133 (vector_builder::operator ==): Likewise.
14134 (vector_builder::finalize): Likewise.
14135 * int-vector-builder.h (int_vector_builder::int_vector_builder):
14136 Take the number of elements as a poly_uint64 rather than an
14138 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
14139 from unsigned int to poly_uint64.
14140 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
14141 (vec_perm_indices::new_vector): Likewise.
14142 (vec_perm_indices::length): Likewise.
14143 (vec_perm_indices::nelts_per_input): Likewise.
14144 (vec_perm_indices::input_nelts): Likewise.
14145 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
14146 number of elements per input as a poly_uint64 rather than an
14147 unsigned int. Use the original encoding for variable-length
14148 vectors, rather than clamping each individual element.
14149 For the second and subsequent elements in each pattern,
14150 clamp the step and base before clamping their sum.
14151 (vec_perm_indices::series_p): Handle polynomial element counts.
14152 (vec_perm_indices::all_in_range_p): Likewise.
14153 (vec_perm_indices_to_tree): Likewise.
14154 (vec_perm_indices_to_rtx): Likewise.
14155 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
14156 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
14157 (tree_vector_builder::new_binary_operation): Handle polynomial
14158 element counts. Return false if we need to know the number
14159 of elements at compile time.
14160 * fold-const.c (fold_vec_perm): Punt if the number of elements
14161 isn't known at compile time.
14163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14165 * vec-perm-indices.h (vec_perm_builder): Change element type
14166 from HOST_WIDE_INT to poly_int64.
14167 (vec_perm_indices::element_type): Update accordingly.
14168 (vec_perm_indices::clamp): Handle polynomial element_types.
14169 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14170 (vec_perm_indices::all_in_range_p): Likewise.
14171 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
14173 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
14174 polynomial vec_perm_indices element types.
14175 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
14176 * fold-const.c (fold_vec_perm): Likewise.
14177 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
14178 * tree-vect-generic.c (lower_vec_perm): Likewise.
14179 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14180 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
14181 element type to HOST_WIDE_INT.
14183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14184 Alan Hayward <alan.hayward@arm.com>
14185 David Sherwood <david.sherwood@arm.com>
14187 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
14188 rather than an int. Use plus_constant.
14189 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
14190 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
14192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14193 Alan Hayward <alan.hayward@arm.com>
14194 David Sherwood <david.sherwood@arm.com>
14196 * calls.c (emit_call_1, expand_call): Change struct_value_size from
14197 a HOST_WIDE_INT to a poly_int64.
14199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14200 Alan Hayward <alan.hayward@arm.com>
14201 David Sherwood <david.sherwood@arm.com>
14203 * calls.c (load_register_parameters): Cope with polynomial
14204 mode sizes. Require a constant size for BLKmode parameters
14205 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
14206 forces a parameter to be padded at the lsb end in order to
14207 fill a complete number of words, require the parameter size
14208 to be ordered wrt UNITS_PER_WORD.
14210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14211 Alan Hayward <alan.hayward@arm.com>
14212 David Sherwood <david.sherwood@arm.com>
14214 * reload1.c (spill_stack_slot_width): Change element type
14215 from unsigned int to poly_uint64_pod.
14216 (alter_reg): Treat mode sizes as polynomial.
14218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14219 Alan Hayward <alan.hayward@arm.com>
14220 David Sherwood <david.sherwood@arm.com>
14222 * reload.c (complex_word_subreg_p): New function.
14223 (reload_inner_reg_of_subreg, push_reload): Use it.
14225 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14226 Alan Hayward <alan.hayward@arm.com>
14227 David Sherwood <david.sherwood@arm.com>
14229 * lra-constraints.c (process_alt_operands): Reject matched
14230 operands whose sizes aren't ordered.
14231 (match_reload): Refer to this check here.
14233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14234 Alan Hayward <alan.hayward@arm.com>
14235 David Sherwood <david.sherwood@arm.com>
14237 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
14238 that the mode size is in the set {1, 2, 4, 8, 16}.
14240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14241 Alan Hayward <alan.hayward@arm.com>
14242 David Sherwood <david.sherwood@arm.com>
14244 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
14245 Use plus_constant instead of gen_rtx_PLUS.
14247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14248 Alan Hayward <alan.hayward@arm.com>
14249 David Sherwood <david.sherwood@arm.com>
14251 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
14252 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
14253 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
14254 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
14255 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
14256 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
14257 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
14258 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
14259 * config/i386/i386.c (ix86_push_rounding): ...this new function.
14260 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
14262 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
14263 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
14264 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
14265 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
14266 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
14267 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
14268 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
14269 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
14270 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
14271 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
14273 * expr.c (emit_move_resolve_push): Treat the input and result
14274 of PUSH_ROUNDING as a poly_int64.
14275 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
14276 (emit_push_insn): Likewise.
14277 * lra-eliminations.c (mark_not_eliminable): Likewise.
14278 * recog.c (push_operand): Likewise.
14279 * reload1.c (elimination_effects): Likewise.
14280 * rtlanal.c (nonzero_bits1): Likewise.
14281 * calls.c (store_one_arg): Likewise. Require the padding to be
14282 known at compile time.
14284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14285 Alan Hayward <alan.hayward@arm.com>
14286 David Sherwood <david.sherwood@arm.com>
14288 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
14289 Use plus_constant instead of gen_rtx_PLUS.
14291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14292 Alan Hayward <alan.hayward@arm.com>
14293 David Sherwood <david.sherwood@arm.com>
14295 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
14296 rather than an int.
14298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14299 Alan Hayward <alan.hayward@arm.com>
14300 David Sherwood <david.sherwood@arm.com>
14302 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
14303 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
14304 via stack temporaries. Treat the mode size as polynomial too.
14306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14307 Alan Hayward <alan.hayward@arm.com>
14308 David Sherwood <david.sherwood@arm.com>
14310 * expr.c (expand_expr_real_2): When handling conversions involving
14311 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
14312 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
14313 as a poly_uint64 too.
14315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14316 Alan Hayward <alan.hayward@arm.com>
14317 David Sherwood <david.sherwood@arm.com>
14319 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
14321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14322 Alan Hayward <alan.hayward@arm.com>
14323 David Sherwood <david.sherwood@arm.com>
14325 * combine.c (can_change_dest_mode): Handle polynomial
14326 REGMODE_NATURAL_SIZE.
14327 * expmed.c (store_bit_field_1): Likewise.
14328 * expr.c (store_constructor): Likewise.
14329 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
14330 and polynomial REGMODE_NATURAL_SIZE.
14331 (gen_lowpart_common): Likewise.
14332 * reginfo.c (record_subregs_of_mode): Likewise.
14333 * rtlanal.c (read_modify_subreg_p): Likewise.
14335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14336 Alan Hayward <alan.hayward@arm.com>
14337 David Sherwood <david.sherwood@arm.com>
14339 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
14340 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
14348 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14349 Alan Hayward <alan.hayward@arm.com>
14350 David Sherwood <david.sherwood@arm.com>
14352 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
14353 in a POINTER_PLUS_EXPR.
14355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14356 Alan Hayward <alan.hayward@arm.com>
14357 David Sherwood <david.sherwood@arm.com>
14359 * omp-simd-clone.c (simd_clone_subparts): New function.
14360 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
14361 (ipa_simd_modify_function_body): Likewise.
14363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14364 Alan Hayward <alan.hayward@arm.com>
14365 David Sherwood <david.sherwood@arm.com>
14367 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
14368 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
14369 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
14370 (expand_vector_condition, vector_element): Likewise.
14371 (subparts_gt): New function.
14372 (get_compute_type): Use subparts_gt.
14373 (count_type_subparts): Delete.
14374 (expand_vector_operations_1): Use subparts_gt instead of
14375 count_type_subparts.
14377 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14378 Alan Hayward <alan.hayward@arm.com>
14379 David Sherwood <david.sherwood@arm.com>
14381 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
14382 (vect_compile_time_alias): ...this new function. Do the calculation
14383 on poly_ints rather than trees.
14384 (vect_prune_runtime_alias_test_list): Update call accordingly.
14386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14387 Alan Hayward <alan.hayward@arm.com>
14388 David Sherwood <david.sherwood@arm.com>
14390 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
14392 (vect_schedule_slp_instance): Likewise.
14394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14395 Alan Hayward <alan.hayward@arm.com>
14396 David Sherwood <david.sherwood@arm.com>
14398 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
14399 constant and extern definitions for variable-length vectors.
14400 (vect_get_constant_vectors): Note that the number of units
14401 is known to be constant.
14403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14404 Alan Hayward <alan.hayward@arm.com>
14405 David Sherwood <david.sherwood@arm.com>
14407 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
14408 of units as polynomial. Choose between WIDE and NARROW based
14411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14412 Alan Hayward <alan.hayward@arm.com>
14413 David Sherwood <david.sherwood@arm.com>
14415 * tree-vect-stmts.c (simd_clone_subparts): New function.
14416 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
14418 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14419 Alan Hayward <alan.hayward@arm.com>
14420 David Sherwood <david.sherwood@arm.com>
14422 * tree-vect-stmts.c (vectorizable_call): Treat the number of
14423 vectors as polynomial. Use build_index_vector for
14424 IFN_GOMP_SIMD_LANE.
14426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14427 Alan Hayward <alan.hayward@arm.com>
14428 David Sherwood <david.sherwood@arm.com>
14430 * tree-vect-stmts.c (get_load_store_type): Treat the number of
14431 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
14432 for variable-length vectors.
14433 (vectorizable_mask_load_store): Treat the number of units as
14434 polynomial, asserting that it is constant if the condition has
14435 already been enforced.
14436 (vectorizable_store, vectorizable_load): Likewise.
14438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14439 Alan Hayward <alan.hayward@arm.com>
14440 David Sherwood <david.sherwood@arm.com>
14442 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
14443 of units as polynomial. Punt if we can't tell at compile time
14444 which vector contains the final result.
14446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14447 Alan Hayward <alan.hayward@arm.com>
14448 David Sherwood <david.sherwood@arm.com>
14450 * tree-vect-loop.c (vectorizable_induction): Treat the number
14451 of units as polynomial. Punt on SLP inductions. Use an integer
14452 VEC_SERIES_EXPR for variable-length integer reductions. Use a
14453 cast of such a series for variable-length floating-point
14456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14457 Alan Hayward <alan.hayward@arm.com>
14458 David Sherwood <david.sherwood@arm.com>
14460 * tree.h (build_index_vector): Declare.
14461 * tree.c (build_index_vector): New function.
14462 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
14463 of units as polynomial, forcibly converting it to a constant if
14464 vectorizable_reduction has already enforced the condition.
14465 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
14466 to create a {1,2,3,...} vector.
14467 (vectorizable_reduction): Treat the number of units as polynomial.
14468 Choose vectype_in based on the largest scalar element size rather
14469 than the smallest number of units. Enforce the restrictions
14472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14473 Alan Hayward <alan.hayward@arm.com>
14474 David Sherwood <david.sherwood@arm.com>
14476 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
14477 number of units as polynomial.
14479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14480 Alan Hayward <alan.hayward@arm.com>
14481 David Sherwood <david.sherwood@arm.com>
14483 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
14484 * target.def (autovectorize_vector_sizes): Return the vector sizes
14485 by pointer, using vector_sizes rather than a bitmask.
14486 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
14487 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
14488 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
14490 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
14491 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
14492 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
14493 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
14494 * omp-general.c (omp_max_vf): Likewise.
14495 * omp-low.c (omp_clause_aligned_alignment): Likewise.
14496 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
14497 * tree-vect-loop.c (vect_analyze_loop): Likewise.
14498 * tree-vect-slp.c (vect_slp_bb): Likewise.
14499 * doc/tm.texi: Regenerate.
14500 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
14502 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
14503 the vector size as a poly_uint64 rather than an unsigned int.
14504 (current_vector_size): Change from an unsigned int to a poly_uint64.
14505 (get_vectype_for_scalar_type): Update accordingly.
14506 * tree.h (build_truth_vector_type): Take the size and number of
14507 units as a poly_uint64 rather than an unsigned int.
14508 (build_vector_type): Add a temporary overload that takes
14509 the number of units as a poly_uint64 rather than an unsigned int.
14510 * tree.c (make_vector_type): Likewise.
14511 (build_truth_vector_type): Take the number of units as a poly_uint64
14512 rather than an unsigned int.
14514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14515 Alan Hayward <alan.hayward@arm.com>
14516 David Sherwood <david.sherwood@arm.com>
14518 * target.def (get_mask_mode): Take the number of units and length
14519 as poly_uint64s rather than unsigned ints.
14520 * targhooks.h (default_get_mask_mode): Update accordingly.
14521 * targhooks.c (default_get_mask_mode): Likewise.
14522 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
14523 * doc/tm.texi: Regenerate.
14525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14526 Alan Hayward <alan.hayward@arm.com>
14527 David Sherwood <david.sherwood@arm.com>
14529 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
14530 * omp-general.c (omp_max_vf): Likewise.
14531 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
14532 (expand_omp_simd): Handle polynomial safelen.
14533 * omp-low.c (omplow_simd_context): Add a default constructor.
14534 (omplow_simd_context::max_vf): Change from int to poly_uint64.
14535 (lower_rec_simd_input_clauses): Update accordingly.
14536 (lower_rec_input_clauses): Likewise.
14538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14539 Alan Hayward <alan.hayward@arm.com>
14540 David Sherwood <david.sherwood@arm.com>
14542 * tree-vectorizer.h (vect_nunits_for_cost): New function.
14543 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
14544 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
14545 (vect_analyze_slp_cost): Likewise.
14546 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
14547 (vect_model_load_cost): Likewise.
14549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14550 Alan Hayward <alan.hayward@arm.com>
14551 David Sherwood <david.sherwood@arm.com>
14553 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
14554 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
14555 from an unsigned int * to a poly_uint64_pod *.
14556 (calculate_unrolling_factor): New function.
14557 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
14559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14560 Alan Hayward <alan.hayward@arm.com>
14561 David Sherwood <david.sherwood@arm.com>
14563 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
14564 from an unsigned int to a poly_uint64.
14565 (_loop_vec_info::slp_unrolling_factor): Likewise.
14566 (_loop_vec_info::vectorization_factor): Change from an int
14568 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
14569 (vect_get_num_vectors): New function.
14570 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
14571 (vect_get_num_copies): Use vect_get_num_vectors.
14572 (vect_analyze_data_ref_dependences): Change max_vf from an int *
14573 to an unsigned int *.
14574 (vect_analyze_data_refs): Change min_vf from an int * to a
14576 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14577 than an unsigned HOST_WIDE_INT.
14578 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
14579 (vect_analyze_data_ref_dependence): Change max_vf from an int *
14580 to an unsigned int *.
14581 (vect_analyze_data_ref_dependences): Likewise.
14582 (vect_compute_data_ref_alignment): Handle polynomial vf.
14583 (vect_enhance_data_refs_alignment): Likewise.
14584 (vect_prune_runtime_alias_test_list): Likewise.
14585 (vect_shift_permute_load_chain): Likewise.
14586 (vect_supportable_dr_alignment): Likewise.
14587 (dependence_distance_ge_vf): Take the vectorization factor as a
14588 poly_uint64 rather than an unsigned HOST_WIDE_INT.
14589 (vect_analyze_data_refs): Change min_vf from an int * to a
14591 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
14592 vfm1 as a poly_uint64 rather than an int. Make the same change
14593 for the returned bound_scalar.
14594 (vect_gen_vector_loop_niters): Handle polynomial vf.
14595 (vect_do_peeling): Likewise. Update call to
14596 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
14597 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
14599 * tree-vect-loop.c (vect_determine_vectorization_factor)
14600 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
14601 (vect_get_known_peeling_cost): Likewise.
14602 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
14603 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
14604 (vect_transform_loop): Likewise. Use the lowest possible VF when
14605 updating the upper bounds of the loop.
14606 (vect_min_worthwhile_factor): Make static. Return an unsigned int
14607 rather than an int.
14608 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
14609 polynomial unroll factors.
14610 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
14611 (vect_make_slp_decision): Likewise.
14612 (vect_supported_load_permutation_p): Likewise, and polynomial
14614 (vect_analyze_slp_cost): Handle polynomial vf.
14615 (vect_slp_analyze_node_operations): Likewise.
14616 (vect_slp_analyze_bb_1): Likewise.
14617 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
14618 than an unsigned HOST_WIDE_INT.
14619 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
14620 (vectorizable_load): Handle polynomial vf.
14621 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
14623 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
14625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14626 Alan Hayward <alan.hayward@arm.com>
14627 David Sherwood <david.sherwood@arm.com>
14629 * match.pd: Handle bit operations involving three constants
14630 and try to fold one pair.
14632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
14634 * tree-vect-loop-manip.c: Include gimple-fold.h.
14635 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
14636 niters_maybe_zero parameters. Handle other cases besides a step of 1.
14637 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
14638 Add a path that uses a step of VF instead of 1, but disable it
14640 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
14641 and niters_no_overflow parameters. Update calls to
14642 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
14643 Create a new SSA name if the latter choses to use a ste other
14644 than zero, and return it via niters_vector_mult_vf_var.
14645 * tree-vect-loop.c (vect_transform_loop): Update calls to
14646 vect_do_peeling, vect_gen_vector_loop_niters and
14647 slpeel_make_loop_iterate_ntimes.
14648 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
14649 (vect_gen_vector_loop_niters): Update declarations after above changes.
14651 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
14653 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
14654 128-bit round to integer instructions.
14655 (ceil<mode>2): Likewise.
14656 (btrunc<mode>2): Likewise.
14657 (round<mode>2): Likewise.
14659 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14661 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
14662 unaligned VSX load/store on P8/P9.
14663 (expand_block_clear): Allow the use of unaligned VSX
14664 load/store on P8/P9.
14666 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14668 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
14670 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
14671 swap associated with both a load and a store.
14673 2018-01-02 Andrew Waterman <andrew@sifive.com>
14675 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
14676 * config/riscv/riscv.md (clear_cache): Use it.
14678 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
14680 * web.c: Remove out-of-date comment.
14682 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14684 * expr.c (fixup_args_size_notes): Check that any existing
14685 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
14686 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
14687 (emit_single_push_insn): ...here.
14689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14691 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
14692 (const_vector_encoded_nelts): New function.
14693 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
14694 (const_vector_int_elt, const_vector_elt): Declare.
14695 * emit-rtl.c (const_vector_int_elt_1): New function.
14696 (const_vector_elt): Likewise.
14697 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
14698 of CONST_VECTOR_ELT.
14700 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14702 * expr.c: Include rtx-vector-builder.h.
14703 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
14704 directly on the tree encoding.
14705 (const_vector_from_tree): Likewise.
14706 * optabs.c: Include rtx-vector-builder.h.
14707 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
14708 sequence of "u" values.
14709 * vec-perm-indices.c: Include rtx-vector-builder.h.
14710 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
14711 directly on the vec_perm_indices encoding.
14713 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14715 * doc/rtl.texi (const_vector): Describe new encoding scheme.
14716 * Makefile.in (OBJS): Add rtx-vector-builder.o.
14717 * rtx-vector-builder.h: New file.
14718 * rtx-vector-builder.c: Likewise.
14719 * rtl.h (rtx_def::u2): Add a const_vector field.
14720 (CONST_VECTOR_NPATTERNS): New macro.
14721 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
14722 (CONST_VECTOR_DUPLICATE_P): Likewise.
14723 (CONST_VECTOR_STEPPED_P): Likewise.
14724 (CONST_VECTOR_ENCODED_ELT): Likewise.
14725 (const_vec_duplicate_p): Check for a duplicated vector encoding.
14726 (unwrap_const_vec_duplicate): Likewise.
14727 (const_vec_series_p): Check for a non-duplicated vector encoding.
14728 Say that the function only returns true for integer vectors.
14729 * emit-rtl.c: Include rtx-vector-builder.h.
14730 (gen_const_vec_duplicate_1): Delete.
14731 (gen_const_vector): Call gen_const_vec_duplicate instead of
14732 gen_const_vec_duplicate_1.
14733 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
14734 (gen_const_vec_duplicate): Use rtx_vector_builder.
14735 (gen_const_vec_series): Likewise.
14736 (gen_rtx_CONST_VECTOR): Likewise.
14737 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
14738 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14739 Build a new vector rather than modifying a CONST_VECTOR in-place.
14740 (handle_special_swappables): Update call accordingly.
14741 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
14742 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
14743 Build a new vector rather than modifying a CONST_VECTOR in-place.
14744 (handle_special_swappables): Update call accordingly.
14746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14748 * simplify-rtx.c (simplify_const_binary_operation): Use
14749 CONST_VECTOR_ELT instead of XVECEXP.
14751 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14753 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
14754 the selector elements to be different from the data elements
14755 if the selector is a VECTOR_CST.
14756 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
14757 ssizetype for the selector.
14759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14761 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
14762 before testing each element individually.
14763 * tree-vect-generic.c (lower_vec_perm): Likewise.
14765 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14767 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
14768 * selftest-run-tests.c (selftest::run_tests): Call it.
14769 * vector-builder.h (vector_builder::operator ==): New function.
14770 (vector_builder::operator !=): Likewise.
14771 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
14772 (vec_perm_indices::all_from_input_p): New function.
14773 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
14774 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
14775 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
14776 instead of reading the VECTOR_CST directly. Detect whether both
14777 vector inputs are the same before constructing the vec_perm_indices,
14778 and update the number of inputs argument accordingly. Use the
14779 utility functions added above. Only construct sel2 if we need to.
14781 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14783 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
14784 the broadcast of the low byte.
14785 (expand_mult_highpart): Use an explicit encoding for the permutes.
14786 * optabs-query.c (can_mult_highpart_p): Likewise.
14787 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
14788 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14789 (vectorizable_bswap): Likewise.
14790 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
14791 explicit encoding for the power-of-2 permutes.
14792 (vect_permute_store_chain): Likewise.
14793 (vect_grouped_load_supported): Likewise.
14794 (vect_permute_load_chain): Likewise.
14796 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14798 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
14799 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
14800 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
14801 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
14802 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
14803 (vect_gen_perm_mask_any): Likewise.
14805 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14807 * int-vector-builder.h: New file.
14808 * vec-perm-indices.h: Include int-vector-builder.h.
14809 (vec_perm_indices): Redefine as an int_vector_builder.
14810 (auto_vec_perm_indices): Delete.
14811 (vec_perm_builder): Redefine as a stand-alone class.
14812 (vec_perm_indices::vec_perm_indices): New function.
14813 (vec_perm_indices::clamp): Likewise.
14814 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
14815 (vec_perm_indices::new_vector): New function.
14816 (vec_perm_indices::new_expanded_vector): Update for new
14817 vec_perm_indices class.
14818 (vec_perm_indices::rotate_inputs): New function.
14819 (vec_perm_indices::all_in_range_p): Operate directly on the
14820 encoded form, without computing elided elements.
14821 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
14822 encoding. Update for new vec_perm_indices class.
14823 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
14824 the given vec_perm_builder.
14825 (expand_vec_perm_var): Update vec_perm_builder constructor.
14826 (expand_mult_highpart): Use vec_perm_builder instead of
14827 auto_vec_perm_indices.
14828 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
14829 vec_perm_indices instead of auto_vec_perm_indices. Use a single
14830 or double series encoding as appropriate.
14831 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
14832 vec_perm_indices instead of auto_vec_perm_indices.
14833 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
14834 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
14835 (vect_permute_store_chain): Likewise.
14836 (vect_grouped_load_supported): Likewise.
14837 (vect_permute_load_chain): Likewise.
14838 (vect_shift_permute_load_chain): Likewise.
14839 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
14840 (vect_transform_slp_perm_load): Likewise.
14841 (vect_schedule_slp_instance): Likewise.
14842 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
14843 (vectorizable_mask_load_store): Likewise.
14844 (vectorizable_bswap): Likewise.
14845 (vectorizable_store): Likewise.
14846 (vectorizable_load): Likewise.
14847 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
14848 vec_perm_indices instead of auto_vec_perm_indices. Use
14849 tree_to_vec_perm_builder to read the vector from a tree.
14850 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
14851 vec_perm_builder instead of a vec_perm_indices.
14852 (have_whole_vector_shift): Use vec_perm_builder and
14853 vec_perm_indices instead of auto_vec_perm_indices. Leave the
14854 truncation to calc_vec_perm_mask_for_shift.
14855 (vect_create_epilog_for_reduction): Likewise.
14856 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
14857 from auto_vec_perm_indices to vec_perm_indices.
14858 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14859 instead of changing individual elements.
14860 (aarch64_vectorize_vec_perm_const): Use new_vector to install
14861 the vector in d.perm.
14862 * config/arm/arm.c (expand_vec_perm_d::perm): Change
14863 from auto_vec_perm_indices to vec_perm_indices.
14864 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
14865 instead of changing individual elements.
14866 (arm_vectorize_vec_perm_const): Use new_vector to install
14867 the vector in d.perm.
14868 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
14869 Update vec_perm_builder constructor.
14870 (rs6000_expand_interleave): Likewise.
14871 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
14872 (rs6000_expand_interleave): Likewise.
14874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14876 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
14877 to qimode could truncate the indices.
14878 * optabs.c (expand_vec_perm_var): Likewise.
14880 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
14882 * Makefile.in (OBJS): Add vec-perm-indices.o.
14883 * vec-perm-indices.h: New file.
14884 * vec-perm-indices.c: Likewise.
14885 * target.h (vec_perm_indices): Replace with a forward class
14887 (auto_vec_perm_indices): Move to vec-perm-indices.h.
14888 * optabs.h: Include vec-perm-indices.h.
14889 (expand_vec_perm): Delete.
14890 (selector_fits_mode_p, expand_vec_perm_var): Declare.
14891 (expand_vec_perm_const): Declare.
14892 * target.def (vec_perm_const_ok): Replace with...
14893 (vec_perm_const): ...this new hook.
14894 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
14895 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
14896 * doc/tm.texi: Regenerate.
14897 * optabs.def (vec_perm_const): Delete.
14898 * doc/md.texi (vec_perm_const): Likewise.
14899 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
14900 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
14901 expand_vec_perm for constant permutation vectors. Assert that
14902 the mode of variable permutation vectors is the integer equivalent
14903 of the mode that is being permuted.
14904 * optabs-query.h (selector_fits_mode_p): Declare.
14905 * optabs-query.c: Include vec-perm-indices.h.
14906 (selector_fits_mode_p): New function.
14907 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
14908 is defined, instead of checking whether the vec_perm_const_optab
14909 exists. Use targetm.vectorize.vec_perm_const instead of
14910 targetm.vectorize.vec_perm_const_ok. Check whether the indices
14911 fit in the vector mode before using a variable permute.
14912 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
14913 vec_perm_indices instead of an rtx.
14914 (expand_vec_perm): Replace with...
14915 (expand_vec_perm_const): ...this new function. Take the selector
14916 as a vec_perm_indices rather than an rtx. Also take the mode of
14917 the selector. Update call to shift_amt_for_vec_perm_mask.
14918 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
14919 Use vec_perm_indices::new_expanded_vector to expand the original
14920 selector into bytes. Check whether the indices fit in the vector
14921 mode before using a variable permute.
14922 (expand_vec_perm_var): Make global.
14923 (expand_mult_highpart): Use expand_vec_perm_const.
14924 * fold-const.c: Includes vec-perm-indices.h.
14925 * tree-ssa-forwprop.c: Likewise.
14926 * tree-vect-data-refs.c: Likewise.
14927 * tree-vect-generic.c: Likewise.
14928 * tree-vect-loop.c: Likewise.
14929 * tree-vect-slp.c: Likewise.
14930 * tree-vect-stmts.c: Likewise.
14931 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
14933 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
14934 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
14935 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
14936 (aarch64_vectorize_vec_perm_const): ...this new function.
14937 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14938 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14939 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
14940 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
14941 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14942 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14943 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
14945 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
14946 check for NEON modes.
14947 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
14948 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
14949 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
14950 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
14952 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
14953 the old VEC_PERM_CONST conditions.
14954 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
14955 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
14956 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
14957 (ia64_vectorize_vec_perm_const_ok): Merge into...
14958 (ia64_vectorize_vec_perm_const): ...this new function.
14959 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
14960 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
14961 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
14962 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
14963 * config/mips/mips.c (mips_expand_vec_perm_const)
14964 (mips_vectorize_vec_perm_const_ok): Merge into...
14965 (mips_vectorize_vec_perm_const): ...this new function.
14966 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
14967 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
14968 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
14969 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
14970 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
14971 (rs6000_expand_vec_perm_const): Delete.
14972 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
14974 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14975 (altivec_expand_vec_perm_const_le): Take each operand individually.
14976 Operate on constant selectors rather than rtxes.
14977 (altivec_expand_vec_perm_const): Likewise. Update call to
14978 altivec_expand_vec_perm_const_le.
14979 (rs6000_expand_vec_perm_const): Delete.
14980 (rs6000_vectorize_vec_perm_const_ok): Delete.
14981 (rs6000_vectorize_vec_perm_const): New function.
14982 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
14983 an element count and rtx array.
14984 (rs6000_expand_extract_even): Update call accordingly.
14985 (rs6000_expand_interleave): Likewise.
14986 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
14987 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
14988 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
14989 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
14990 (rs6000_expand_vec_perm_const): Delete.
14991 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
14992 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
14993 (altivec_expand_vec_perm_const_le): Take each operand individually.
14994 Operate on constant selectors rather than rtxes.
14995 (altivec_expand_vec_perm_const): Likewise. Update call to
14996 altivec_expand_vec_perm_const_le.
14997 (rs6000_expand_vec_perm_const): Delete.
14998 (rs6000_vectorize_vec_perm_const_ok): Delete.
14999 (rs6000_vectorize_vec_perm_const): New function. Remove stray
15000 reference to the SPE evmerge intructions.
15001 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
15002 an element count and rtx array.
15003 (rs6000_expand_extract_even): Update call accordingly.
15004 (rs6000_expand_interleave): Likewise.
15005 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
15006 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
15008 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
15010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15012 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
15013 vector mode and that that mode matches the mode of the data
15015 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
15016 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
15017 directly using expand_vec_perm_1 when forcing selectors into
15019 (expand_vec_perm_var): New function, split out from expand_vec_perm.
15021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15023 * optabs-query.h (can_vec_perm_p): Delete.
15024 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
15025 * optabs-query.c (can_vec_perm_p): Split into...
15026 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
15027 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
15028 particular selector is valid.
15029 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15030 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15031 (vect_grouped_load_supported): Likewise.
15032 (vect_shift_permute_load_chain): Likewise.
15033 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15034 (vect_transform_slp_perm_load): Likewise.
15035 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15036 (vectorizable_bswap): Likewise.
15037 (vect_gen_perm_mask_checked): Likewise.
15038 * fold-const.c (fold_ternary_loc): Likewise. Don't take
15039 implementations of variable permutation vectors into account
15040 when deciding which selector to use.
15041 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
15042 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
15043 with a false third argument.
15044 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
15045 to test whether the constant selector is valid and can_vec_perm_var_p
15046 to test whether a variable selector is valid.
15048 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15050 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
15051 * optabs-query.c (can_vec_perm_p): Likewise.
15052 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
15053 instead of vec_perm_indices.
15054 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
15055 (vect_gen_perm_mask_checked): Likewise,
15056 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
15057 (vect_gen_perm_mask_checked): Likewise,
15059 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
15061 * optabs-query.h (qimode_for_vec_perm): Declare.
15062 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
15063 (qimode_for_vec_perm): ...this new function.
15064 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
15066 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
15068 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
15069 does not have a conditional at the top.
15071 2018-01-02 Richard Biener <rguenther@suse.de>
15073 * ipa-inline.c (big_speedup_p): Fix expression.
15075 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15078 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
15081 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
15085 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
15086 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
15087 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
15088 cond_taken_branch_cost 3->4.
15090 2018-01-01 Jakub Jelinek <jakub@redhat.com>
15092 PR tree-optimization/83581
15093 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
15094 TODO_cleanup_cfg if any changes have been made.
15096 PR middle-end/83608
15097 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
15098 convert_modes if target mode has the right side, but different mode
15101 PR middle-end/83609
15102 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
15103 last argument when extracting from CONCAT. If either from_real or
15104 from_imag is NULL, use expansion through memory. If result is not
15105 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
15106 the parts directly to inner mode, if even that fails, use expansion
15109 PR middle-end/83623
15110 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
15111 check for bswap in mode rather than HImode and use that in expand_unop
15114 Copyright (C) 2018 Free Software Foundation, Inc.
15116 Copying and distribution of this file, with or without modification,
15117 are permitted in any medium without royalty provided the copyright
15118 notice and this notice are preserved.